Cdnj04820enc 001

Proceedings
of the
First International Conference on
STRUCTURAL
MECHANICS
IN REACTOR M S
TECHNOLOGY
Berlin, Germany 20-24 September 1971
Vol.
5 REACTOR PLANT STRUCTURES

A N D CONTAINMENT
Part K Seismic Response Analysis of Nuclear

Power Plant Systems
compiled by. Thomas A. Jaeger
Organized by:
Commission of the European Communities, Brussels

Bundesanstalt fr Materialprfung (BAM) , Berlin
in cooperation with.
FIRL
IKT
IPPT
ISD
NIS
NUS
SDK
Conference Organisation by:

Bundesanstalt fr Materialprfung (BAM), Berlin, Germany
in cooperation with:
The Franklin Institute Research Laboratories (FIRL), Philadelphia, Pa.
Institut fr Kerntechnik (IKT), Technische Universitt Berlin
Instytut Podstawowych Problemw Techniki (IPPT),
Polska Akademia Nauk. Warsaw
Institut fr Statik und Dynamik (ISD), Technische Universitt Stuttgart
Nuklear-Ingenieur-Service GmbH (NIS), Hanau
Nuclear Utilities Service (NUS), Rockville, Md.
Ingenieurunternehmen fr spezielle Statik, Dynamik und Konstruktion
Lrrach
Sponsoring Societies :
American Concrete Institute (ACI)
The American Society of Mechanical Engineers (ASME),
Nuclear Engineering Division
The American Nuclear Society (ANS)
Atomic Energy Society of Japan (AESJ)
The British Nuclear Energy Society (BNES)
Kerntechnische Gesellschaft (KTG),
incorporated in the Deutsches Atomforum e.V.
Schweizerische Vereinigung fr Atomenergie (SVA)
Vereinigung der Grosskesselbetreiber e.V. (VGB)
Executive Committee :
Thomas A. Jaeger, General Chairman
John H. Argyris, for ISD
Hartwig Benzler, for Commission of the European Communities
Klaus Brandes, for BAM
William J . Gallagher, for NIS
Charles F. Jones, for NUS
Hans H. Hofmann, for SDK
Antoni Sawczuk. for IPPT
Zenons Zudans, for FIRL
Representatives of Sponsoring Societies :
Yoshio Ando for AESJ
Ian Davidson, for BNES
Claus-Benedict von der Decken, for KTG
Nathan M. Newmark, for ACI
Karl Wirtz, for ANS
Otto Schaub. for SVA
K. E. Schroeter, for ANS Local Section in Central Europe
Ottmar Schwarz, for VGB
J. E. Zerbe, for ASME Nuclear Engineering Division
Patronage : Der Bundesminister fr Wirtschaft und Finanzen
General Support : Senate of the City of Berlin (West)
Editor : Commission of the European Communities
DG XIII
Center for Information and Documentation - CID
Luxembourg
(SDK),
Proceedings
of the
First International Conference on
STRUCTURAL
IN REACTOR
TECHNOLOGY
Berlin, Germany 20-24 September 1971
Vol.
5 REACTO R PLANT STRUCTURES

A N D CO NTAINMENT
Part K Seismic Response Analysis of Nuclear

Power Plant Systems
EUR 4820 d-ff-e
compiled by: Thomas A. Jaeger
Organized by.

Bundesanstalt fr Materialprfung (BAM) , Berlin
in cooperation with:
FIRL
IKT
IPPT
ISD
NIS
NUS
SDK
PREFACE
The purpose of the First International Conference on STRUCTURAL
MECHANICS IN REACTOR TECHNOLOGY was to bring together
engineers and scientists who are actively engaged in solving structural
mechanics problems in the field of reactor technology and fundamentalists
in the general field of engineering mechanics to present and discuss
applied and fundamental papers on structural mechanics problems
in reactor technology.
The meeting of more than 800 reactor technologists and engineering
mechanicians from 33 countries all over the world has brought together
a wealth of information and inspiration for the benefit of both reactor
technology and structural mechanics science.
All contributed original papers together with the discussions, as well
as the summaries of the invited lectures (designated by *), are contained
in the present Proceedings published by the Commission of the European
Communities. A survey of the topical grouping of the Proceedings
is given on the next pages. The detailed overall list of contents of the
Proceedings is given in Volume 1.
The 70 invited lectures have been published in special conference
issues of the journal NUCLEAR ENGINEERING AND DESIGN, NorthHolland Publishing Co., Amsterdam. The arrangement of contents
of these issues (Vol. 18, No. 1, 2, 3; Vol. 19, No. 1, 2; Vol. 20, No. 1, 2;
(1972)) is given in Volume 1 of the present Proceedings.
Thomas A. JAEGER,
General Chairman of the Conference
Berlin, March 1972
CONTENTS
Pages
Session K 1
Survey Lectures : Earthquake

and Aseismic Design
Response Analysis
Chairmen :
R.J. SCAVUZZO,
Department of Mechanical Engineering, Rensselaer Polytechnic
Institute of Connecticut, Hartford Graduate Center, East Windsor
Hill, Connecticut, U.S.A.
H. SHIBATA,
Institute of Industrial Science, University of Tokyo, Tokyo, Japan
K 1/1 *
Earthquake
Response Analysis of
Reactor
Structures
N.M. NEWMARK,
Department of Civil Engineering, University of Illinois, Urbana,
Illinois, U.S.A.
Discussion
K 1/2*
Evaluation of the Requirements of Nuclear Systems for

Accomodating Seismic Effects
T.W. PICKEL, Jr.,
Reactor Division, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, U.S.A,
Discussion
K 1/3*
Philosophy and Practice of the Aseismic Design of

Nuclear Power Plants: Summary of the Guidelines in
Japan
T. HISADA,
Kajima Institute of Construction Technology, Kajima Corporation,
Tokyo
K. AKINO,
The Japan Atomic Power Co., Tokyo
T. IWATA,
The Kansat Electric Power Co.
0. KAWAGUCHI,
Power and Nuclear Fuel Development
K. OMATSUZAWA,
The Tokyo Electric Power
Tokyo
Corporation
Co., Nuclear Power
Department,
H. SATO,
Institute of Industrial Science, University of Tokyo, Tokyo
Y. SONOBE,
Chiba Institute of Technology
H. TAJIMI,
Nihon University, Japan
Discussion
VI
Pages
K 1/4*
Problmes de sismes : techniques utilises pour les

racteurs nuclaires en France
D. COSTES et al..
Dpartement des Etudes de Piles, C.E.A., Centre d'Etudes
Nuclaires de Saclay, Gif-sur-Yvette, France
Discussion
Session K 2
11
12
Aseismic Design of Nuclear Power Plant Structures
Chairmen :
J.M. BIGGS,
Department of Civil Engineering, Massachusetts Institute of
Technology, Cambridge, Massachusetts, U.S.A.
N.M. NEWMARK,
Illinois, U.S.A.
K 2/1 *
K 2/2
K 2/3
K 2/4
The Earthquake Response Analysis for a BWR Nuclear

Power Plant Using Recorded Data
15
K. MUTO,
Muto Institute of Structural Mechanics, Inc., Tokyo
K. OMATSUZAWA,
The Tokyo Electric Power Company, Inc., Nuclear
Department, Tokyo, Japan
Discussion
16
Power
Strong Motion Earthquakes and their Effects on Nuclear

Power Plants
R.B. MATTHIESEN,
School of Engineering and Applied Science, University of
California, Los Angeles, California
C.B. SMITH,
Norman Engineering Co., Los Angeles, California, U.S.A.
Discussion
Earthquake Calculations : their Importance w i t h Respect
to Areas of Average and Low Seismic Activity and the
Application of Computer Oriented Methods
A.E. HUBER, P.O. SCHILDKNECHT,
SDK Ingenieurunternehmen fr spezielle Statik, Dynamik und
Konstruktion GmbH, Lrrach, Germany
Discussion
Aseismic Design of Structures w i t h Nuclear Reactors Method of Earthquake Response Analysis for Composite
Structures Evaluated for Damping Efficiencies by Material
and Structure Type
Y. TSUSHIMA, J. JIDO,
Takenaka Komuten Co. Ltd., Technical Research Laboratory,
Tokyo, Japan
Discussion
19
34
35
67
69
91
VII
Pages
K 2/5
K 2/6
K 2/7
Aseismic Design of Nuclear Reactor Building - Stress

Analysis and Stiffness Evaluation of the Entire Building
by the Finite Element Method
Y. TSUSHIMA, Y. HAYAMIZU, K. NISHIYAMA,
Takenaka Komuten Co. Ltd., Technical Research Laboratory,
Tokyo, Japan
Discussion
Aseismic Design for Japan Experimental Fast Reactor
(Joyo)
K. AKINO, M. KATO,
The Japan Atomic Power Company, Tokyo, Japan
Discussion
Berechnung der Erdbebenschwingungen von Strukturen
mit der Finite-Element-Methode - Mechanische Modelle
von Kernkraftwerken mit Einbauten
K. MARGUERRE, M. SCHALK, H. WLFEL,
Institut fr Mechanik, Technische Hochschule
Darmstadt,
Darmstadt, Germany
Discussion
Session K 3
K 3/2
K 3/3
108
109
122
123
139
Seismic Loading and Interaction Effects
Chairmen :
M. BENDER,
General Engineering Division, Oak Ridge National
Oak Ridge, Tennessee, U.S.A.
K. MARGUERRE,
Institut fr Mechanik, Technische Hochschule
Darmstadt, Germany
K 3/1
93
Laboratory,
Darmstadt,
Statistical Treatment of Seismic Loading on Reactor

Buildings and Equipments
A. AMIN, A.H.-S. ANG,
Illinois, U.S.A.
Discussion
The Response Spectrum Analysis to an Artificial
Earthquake w i t h T w o Ground Predominant Periods
H. SATO,
Discussion
The Influence of Seismic Pulse Time on Structure-Foundation Interaction
R.J. SCAVUZZO,
Department of Mechanical Engineering, Rensselaer Polytechnic
Institute, Hartford Graduate Center, East Windsor Hill, Connecticut
R.R. LITTLE,
Department of Mechanical Engineering, The University of Toledo,
Toledo, Ohio, U.S.A.
Discussion
141
154
155
173
175
193
VIII
Pages
K 3/4
Analysis of Soil-Structure Interaction Effects under

Seismic Excitation
C.J. COSTANTINO,
School of Engineering, The City College of the City University
of New York, New York, U.S.A.
K 3/5
Soil-Foundation Interaction of Reactor Structures Subject

to Seismic Excitation
T.H. LEE, D.A. WESLEY,
Gulf General Atomic, San Diego, California, U.S.A.
Discussion
K 3/6
K 3/7
Dynamic Calculations Using a Framework Analogy to

Predict the Seismic Response of a Nuclear Reactor
D.A. JOBSON,
United Kingdom Atomic Energy Authority, Reactor Group, Risley,
Warrington, United Kingdom
Discussion
Parametric Analysis of Soil-Structure Interaction for a
Reactor Building
R.V. WHITMAN, J.T. CHRISTIAN, J.M. BIGGS,
Discussion
Session K 4
195
211
232
235
256
257
278
Aseismic Design of Nuclear Power Plant Piping

and Equipment
Chairmen :
K. AKINO,
C.B. SMITH,
School of Engineering and Applied Science, University of California, Los Angeles, California, U.S.A.
K 4/1 *
Development of Aseismic Design of Pipings Vessels and

Equipments in Nuclear Facilities
H. SHI BATA, A. WATARI, H. SATO, T. SHIGETA,
Institute of Industrial Science, University of Tokyo, Tokyo
A. OKUMURA,
Faculty of Science and Engineering, Waseda University,
S. FUJII, M. IGUCHI,
Faculty of Engineering, University of Tokyo, Tokyo, Japan
Discussion
K 4/2
Seismic Design Coefficients of Equipment in Nuclear

Power Plants
C.-W. LIN,
Westinghouse Electric Corporation, Power Systems, Pittsburgh,
Pennsylvania, U.S.A.
Discussion
281
282
283
298
IX
Pages
K 4/3
Paper deleted
K 4/4
Aseismic Design of Asymmetric Structures and

Equipment Contained
CH. CHEN,
Gilbert Associates Inc., Reading, Pennsylvania, U.S.A.
the
Discussion
K 4/5
Paper deleted
K 4/6
Dynamic Analysis of Vital Piping Systems Subjected to

Seismic Motion
CH. CHEN,
Discussion
K 4/7
Seismic Response Spectra for Equipment Design in

Nuclear Power Plants
J.M. BIGGS,
Discussion
299
317
319
328
329
343
Topical Grouping of the Proceedings

of the First International Conference on
Structural Mecanics in Reactor Technology
Vol. 1 .
SURVEY OF THE CONFERENCE: REACTOR TECHNOLOGY

Prefaces : Opening Address
Topical Scope of the Conference
Part A.
General Lectures
On the Dissemination of Scientific Information
Power Reactor Development Strategies
Part B.
Power
Reactor
Mechanics
BO/1 *
BO/2 *
BO/3 *
BO/4*
Vol. 2.
Development
and
Structural
Pressurized Water Reactor

Development and Its
Mechanical-Structural Requirements and Problems
Boiling Water Reactor Development and Its Mechanical
Structural Requirements and Problems
High-Temperature Gas-Cooled Reactor Development and
Its Mechanical-Structural Requirements and Problems
Liquid-Metal Cooled Fast Breeder Reactor Development
and Its Mechanical-Structural Requirements and
Problems
Overall List of Contents of the Proceedings
About the Organizing Institutions
List of Conference Participants
REACTOR CORE
Part C.
Fuel and Cladding
C1
C2
C3
C4
Part D.
D
D
D
D
Mathematical-Physical Models for Nuclear Fuels Materials

Stress and Deformation Analysis of Fuel Element Claddings
Fuel-Cladding Interaction Effects
Structural Analysis of Fuel Rods
Fuel Elements
1
2
3
4
Structural Analysis of Fuel Elements

Stress Analysis of Graphite Fuel/Moderator Elements I
Stress Analysis of Graphite Fuel/Moderator Elements II
Pebble Bed Reactor Mechanics
XII
Vol. 3.
REACTOR
COMPONENTS
Part E.
Shock and Vibration Analysis of Reactor Components
E1
E2
E3
E4
Part F.
F
F
F
F
F
Vol. 4.
Thermal Shock, Pressure Pulse, and Impact Response

Analysis
Dynamics of Fast Reactor Excursion and Containment
Fuel Rod Vibrations in Parallel Flow
Reactor Component Vibrations
Structural Analysis of, Core Support
Circuit Structures
1
2
3
4
5
Structural
Structural
Structural
Structural
Structural
and Coolant
Analysis of Reactor Core Support Structures

Analysis of Miscellaneous Reactor Components
Analysis of Coolant Circuit Components I
Analysis of Coolant Circuit Components II
Analysis of Coolant Circuit Components III
REACTOR PRESSURE VESSELS

Part G.
G1
G2
G3
G4 *
G5
G6
G7
Part H.
Steel Pressure Vessels

Stress Analysis of Steel Reactor Pressure Vessels: Shells
of Revolution
Stress Analysis of Steel Reactor Pressure Vessels: Shell
Intersections
Stress Analysis of Steel Reactor Pressure Vessels: Vessel
Flanges
Survey Lectures: Fracture Mechanics Analysis of Steel
Reactor Pressure Vessels and Piping
Fracture Mechanics Analysis of Steel Reactor Pressure
Vessels: Materials Behaviour
Fracture Mechanics Analysis of Steel Reactor Pressure
Vessels: Methods of Analysis
Crack Detection Measurements
Prestressed Concrete Pressure Vessels
H 1
Mathematical-Physical Characterization of Concrete I
H2
Mathematical-Physical Characterization of Concrete II
H 3
Methods for Structural Analysis of Prestressed Concrete

Reactor Pressure Vessels
Design and Structural Analysis of Prestressed Concrete
Reactor Pressure Vessels
Comparison of PCPV Model Test Results with Theory
Structural Analysis of PCPV Liner and Insulation
H4
H 5
H 6
XIII
Vol. 6.
REACTOR PLANT STRUCTURES A N D CONTAINMENT

Part J.
J
J
J
J
Analysis of Shell Structures; Containment

1
2
3
4
Part K.
Seismic Response Analysis of Nuclear Power Plant

Systems
1*
2
3
4
Vol. 6.
Analysis of Thin-Shell Structures I

Analysis of Thin-Shell Structures II
Containment of Power Reactor Plants I
Containment of Power Reactor Plants II
Survey Lectures: Earthquake Response Analysis and

Aseismic Design
Aseismic Design of Nuclear Power Plant Structures
Seismic Loading and Interaction Effects
Aseismic Design of Nuclear Power Plant Piping and
Equipment
STRUCTURAL ANALYSIS A N D
Part L.
L
L
L
L
L
L
DESIGN
Thermal and Mechanical Analysis

1*
2
3
4
5
6
Part M .
M 1*
M2*
M3*
M4
M5
M6
M 7
Survey Lectures: Solid Mechanics Theory

Heat Generation and Conduction Analysis
Inelastic High-Temperature Behaviour of Metals
Inelastic Analysis of Metal Structures
Low-Cycle Fatigue and Shakedown Analysis
Fracture Mechanics: Special Topics
Design, Reliability, Computation Methods
Survey Lectures: Failure Experience, Reliability
Survey Lectures: Computer Methods, Mathematical
Modeling, Education
Survey Lectures; Applied Mechanics and Design
Reliability Analysis of Mechanical Reactor Components
Finite Element Methods for Structural Analysis I
Finite Element Methods for Structural Analysis II
Mathematical Models and Methods
K 1/1*
EARTHQUAKE RESPONSE ANALYSIS OF REACTOR STRUCTURES

N.M. NEWMARK,
Department of Civil Engineering,
University of Illinois, Urbana, Illinois, U.S.A.
The general nature of the principles upon which earthquake resistant

design is based is described with particular reference to components
and elements of nuclear reactor facilities. Special attention is
paid to the response and design criteria of items of equipment or
of components that are mounted on or attached to responding elements,
and basic procedures are developed to bound the dynamic response of
such items.
Consideration is given to vertical as well as horizontal excitation,
and the combination of the effects of the various excitations. Suitable approximations are developed for inelastic response estimates.
One section of the paper is devoted to relative motions of points
some distance apart, and to bounds for such relative motions.
Recommendations are made for the general criteria governing the
design of nuclear facilities, including the basic parameters governing response characteristics and energy absorption.
' Published in Nuclear Engineering and Design 20 (1972).
DISCUSSION
R. SCHNEIDER, Germany
Your r e s e a r c h is mainly based on acceleration r e c o r d by El-Centro-Earthquake
(magnitude M 7.0). In Central Europe the maximum magnitude is about M = 6-6.5. What
about the s i m i l a r i t y between ground and design s p e c t r a between a M s: 6.0-earthquake and
the 7.0 magnitude earthquake ? Can one use the design s p e c t r a cited in your lecture also for
Central Europe maximum earthquakes just by multiplying the s p e c t r a by diminishing factors ?
N. M. NEWMARK, U. S. A.
F i r s t , my r e s e a r c h is not based mainly on the E l - C e n t r o 1940 record, but on the
r e c o r d s of many other earthquakes and of motions due to blasting and impact as well. The
general conclusions a r e applicable, r e g a r d l e s s of earthquake magnitude. I would recommend
that a ground velocity of 1 5 c m / s e c or in. / s e c be used as a minimum design value, and
this c o r r e s p o n d s to about 0.1 g acceleration. The use of anything less would in my opinion
be distinctly unsafe. I believe strongly that the frequency range for design for earthquakes
of any intensity, on moderately firm ground, should be s i m i l a r to the design spectrum range
shown in my paper. This needs modification by a slight shift to higher velocities for a given
acceleration for very soft soils, and a slight shift to higher accelerations but unchanges
velocities for hard rock.
Ch. CHEN, U. S. A.
1. How should the response from vertical response spectrum and horizontal response spect r u m be combined ? It seems that the absolute addition is too conservative.
2. Since there is no normal modes in plastic region for multidegree of freedom system, is
there any way to apply the inelastic single degree of freedom spectrum to multidegree of
freedom s y s t e m in inelastic region ?
J^
N.M. NEWMARK, U.S.A.
1. The addition of the absolute values of response to vertical and horizontal excitation is
much too conservative. I have recommended in e a r l i e r papers that the individual responses
in the two horizontal directions and the vertical direction be combined by taking the square
root of the sums of the squares of the component responses.
2. Since there a r e no modes in the responses in the non-linear range, for the plastic region
the modified response s p e c t r a 1 have described can be used only as an approximation for
m u l t i - d e g r e e of freedom s y s t e m s . With some judgment in the selection of the modes of the
elastic s y s t e m that have the major inelastic response, a reasonably good approximation can
be developed for use in design, however.
- 3 -
H. SATO, Japan
We have not been successful to know acceleration and displacement simultaneous-
ly by instrumentation. Displacement motion might be l a r g e r than that we know at the moment

when it contains longer period beyond the capability of the a c c e l e r o m e t e r . Would you please
make any comment on the accuracy of the proposed maximum value of displacement ?
.
A
In general large displacements a r e not controlling factors in the design of nuclear
power plant facilities and components except possibly for very long period e l e m e n t s , with
periods longer than 5 seconds. However, even for these, one can use the maximum ground
velocity as a m e a s u r e to obtain a conservative estimate of the r e s p o n s e . In the case of fault
motions at the surface near the epicenter, values have been r e c o r d e d as much as 20 ft for
very large earthquakes, but the ground velocity associated with them is v e r y much less than
the maximum ground velocity. Hence it is not a serious m a t t e r that we do not have good r e c ords of maximum transient ground displacement if we have r e c o r d s of maximum ground v e locity one can cipher these values from m e a s u r e m e n t s of maximum ground acceleration.
H. SHIBATA, Japan
Dominant periods of liquid surface in torus of BWR type r e a c t o r s might be m o r e
than 15 seconds. So the estimation of displacement of such long period ground motion s e e m s
to be also important for actual design.
The importance of the response of the liquid in the torus is great, but the r e s p o n s e
is affected also by ground velocity as well as displacement. The comments made in my reply
to Dr. Sato a r e also applicable in this c a s e .
K 1/2*
EVALUATION OF THE REQUIREMENTS OF NUCLEAR SYSTEMS

FOR ACCOMODATING SEISMIC EFFECTS
T.W. PICKEL, Jr.,
Reactor Division,
Oak Ridge National Laboratory, Oak Ridge, Tennessee,
U.S.A.
A logical approach to the evaluation of nuclear .system requirements for accomodating seismic effects, is presented in this paper.. This approach involves selection of the type of analysis best suited to the circumstances of
a specific nuclear system, selection of a method of solution for the equations resulting from the analysis method selected, development of a suitable
mathematical model wherein the component parts of the system are adequately
defined, and definition of the interactions of forces among these components
of the system.
Engineering approaches to seismic design analysis are based on lumped-mass
modeling techniques. However, in some situations such as soil-structure interactions, finite element methods have value. The structural model most frequently used is one based on simple beam theory, but the membrane shell theory is more suitable for some purposes. Some criteria for judging the applicability of each of these methods to specific circumstances are presented
in this paper.
Three methods have been used to solve the equations which result from the
lumped-mass model. These methods are direct integration, response spectrum
modal analysis, and time-history modal analysis- The principles are well
established, and the current work with these methods is directed primarily
toward improving computational-techniques. The advantages and disadvantages
of each of these methods are compared, and their use is illustrated in this
paper.
An important problem in seismic analysis that needs further consideration is
selecting the mathematical model for the system. This problem is common to
each of the methods of solution previously mentioned. Because of their complexity, nuclear systems must be divided into parts and seismic analyses made
of each part. Interactions among parts of the system must be considered. Major
points of concern are the selection of the degrees of freedom to be considered,
selection of parts of the system that can be assumed to be decoupled or eliminated by condensation, and the amount of damping that exists in the system.
The considerations involved in defining a mathematical model for nuclear
systems are discussed in this paper.
The factors affecting transmission of vibratory forces through a nuclear system include natural vibrational characteristics, damping phenomena, and
vibrational control mechanisms. As the effects of vibratory ground motion are
transmitted through a nuclear system, each part of the system acts as a filter
to predominately transmit selected frequencies to subsequent parts of the
system. This transmission of selected frequencies magnifies the response of
the subsequent parts which have natural frequencies close to those selectively transmitted. The transmission of forces through a system can be limited
or altered by the use of appropriate design factors such as damping, stiffness of connecting members, location of components, and vibrational control
devices. The use of these design factors is discussed in this paper.
Published in Nuclear Engineering and Design 20 (1972).
- 6
DISCUSSION
M. S. RAO, India
1. What would happen when the epicentre of the earthquake coincides with the reactor site ?
2. In India there was an earthquake at Koyna dam Site about 5 y e a r s ago which was suspected
to be due to the impounding of a large m a s s of water. Some of the nuclear stations in India
a r e being planned near large hydro power stations to operate as base load plants. This question
has particular relevance to this thinking.
M. BENDER, U.S.A.
1. This is so unlikely that it should be ignored for evaluation purposes.

2. The triggering of earthquakes is not well understood but surface effects from water p r e s
sure a r e unlikely to have significance. High p r e s s u r e injected into the substructure might be
of significance as indicated by effects of injection wells in Colorado.
R . J . SCAVUZZO, U.S.A.
Seismic snubbers, which allow for thermal deflection of piping but support piping
during seismic loads, a r e being considered in some designs. Could you comment on the use
of these components ?
M. BENDER, U. S. A.
Snubbers a r e an essential device to decouple the large s t r u c t u r e s dynamically
by providing an energy absorption means of dampening. The effects of t h e r m a l expansion
have to be considered and this usually can be worked out because t h e r m a l effects a r e shown
acting and can be mechanically managed in the snubber system.
H. SHIBATA, Japan
Could you give me an idea how to describe the design c r i t e r i a of puinps and valves
in your code ?
M. BENDER, U. S. A.
With respect to pumps, valves and other dynamic equipment, consider functional
capability and r e q u i r e m e n t s . For pumps usually shaft deflection, seal response and impeller
rubbing a r e pertinent effects. For valves it is usually the question of operator response,
packing gland r e s t r a i n t , and seat alignment that need study.
K 1/3*
PHILOSOPHY AND PRACTICE OF THE ASEISMIC DESIGN

OF NUCLEAR POWER PLANTS :
SUMMARY OF THE GUIDELINES IN JAPAN
T. HISADA,
Kajima Institute of Construction Technology, Kajima Corporation, Tokyo,
K. AKINO,
The Japan Atomic Power Co., Tokyo,
T. IWATA,
The Kansai Electric Power Co.,
O. KAWAGUCHI,
Power and Nuclear Fuel Development Corporation,
K. OMATSUZAWA,
The Tokyo Electric Power Co., Nuclear Power Department, Tokyo,
H. SATO,
Institute of Industrial Science, University of Tokyo, Tokyo,
Y. SONOBE,
Chiba Institute of Technology,
H. TAJIMI,
Nihon University, Japan
This report is summarizing guidelines for aseismic design of nuclear power

plants. The guide-line was prepared by the Japan Electric Association in
April, 1970, after laborious three year's works.
In ohapter
1, philosophy and criteria are described. All components in plant
should be classified into three classes in accordance with their contributions
to reactor safeties. Design to earthquake loadings should be based on "Design
Basis Earthquake" which is decided in consideration of local seismisity.
In chapter
2, site selection and review for ground are described in sense of
seismic aspect.
(
In chapter
3, it is discussed how to decide on the earthquake motion for de
sign. In Japan, semi-statistical approaches are used in usual practice.
In chapter
1, design philosophy and practice of building structures and con
tainment vessels are described. They are designed under statical seismic
forces, and the design of the class "A" structures should be checked by dy
namic response technique.
In chapter
5, design philosophy and practice of piping, vessels and equip
ment are described. Those which belong to class "A" items should be designed
in dynamic sense. Several programs for dynamic analyses of these items are
prepared. Allowable stress under earthquake conditions is discussed in re
lation to other codes, for example, ASME Section III.
The most part of the philosophy and design criteria have been adopted to all
nuclear power plants which were built and being built in Japan.
Published in Nuclear Engineering and Design 20 (.72).
DISCUSSION
H. BARNERT, Germany
Is there any activity in the field of forecast of earthquakes in Japan ? Work on
these questions might be important if design c r i t e r i a turn out to be too hard, for example
for the shutdown system.
K. OMATSUZAWA, Japan
Yes. An organisation has been established to study how to forecast the earthquake
The activity of this organisation is as follows:
1. Accurate m e a s u r e m e n t of level of ground surface in a s h o r t e r period.
2. Measurement of m i c r o - e a r t h q u a k e near the a r e a where large earthquake is expected in
comparatively near future.
3. Accurate m e a s u r e m e n t of horizontal movement of the ground by l a s e r beam.
At p r e s e n t , however, it is not p r a c t i c a l to forecast the large earthquake before a few minutes or seconds for shutting down the r e a c t o r , although under consideration in other field
of engineering seismology.
H. WOLFEL, Germany
When performing a time history analysis,which time duration of the earthquake
and which time histories do you use ?

Time duration of input earthquake wave for dynamic analysis depends on the
selected waves. At the very beginning of analysis of a certain s t r u c t u r e , we use full r e c o r d
of earthquake waves, and then we will find the time when the maximum response will occur.
After that we use the waves a little longer than the time of maximum response.
H. WOLFEL, Germany
For determining the s t r e s s e s of the building do you take the acceleration multi-
plied with the m a s s distribution as dynamic loading or do you take the forces of your m e chanical model ?
For determining the s t r e s s for member design, we use the force, such as shear
force and bending moment, derived from the time history analysis. Acceleration of the
building is only used for determining the s e i s m i c loading of equipment and piping installed
at each point of r e a c t o r building.
L. ESTEVA, Mexico
You mentioned some numbers by which you multiply the l a t e r a l force
coefficients
of ordinary s t r u c t u r e s in order to obtain the lateral force coefficients applicable to the

various types of s t r u c t u r a l elements of nuclear r e a c t o r s . On the other hand, you also showed
charts of ground velocities and accelerations for given r e t u r n periods. Have those coefficients
been derived from the mentioned chart ? If so, in what manner ? What is the quantitative
justification of those numbers ?
In the design of "A" class i t e m s , l a t e r a l forces a r e determined by both static and

dynamic analysis. And the m e m b e r s of s t r u c t u r e s a r e also determined by the l a r g e r forces.
Static seismic coefficient is based on the building standard law of Japan, and the dynamic
analysis is more scientifically determined by the local seismicity.
So, the earthquake forces in a low seismicity a r e a , being determined by the static seismic
coefficient (factor of 3), a r e controlling the design, and on the c o n t r a r y , in a high seismicity
a r e a by the dynamic response analysis.
N.N. KULKARNI, India

In some reactor buildings very heavy equipment like p r e s s u r e v e s s e l and s t e a m
raising units a r e directly supported on foundations or floors which a r e rigid.

But there a r e r e a c t o r buildings like CANDU r e a c t o r s in which the equipment is suspended in
the form of a pendulum type support.
How do these s y s t e m s compare from earthquake response point of view ? What will be your
recommendation to support and locate such load ?
The way to minimize the response of acceleration, or of force, for the equipment
in a r e a c t o r building during an earthquake is as follows:
1. to lower the input acceleration,
2. to prevent the resonant phenomenon,
3. to increase damping.
To support the equipment at a higher level as you asked is not good for item 1. In this c a s e ,
however, if the frequency is much lower than the supporting s t r u c t u r e because of its pendulum system, then the responsive acceleration or force of such items will become low, while
the relative displacement will i n c r e a s e .
D. LUTOSCH, Germany
You mentioned the importance of damping. What a r e the damping factors you a r e
considering in Japan in your nuclear power plants ?
10 -
The damping factors used for our design purposes of r e a c t o r facilities a r e as
follows :
Reinforced concrete
Welded steel s t r u c t u r e
Riveted or bolted steel s t r u c t u r e

Vital piping s y s t e m
2
0. 5
Control rod drive m e c h a n i s m
3. 5
Fuel a s s e m b l i e s in water
7. 0
K 1/4*
PROBLMES
DE SISMES
: TECHNIQUES
POUR LES RACTEURS NUCLAIRES EN
UTILISES
FRANCE
D. COSTES et al.,
Dpartement des Etudes de Piles,
C.E.A.,Centre d'Etudes Nuclaires de Saclay, Gif-sur-Yvette, France
Les tudes de comportement des structures sous sismes comportent les postes
suivants:
- dfinition du niveau d'intensit du sisme prendre en compte, avec
analyse de probabilits,
- schmatisation des mouvements du sol: accelrogrammes rels ou algbriques, analyse de bruit,
- calcul des rponses de structures en calcul lastique,
- intervention de la plasticit et des dgradations; programme de calcul
correspondants,
- rglementation.
La communication fera le point des techniques utilises pour les racteurs
nuclaires en France.
* Published in Nudar
Engineering and Design 20 (1972).
12
DISCUSSION
R. SC HNEIDER, Germany
Pour vos projets en F r a n c e e s t c e que vous avez utilis pour la dtermination
de vos s p e c t r e s seulement des chocs a m r i c a i n s comme " E l C e n t r o " ? Quelles acclrations

avezvous choisi ?
D. C OSTES, F r a n c e
Le s i s m e algbrique a t dtermin essentiellement d'aprs des s i s m e s a m
r i c a i n s , parce que les e n r e g i s t r e m e n t s franais correspondaient seulement des m i c r o
s i s m e s , beaucoup moins riches en frquences diverses qu'un grand s i s m e .
Les intensits nominales choisies pour les r a c t e u r s franais sont de 7 ou 8 selon les rgions;
les s i s m e s majors s'en dduisent.
D. LUTOSC H, Germany
How does the "intensit nominale" correspond to the acceleration of the ground ?
The correlation between intensity and maximal acceleration is dubious. I took in
my paper the relation :
A [ m / s 2 j = 3 21"
o
A
Q
H. SHIBATA, Japan
Hew many major earthquakes have you had for the last one hundred years ?
I may give the following figure :
1 magnitude 6.25 between 1901 and 19SS.
H. SATO, Japan
1. Does the algebraic earthquake or the simulated earthquake include the natural period of
ground ?
2. How did you control the duration of the simulated earthquake ?
^
D. C OSTES, France
1. Le s i s m e algbrique est centr sur une frquence fondamentale obtenue par le jeu d'un
paramtre.
13
J ' a i montr l'adaptation au s i s m e d'El-Centro. Pour une collection importante de s i s m e s ,

on peut trouver une frquence fondamentale moyenne un peu modifie.
2.
L'application du s i s m e algbrique une collection de r s o n a t e u r s montre le temps au
bout duquel on obtient la rponse maximale de chaque mode, c a r a c t r i s comme un r s o n a teur.
K 2/1*
THE EARTHQUAKE RESPONSE ANALYSIS

FOR A BWR NUCLEAR POWER PLANT USING RECORDED DATA
K. MUTO,
Muto Institute of Structural Mechanics, inc., Tokyo,
K. OMATSUZAWA,
The Tokyo Electric Power Company, Inc.,
Nuclear Power Department, Tokyo, Japan
The aim of this paper is to analyze the results of observed vibrations caused
by a vibrator and an actual earthquake in various parts of a BWR nuclear
power plant, and then to draw conclusions as to that plant's earthquake vibration character.
To conduct an earthquake response analysis, the most important thing is to
evaluate such input data as masses, rigidities, damping factors, etc. for
each part of the vibration model. The reasons why nuclear power plants are
considered to have among the most complicated of vibration systems are 1) that
they consist of both a reinforced concrete building and a shielding wall where
can be found ite! containment, eouinment, pioina and instruments; and 2)
that they are supported by a foundation with a large damping.
From this point of view, it therefore becomes necessary to consider different
damping factors for each structural part of the plant. During the past several
years, we have been engaged in the development of computer programs which in
conjunction with observed earthquake and vibration test data have then been
used for analysis of the vibration model. Fortunately, in November of 1969,
we were able to perform a forced vibration test on the Tokyo Electric Power
Company's nuclear power plant located in Northern Japan's Fukushima prefecture. As a result, we were able to arrive at that plant's vibration character
and then to evaluate the rigidities and damping factors in foundation, building, roof, etc..
After that, sets of seismographs were installed in various parts of this
plant. On May 26, 1970, the acceleration waves from an earthquake with
Fukushima offshore epicenter were thereby recorded on each floor, in the
foundation, roof, containment, circular wall, etc.. We then applied the resulting data to the above-mentioned computer programs and were able to determine to what extent our method of analysis had been correct.
16 -
DISCUSSION
J. P. LAFAILLE, Belgium
When studying the response of a building with a 2-dimensional model it is impos
sible to detect any response of the building in a direction different from the excitation. This
coupling effect would occur if the excitation did not occur in a principal direction of inertia.
Were there checks made to verify that there were not such effects ?
K. MUTO, Japan
The earthquake in May 1970 had ground motion in both NS and EW directions, and
coupled effects would have o c c u r r e d in the building. Our m e a s u r e m e n t was made in NS d i r e c
tion only because of economy. Analysis was then limited in this direction. But the fairly
accurate coincidence of theory and m e a s u r e m e n t was taken.
In future, for a three-dimensional approach, a lot of instruments and m e a s u r e m e n t s would
be desirable as well as relevant analysis.
C. B. SMITH, U. S. A.
Could you tell us about the experimental methods used in your work ?
K. MUTO, Japan
A large vibration machine was installed on the 5th floor of the reactor building.
The displacement was m e a s u r e d at many points of the building at every small step of frequen
cy, then the response curve was taken due to the sinusoidal excitations.
Ch. CHEN, U. S. A.
The elegant complex eigenvalue problem was used to consider different damping
ratios in the same model. The engineers always have the tendency to use a simplified method
so long as the difference is within engineering tolerance. 1 wonder if the authors compared
these r e s u l t s with other simplified r e s u l t s based on weighted damping ratio. It seems that
there is some basic agreement between these two approaches as shown on the slides, namely
modes 1 and 7 have higher damping coefficients.
K. MUTO, Japan
In the 1st, 4th and 7th modes, the movements of soil are very pronounced. On the
basis of the analysis, we consider a large damping coefficient of 0.04 sec. for the soil part
of the vibration model. So it may be natural that the mode in which the soil movement is
pronounced has a large damping factor.
- 17
A. H. HADJIAN, U. S. .
Viscous damping is a equivalence and generally determined by tests of actual

s t r u c t u r e s . How a r e then the damping coefficients
coefficients 'V. determined and why was the damping
matrix made proportional to the stiffness matrix ?
K. MUTO, Japan
A
The damping coefficient V of the vibration element is computed from both the
natural period (T) and the damping factor (h) which a r e usually m e a s u r e d by the vibration
test. In case of one m a s s system, the V may be e x p r e s s e d as follows:
V = Th/K
As for the definition of V , refer the equation (1):
{Fli=[B].|v). +
[vB].
W i
in which
|Fj
= external force vector
IB] = stiffness matrix of the element

(v) = displacement vector.
K 2/2
STRONG MOTION EARTHQUAKES AND THEIR EFFECTS

ON NUCLEAR POWER PLANTS
R.B. MATTHIESEN,
School of Engineering and Applied Science,
University of California, Los Angeles, California,
C.B. SMITH,
Norman Engineering Co., Los Angeles, California, U.S.A.
ABSTRACT
For the past four years, UCLA has been studying the effects of earthquakes on nuclear power plants. Structural vibrators, hydraulic rams, and explosive blasts have been used to excite reactor structures and equipment.
Many of the systems we have tested respond nonlinearly at high force levels. Since linear elastic theory is commonly employed to predict earthquake
response, knowledge of nonlinear behavior at high force levels is an important
aspect of reactor safety studies. Experimental data on full-scale systems is
limited due to the difficulty of inducing high level response.
We have tried two methods for exciting reactor equipment to levels of response approaching strong motion earthquakes. One method uses a .hydraulic ram
to produce a large static displacement of equipment such as a steam generator
or a pipe. When the displacing force is suddenly removed, the equipment undergoes large amplitude free vibrations.
Another technique uses explosives placed in the soil adjacent to the reactor containment building. Tests have been performed with up to one ton of
high explosive detonated within one hundred meters of a reactor containment
building.
The 9 February 1971 San Fernando earthquake (M = 6.5) tripped seismic instrumentation at the UCLA and San Onofre reactors. Since tests have been made
at both reactors, we now have an opportunity to compare vibration test results,
analytical results, and response measured during the earthquake.
1. INTRODUCTION
During the past four years, UCLA has been conducting research to study
the effects of earthquakes on nuclear power plants. In addition to theoretical and model studies, we have performed experimental tests at five reactors:
the UCLA research reactor, Los Angeles, California; the Experimental GasCooled Reactor, Oak Ridge National Laboratory; the Carolinas-Virginia Tube Reactor, Parr, South Carolina; the Enrico Fermi Fast Breeder Reactor, Monroe,
Michigan; and the San Onofre Nuclear Generating Station, San Onofre, Califor-
nia (see Matthiesen and Smith [1], [2], [3]; Bleiweis, Hart, and Smith [4];
Ibanez, Matthiesen, Smith, and Wang [5] ) . The tests have used structural vibrators, hydraulic rams, and explosive blasts to excite structures and equipment.
The results of these tests have been used to develop mathematical models
which are considered valid for the level of response in the tests when they
reproduce the experimental data. The mathematical models are then used to
predict the response of the system to various digitized earthquake records.
The experimental tests have provided new information concerning the dynamic
properties of the large structures and components used in nuclear power plants.
Data have been obtained on reactor containment buildings, stacks, water towers,
steam generators, pressure vessels, cores, primary coolant pumps, pressurizers,
and other items. In the few cases where information has been available, we
have compared experimentally determined parameters (natural frequencies, damping, effective masses, and mode shapes) with theoretical studies published by
others.
A number of techniques have been used to excite the structures tested.
These will be listed in order of increasing force capability. Ambient vibrations are caused by low level forces such as wind motion and natural ground
vibrations. These can be detected with sensitive instruments and can be used
to measure the dynamic properties of structures. Comparative studies have
shown that a structure may respond differently when excited by ambient vibrations than when excited by forced vibrations (see Schmitt [6] ) . Nevertheless,
this method is useful in some cases and can be used for preliminary surveys
in advance of high force level tests.
At the intermediate force level we use structural vibrators to cause motion of buildings and equipment (Figure 1 ) . We have used two types of vibrators. The large units operate in the frequency range up to about 9.5 llz. The
small units operate up to about 55 liz. The force output can be varied in the
large units by adjusting the weights placed in rotating "baskets". The maximum force output for the two large machines is 5000 lbs each. The force output of the small vibrators is varied by adjusting an eccentric rotor. They
have a maximum force output of 10,000 lbs each. Either set of machines can be
operated by control consoles and drive units which include servo-operated
feedback control systems for maintaining frequency constant at a given setting
within O.lli (0-10 Hz) or 0.61 (0-60 llz).
Dynamic response is measured using two types of accelerometers. One type
uses a four-arm strain gage bridge as the sensitive element. The characteristics of these devices have been published by Ibanez, et. al. [5]. The other
accelerometer is a servo type with a flat frequency response to 500 llz. The
accelerometers are mounted on the structure at appropriate locations and the
signals are recorded with strip chart recorders. The strip chart records are
processed manually and punched onto cards before being analyzed using a digital computer.
Two methods have been developed for exciting reactor equipment to levels
- 21 of response approaching strong motion earthquakes. In one method we use a hydraulic ram to produce a large static displacement of equipment such as a
steam generator or primary coolant pipe. When the displacing force is suddenly removed, the equipment undergoes large amplitude free vibrations.
Another technique makes use of explosive charges placed in the soil adjacent to the reactor containment building. The explosives have been placed in
bore holes located at distances from 100 to 1000 feet from the containment
building (Figure 2 ) . Tests have been performed with varying quantities of
high explosive, ranging from one pound up to 2000 pounds (Figure 3 ) . Plans
are being made for additional tests using 25,000 pounds of explosive. The results of theSie tests indicate that explosive blasts are a useful tool for dynamic testing. Excellent agreement has been obtained in comparisons of forced
vibration test data and blast data.
We use the same type of instrumentation for recording the response due to
the blasts. The duration of the blast excitation is shorter, so in addition
to the strip chart recorders we employ FM magnetic tape recorders (Figure 4 ) .
The electronic records can be processed automatically, first using a subroutine to digitize them and then analyzing the digitized records using Fourier
analysis programs. The digitized data are convenient in that other manipulations can be performed; for example, filtering of the records is sometimes
useful to improve the signal-to-noise ratio.
2. NONLINEAR RESPONSE
There are several deficiencies in the simulation methods discussed above.
The first deficiency is that site conditions and soil-structure interaction
play a part in the response of a nuclear power plant to strong motion earthquakes. We do not model these effects at the present time, except in the
sense that the soil was present and influenced the structure during the time
the test data were acquired. If we assume that the influence of the soil on
the structure is the same whether the excitation comes from the soil (as in an
earthquake) or from the structure (as in our forced vibration tests), then the
interaction effect has indeed been simulated. In the blast tests, the excitation does come through the soil, so if there is a soil-structure interaction,
its effect is included. However, there are differences in the frequency content, duration, and direction of propagation of the blast waves when compared
to earthquake waves, and this should be considered in interpreting test results .
Our mathematical models have been subjected to representative digitized
earthquake records, but these have always been ground motion records recorded
at other sites where there were no reactor-like structures. Future work is
planned to study soil-structure interaction, both theoretically and experimentally, so ground motion records can be coupled to the structure in a more
meaningful way. This work will include a study of the variation of dynamic
soil characteristics when subjected to large stresses such as those produced
by strong motion earthquakes.
- 22 The second deficiency in the simulation methods is that the frequency

spectrum of the exciting force is different from strong motion earthquakes.
The ambient methods use "broad band noise" to excite the system. The forced
vibration tests excite the system with energy at a single frequency which is
incrementally varied over the range of interest. The impulse tests excite the
system with a combination of high frequency motion and motion at the natural
frequency of the excited component. The blast test produces a "fairly narrow
band" (0-100 Hz) excitation, but the duration of ground motion is short compared to earthquakes and the dominant frequencies in the spectrum are generally higher than the dominant frequencies in an earthquake.
The third deficiency concerns the level of excitation used in the simulation tests. Typically, the equipment and structural response we measured in
the structural vibration tests is one-hundredth to one-thousandth of the value
that would result from a strong motion earthquake. We have observed in our
work that reactor structures respond differently under ambient vibrations
(one-millionth to one-ten thousandth of strong motion earthquake motion) than
during the forced vibration tests. Even over the limited range of forces possible with the structural vibrators, we have observed significant departures
from linear elastic response. Further differences can be expected (compared
to forced vibration tests) when strong motion earthquakes provide the exciting
Iforce.
In the seismic analysis of nuclear power plants, it is common to assume
that linear elastic theory holds, and that theoretical models or mathematical
models are "linear". This is necessary because nonlinear analyses are difficult and expensive to perform and the nature of the nonlinearities is not understood. We have made the same assumption in extrapolating calculated responses based on our low- and intermediate-level test data to "predictions" of
strong motion earthquake response. In each case, we have called attention to
this assumption by referring to such predictions as an "earthquake-like" response.
Our concern with this problem led us to the concept of dynamic tests
using explosive blasts to simulate strong motion earthquake response. Blasts
as large as one ton of high explosive have been detonated in the soil at a
distance less than 100 meters from the center of the containment building.
These tests produced peak horizontal ground accelerations greater than lg
and caused the reactor components to exceed seismic design limits.
As a means of characterizing reactor equipment response to strong motion
earthquakes, two classifications have been proposed:
Softening system:
a structure or system in which the response

at high levels of excitation is greater than
what would be expected based on a linear extrapolation of low level response.
Hardening system:
a structure or system in which the response

at high levels of excitation is less than
what would be expected based on a linear extrapolation of low level response.
- 23 Test results analyzed to date have given positive evidence that these
nonlinear effects occur at higher levels of excitation. Significant changes
have been observed; these can be correlated with the effective stiffness and
with the effective damping of the system examined.
In the above definitions we refer to the total system, so that "harden
ing" and "softening" refer to a combined effect of both stiffness and damping.
Most of the results we have observed fall in the softening system category.
3. USE OF BLAST TESTS
To illustrate the use of explosive blast testing, we shall outline one
such test and describe some of the pertinent results which have been obtained.
In this test, explosives were detonated in bore holes that were typically'
20 m deep. The distance and depth varied slightly from test to test. The re
actor containment building, pressure vessel, piping, core, and steam'generator
were instrumented with accelerometers. Other accelerometers were placed on
the soil away from the building. In addition, two three-component bore hole
seismometers were located at the bottom of borings placed between the blasts
and the containment building.
Soil response was obtained with the bore hole seismometers and with the
"free field" accelerometers. This information is being used to compute and
S wave velocities for the reactor site. Wave velocities have also been ob
tained using geophones and hammer blows.
These tests have led to a tentative identification of wave velocities for
the site:
Dynamic
Velocity, ft/sec
Stratification
(depth in feet)
Wave
S Wave
0 - 25
4400
3000
25 - 63
6800
4700
Below 63
9300
6400
Additional soils information will be derived from the blast data. One aspect
of the study which appears promising is an examination of the effect of soil
strain level on the velocity attenuation parameter. From this study we hope
to gain insight into the behavior of typical nuclear power plant soils when
subjected to high level excitation.
Structure (containment building) response and soil-structure interaction
are being studied by comparing the records obtained on the free field, in the
basement of the building, and on the operating floor (4th floor) of the build
ing. The response of the containment building during the blasts is complex.
During the initial ("forced vibration") portion of the blast record,
significant response at -12 Hz was observed. This is interpreted to be rock
ing of the structure on the soil. After the initial excitation of the blast
dies away, the structure responds with what we call "free vibrations". In
this case, the observed vibrations were seen to be those corresponding to
known lateral and torsional modes of vibration. Since the blasts excite sev-
- ?4 eral modes of vibration simultaneously, careful placement of accelerometers is

a necessity. The resulting records must be Fourier analyzed if modal coupling
is present.
ReactOT equipment and piping responses were also obtained during the
blast tests. The data are be in," used to obtain the dynamic parameters of each
equipment item, and also to sec how the parameters vary with the level of excitatioT'..
Figure 5 is a plot of the response of the top of a steair, generator to a
series o' blast tests. Xote that the initial response during the blast (the
"forced vibrations") occurs at u high frequency. After approximately one
second, the forced vibrations end and the steam generator undergoes free vibrations at its natural frequency. In the largest tests, the lateral accelerations of the steam generator exceeded the design seismic loading.
The blast data can yield information regarding modal frequencies, mode
shapes, and modal damping. IVc are studying each piece of equipment as well as
the primary coolant piping. When possible, blast test da'ta are compared to
other data and calculated values.
We are also examining how the equipment parameters change at higher force
levels. Figure 6 indicates how the modal frequency of a steam generator
shifts as the force level increases. The behavior shown is characteristic of
a softening system. Figures 7 and 8 are for a large reactor pressure vessel.
Note (in Figure 7) that the vessel natural frequency is characteristic of a
softening system. In Figure 8, however, we see that the damping is increasing
with the higher level of excitation.
Analytical models for describing these nonlinear parameter changes is the
subject of current work at UCLA.
When the frequencies observed during the blast tests were compared with
the structural vibrator test results, good agreement was found. At the same
level of excitation, the results were in close agreement, and in many cases
in exact agreement.
These results are typical (tests not all at same force level):
Structural
Vibrators
Blasts
Steam generator
EW
NS
5.92 Hz
5.8
5.94
5.75
Containment building
EK
NS
Torsion
4.65
4.2
8.2
Reactor core
l.V,
Nh
4.5
3.9
4.62
4.0
8.5
4.3
3.8
Test
Reference
T3R2
T3R3
T1R2
T3R2
T3R2
T2R3
T2R3
In conclusion, blast tests arc a useful tool for dynamic testing. By

suitably varying the detonation rate of the explosive, the depth and distance
of the charge, and the length of delay between successive charges, records of
different lengths and differing frequency content can be produced.
- 25 4. COMPARISON WITH EARTHQUAKE RESPONSE

Only a limited number of nuclear reactors in the United States have
strong motion seismographs and until February 1971 none of these had been
subjected to anything other than small earthquakes. In California, the UCLA
research reactor is equipped with a USCGS strong motion instrument. In addition, the reactor structure is instrumented with accelerometers which are recorded when the strong motion instrument trips.
Also in California the San Onofre Nuclear Generating Station has a Teledyne MTS-100 system with sensors on the containment structure basement, the
steam generator, and the pressurizer. The station has an AR-240 strong motion accelerograph, as well as several passive "peak recording seismometers".
Several records have been obtained at San Onofre, including the earthquakes of 7 August 1966, 8 April 1968 (Borrego Mountain), 12 September 1970
(Lytle Creek) and most recently 9 February 1971 (San Fernando).
Copies of records from the recent earthquakes were made available to the
authors by the Southern California Edison Company. These were the records
from 12 September 1970, when an earthquake near Cajon Pass triggered the
strong motion earthquake recording system, and records obtained during the
9 February 1971 San Fernando earthquake. The records were digitized at UCLA
and then Fourier analyzed. The results show significant differences when
compared to the results of forced vibration tests (see Ibanez, et. al. [S],
and Matthiesen, Ibanez, Selna, and Smith [10] ) .
Figures 9 and 10 show comparative data taken from the forced vibration
tests and earthquake results for the pressurizer and steam generator respectively.
For the pressurizer, the resonant frequency is less when obtained from
the earthquake record than when obtained from the forced vibration test.
The modal damping is approximately the same in both cases. Although the
September earthquake was small, the displacement due to the earthquake motion is nearly 100 times as large as the displacement produced by the forced
vibration test. The pressurizer might be classified as a "softening system".
The steam generator responded at a higher frequency during the earthquake than during the forced vibration tests, and the damping is significantly greater also. In the case of the steam generator, the earthquakeinduced displacement is about 50 times as great as the forced vibration test
displacements. The steam generator appears to respond as a "hardening system".
The method of supporting each of these components provides insight into
why they respond the way they do.
We have observed a softening behavior
in other large components which are essentially cantilevered beams fixed at
the base. An example is the EGCR steam generator, which was discussed previously.
The San Onofre steam generator is supported by hangers. As the connecting piping is stressed, we might expect the steam generator to be stiffer when
- 26 undergoing large displacements. Further study is needed to assess the influence of the connecting piping on the steam generator response before a
definite conclusion can be reached.
Shortly after the September earthquake data had been analyzed, Southern
California experienced the 9 February 1971 San Fernando earthquake. In terms
of damage and destruction, this M = 6.5 earthquake was the worst one in California since the 1933 Long Beach earthquake. Strong motion accelerograph
records were obtained at the San Onofre Nuclear Generating Station, at the
UCLA reactor, and at numerous structures in Los Angeles.
Figure 11 shows the San Onofre steam generator record. Peak accelerations of approximately 0.2g were recorded at the top of the steam generator.
It is interesting to note that the frequency in the time trace where the
largest amplitude vibrations occurred is approximately 2.9 Hz. From the work
of Ibanez, et. al. [S], this is seen to correspond not to the steam generator
but to the interaction of the steam generator with the pump. Additional analysis of the San Onofre records is underway.
At UCLA the strong motion instrument tripped during the February earthquake and again on four subsequent aftershocks during the day of the earthquake and the day following. The reactor instrumentation failed to trigger
during the initial event, but was turned on and left on for most of the day
and recorded nearly one hundred aftershocks. Several of these were large;
one caused the top of the reactor shield to reach O.Olg, or the same level of
acceleration produced during structural vibrator tests.
The earthquake record time traces show a component at -16 Hz in several
cases. This compares favorably with the value of 16.6 Hz found in the forced
vibration tests. The NS accelerations of the reactor were consistently
greater than the EW; this is consistent with the vibrator tests and with the
direction of propagation of earthquake energy.
Finally, we are happy to report that no vibration effects were observed
during the earthquake which could have led to operational problems at either
the research reactor or at San Onofre. In fact, San Onofre operated during
the earthquake without problems.
5. CONCLUSIONS
Nonlinear behavior of nuclear power plant components, which certainly can
be expected when plants are subjected to strong motion earthquakes, is not yet
well defined. Additional work is needed to understand the mechanisms which
give rise to the nonlinearities, to develop calculational procedures for computing response at high forcing levels, and to develop design criteria which
will enable power plant designers to accommodate high level response in an
economic and reliable way. Explosive blast tests are a useful tool for this
work. Experimental tests, when combined with analysis and data obtained during actual earthquakes, will give nuclear power plant designers information
needed to insure public safety.
- 27 REFERENCES
[1]
MATTHIESEN, R.B., SMITH, C.B., "A Simulation of Earthquake Effects on the

UCLA Reactor Using Structural Vibrators," UCLA Department of Engineering
Report NEL-105 (October 1966).
[2]
SMITH, C.B., MATTHIESEN, R.B., "Forced Vibration Tests of the Experimental Gas-Cooled Reactor (EGCR)," UCLA Engineering Report #69-42 (August
1969).
[3]
MATTHIESEN, R.B., SMITH, C.B., "Forced Vibration Tests of the CarolinasVirginia Tube Reactor (CVTR)," UCLA Engineering Report #69-8 (February
1969).
[4]
BLEIWEIS, P., HART, G.C., SMITH, C.B., "Enrico Fermi Nuclear Power Plant
Dynamic Response During Blasting," ANS Transactions 13, 1, pp. 231-232
(June 1970).
[5]
IBANEZ, P., MATTHIESEN, R.B., SMITH, C.B., WANG, G.S.C. , "San Onofre
Nuclear Generating Station Vibration Tests," UCLA-ENG-7037 (August 1970).
[6]
SCHMITT, R.C., "Evaluation and Comparison of Structural Dynamics Investigation of the Carolinas Virginia Tube Reactor Containment," Report
IN-1372, Idaho Nuclear Corporation (May 1970).
[7]
Personal communication, B.J. MORRILL (USCGS) to R.B. MATTHIESEN, November

3, 1966.
[8]
HOUSNER, G.W., WEST, P.J., JOHNSON, C G . , "Analysis of Ground Motions at

San Onofre Nuclear Generating Station, April 9, 1968," unpublished
report.
[9]
Personal communication, J.D. HORNBUCKLE to C.B. SMITH, transmitting

records obtained at San Onofre of the 12 September 1970 earthquake.
[10] MATTHIESEN, R.B., IBANEZ, P., SELNA, L.G., SMITH, C.B., "San Onofre
Nuclear Generating Station Supplementary Vibration Tests," UCLA-ENG-7095
(December 1970) .
28 -
FIG I
STRUCTURAL VIBRATOR AND CONTROL PANEL
FIG 2 DRILL RIG
- 29
FIG 3 LOADING EXPLOSIVE IN BORE HOLE
FIG 4 RECORDING INSTRUMENTATION
30
o.i
io
r""
loo
FREQUENCY (hz)
Fig 5
EA STWEST RESPONSE OF TOP OF STEAM GENERA TOR
>
o
UI
U.
5
IO
50
ACCELERATION (thousand!he of g )
Flg. 6 FREQUENCY ACCELERATION RELATION FROM EASTWEST RESPONSE OF
EG CR
STEAM GENERATOR
200
31
37.4
OÔ
37.2
D~~.
37.0
>
o
36.8
o bio.t tsst west
or
u.
\
\
D blat tett north
36.6
364
IO
ACCELERATION
SO
100
( thousandths of a g )
400
Fig, 7 FREQUENCYACCELERATION RELATION FROM FREE VIB RATION RESPONSE OF

EG CR REACTOR VESSEL
38.0
O blast test north
D blast tost west
1.0
15
2.0
DAMPING (percent of critical)
Fig. 8
FREQUENCYDAMPING RELATION FROM RESPONSE

OF EG CR REACTOR VESSEL
32
3.0
{ I
2 9
otlon tests
earthquake of 12 sept 7 0
2.8
N
f
>
g
2.7
UI
UI
2.6
2.5
0.2
OS
1.0
5.0
IO
50
100
200
DISPLACEMENT (thousandths of an Inch)

Fig.9
FREQUENCYDISPLACEMENT RELATION FROM NORTHSOUTH RESPONSE

OF SONGS PRESSURIZER
2.2
O forced vibration t e t t i
2.1
O earthquake of 9 feb 71
D-
j 2.0
UI
UI
u.
>
1.8
1.7
0.2
0J5
1.0
50
10
50
DISPLACEMENT (thousandths of an inch)
100
Fig. 10 FREQUENCY-DISPLACEMENT RELATION FROM EAST-WEST RESPONSE

OF SONGS
STEAM GENERATOR
200
33
'-** u
ryv*^>V*yvV^yy^
??Atl>*V|^*S
'**r*JJ'+*r*<* *~**
^^\^/\<(('\
9HHs^y^^BSBîsfl
FIG II TRA CE OF SA N ONOFRE STEA M GENERA TOR

EARTHQUAKE RESPONSE
- 34 DISCUSSION
N. J. M. REES, U. .
Did you consider the effect of air blast on your s t r u c t u r e s from your buried
explosions since from your slides they appeared to have all been vented ones ?
We did not m e a s u r e the air blast contribution. On a guard shack about 100' from
the b l a s t s , no windows were broken during any of the t e s t s .
J. D. STEVENSON, U.S.A.
Did you try to determine the source of non-linear effects ? Slippage of supports
vs. change in s t r e s s level in components.

A
We have not made a thorough study of the source of these effects. In the case of
the EGCR steam generator, we believe they a r e due to slippage of bolted connections at the
base. At San Onofre, the hardening effect is thought to be due to stretching of the connecting
piping. The San Onofre steam generator is supported by hangers.
K. AKINO, Japan
I have doubts about the blast test, whether that excitation method can produce
prteper vibration for equipment or not. How do you think about this point ?
The question refers to whether or not the blast impulse duration (*- 1 second) in
sufficiently long to excite equipment.
We found that the blasts produced r e c o r d s as long as 20 seconds on some pieces of equipment.
Soil r e c o r d s had s e v e r a l cycles of vibration but were usually only about one second long.
Longer r e c o r d s can be produced by using s e v e r a l charges with a time delay. Blasts lasting
up to 10 seconds could be produced this way. Because of the excellent agreement between
our forced vibration test results and the blast r e s u l t s , we did not use longer blasts.
At some other plant, this would depend on the damping and m a s s of the equipment items as
c o m p a r e d to the s t r u c t u r e .
K 2/3
EARTHQUAKE CALCULATIONS : THEIR IMPORTANCE

W I T H RESPECT TO AREAS OF AVERAGE
AND LOW SEISMIC ACTIVITY AND THE APPLICATION
OF COMPUTER ORIENTED METHODS
A.E. HUBER, P.O. SCHILDKNECHT,
SDK Ingenieurunternehmen
fr spezielle Statik, Dynamik und Konstruktion

Lrrach, Germany
GmbH,
ABSTRACT
In the past design criteria for earthquake resistant conventional structures
in areas of low and medium seismic activities have been semi-empirical. Even
though these estimates are adequate for motst conventional structures they
appear not to be sufficient to more involved structures such as nuclear power
plants. It is therefore appropriate to undertake a more thorough investigation
for this class of structure.
A critical appraisal of earthquake design suggests the consideration of two
main points which should be kept in mind throughout every rigorous analysis.
On the one hand there is the potential danger of nuclear power plants. This
fact becomes more important if the site is located in a populated area. Therefore, in order to reduce this risk to an absolute minimum, safety requirements
have to be high. On the other hand, however, unreasonably high safety factors
should be avoided from the economical point of view.
A more thorough analysis based on the general dynamic problem is presented in
this paper. The general formulation includes the response spectrum modal analysis and time history modal analysis in order to solve seismic problems. A
survey on the generation of response spectra and time histories, especially
with respect to areas of low and medium seismic activities, is then presented.
The necessity of seismic response analysis is demonstrated by response spectra
and by a numerical example.
A useful method is then presented in order to solve the general eigenvalue
problem arising in structural dynamics. The method is an energy minimization
method and is therefore very efficient for large systems stemming e.g. from
a finite element analvs-s.
36
1.0 Dynamische Berechnung von Konstruktionen
1.1 Berechnung des Bewegungsablaufs (Time History Modal Analysis)
Wir betrachten eine linear elastische Konstruktion mit i Freiheitsgraden,
die mit i Massen behaftet sind.
Das verallgemeinerte Eigenwertsproblem eines solchen Systems und dessen
Lsung, nmlich die Eigenwerte und Eigenvektoren, werden hier als bekannt
vorausgesetzt. Auf dieses Problem wird im letzten Teil eingegangen.
Betrachten wir zuerst eine statische Belastung des Systems. Diese lsst
sich nach den Eigenvektoren entwickeln. Die Deformation, welche dem norma
lisierten Eigenvektor A. entspricht, erhlt man durch eine statische
3
im
Lastgruppe L. , die die Massenkrfte eines frei schwingenden Systems

ersetzt.
L.
im
i
m
w
= w 2 A. = Masse Beschleunigung
(1)
m g
im
= Nummer des Freiheitsgrades
= Nummer der Eigenfrequenz bzw. des Eigenvektors
= Eigenfrequenz
= Masse
g
A = Eigenvektor (normalisiert)
Im folgenden wird eine einzelne Last L. als charakteristische Last be
im
zeichnet, eine Lastgruppe L. (1=1...ie) als charakteristische Lastgruppe.
Eine beliebige, in den gewhlten Freiheitsgraden i angreifende Lastgruppc
kann durch Superposition der charakteristischen Lastgruppen darge

stellt werden.
, , rn^rne (1)
1
F
= ^ ^. B 1m L.im
m=l
wobei den Beteiligungsfaktor der charakteristischen Lastgruppe
L. (1=1...le) bedeutet.
Die Beteiligungsfaktoren
erhlt man aus dem Gleichungssystem
(2)
37
zu
\#>\ -
Mit H i l f e
~ >' , 1I UJ I
TL. 1
l im J
Imi
(4)
l i
der Orthogonalittsbedingungen
(D = J1
B'"
m
w
m
vergi.
werden,
dass
" .
im
^ ^
fr=1
Ill
kann g e z e i g t
(5)
S.94
Betrachten wir eine Lastgruppe P., bei der an der Stelle i=j die Last
P. = P. angreift und alle brigen Lasten P. = O (fr i j] sind, so
erhlt man aus (5)
= 4 P. .
(6)
Zur eindeutigen Kennzeichnung der Lastgruppe gengt dabei der Index (j)
anstelle von (1).
Da die Beteiligungsfaktoren sowohl fr Belastungen als auch fr die Ver
schiebungen gelten, erhlt man die Verschiebungen infolge Belastung P. zu:
m=me
qP1 = . 01
m A.im
m=l
(7)
Betrachtet man eine zeitlich vernderliche Belastung .(t), die im i=j

angreift
p.(t)
mit
EP'
m
= p.
f.(t)
= f.(t)
m=me
:
: Bp>L.m
= 4 p . A.
vi
*]
]m
so lsst sich nach 121 , S. 22 die Verschiebung in folgender Form dar

stellen:
m=me
4- (t) = 3 > Z D01 (t) EP' .

^
'
'
Dabei ist cP {t) die Lsung der Gleichung
(9)
'
38
1
2
2
'
D1* (t) = wm
fj (t)
m (t) + wm m
(10)
Bercksichtigt man viskose Dmpfung, so ist rP' (t) die Lsung der
Gleichung
"cP1 (t) + 2b m PJ (t) + w 2 LP1 (t) = w 2 f. (t)
m m
mit b
m m
(11)
als Dmpfungsfaktor.
Die Lsung der Gleichung (11) lautet:

t
rpl (t)
2
bm
\Rm
V ff. ,^
"bbi
(t) e~
(t
~' . sin \/wmb^ . (tt)dt
(12)
'0
Wenn wir die Belastungen in allen Punkten j=l...je betrachten, erhlt man
durch Summation ber j aus Gleichung (8) und (9):
j = j j e m=me
q<
(t)
=
j=l
q.
î
(t)
X ^
m=l
(t) = ^_ .
m=l
w4 ^ ^. Z,
m j=l
rJmjl (t) rp3. A im

.
J
J
(13)
Entsprechende Formeln lassen sich fr beliebige Anfangsbedingungen her

leiten. Damit sind die zeitlichen Verschiebungen fr eine beliebige, zelt
lich vernderliche Belastung bekannt. Aus den Verschiebungen lassen sich
nach bekannten Methoden die Schnittkrfte und Spannungen bestimmen.
Fr eine Erdbebenbelastung erhlt man aus der Beschleunigung in Richtung k
des als starrer Krper betrachteten Systems die entsprechenden Lasten zu:
W.
P. (t) = p. f. (t) = 6., p. 1 'i (t)
3
]k ^3
(t) = Grundbeschleunigung, z.B. aus Accelerogramm

6.,
= 11 wenn
wenn der
der Freiheitsgrad
Freiheitsgrad j die gleiche Richtung wie die
Erdbebenbeschleunigung hat.
6 .,
= O in allen anderen Fllen
(14)
39
Mit dieser Belastungsfunktion und mit b
3
m=me
q, = t>
A.
j=je
W.
~z>~ 6., p. 1 A.
ifj jk *} g
im
= c w erhlt man aus Gl. (13):

m m m
. y
]m
(15)
dabei ist:
t
= \
m wm l
r1
(t) e " c m w m
(t
"' sin w
m (tt)dt

(16)
to
Man kann also den Bewegungsablauf (time history) fr eine beliebige Grund
beschleunigung berechnen.
In der Praxis interessieren in Bezug auf die Festigkeitsanforderungen nur
die Maximalwerte der Verschiebungen bzw. der dynamischen Belastungen.
(Um die bersichtlichkeit zu wahren, werden diese Betrachtungen auf die
Verschiebungen beschrnkt. Sie lassen sich analog fr dynamische Be
lastungen anstellen.) Die Ermittlung dieser Maximalwerte aus dem Bewe
gungsablauf, Gl. (15), ist allerdings sehr aufwendig.
Eine Abschtzung der maximalen Verschiebungen erhlt man durch betrags
mssige Superposition der Maximalwerte y
, Gl. (16), fr jede einzelne
Eigenfrequenz. Die Abschtzung liegt auf der sicheren Seite. Um starke
berschtzungen zu vermeiden, werden hhere Frequenzen nur mit der mitt
leren Quadratwurzel der max. Verschiebungen bercksichtigt. Damit redu
ziert sich das Problem auf die Berechnung der Maximalwerte des Integrals,
Gl. (16), was analytisch oder numerisch kein Problem darstellt.
Dieses Berechnungsverfahren msste bei Erdbebenberechnungen fr verschie
dene, charakteristische Accelerogramme und verschiedene Bauwerksparameter
durchgefhrt werden und ist deshalb relativ aufwendig.
Um ein fr die Praxis zweckmssiges und auf der sicheren Seite liegendes
Berechnungsverfahren zu erreichen, hat es sich als vorteilhaft erwiesen,
sogenannte "Response Spektren" zu verwenden. Diese Methode wird im nch
sten Abschnitt erlutert. Sie beruht im wesentlichen auf den bisherigen
Ableitungen.
40
1.2 Berechnung mittels Response Spektren (Response Spectrum Modal Analysis)
Betrachtet man ein gedmpftes System mit einem Freiheitsgrad, das einer
Grundbeschleunigung unterworfen wird, so erhlt man die Relativverschie
bung zu: (vergi. 13] , S. 85)
y (t, w, c) = i
m
t
\ (t) e" c
Wm
(t
~y
sin w m (tt)dt
(17)
'T
= Grundbeschleuniqunq eines Erdbebens

= Eigenfrequenz des gedmpften Systems, fr die Praxis
kann sie durch die ungedmpfte Eigenfrequenz ersetzt
werden.
= Relativverschiebung
= Dmpfungsfaktor
Es sei hier vermerkt, dass diese Relativverschiebung gleich dem in Gl.(16)

angegebenen
Wert y'm ist.
5 3
Der Maximalwert der Verschiebung y (t, w, c) wird "Displacement Response"
genannt. Diese Maximalwerte knnen bei einem gegebenen Accelcrograiiun
fr verschiedene Eigenfrequenzen und verschiedene Dmpfungsfaktoren
mittels Gl.(17) ausgewertet werden. Trgt man diese Displacement Response
Werte S, in einem Diagramm ber der Frequenz auf, so erhlt man ein
"Response Spectrum", in diesem Falle das "Displacement Response Spectrum".
Zur Veranschaulichung knnte man sich die Ermittlung eines Response
Spektrums so vorstellen:
Man unterwirft eine Reihe von Einmassensystemen mit verschiedenen Eigen
frequenzen und verschiedenen Dmpfungsfaktoren der gleichen Grundbeschleu
nigung (z.B. einem Accelerogramm), misst die maximale relative Auslenkuny
und trgt sie in einem Diagramm ber den Eigenfrequenzen auf.
In analoger Weise kann man "Response Spektren" S und , fr die maximale
Geschwindigkeit und die maximale Beschleunigung aufstellen. Sie lassen
sich auch aus S, durch folgende Beziehungen herleiten:
41
i m y 2 = = k y 2 und my = k . y
erhlt man mit w 2 =
m
y
und somit:
= w y = w
= S
= w Sd
Durch diese Verknpfung lassen sich alle drei Response Spektren in einem
Diagramm mit logarithmischer Teilung darstellen, (siehe Abb. 1)
Zusammenfassend kann man folgende Definition treffen:
Das Response Spektrum fr ein vorgegebenes Erdbeben ist ein Diagramm, das
die Vernderung des maximalen Ansprechens (max. Verschiebung, Geschwin
digkeit, Beschleunigung) eines Einmassensystems mit der Eigenfrequenz
zeigt, wenn es einer dem gegebenen Erdbeben entsprechenden Grundbeschleu
nigung unterworfen wird.
Da die Gleichungen (16) und (17) bereinstimmen, kann man den Response
Spektren den Maximalwert y , Gl.(16), fr beliebige Eigenfrequenzen ent
nehmen. Damit knnen die maximal mglichen Verschiebungen nach Gl.(15)
berechnet werden. Dabei kann wiederum fr hhere Eigenfrequenzen nur die
mittlere Quadratwurzel der Verschiebungen bercksichtigt werden, um
starke berschtzungen zu vermeiden.
Soweit entspricht diese Methode der im vorhergehenden Abschnitt erwhnten
Time History Modal Analysis mit einer Abschtzung der maximal mglichen
Verschiebungen.
Der Vorteil der Response Spectrum Modal Analysis besteht nun darin, dass
man Response Spektren entwickeln kann, die ein breites Spektrum von mg
lichen Erdbeben und mglichen Parametern des Systems abdecken. Damit
entfllt die mehrmalige, dynamische Berechnung fr verschiedene Accelero
gramme und Systemparameter. Zustzlich knnen Sicherheitsbetrachtungen
in das Response Spektrum integriert werden.
Abb. 1 zeigt das Response Spektrum fr ein registriertes Erdbeben, wobei
die stark ausgebildeten Minima und Maxima charakteristisch sind. Daraus
ist ersichtlich, dass gewisse Frequenzen sehr stark vertreten sind, wh
rend andere kaum in Erscheinung treten.
- 42 In Abb. 2 ist ein Entwurfs-Spektrum dargestellt, das durch Normierung und

Mittelung aus vier verschiedenen Accelerogrammen gewonnen wurde.
Dieses Response Spektrum zeigt geglttete Kurven, die eine Reihe von
Accelerogrammen abdecken.
2.0 Mglichkeiten zur Bestimmung von seismischen Eingabeparametern fr die
dynamische Berechnung von Konstruktionen
2.1 Allgemeines
Die folgenden Abschnitte sollen kurz die wesentlichen Verfahren skizzieren, mit denen man aus den Ergebnissen einer seismologischen
Untersuchung
die seismischen Eingabeparameter fr eine dynamische Berechnung von Konstruktionen bestimmen kann. Entsprechend den vorangegangenen
Ausfhrungen
sind also fr den Standort charakteristische Time Histories oder Response

Spektren zu bestimmen. Aus einer seismologischen Untersuchung knnen u.a.
folgende Angaben fr den betreffenden Standort zur Verfgung stehen:
a) Intensitt
b) maximale Grundverschiebung, -Geschwindigkeit,
-Beschleunigung
c) Frequenzgehalt oder pauschale Angaben ber vorherrschende Frequenzen der Grundbewegung
d) Seismogramme und speziell Accelerogramme von
registrierten Erdbeben.
Die Intensitt und die Maximalwerte von Verschiebung, Geschwindigkeit
und Beschleunigung der Fundamentsohle bzw. des festen Untergrundes
(Fels)
am Standort kann man beispielsweise aufgrund einer statistischen Auswertung der verfgbaren seismischen Daten bestimmen
(vergi. L. Esteva (41).
Oft werden jedoch fr die Auslegung nur einer oder zwei dieser Ausgangswerte, nmlich die Intensitt und/oder die maximale
bzw. Grundverschiebung
Grundbeschleunigung
angegeben. Weitere, auf direkten
seismischen
Messungen beruhende Angaben knnen in vielen Fllen wegen
ungengender
Unterlagen nicht gemacht werden. An zustzlichen Angaben knnen geologische Daten fr den Standort, wie z.B. Bodenkennwerte, Entfernung von
Erdbebenherden etc., zur Verfgung stehen.
Wenn im Rahmen dieser Betrachtungen von Bodenkennwerten und Grundverschicbungen, Grundbeschleunigungen etc. die Rede ist, so beziehen sich diese
Angaben auf den festen Untergrund
(Fels) .
- 43 Fr das Problem der Verstrkung von Erdbeben durch berlagernde, weichere

Baugrundschichten sei auf entsprechende Literatur verwiesen. Es soll hier
jedoch der Vollstndigkeit halber erwhnt werden, dass die lokalen Baugrundverhltnisse das dynamische Verhalten der Konstruktionen sehr wesentlich beeinflussen knnen.
2.2 Maximale Grundbeschleunigung a, maximale Grundgeschwindigkeit v
und maximale Grundverschiebung d
Sofern gengend Seismogramme mit entsprechendem Auflsungsvermgen existieren, knnen a, v und d daraus sehr genau ermittelt werden. In diesem
Fall wird man allerdings die spter erwhnten, exakteren Methoden whlen.
2.21 Intensitt I und maximale Grundverschiebung d vorgegeben
Wir wenden uns nun dem Fall zu, dass lediglich die Intensitt I und die
maximale Grundverschiebung d vorliegen. (Auf die Bedeutung einer Angabe
von d fr mitteleuropische Verhltnisse wird spter eingegangen; fr
amerikanische Verhltnisse kann man auf diese Angabe verzichten.)
Die Intensittsskalen lassen sich nherungsweise mit der maximalen Grundbeschleunigung a und der maximalen Grundgeschwindigkeit v in Beziehung
setzen. Eine Beziehung zwischen modifizierter Mercalli Intensitt und
der maximalen Grundbeschleunigung ist z.B. 151 zu entnehmen. Die Beziehungen zwischen Intensitt I und maximaler Grundbeschleunigung a
haben sich jedoch als sehr grob erwiesen (vergi. 171). Aus verschiedenen
theoretischen und empirischen Grnden ist ein besserer Zusammenhang
zwischen Intensitt I und maximaler Grundgeschwindigkeit v festzustellen
(vergi. 181 ) .
Eine entsprechende empirische Beziehung lautet:
* = ^Tc^F
I = modifizierte Mercalli Intensitt
v = maximale Geschwindigkeit
Bei vorgegebenem I erhlt man aus dieser Beziehung v.
2.22 Intensitt I und Schwchungsgesetze vorgegeben
Ein besseres Verfahren zur Bestimmung von a und d bei vorgegebener
Intensitt I besteht darin, Schwchungsgesetze fr a, v und d zu benutzen.
C)
44
Solche Schwchungsgesetze sind z.B. bei L. Esteva 141 zu finden und
haben die Form:
a = C, e 2 1 (R + C3 ) ' C ^
<20a)
= Ki e 1 (R + K 2 eK3l)"K'
(20b)
K,C = Konstanten
I
= Intensitt
R
= Abstand des Hypozentrums
Ein entsprechendes Schwchungsgesetz kann man fr die maximale Grundver
schiebung d formulieren.
Die Konstanten sind im wesentlichen abhngig von lokalen Bodenbedingun
gen, von der Art der geologischen Formation, die von den Schockwellen
passiert werden, von den Schockmechanismen etc. Durch lokale seismische
Aufzeichnungen und/oder durch geeignete Wahl von Aufzeichnungen in Ge
bieten mit hnlichen geologischen Verhltnissen lassen sich diese Kon
stanten hinreichend genau abschtzen. Bei Verwendung dieser Schwchungs
gesetze ist allerdings die Kenntnis von aktiven Verwerfungen in der Um
gebung des Standortes erforderlich, um den Abstand der mglichen Hypo
zentren bestimmen zu knnen.
Da der relative Frequenzgehalt durch die relativen Werte von a, und d
festgelegt wird, kann man mit dem aus Gl.(19) ermittelten und den ge
nannten Schwchungsgesetzen a und d so bestimmen, dass der relative
Frequenzgehalt mit den mglichen Erdbebenquellen und den lokalen geolo
gischen Gegebenheiten bereinstimmt.
2.3 Grundspektrum
Die drei genannten Werte a, v und d charakterisieren die Grundbewegung am
Standort. Trgt man diese Werte in einem vierfach logarithmischen Dia
gramm (Abb. 4) auf, so erhlt man eine polygonartige Eingrenzung des so
genannten "Grundspektrums". Damit ist auch der relative Frequenzgehalt
des Grundspektrums festgelegt.
Um den lokalen Baugrundverhltnissen Rechnung zu tragen, kann ein solches
Grundspektrum durch entsprechende Verstrkung im betreffenden Frequenz
bereich variiert werden.
45
2.4 Bestimmung von Response Spektren aus a, und d
Aus einem Grundspektrum lassen sich Response Spektren entwickeln. Empiri
sche Untersuchungen haben ergeben, dass man Response Spektren einfach
durch Multiplikation von a, v und d des Grundspektrums abschtzen kann,
(vergi. 191).
Dieser Zusammenhang ist durch die empirisch festgestellte Tatsache be
grndet, dass das Response Spektrum fr 2025% Dmpfur.gsrate nherungs
weise mit dem Grundspektrum zusammenfllt (vergi. 1141).
Ausgehend von dieser empirischen Feststellung und von einem vorgegebenen
Grundspektrum kann man diese Faktoren mit Hilfe stochastischer Bewegungs
modelle auf mathematischphysikalischem Wege bestimmen. Auf stochastische
Bewegungsmodelle wird spter noch eingegangen.
Ein Satz solcher Multiplikationsfaktoren fr elastisches Materialverhal
ten und verschiedene Dmpfungsraten ist beispielsweise in 1101 zu finden,
(vergi. Abb. 3 ) . Diese Faktoren basieren auf dem El Centro Erdbeben von
1940. hnliche Faktoren existieren fr nichtlineares Materialverhalten.
Der Vollstndigkeit halber sei noch erwhnt, dass fr extrem hohe Eigen
frequenzen die maximale ResponseBeschleunigung gleich der maximalen
Grundbeschleunigung ist; bei extrem niederen Frequenzen ist die maximale
ResponseVerschiebung gleich der Grundverschiebung. (Einzelheiten siehe
1101 ) .
2.5 Bestimmung von Response Spektren aus Seismogrammen
Eine direkte Methode zur Bestimmung von Response Spektren besteht darin,
aus fr das betreffende Gebiet charakteristischen Seismogrammen direkt
die Response Spektren zu berechnen. Dazu werden aus den einzelnen Accele
rogrammen z.B. entsprechend Gl.(17) die entsprechenden Response Spektren
ermittelt. Diese Response Spektren normiert man auf eine bestimmte Maxi
malbeschleunigung bzw. auf eine bestimmte Intensitt. Durch Mittelwert
bildung bzw. durch Eingrenzung dieser normierten Response Spektren er
hlt man ein einziges Response Spektrum, das eine Reihe von mglichen
Erdbeben abdeckt. Dieses Verfahren wurde erstmals von G. Housner auf vier
Erdbeben an der amerikanischen Westkste angewendet (vergi. Abb. 2 ) .
Durch Normierung eines solchen Response Spektrums auf eine vorgegebene
Erdbebenintensitt bzw. auf eine vorgegebene maximale Grundbeschleunigung
erhlt man ein Response Spektrum, das der dynamischen Berechnung der An
lage zugrunde gelegt werden '.ann.
46
2.6 Bewegungsablauf
(Time Histories)
Ausgehend von einem
(oder mehreren) aufgezeichneten Accelerogramm, das
als charakteristisch fr die Umgebung des Standortes angesehen werden

kann, wird dieses Accelerogramm entsprechend der vorgegebenen
Intensitt
oder der maximalen Grundbeschleunigung normiert. Um individuellen Fluktu

ationen Rechnung zu tragen, kann man weitere Accelerogramme durch Ampli
tudenmodulation und zeitliche Verzerrung entwickeln.
Aus diesen Accelerogrammen kann man einerseits ein als Berechnungsgrund

lage dienendes Response Spektrum berechnen, andererseits knnen sie als
direkte Eingabewerte fr die dynamische Berechnung dienen. Die Erzeugung
von knstlichen Accelerogrammen mit Hilfe stochastischer
Bewegungsmodelle
wird im folgenden erwhnt.
2.7 Stochastische
Bewegungsmodelle
Einschlgige Untersuchungen haben ergeben, dass sich Erdbeben mittels

stochastischer Bewegungsmodelle sehr gut reprsentieren lassen,
(vergi. z.B.
141 S. 335,
1111,1121,1131).
Diese Modelle beruhen im wesentlichen auf einem stationren Gauschen

Prozess, der mit Hilfe einer einhllenden, zeitabhngigen
Intensitts
funktion zum instationren Prozess abgewandelt werden kann.
Physikalisch bedeutet dies, dass eine Reihe von mglichen Erdbebenanre

gungen durch eine gengend grosse Anzahl von zuflligen Anregungsfunk
tionen darstellbar ist. In der Summe mssen diese Anregungsfunktionen
allerdings gewissen statistischen Gesetzen gengen. Fr eine detaillierte
Darstellung sei z.B. auf
14 1 S. 335 verwiesen. Mit Hilfe dieser stocha
stischen Bewegungsmodelle knnen sowohl Accelerogramme als auch Response

Spektren bestimmt werden. Diese Methode bietet nicht nur die Mglichkeit,
z.B. ein konkretes Response Spektrum zu entwickeln, sie kann auch Auf
schluss darber geben, mit welcher Wahrscheinlichkeit ein konkretes
Response Spektrum berschritten wird. Dadurch lassen sich z.B. die erfor
derlichen Sicherheitsfaktoren und das Risiko auf ein Optimum reduzieren.
Als Eingabedaten fr diese stochastischen Berechnungen sind
lediglich
erforderlich :
1) die konstante Spektraldichte So
2) die relative Spektraldichte S ()
3) die einhllende Intensittsfunktion der in Betracht
zu ziehenden Erdbeben (Accelerogramme).
47
Die konstante Spektraldichte So bedeutet fr praktische Anwendungen ledig
lich eine Normierungskonstante. Es bestehen unter anderem folgende Mg
lichkeiten, So zu bestimmen:
1) durch Normierung auf eine vorgegebene Intensitt
2) durch Normierung auf eine vorgegebene maximale Grundbe
schleunigung bzw. Verschiebung
3) durch eine derartige Normierung, dass ein vorgegebenes
Grundspektrum mit dem ermittelten Response Spektrum fr
ca. 25% der kritischen Dmpfung im Mittel mglichst gut
bereinstimmt.
Fr die relative Spektraldichte stehen sehr anpassungsfhige und theore
tisch fundierte Nherungsformeln zur Verfgung (vergi. 13) S. 3 3 9 ) , z.B.:
S. () =
- ^
/
ug
(21)
4g2
= relative Spektraldichte
= charakteristische Grundfrequenz
,g = charakteristische Dmpfungsrate
des Grundes
f
-3
Diese Formel ist sehr leicht an rtliche geologische Bedingungen anpass
bar. Die Parameter
Pramete) und knnen aus aufgezeichneten Seismogrammen ab
g
g
geschtzt werden.
Eine experimentelle Bestimmung dieser Parameter sollte man in Betracht
ziehen, kann fr bestimmte geologische Bedingungen von anderen Gebieten
bernommen werden. Fr eine Abschtzung von gengt die Kenntnis von
vorherrschenden Grundfrequenzen (Frequenzen des B o d e n s ) .
Es ist allerdings einer gewissen Fluktuation dieser Werte Rechnung zu tra
gen. Entsprechende Anhaltswerte fr und sind in der Literatur zu
g
g
finden. Es sei noch darauf hingewiesen, dass bei Annahme einer konstanten
relativen Spektraldichte fr in der Praxis vorkommende Dmpfungsraten b e
reits eine relativ gute bereinstimmung mit Response Spektren, die aus
Accelerogrammen ermittelt wurden, besteht.
- 48 Die einhllende Intensittsfunktion der in Betracht zu ziehenden Accelerogramme kann fr einen bestimmten Standort aus (wenigen) aufgezeichneten
Accelerogrammen abgeschtzt werden. Ihr Einfluss auf das Response Spektrum
ist fr in der Praxis vorkommende Dmpfungsraten relativ gering, so dass
eine Nherung dieser Funktion bereits gute Ergebnisse liefert.
2.8 Anmerkungen
In den vorangegangenen Abschnitten wurden die wesentlichen Mglichkeiten
zur Festlegung von Entwurfsparametern (Accelerogramme, Response Spektren)
schematisch aufgezeigt. Diese Mglichkeiten knnen selbstverstndlich auf
verschiedene Weise variiert und kombiniert werden. Zu Kontrollzwecken wird
man in der Praxis verschiedene, weitgehend voneinander unabhngige Wege
beschreiten.
3.0 Mglichkeiten zur Bestimmung von seismischen Eingabeparametern in
Gebieten mit mittlerer und geringer seismischer Aktivitt
3.1 Allgemeines
Der Rahmen dieser Betrachtungen soll hier auf mitteleuropische Verhltnisse bezogen werden. Die Verhltnisse in anderen Gebieten drften jedoch
hnlich liegen.
Diese Gebiete mit mittlerer und geringer seismischer Aktivitt zeichnen
sich durch sprliche Informationen ber relativ starke Erdbeben aus, die
bei einer Auslegung von Kernkraftwerken zugrunde gelegt werden mssen.
Wenn schon z.B. in amerikanischen und japanischen Gebieten mit hufiger
seismischer Aktivitt und versierten Apparaturen ber unbefriedigende Informationen geklagt wird, so kann man in den mitteleuropischen Gebieten
von nahezu mangelhaften Informationen sprechen. Es stellt sich also die
Aufgabe, mit den verfgbaren Informationen unter Bercksichtigung der berechnungstechnischen Mglichkeiten diese Lcke zu schliessen.
In diesen Gebieten besteht zwar ein relativ dichtes Netz von seismischen
Stationen. Charakteristisch ist jedoch, dass die vorhandenen Seismographen
einerseits zu empfindlich sind und bei strkeren Erdbeben aus den Angeln
springen, andererseits besitzen sie ein zu geringes Auflsungsvermgen,
um detaillierte Aussagen ber den Frequenzgehalt liefern zu knnen. Eine
direkte Ermittlung von charakteristischen Accelerogrammen und Response
Spektren aus den aufgezeichneten Verschiebungsdiagrammen scheint deshalb
praktisch nicht mglich zu sein.
49
Aus den vorhandenen Aufzeichnungen kann man jedoch im wesentlichen folgen
d e , verwertbare Informationen sammeln:
1) Hufigkeit von Erdbeben
2) maximale Grundverschiebung d
3) Intensitt der Erdbeben I
4) Lage der Erdbebenherde
5) Pauschale Angaben ber die vorherrschenden
Frequenzen
Aus diesen Angaben lassen sich weitere, wichtige Daten abschtzen.
3.2 Maximale Grundbeschleunigung a, maximale Grundgeschwindigkeit
und maximale Grundverschiebung d

Mit den in 2.21 genannten Beziehungen kann man a und aufgrund der fr
den Standort ermittelten Intensitt abschtzen. Eine Bestimmung der Kon
stanten von Schwchungsgesetzen fr a, und d entsprechend 2.22 ist damit
ebenfalls mglich. Man sollte sich aber bewusst sein, dass diese Schw
chungskonstanten aufgrund der oben genannten Abschtzungen fr v und a
bestimmt werden mssen, sofern nicht genauere Grundlagen aus Gebieten mit
hnlichen geologischen Verhltnissen vorliegen. Der Vorteil einer besseren
Wiedergabe des Frequenzgehaltes bei Anwendung von
Schwchungsgesetzen
bleibt jedoch wegen der Betrachtung von verschiedenen mglichen Erdbeben

quellen erhalten.
Mit den in 2.21 und 2.22 skizzierten Methoden ist also zumindest eine A b
schtzung von a, v und d mglich.
3.3
Grundspektrum
Das Grundspektrum lsst sich mit den Werten d, v und a entsprechend
2.3
abschtzen.
3.4 Bestimmung von Response Spektren aus a, v und d

Prinzipiell knnen Response Spektren entsprechend der Darlegung in 2.4
bestimmt werden. Die Grsse der Multiplikationsfaktoren ist im wesent
lichen abhngig von der Art der zu berechnenden Konstruktion, von der Dau
er und dem zeitlichen Amplitudenverlauf der Erdbeben. Fr gleichartige
Konstruktionen ndern sich die Multiplikationsfaktoren nicht. Die Dauer
und der zeitliche Amplitudenverlauf der Erdbeben haben keinen wesentlichen
Einfluss fr in der Praxis vorkommende Dmpfungsraten und Erregungszeiten.
In welchem Masse die Multiplikationsfaktoren vom relativen Frequenzgehalt

abhngen, kann mittels stochastischer Bewegungsmodelle bestimmt werden.
Da der relative Frequenzgehalt primr durch das Grundspektrum
festgelegt
- 50 wird (und sekundr durch die Eigenschaften der zu berechnenden Konstruktion) , kann man annehmen, dass sich diese Faktoren mit dem relativen
Frequenzgehalt nicht wesentlich ndern. Es scheint damit gerechtfertigt,
die in 2.4 zitierten numerischen Werte der Multiplikationsfaktoren zumindest fr eine Abschtzung des Response Spektrums zu bernehmen (vergi.
Abb. 3 ) . Eine berprfung dieser Multiplikationsfaktoren fr lokale Gegebenheiten aufgrund weniger, registrierter Accelerogramme ist im Bereich
des vorhandenen Auflsungsvermgens der Seismogramme mglich und sollte
vorgenommen werden. Eine vollstndige berprfung kann man mit wenigen,
zuknftigen Aufzeichnungen von hherem Auflsungsvermgen erreichen.
3.5 Direkte Bestimmung von Response Spektren aus Seismogrammen
und Time Histories
Diese Mglichkeiten scheiden fr den Frequenzbereich, der das Auflsungsvermgen der Seismogramme bersteigt, praktisch aus. Dieser Frequenzbereich ist allerdings fr die mitteleuropischen Erdbebengebiete von Bedeutung. Es ist deshalb anzuregen, die Seismographen in Zukunft so auszulegen, dass dieser Weg beschritten werden kann.
3.6 Stochastische Bewegungsmodelle
Diese Mglichkeit, die weitgehend von empirischen Fakten gelst ist und
vorwiegend auf theoretisch-mathematischer Basis aufbaut, bietet eine echte
Alternative, um aus dem Dilemma mangelnder Daten herauszukommen. Sie erfordert nur wenige, grundlegende Parameter. Die Eingabewerte fr diese
Methode knnen auch fr Regionen mit mangelnden seismischen Daten hinreichend genau bestimmt werden. Damit ist auch fr die genannton Gebiete mit
mittlerer und geringer seismischer Aktivitt und lckenhaften seismischen
Daten nicht nur der Weg ber Response Spektren mglich. Mit Hilfe stochastischer Bewegungsmodelle erschliesst sich die Mglichkeit, dynamische
Berechnungen mittels Time Histories vorzunehmen.
3.7 Anmerkung
Der Vollstndigkeit halber sei noch erwhnt, dass sich die genannten Mglichkeiten nicht auf lineares Materialverhalten der zu berechnenden Konstruktionen beschrnken, sie lassen sich genauso auf nicht lineares Materialverhalten anwenden.
- 51 4.0 Vergleich von mitteleuropischen Response Spektren mit

amerikanischen Response Spektren und Minimalwerten
In Abb. 4 sind charakteristische Grundspektren und Response Spektren von
2% kritischer Dmpfung fr amerikanische und mitteleuropische Verhltnisse aufgetragen. Das angegebene Grundspektrum fr mitteleuropische
Verhltnisse entspricht etwa einer Intensitt VI (Mercalli-Sieberg).
Der Frequenzgehalt entspricht in Mitteleuropa registrierten Werten.
Zustzlich sind in Abb. 4 die fr die Auslegung von Kernkraftwerken entsprechend 10 empfohlenen minimalen Grund-und Response-Spektren gezeigt,
die auch in Gebieten, in denen Erdbeben unwahrscheinlich sind, angesetzt
werden sollten.
Aus dem unterschiedlichen Frequenzgehalt ist ersichtlich, dass sich die
Messungen in Mitteleuropa nur auf kleinere Erdbeben beziehen, die in
relativ grosser Entfernung vom Erdbebenherd registriert wurden. Da diese
schwachen Erdbebenintensitten nicht charakteristisch fr strkere Erdbeben sind, ist eine Auslegung entsprechend den in lOJ empfohlenen
Werten sinnvoller.
5.0 Vergleiche von dynamischen Berechnungen mit Berechnungen

nach Normen und Codes
In Abb. 5 ist ein akutes Response Spektrum und das dem Uniform Building
Code entsprechende Response Spektrum dargestellt. Daraus ist ersichtlich,
dass eine Bemessung entsprechend dem Uniform Building Code eine Unterschtzung um den Faktor 12 ergeben kann, sofern die Plastizittsreserven
des zu berechnenden Bauwerks nicht in Anspruch genommen werden drfen.
Ist eine volle Beanspruchung des plastischen Bereichs erlaubt, so knnen
immer noch Unterschtzungen um einen Faktor von ca. 2 auftreten. Die
genannten Faktoren beziehen sich auf den ungnstigsten Frequenzbereich.
Abb. 6 zeigt die bereits aufgefhrten Response Spektren
pische Verhltnisse und das der DIN 4149 entsprechende
fr Erdbebenzone I und festen Baugrund. Daraus kann man
nach DIN sowohl eine berschtzung als auch eine starke
der Erdbebenwirkung mglich ist.
fr mitteleuroResponse Spektrum
entnehmen, dass
Unterschtzung
6.0 Seismische Berechnung eines Reaktorcontainments mit Einbauten

Siehe hierzu 9.0
7.0
Zusammenfassung
Den vorhergehenden Ausfhrungen ist zu entnehmen, dass die Lcke von

sprlichen seismischen Ausgangsdaten fr dynamische Berechnungen von
Konstruktionen einerseits mit Hilfe von Abschtzungen teilweise ge
schlossen werden kann, andererseits bieten die
mathematischphysikalisch
fundierten, stochastischen Berechnungsmethoden eine echte Alternative,

diese Lcke vollstndig zu umgehen.
An Hand von Response Spektren wurde gezeigt, dass quasistatische Berech

nungen, wie sie beispielsweise nach entsprechenden Normen
vorgeschlagen
werden, sowohl eine berschtzung als auch eine starke Unterschtzung

der seismischen Wirkungen in bestimmten Frequenzbereichen ergeben.
Dynamische Berechnungen knnen also helfen, einerseits das hohe Sicher
heitsbedrfnis bei Kernkraftwerksanlagen
in realistischer Weise zu b e
friedigen, andererseits knnen sie eine unwirtschaftliche
Dimensionierung
verhindern. Insbesondere ist es in vielen Fllen mglich, aufgrund dyna

mischer Analysen Massnahmen zu ergreifen, die ein Ansprechen der Kon
struktionen auf seismische Erschtterungen enorm reduzieren oder ber
haupt ausschliessen.
.0 Eigenvalue Problems by the Method of C onjugate Gradients
The general eigenvalue problem arising in structural dynamics can be

represented with the matrix equation
where
= master stiffness matrix
= master mass matrix
= generalized displacement
= the eigenvalue
(1)
vector
In vibration analysis the order of the stiffness matrix and the mass
matrix respectively is so high that it is too expensive to obtain a
complete eigensolution. For most vibration problems, however, a partial
eigensolution will be satisfying. The importance of the need of an
efficient scheme for the solution of the eigenvalue problem is evident.
- 53 The available solution techniques for eigenvalue problems can be classi

fied into Transformation Methods and Iterative Methods. Among the trans
formation methods the Jacobi method and the Givens method are the most
wellknown. The Jacobi method performs orthogonalization on the original
matrix to achieve a diagonal matrix with the eigenvalues of the matrix on
the diagonal. The Givens method, similar to the Jacobi method, yields a
tridiagonal matrix by transformation. The advantage is that the complete
eigensolution is performed with the transformation methods. A big dis
advantage is that the solution is accomplished by operating on the system
matrices, i.e. inversion of the matrix using the Jacobi method. The facts
clearly make the transformation methods less attractive than the widely
used iterative methods, particularly the conjugate gradient method.
The conjugate gradient method is the widely used iterative method. This
scheme to minimize a function was first developed by Hestenes and Stiefel
( 115) ). Bradbury and Fletcher ( 116) ) developed from that the Rayleigh
Quotient Solution. Also Prato (1171) presents an application of the con
jugate gradient method as a solution technique for static, vibration and
stability analysis. An extension of the technique and algorithm developed
by Bradbury and Fletcher ( 1161) is presented by Fox and Kapoor ( 118)) .
The conjugate gradient method is highly recommended since the algorithm
does not require assembled system matrices. Starting with the lowest
eigenvalue and eigenvector the higher modes are calculated insuring
orthogonality on the previous eigenvectors. It is easily understood that
an accumulation of round-off errors takes place as the number of the eigen
value and its associated eigenvector becomes much bigger than the number
of the first or lowest eigenvalue.
The solution described here uses the wellknown Rayleigh quotient
R () = 4 ^
a us
<2)
which is equal to the eigenvalue when the eigenvector is substituted

into equation (2). The Rayleigh quotient is minimized to obtain the lowest
eigenvalue and the associated eigenvector. For all the higher eigenvalues
and eigenvectors the Rayleigh quotient is again minimized, however, with
the restriction to be in a subspace which is M-orthogonal to the previous
eigenvectors. Assuming that the matrices and M stem from a finite ele
ment analysis, it is seen that an assemblage of the element matrices to
system matrix is not necessary.
- 54 The Rayleigh quotient implies a numerator which is twice the strain energy
and a denominator which is twice the maximum kinetic energy of the struc
ture. The Rayleigh quotient can therefore be accomplished by taking the
sum of the potential and the kinetic energies of the individual elements.
Equation
(2) can therefore be rewritten as:
. 6. k. 6.
R
() = i-i
where
(3)
. m. .
k = element stiffness matrix

m = element mass matrix
i = decomposed generalized
displacement
r = number of discrete elements

The initially estimated vector 0 is changed in each iteration cycle to
wards the eigenvector . Theoretically, the solution is achieved with the
conjugate gradient method after cycles, where is the number of system
degrees of freedom.
The function must be quadratic, which is true for the Rayleigh quotient.
Rounding-off errors, however, will probably require more than steps to
get to a satisfactory convergence. To minimize a function of several
variables the gradient method uses local information about the rate of
change of the function with respect to the changes in the variables and
requires the evaluation of the gradient vector R. The Rayleigh quotient
is a function of variables
(equation 3 ) . It is differentiable and its
gradient vector is
VR
2KA-2RMA
(4)
Hence, the minimization algorithm can be written as:

= arbitrary
(5)
Po =
(6)
R ()
qo = "Po
+ = + "
(7)
(8)
55
Pi+1
= VR
(. + 1 )
_ Pj+iT
i =
(9)
i+i
<10)
p~7~
p
i+i = p i + i
i
+
(11)
From any approximation to a local minimum ., a search is made along a

direction q. to find a better approximation . .. The step length a. in
the q. direction must now be chosen such that the function is minimized in
^1
that direction.
Rewriting equation
(2) using the expression of equation
R (.+.q.) =
(4) yields
(.+ q ) ( + q )
1
1 1
1
=
(.+a.q.)1 M(Ai+aiqi)
(12)
The minimum of the Rayleigh quotient satisfies the equation
tl3)
as: =
Performing this differentiation, equation
uct.
+ . + w = O
where
(13) has then the form
(14)
M q i ) - ( A i T K q ^ IqJ
u = (qj
K q ^ (&J
= ( qi T
Kq i ) ( A i T 1 ) - ( 1 1 ) ( qi T
w = ( 1 K q i ) (j
The two roots of equation
Mq)
(15)
(16)
1 ) - ( 1 .) ( 1 M q i }
(1?)
(14) correspond to the maximal and minimal points
of the Rayleigh quotient in the direction q.. The criterion for a. found
from performed test examples
(stability problems) contradicts with the
idea that the positive a. corresponds to the minimum of the quotient,

while the negative a. corresponds to the maximum.
- 56 Since both roots often appear positive it is suggested to actually calcu

late for both values of a. the Rayleigh quotient. This is done without
any additional computer time since the parameters of (15), (16) and (17)
were already
available and valid for both a.'s. Then the a. associated
J
1
with the lower of the two Rayleigh quotients was chosen to estimate a new
vector . ,.
The second eigenvalue and its associated eigenvector can therefore be
determined by posing a new minimization problem.
2 2
~
2 2
R (2)
is a minimum subject to the orthogonality condition

?
=0
In this restricted subspace the Rayleigh quotient has a unique minimum

equal to 2. The eigenvectors are said to be M-orthogonal to each other.
In general the orthogonality of the eigenvector and' all the previously
calculated eigenvectors ,..., , defining a subspace is established by
means of the so called projection matrix.
The vector set can then be written
and hence
V. = 0
q
i=l, .. . , q-1
^
The projection matrix consists basically of all the previously determined

eigenvectors 1,..., q-1 in form of a more involved matrix, (see ref. 18).
Premultiplication of the gradient vector with the projection matrix
eliminates from the vector the non-orthogonal components. Hence the re
sulting gradient vector is orthogonal to the subspace.
The starting point of the iteration as well as the gradient vector are
projected on to that subspace in order to find that specific intermediate
eigenvalue.
- 57 9.0 Seismische Berechnung eines Reaktorcontainments mit Einbauten

Abb. 7 zeigt das System, das aus Containment, Reaktorsttzung,
Druckbehlter, biologischem Schild und Zwischendecke besteht.
In Abb. 8 ist das dynamische Modell dargestellt.
Die Berechnung basiert auf dem Response Spektrum fr mitteleuropische
Verhltnisse (Abb. 4 ) .
Die ersten vier normalisierten Eigenvektoren und die zugehrigen Eigenfrequenzen sind aus Abb. 9 zu ersehen.
Abb. 10 zeigt die Beschleunigungen fr jede einzelne Frequenz.
Literature
1
NORRIS, C.H., HANSEN, R.J. etc., "Structural Design for Dynamic Loads",
McGraw Hill, New York, Toronto, London (1959)
ZUDANS, ., FISHMAN, H.M., REDDY, G.V.R., CHOW, T.Y., "Technical Report:

Lums, Manual for the Dynamic Response of Lumped Mass Systems Program",
The Franklin Institute Research Laboratories
WIEGEL, R.L., "Earthquake Engineering", Prentice-Hall, Inc., Englewood

Cliffs., N.Y. (1970)
HANSEN, R.J., "Seismic Design for Nuclear Power Plants", The M.I.T.
Press (1970)
"Nuclear Reactors and Earthquakes", TID-7024, USAEC (1963)
HILLER, W., ROTHE, J.P., SCHNEIDER, G., "The Rhinegraben Progress

Report 1969"
IIERSBERGER, J., "A Comparison of Earthquake Accelerations with Intensity

Ratings", Bull. Seism. Soc. Am. V. 46, 1656
NEUMANN, F., "A Broad Formula for Estimating Earthquake Forces in

Oscillators", 2nd WCEE, Tokyo, Japan, 1960
VELETSOS, A.S., NEWMARK, N.M., CHELAPATI, C.V., "Deformation Spectra for

Elastic and Elastoplastic Systems Subjected to Ground Shock and Earth
quake Motions", Proc. 3rd WCEE, New Zealand, 1965
10
NEWMARK, N.M., HALL, W.J., "Seismic Design Criteria for Nuclear

Facilities", 4th WCEE, Santiago, Chile, 1969
11
PENZIEN, J., LIU, S.C., "Nondeterministic Analysis of Nonlinear

Structures Subjected to Earthquake Excitation", 4th WCEE, Santiago,
Chile, 1969
12
JENNINGS, P.C., HOUSNER, G.W., , N.C., "Simulated earthquake

Motions for Design Purposes", 4th WCEE, Santiago, Chile, 1969
13
GOTO, ., TOKI, K., "Structural Response to Nonstationary Random

Excitation", 4th WCEE, Santiago, Chile, 1969
- 59 14
ESTEVA, L., ROSENBLUETH, ., "Espectros de Temblores a Distancia

Moderados y Grandes", Boletin, Sociedad Mexicana de Ingenieria Sismica,
V. 2, No. 1, March 1964
15
HESTENES, M.R., STIEFEL, E., "Methods of Conjugate Gradients for Solving

Linear Systems", Journal of Research of the National Bureau of Standards
Vol. 49, No. 6, Research Paper 2379, December 1952
16
BRADBURY, W.W., FLETCHER, R., "New Iterative Methods for Solution of

the Eigenproblem", Numerische Mathematik 9, pp. 259-267, 1966
17
PRATO, C A . , "Plate and Shallow Analysis by Conjugate Gradients",

Ford Foundation Research Report, R 69-53, MIT, September 1968
18
FOX, R.L., KAPOOR, M.P., "A Minimization Method for the Solution of the
Eigenproblem Arising in Structural Dynamics", Case Western Reserve
University, Cleveland, Ohio, September 1968
19
SCHILDKNECHT, P.O., "Large Order Simultaneous Equations and Large Order

Eigenvalue Problems by the Method of Conjugate Gradients", M.S. Thesis,
Cornell University, Ithaca New York, September 1970
60
Response spectrum
El Centro earthquake 190
from [10]
0j05
0.1
Q2
Frequency, cps
Fig. 1
from [3]
Period of vibration. (sec)
Fig 2
320
fil
Relative Values Of Spectrum Amplification Factors

Percent Of Critical
Damping
0
0,5
1
2
5
7
10
20
Amplification Factor For

Displacement Velocity Acceleration
2,5
2,2
2,0
1.8
1,2
1,1
1.0
from [10]
Fig. 3
.0
3,6
3,2
2.8
1,9
1,5
1,3
1,1
6,4
5,8
5,2
4.3
2,6
1,9
1,5
1,2
fi
>^
O
C
O)
:J
it
DSS/UI
Fiq
')|/\
S3 -
O
O
O
in
o
(NI
Ju
ZI
cr
oas/u; )||/\
from [4] p. 66
Fig. 5
64
8
o
S
in
S
Osi
o
o
o
LT)
o
CM
LT)
O
aas/u }!|/\
Fig.
FIG.7
REACTOR CONTAINMENT
WITH INTERIOR STRUKTURES
FIG.8 DYNAMIC MODEL
MODE 1(3.78 CPS)
*rt
M0DE1 (3.78CPS)
MODE2(7.60CPSI
MODE3(13,68CPS]
MOOE3I13.68CPS)
MODE4 (1856CPS1
FIG.9 MODE SHAPE (NORMALIZED)
MODE 5 11990 CPSl
MODE 6 (28.80 CPS)
FIG.10 ACCELERATION IN g
- 67 DISCUSSION
-^
K. ZILCH, GermanySome comments on the input data of a s e i s m i c design.
May I emphasize that the results of our aseismic calculation can only be as exact as the
input. That means, we have to try to get good input data as well as make p r o g r e s s in the
calculation methods. In a r e a s of low seismicity the great problem is the lack of satisfactory
seismic data. Therefore, all available mathematical and engineering tools should be used to
utilize the data given. In this context.I want to refer to well known probabilistic analysis of
seismic data, for example relationships between the expected r e t u r n periods of earthquakes
and the magnitude (1), (2).
For example: log N = a-bM
N is the mean yearly number of earthquake magnitudes greater than M.
Such methods give at least some hints for a design, and only information on m o r e than one
specified design earthquake enables the engineers to make a r e a l risk analysis resulting in
definitive numbers of reliability and to avoid that some given definitions remain subjective
and open to individual interpretation.
References :
(1)
J. F. Borges and M. Castanheta, "Structural safety", 2nd Edition, Lisbon, Laboratorio

Nacional de Engenharia Civil, 1971.
(2)
L. Ahorner, "Seismizitt und junge Tektonik in den Rheinlanden", Umschau, S. 259,

1967.
H. SHIBATA, Japan
Response analysis to pseudo-earthquakes produced by random noise oscillator
is more reliable than to natural earthquakes for design, if we can find an adequate filter for
ground.
Because through such analysis we can predict the confidence limit of the r e s u l t s of response
analysis. In my personal opinion, the pseudo-earthquake approach for the design analysis
should be more used than poor natural earthquake analysis.
K. UCHIDA, Japan
How do you estimate the stiffness of the cylindrical walls in the r e a c t o r building ?
Are there any differences between the stiffness computed according to beam theory and the
one computed by a shell theory ?
A. E. HUBER, Germany
The stiffness of the cylindrical wall in the reactor building and of the containment
- 68 shell were computed according to shell theory. The difference between shell theory and beam
theory for the r e a c t o r support and the r e a c t o r shield is negligible.
H. RIEKERT, Germany
Are the comments on the advantage of conjugate gradient methods over transform
methods based on theoretical aspects or on practical comparisons. Conjugate gradient methods converge theoretically in n s t e p s , but usually not in practice. So it would be of interest
to know, whether t r a n s f o r m methods as for instance the QR-method or the Householdermethod could not be applied h e r e where advantage could be taken of the s y m m e t r y of the m a t r i x problem.
P. O. SCHILDKNECHT, Germany
As I have pointed out, the convergence of the conjugate gradient method depends
to a high degree on the assumption of the starting vector.
Since we a r e able to make r e a s o n -
able assumptions for this vector as long as we deal with idealized two-dimensional dynamic
p r o b l e m s , the conjugate gradient method (as an energy method) is favorable for large systems
over the transformation methods.
K. AKINO, Japan
In West Germany do you have an actual project to apply containment and vessel
support s t r u c t u r e s which were shown on your slides ?
The slides refer to a feasibility study of the shown s t r u c t u r e in Europe, not in
Germany.
Do you have some plan to include the soil spring under the building in future ?
With our p r o g r a m s it is possible already to include soil foundation interactions
K 2/4
ASEISMIC DESIGN OF STRUCTURES W I T H NUCLEAR REACTORS METHOD OF EARTHQUAKE RESPONSE ANALYSIS
FOR COMPOSITE STRUCTURES EVALUATED
FOR DAMPING EFFICIENCIES
BY MATERIAL AND
STRUCTURE
Y. TSUSHIMA, J. JIDO,
Takenaka Komuten Co. Ltd.,
Technical Research Laboratory, Tokyo,
TYPE
Japan
ABSTRACT
The purpose of this paper Is to summarize analytical procedures which

have been employed in the evaluation of dynamic properties and dynamic
response values for earthquake motions putting emphasis on estimating the
damping capacities of a special structure like a nuclear power plant which
is made of various materials such as concrete, steel, special alloys, etc.,
and various structural types such as walls, shells, frames, trusses, etc.,
considering the interaction between foundation of structure and ground at
time of earthquake motion.
The formulation of equations is discussed including the procedures for
defining the complex stiffness matrix by which the damping capacities of
structures can be estimated as hysteretic damping.
This paper further elaborates on techniques for solving the eigenvalue
equation and the proof of orthogonality properties of the complex stiffness
matrix.
Finally, the numerical results of dynamic properties and dynamic
response analyses are obtained by using computer programs based on the
abovementloned theory using some different earthquake motions for a nuclear
power plant to be constructed in Japan and further considerations and dis
cussions of usage are presented.
1.
INTRODUCTION
structure with a nuclear reactor consists of Nuclear Reactor at the

center, steel Reactor Pressure Vessel, steel concrete Shield Wall, flask
shaped Dry Well and Reactor Building of reinforced concrete walls. The
dynamic properties of this structure are characterized as follows:
(a)
These are composed of different structures and having a common founda

tion, they are interconnected which makes analysis very complicated.
Therefore, these structures of which dynamical models are different,
Interact on each other at time of motion.
(b)
Each has a different weight, stiffness and damping.capacity.
- 70 Especially, Reactor Building has heavy weight in comparison with others.

Response values of Reactor itself, which is the most important part of
this structure, are affected by the dynamical properties of Reactor
Building.
(c)
Reactor Building being spatially constructed of many walls which react

with the external force acting in various directions makes calculation
of stiffness of the building difficult.
(d)
Reactor Building being massive and having a short period, the dynamic
properties (natural period, damping capacity) are.affected by the
interaction between foundation and ground. There is theoretically no
difference in dynamic properties at time of earthquake motion between a
structure like this and normal structures but it must be analyzed by the
best method which can be considered for its dynamic properties.
(e)
Reactor Building has heavy weight and rigid stiffness of which response
values are quite large. On the other hand, Nuclear Reactor has light
weight and less rigid stiffness of which response values are very small.
Thus the difference in these two structures causes several problems at
the time of analysis together. For the reasons mentioned above, the
dynamic analysis must be done in the following two steps:
1st step : dynamic analysis based on structures of heavy weight
2nd step : dynamic analysis based on structures of light weight
using results of 1st step
Considering the abovementioned dynamic properties of structures, a

study has been made which will be explained hereinbelow.
(a)
The dynamic analysis is performed using the model analysis method based
on the lumped mass-spring system.
(b)
The stiffness of Reactor Building and Dry Well are calculated by the
finite element method (F.E.M.).
(c)
Damping capacity is evaluated as hysteretic damping, which can be

evaluated by expressing with complex numbers. Dynamic analysis and
calculation of eigenvalues are performed using the complex stiffness.
(d)
The stiffness and damping capacity of ground are analyzed according to

Taj imi's theory.
Here a series of programs for dynamic analysis of earthquake motions

are developed using an IBM computer S/360 J-75.
2.
STIFFNESS OF STRUCTURES
Reactor Building
Reactor Building is spatially constructed with plain and shell shaped
reinforced concrete walls. Therefore, stiffness of these walls are calculated by F.E.M., which Is the best and most popular method for evaluating
stiffness of structures.
- 71 Dry Well
Dry Well Is a typical thin shell structure of which dynamic
proper
ties must be calculated considering movements in bending and shearing,
which is similar to movement of free end of cantilever with change in shape
of shell. Stiffness of this is also calculated using F.E.M..
Other Structures
Stiffnesses of
other structures like Reactor Pressure Vessel and
Shield Wall are calculated using the bending shear deflection theory.
Details about the stiffness calculation of structures are explained in
the paper, "Aseismic Design of Nuclear Reactor Building Stress Analysis
and Stiffness Evaluation of the Entire Building by the Finite Element
Method," [1],
3.
DAMPING PROPERTIES OF STRUCTURES
Generally, the damping properties of structures are usually assumed as

follows :
(a)
(b)
(c)
(d)
viscous damping due to viscous properties of materials molecules

hysteretic damping due to imperfect elasticity of members of
structures
structural damping due to slip motion at joints of members of structures
radiation damping due to interaction of structure and ground at time of
motion
In conventional response analyses of structures In general, the damping

mechanism of a structure has been evaluated as viscous damping proportional
to the motion velocity of the structure. There are various reasons for this,
but one would be that this can be readily expressed in mathematical terms.
In recent years, vibration test data of a large number of actual buildings
have been obtained, and upon analyses of these data, it is found that in some
structures the ratios (h/) between natural circular frequencies () and
damping ratios (h) are not necessarily constant.
From recent experimental study on dynamic properties of reinforced conc
rete frames the damping properties of reinforced concrete structures have
been reported as hysteretic damping.
From analyses of the above data, it is theorized that the damping
mechanism of a structure does not consist only of viscous damping, but also
of hysteretic damping due to imperfect elasticity of the frame Itself and
radiation damping caused by the interaction between foundation of structure
and ground. A special structure such as one containing a nuclear reactor,
from the very fact that materials and structural types differ, would possess
complicated damping ratios;analysis of earthquake response of such a
structure must be performed by methods allowing adequate evaluation of the
above damping ratios. For the reason that the requirements of such response
analysis are satisfied, and the reason that, according to the dynamic theory
of elastic semi-infinite medium, the radiation damping can be evaluated by
- 72 expressing the stiffness of ground by complex numbers, an experiment was made

expressing the stiffnesses of structure and ground by complex numbers
(KR + 1 K T ) and evaluating the damping ratios of the structure as hysteretic
damping.
If the stiffness matrices are expressed by complex numbers, the equation
of single freedom motion for the dynamic analysis of the structures can be
expressed as follows:
Mi + (KR + 1 K T ) X = -Mx0
(1)
where M : mass, : acceleration of earthquake motion.
Details of the equation of motion will be explained elsewhere.
Imaginary Parts of Stiffness Matricest^J
Therefore, in Eq. (1), the solution of equation may also be assumed as
follows :
. -huit
ijt
, ~\
= Ae
e
(2)
Substituting Eq. (2) into Eq. (1), the following equation is obtained.
,,
KR +
,.
(- + 1 ) 2 + jj
i = 0
(i)
Separating the above equation into real part and imaginary part:
2(1 h 2 ) =
2 2 I
(il)
and assuming the damping ratio (h) to be negligibly small, the approximate
formula can be introduced as follows:
KR
Ki
2
=
h =
(5)
M"
2 K
Therefore, damping ratio (h) is apparently connected with only KR and Kj and
the Imaginary parts of complex stiffnesses are obtained by the following
equation
Ki 2hKR
(6)
Properties of Hysteretic Damping Ratios
On the assumption that imaginary parts of complex stiffness may be
proportional to the real part
KR + IKi = (1 + 12h)KR
(7)
in which the sign h expresses the damping ratio to be established for the
particular system. Substituting the equation = ue
into Eq. (7), the
original equation can be expressed as follows:
(1 + 12H)KRU = X2MU
(8)
therefore, K R U = to02MU where 0 : undamped natural circular frequency.

Expressing the undamped eigenvalue of Jth mode by the sign u0j> the
damped eigenvalue of jth mode (XJ) is obtained by Eq. (8) a:; follows:
Xj = (1 + i2h)u)0j
(9)
Substituting Xj = u>j + iwjhj Into Eq. (9) and separating the resultant
equation into real part and Imaginary part, the following two equations are
obtained :
2/(1 (10a)
j
=
= ">j-/^
- nh)
j
2 _
/,,
,.2
73
hj = hu 0 j 2 Aj 2
(10b)
Eliminating the sign 0< from the above two equations, the damping ratio of
Jth mode can be expressed by the following equation:
hj = (1 + /l+Jh2)/2h
(11)
The damping ratio (hj) is constant by hypothesis and assuming the damping
ratio (h) to be small enough compared with 1, It can be concluded that the
damping ratio of each mode is equal to the damping ratio established by
material and structural type.
Complex Stiffness Matrices of MultiDegrees of Freedom
In calculating the complex stiffness matrices of a composite structure
such as one containing a nuclear reactor, at first, the entire structure is
divided into a number of groups (G) by damping ratios expressed by the sign
G H and the complex stiffness matrices (KR + 1 K T ) Q in the local coordinate
system for each group are prepared. The Individual complex stiffness matrix
contributing to each group is calculated by Eq. (12) from the individual
real stiffness matrix (QKR) which is calculated by F.E.M. and other methods.
[KR + IKIJG = (1 + 12 G H)[ G K R J
(12)
In the second stage, the complete complex stiffness matrix can be

obtained by superposing of the individual complex stiffness matrix in the
global coordinate system of the entire structural assemblage by means of the
following equation.
[KR
+ KTJS
= r[KR
+ IKTJG
(13)
G
As a result, the equation of motion of the entire structure can be obtained
as follows:
[MJs} + [KR + 1 K T J S { X } = [MJ{x0)
(Ik)
k.
INTERA
C TION BETWEEN FOUNDATION AND GROUND
Foundation Ground System

The ground may act in the two directions of propagating earthquake
motions and supporting structures at time of motion. Therefore, It assumes
radiation energy caused by interaction between foundation and ground. The
effect of the interaction mentioned above must be considered during seismic
analysis of important structures such as a nuclear reactor building. In
dynamic analysis of the foundationground system, two effects, one on the
underside and the other on the lateral side of foundation must be considered.
Consequently, the foundationground system must be idealized by a method
which can estimate the stiffness of ground analytically with accuracy at the
time of the abovementioned effects. To accomplish the analysis required, the
foundation assumed to be attached to the ground by means of some springs and
dashpots which consist of lateral springdashpots and a rotational spring
dashpot as shown in Fig. 1 and Fig. k.
The analysis in this paper is carried
out with the application.
Calculation of Stiffness of Ground
74
In preceding studies, .there have been two well known techniques for
calculation of stiffnesses: one of them deals with the ground statically and
the other deals with the ground dynamically being based on the application
of the wave propagation theory. Aiming for practical uses in dynamic
analyses of structures, the coefficient of subgrade.reaction, from which the
stiffnesses can be calculated, Is defined as a linear relation between stress
and strain of soil and expressed analytically as a function of stress of
ground, the shape of a foundation and its area. In addition to effect of
spring, the interaction also contains the effect which is known as dissipa
tion of energy and by which the effect of spring is decreased and effect of
damping resultingly grows. For the'purpose of estimating the effect of
damping capacity, the stiffness must be calculated considering dynamic
properties of the interaction.
This problem of dynamic properties of interaction has been solved using
the experimental results of forced vibration tests by means of exciter.
This has been also analytically and numerically expressed based on the theo
retical displacement of foundation on an elastic semiinfinite caused by a
harmonic force as expressed by H. Tajiml and T. KoboriR. Minai.
From the abovementioned investigations, the stiffness of ground can be
expressed in the form of a complex number as shown below:
(15)
F K () = () + iFC(i)
The complex number () can be analytically obtained as a function of
the exciting natural circular frequency () of which () and () are
respectively described in the form of two curved lines as shown In Fig. 5 .
As it is very difficult to calculate the eigenvalues of structures In the
form of Eq. (15) considering the interaction at time of earthquake motion,
Eq. (15) must be changed into the same complex number form as the stiff
nesses of upper structures which are independent of the excitation by
natural circular frequency. Therefore, firstly the complex stiffness Is
calculated for the fundamental natural circular frequency (mj) by the
equation below.
( ) = () + lpC(un)
(16)
By using this result, eigenvalues can be approximately calculated. Hut
estimating the stiffness being independent of natural circular frequencies,
the radiation damping of ground for higher order modes is usually evaluated
to be less than the actual value.
Actually, as ut has not yet been determined at time of calculation of
the stiffness, at first, the stiffness must be calculated for the approxi
mate frequency and further approximate eigenvalues can be calculated using
this result.
After these successive procedures are iterated several times, elaborate
eigenvalues can be obtained. As the abovementioned procedures require a
great deal of labor and time, for practical purposes of dynamic analysis,
It Is better to use <m which was calculated neglecting damping capacities.
7 )
As the other method for calculation of damping capacity without theoreti

cal calculations, some values of damping ratio are determined by engineering
Judgment and using this result the stiffness and damping capacity are commonly
calculated.
Because of lack of sufficient space to describe in detail the theory of
complex stiffness expressed as a function of exciting frequencies, only the
numerical results
calculated by Tajiml's theory will be introduced. The
abovementioned Investigations made by Tajiml and Koborl - Minai, described
the interaction between foundation and ground on the underside of a founda
tion, but theories regarding effect of surrounding ground on the lateral
sides have not yet been established.
Therefore, for practical purposes the resistance effect of surrounding
ground must be commonly estimated by means of coefficient of subgrade reaction
and some values of damping ratio are determined by engineering judgment and
stiffness and damping capacity must be calculated using these damping ratios.
5.
EQUATION OF MOTION
The structures are idealized by the lumped mass-spring system shown in

Fig. 1. The equation of motion of the multi-degrees of freedoms system,
which Is allowed to sway and rock under earthquake motions ( x 0 ) , can be
mathematically expressed using the coordinate system and parameters indicated
in Fig. 1 as follows:
[M](x) + [K](x) = -[M]{C}x0
(17)
where Eq. (17)

[M]
[K]
Ix)
(x)
(C)
where the matrix [M]
mass matrix
complex stiffness matrix
complex displacement vector
complex acceleration vector
vector of external excitation
mass of point "i"
total mass of foundation
inertia moment of foundation
around axis at point "BE"
inertia moment of i-th mass
around axis at its gravity center
mi
mB
in,,
[M]
(18)
mi
IB
li
mrr
where t h e m a t r i x
[K]
( -
BE)2
+
1=2
I]
(19)
76
"rr
+ i^sr
k2s + i k 2 r
kr.?
k 2
i*
2
r
k S2 + ik S2
K kr,!
iki
ri *+ l*ri
k s l + iki
k 2 2 + ik 2
T+
i^rs
k S s + ik s
k
2s + iks
2i
k r n + ik n
i 2i
2n + l k 2 n
[KJ
(20)
k
ir
n r,
nr
i 'ir
,,
i"nr
K*^n<z T "*" l^rit;

K
K *^5T **" K^n
is + i is
ns
i ns
l2
i i2
ii
ri
J o
kss + i k ss
ln
l in
K
^ril "*"K ^ r
i n2 " ~ K ni
l ni ~ "
nnkr iKnn
n2
rr+i k rr = k R +ik R +/ (ks(y ) + ^
krs + ikfs
i ii
(y ) ) (yH BE ) 2 ) dy+
(k r2rl tk' 2I ,, )*
r,=2 r,=2
(21)
(H r 2 H B E )(H r l H B E )
fHBn
fHBn
= / (ks(y)+ik(y)(yHBE))
dy+
(kr2ri
r, = 2 r,=2
ik r 2
^*
( 2 )
fHBE
/(kg(y)+1ks(y)) dy+
(kr2ri+ikr2ri)
J
r2=2 r,=2
(22)
(23)
si+i k i = Kis+iki,
(21)
k r,i
r,=2
kri + i k ri = k ir + i k ir = k ril ( H H B E )
r,=2
where vectors {}, () and (C)
+ i^
BE + i^BE
x2 + Z
(26)
xi + ix{
_xn + ixn
6.
i '
*BE + i*BE
x 2 + ix2'
(]
(25)
(x)
. (27)
(28)
(C )
+ 1*1
*n
l*'n
EIGENVALUE PROBLEM
Eigenvalue C alculation
For determination of dynamic properties, the damped free vibration
equation (29), obtained from Eq. (17), needs to be anallzed as follows:
[MJ(x) + [KJ(x) = (0)
(29)
In order to obtain the eigenvalues and the eigenvectors, a olution for Eq.
(29) may be made in the following form:
(x) = l x ) e U t
(30)
where the sign expresses the complex natural circular frequency and the
77
sign () expresses the complex eigenvectors.
Substituting Eq. (30) and the
_1
equation {x) = [MJ 2 (Y) into Eq. (29), Eq. (29) is changed as follows:
_1
_1
[KJ[MJ 2 {Y) = A2[MJ[MJ 2 {Y)
''
(3D
1
~2
Premultiplying both sides by [MJ :
_1
_1
_1
_1
[MJ 2[KJ[MJ 2 (Y) X2[MJ 2[MJ[MJ 2 {Y)
or in other form
[KJ{} = x 2 {Y)
(32)
here [KJ is symmetric and [KJ is also symtrie since
_1
_1
_1
_1
[KJT = ([MJ 2[KJ[MJ 2 ) T = [MJ 2[KJT[MJ 2 = [KJ
now in order to obtain the solution, Eq. (32) must be analyzed as follows:
The method of analyzing Eq. (32) consists of two parts;t5J
(a) 1st step: The given matrix [KJ is reduced to almost triangular
(Hessenberg) form [Hj by elementary similarity transformations.
It follows that
[TJ_1[KJ[TJ = [HJ
(33)
Substituting Eq. (33) into Eq. (32), Eq. (3 is obtained as
follows :
[HJ {Y} = x 2 [l] {}
(3D)
(b) 2nd step: The iterative search for the eigenvalues of [Hj is
performed. A natural extension of Hyman's method may be used to
evaluate ( 2 ) = det ( - X 2 I) and any number of derivatives in
an accurate and stable way. By hypothesis, eigenvalues of matrix
[Hj are correspondent to those of matrix [MJ and [KJ.
The j-th eigenvalue (Xj) can be solved through the above procedure and
can be expressed by a real part (ARJ) and by an imaginary part (Xij) as
follows:
Xj = XRj + IXIJ
(35)
Therefore, the J-th natural frequency (mj) and the j-th damping ratio (hj)
are given as follows:
"j = XRJ
(36)
hj = *Ij/XRj
(37)
Eigenvector Calculation
By replacing successively the eigenvalues (Xj) in Eq. (34), the eigen
vectors (xj) are obtained by analyzing the simultaneous linear equation
through a procedure known as the Gauss reduction. The J-th eigenvectors
(j) can be expressed by a real part (xRj) and by an Imaginary part ( X T J )
as follows:
(xj) = (XRj) + i U l j )
(38)
78
Orthogonality Property of Complex Matrix!^ J
Matrix [MJ and [KJ are respectively real and complex symmetric matrix,
let Xi and Xj be eigenvectors and let (x 1 ) and {j) be complex eigenvectors
corresponding respectively to Xi and Xj. Here 1 / j
Then
[Kj(xi) = Xi[Mj{xi) and [KjUjl = Xj[MJ(xj)
(39)
If the first equation in premultiplied by (x|)'r and the second by {xj) T ,
the following are respectively obtained:
{Xj}T[KJ{Xi) = Xi(Xj}T[MJ{xi)
(40)
and
{Xi)T[KJ{Xj) = Xj(Xi)T[MJ(Xj)
(Hi)
Now, if transpose of each side of Eq. (40) is taken, remembering that [KJ
and [MJ are [KJT = [KJ and [MJT = [MJ, the following equation is obtained:
(Xi)T[Kj(xj) = Xi{Xi)T[Mj{Xj)
(12)
Finally, substractlng Eq. (Il) from Eq. (42), the following equation is
obtained.
(XiXj){x1)T[MJ(xj) = 0
(43)
Therefore, since X^ f* X j , by hypothesis, it follows that
XijTtMjXj) = 0
1 fi J
(44)
On the other hand, since X^ = x., by hypothesis, it follows that
{Xi)T[Mj(Xj) = c'
1 = j
(45)
where C Is a complex number.
If the matrix [KJ is a symmetric real matrix and the matrix [MJ ir a
unit matrix, C is equal to 1.
7.
RESPONSE ANALYSIS
The dynamic response analysis of a structure is performed by solving

the equation of motion Eq. (17). There are two different techniques for the
dynamic response analysis. One of them is step by step direct integration of
Eq. (17) and the other is the mode superposition method well known as the
modal analysis method. The dynamic response analysis in tills study is per
formed by the modal analysis method, using eigenvalues and eigenvectors.
The basic concept of the modal analysis method is that the displacement
vectors (x) and the earthquake motion vectors (C ) can be expressed as a line
ar combination of vibration modes ar, follows:
+
+ q
( l(,)
(x) = q,(x,} + q 2 (x 2 ) +
V*.]1 +
ml { x ml J + W
'
(C) = i(xi) + e 2 fx 2 ) +..+ j {x j ) +....+ m_](xm_i) + mfXrn)
(17)
In the above equations qj and < represent a modal displacement and a
participation coefficient of jth mode. The modal amplitudes consisting of
acceleration, velocity and displacement are functious of time (t) and j is
expressed by the following equation:
j =
r=l
rxrj
r=l
rxj2
(/,J)
79
Substituting Eqs. (16) and (47) into Eq. (17)
m
m
[Mj(x,)q + [Kjtxjjq, = x 0 [Mjx^JS,
(49)
J
J
J=i
j=i
j=i
Then premultlplying by the transpose of an arbitrary modal vector Xi T which
is not the same as the jth mode and taking advantage of the orthogonality
properties, a single uncoupled equation of motion for jth mode is obtained.
q, + X 2 q] = BjXo
(50)
or
q\0 + X,2 = x
(51)
where
q j 0 = Qj/j
For arbitrary loading earthquake motions, the solution of each modal
response Eq. (51) can be performed by the Duhamel Integral.
When the modal responses consisting of accelerations, velocities and
displacements for all significant modes have been determined at any time
"t", the response values of mass points at this time are then obtained by
Eqs. (52), (53) and (51).
m
.
m
.
() = j q J o (xj)
(52)
() = j q J o ( Xj }
(53)
() =
j q'j0 (Xj) + x0
(54)
If the stiffness matrix [KJ of Eq. (17) is complex, accelerations,

velocities and displacements which are obtained through the above procedures
are also expressed by complex numbers. In practice, the response values are
expressed by the absolute values of complex numbers. It may be necessary to
Include all mode shapes to obtain adequate accuracy by modal superposition,
but the response values can be approximated reasonably well by superposing
the first few modes essential for calculation by judging wnether the partici
pation functions are large or small for each structure of the entire structure.
8.
NUMERI
C AL RESULTS
For further consideration and the usage of the abovementioned theory,

the following analyses were performed using computer programs based on the
abovementioned theory for a nuclear power plant to be constructed in Japan.
Model for Dynamic Analysis
Fig. 2 shows a section (xx section) of a typical nuclear power plant.
An equivalent lumped massspring system, shown In Fig. 3, is constructed to
approximate many sizes of structures and another consisting of two lateral
springsdashpots and a rotational sprlngdashpot shown In Fig. 4, is also
constructed to approximate the interaction effect. According to the com
puter programs, the entire structure Is divided into a number of blocks (B)
and each block is further divided into a number of groups G) according to
damping capacities by the use of damping ratios ( Q H ) .
This nuclear power plant was divided Into one block and seven groups as
shown In TableIV and Fig. 3
- 80 Calculation of Dynamic Properties

The stiffness matrix of each group and stiffnesses of ground were indi
vidually calculated as follows:
(a) The stiffness matrices (real part) of Reactor Building (G-l) and Dry
Well (G-2) were calculated by F.E.M. and those of Shield Wall (G-3),
Truss (G-4), Skirt (G-5), Reactor Pressure Vessel (G-6) and Stabilizer
(G-7) were calculated by the bending shear deflection theory.
(b) The complex stiffness matrix of each group was calculated by Eq. (12)
putting the damping ratio (QH) of each group shown in Table IV into
this equation. The damping ratio of each group can be estimated from
the damping ratios shown in Table III considering the material and
structural type of each group.
(c) The complex stiffness of ground was calculated in two ways.
o Case A : The real part of complex stiffness was approximately calcu
lated by the method based on Tajimi's theory and the imaginary part of
this was done putting the damping ratios into Eq. (12). The damping
ratios used in this calculation are shown in Table III.
o Case : The complex stiffness was theoretically calculated by the
method based on Tajimi's theory putting the fundamental natural circular
frequency into Eq. (16). These are shown as case 3 in Table III.
These numerical results are shown in Tables I and II respectively.
Calculation of Natural Periods, Damping Ratios and Participation Functions
The calculation was performed in three ways,
case 1 : neglecting swaying and rocking of foundation (FIX)
case 2 : considering swaying and rocking of foundation (R&S) and using the
stiffness of ground shown In Table I.
case 3 : same as case 2 but using the stiffness shown in Table II.
The numerical results are shown in Table V to Table VII and In Fig. 6
to Fig. 9. Table VII shows the natural periods and damping ratios of case 1
and case 3. Judging from the results, the damping ratio of each mode, of
which mode shape is largest in each mode, is nearly equal to the damping
ratio ( Q H ) of the group which has the most large mode shape. If two groups,
of which damping ratio (G1*) a r e n o t equal, have a large mode shape respec
tively in the same mode, the damping ratio of this mode is between these
two damping ratios ( Q H ) of groups according to their motions. This is the
reason why the damping ratios of higher modes.do not amount to overdamping.
Table V shows the first five periods and damping ratios of which mode
shapes are especially large in all modes. It is shown in Table V that the
damping ratios of each mode in case 3 increases gradually as mode number
increases from 1st mode to 3rd mode.. It may be the reason why the damping
ratio of higher mode is affected by radiation damping caused by the inter
action. But it is also shown in Table V that the damping ratio of case 3
does not increase as mode number increases higher, than number 3. This may be
explained by the fact that the stiffness of ground was evaluated using the
- 81 approximate fundamental natural circular frequency by hypotheses, and that

the stiffness of ground is constant as a result.
Table VI shows the comparison of natural periods, and damping ratios
among three cases of whole structures.
Fig. 6 shows the first three participation...functions of Reactor Building.
Fig. 7, Fig. 8 and Fig. 9 show those of the entire structure for case 3.
Calculation of Dynamic Response Values for Earthquake Motions..
1
The dynamic response values were calculated for case 2 and case 3 to
make -clear the relationship between the damping ratio and the response.
(a) El Centro I9I0 N-S component (max. acceleration = 300 gal)
(b) Taft
1952 -W component (max. acceleration = 300 gal)
Fig. 10 and Fig. 11 show the max. values of displacement and overturn
ing moment for case 3 respectively. Fig. 12 shows the comparison between
case 2 and case 3 for the overturning moment to the Taft 1952 -W component.
Fig. 12 shows that the values of case 3 are smaller by about IO? than
those of case 2.
9.
CONCLUSIONS
(a)
In short, this method can be described in the following 3 steps.

Evaluate the damping capacities of individual structures as hysteretic
damping by material and by type of structure and evaluate the radiation
damping of interaction between foundation and ground.
(b)
Calculate the natural periods and damping ratios of each mode for the
entire structure by using the complex stiffness matrix based on step one.
(c) Make the dynamic analysis of entire structure for earthquake motions by
using the results obtained in step two.
Up till now, analytically and theoretically, it has been very difficult
to evaluate damping capacities which are more accurate and satisfactory; by
this method it may be seen this has become possible.
ACKNOWLEDGEMENT
The authors wish to thank Dr.. Hiroshi Taj imi, Professor, of Building
Structure, Nihon University and Mr.. I. Funahashl, Research. Manager of
Takenaka Technical Research Laboratory for their valuable advices,_ and
the cooperation of structural engineers ..engaged, in the design of the nuclear
reactor building mentioned in this paper.
[1J
[2j
REFERENCES
Authors: Y. Hayamizu, Y. Tsushima and K. Nishiyama
. Tajimi "Theories of Vibration of Buildings," Building Structures
[3J
Series. Vol. 17, May I965, Corona Publishing Co. (Japanese)

H. Tajimi, "Interaction of Building and Ground," Earthquake Engineering,
[1]
[5J
November I968, Shokokusha Publishing Co. (Japanese)

C. R. Wylie, Jr. "Advanced Engineering Mathematics"
B. Parlett "Laguerre's Method Applied to the Matrix Eigenvalue
Problem," I963
82
TABLE I
OASEA
TABLE OASEB :
STIFFNESS of GROUND
STIFFNESS of GROUND
(K(txk)
fKe(t/d)
( t<n/rnd))
STIFFNESS
KR( t o r r a d )
STIFFNESS
KM 341'+;541'
KB?
Z S 4 X 1 0 ' + 23 4 X 1 0 1
KSI
1 0 * 1 0 5 + 3.06X10*
Kat
S24X105 + . W 8 X 1 0 *
KR
2.44X10 l l + i Z 4 4 X 1 0 "
Z42X10U+i 250X1 0U
TAULE H
DAMPING
RATIOS
by MATERIAL and STRUC TURE TYPE
MATERIAL and S'lRUC TUKE TYPE
for
CASE CASE : A S E
1
2
J
REINFORC ED C ONC RETE

0JD5 005 05
STRUCTC RE
for
DAMP I N G
RATIO
(A )
STEEL STRUC TUHE
for INIERnON
tEJWEEN
FOUNDATION
and
GROUND
LATERAL NOTION
of
S I D E GROUND
LATERAL MOTION
of
BA SE GROUND
VERTICAL NOTION
of BA SE GROUND
O01 0J01 0.01
005 005
0O5 0259
005 0051
83
TABLE IV DAMPING RATIOS of GROUPS
GROUP STRUCTVRE NAMES of
NAME NVCLAR POAQSR PLANI
Wf
skwn
G0
FOUNDATION
G1
REACTOR B U I L D I N G
0.0 5
G2
DRY WELL
0.01
G3
SHIELD WALL
aos
G4
TRASS
0.0 1
G5
SKIRT
0.01
G6
REACTOR PRESSURE
VESSEL
0.01
G7
STABILIZER
TABLE V C OMPARISON TABLE
TAB. m
PERIODS and DAMPING RATLOS for

CASE
PERIOD
(SEC)
DAMPING
PERIOD
RATIO
MODE
(SEC)
(A )
CASE
REAC TOK BUILDING
CASE
MODE
A
VODE
At
DAMPING
RATIO
( A )
MODE PERIOD
(SEC)
3
DAMPING
RATIO
( A
0.1927
0.0498
0.2488
0.0498
02421
0.0 7 9 9
0.1032
0.0495
0.1223
0.0498
0.1 1 9 1
0.0 9 3 4
0.0690
0.0 4 8 9
0.0835
0.0 4 9 1
0.0812
0.1232
14
0.0483
0.047 8
0.0 7 6 1
0.0 4 8 2
1 1 0.0595
0.0 7 3 8
16
0.0 3 4 9
a0499
1 1 0.0597
0.0485
16
0.0556
MODE NUviBER OF REAC TOR BUILDING

MODE NUMBER OF THE ENTIRE STRUC TURE
0.0466
84
TABLE \J
DOMPA RISON TA BLE :
PERIODS and DA MPING RA TIOS for

CASE
CASE
THE ENTIRE STRUC TURE
CASE
MODE
P E R I O D DAMPING
RATIO
(SEC)
(A)
AS.
PERIOD
(SEC)
DAMPING
RATIO
(A)
PERIOD
(SEC)
DAMPING
RATIO
(A)
0.1927
.0498
02488
00498
02421
00799
Or032
O0 4 9 5
01223
O0498
01 19 1
O0934
0.0850
0035?
O0852
O0364
O0851
O0362
0.0799
aa
O0835
O049 1
O0812
01232
0.0690
a0488
O0799
ao KM
O0 79 9
O0 1 0 3
MODE NUMBER OF tHBL'ENT.1 RB ! STEUGTURE
TABLE VI
ODE
SUPERIOR GROUP of EAC H MODE in C ASE

for THE ENTIRE STRUC TURE
1 and C ASE 3
CASE
(SEC )
1
k
01927
O0498
O1032
O0495
O0850
O0359
0.0799
O0 1 0 1
5'
O0 6 90
O0488
SUPERIOR GROUP
1 2
O
o m
m
tf 0
4
O
O
CASE
7 (SEC )
O0 o
0 0
0.2 4 21
O0799
0.1 1 91
O0934
o o o
0.0851
0.0362
O0812
01232
0.0 7 9 9
0.0 1 0 3
o o o
F irst five madet of all modes are shown
S U P E R I O R GROUP
1 2
m
m
*
' )
o o
o o o
o o o o
85
Xn
nRWP
AXn
2 ~ n
MA.SS NUMBERS o/ THE STRUC TURE
HE
ORIGIN of C OORDINATE for fC C KING MOTION
KR
HEIGHT of THE POINT
(m)
HB
(cJVrad)
BE
ROCKING STIFFNESS of GROUND
(rad)
ANGULAR DISPLAC EMENT of ROC KING

MOTION
()
EARTHQUAKE DISPLAC EMENT

GRAVITY C ENTER of FOUNDATION
HEIGHT of THE POINT
(TO)
Ke ( y ) ( M)
IB
H
FIG.
SWAYING STIFFNESS of GROUND
(U)
DISPLACEMENT of THE POINT
(m)
HIGHT of THE l-tk
BE
MASS POINT
COORDINATE and PARAMETERS of MOTION SYSTEM
86 -
12000__
llL,
s-:
;.- l> . 1/ .M M V * _ > L V _ M | ,
G1
REA CTOR
BUILDING
G2
DRY W E L L
G3
SHIELD
G6
RE
A CTOR
WA LL
PRESSURE
VESSEL
FIG.
TRANSVERSE SEC TION
TPWnlFl 46.0)
(XX SEC TION)
o f ENTIRE STRUC TURE
60
TPt?t(n iti)
G-1
5 70
TP 7(Fl 9.7)
REACTOR
BUILDING
23
__IP2SD(F122B)
G-6
REACTOR PRESSURE
VESSEL
G-7
STABILIZER
,
- ^
5,80
._tP_2.2(FL16.2)
. TP 1-.5IFI 8.5)
TP 6JK Fl 0.00)
le-iliF^-ttJ)
4 00
I07IF-I6.7)
FIG.
EQUIVALENT LUMPED MASS SPRING SYSTEM of ENTIRE STRUCTURE
87
Kv^KaiKa 2
Ksi i Ksi
g.,g t
BASE GROUNO
FIG.
EQUIVALENT FOUNDATIONGROUND SYSTEM
U):
N
A TUR
A L CIRCULA R
FREQUENCY
UJi:
1ST MODE NA TURA L

CIRCULAR FREOENCY
FKUJ) .
Fk(.uj).i F Ctw)
FK(UA) = Fk(m).i F CCm)
FIG. 5
COMPLEX STIFFNESS of GROUND
PARTICIPATION
FUNCTION
FIG. 6
CASE 1
CASE 2
* C ASE 3
C OMPARISON FIGURE: PARTICIPATION FUNC TIONS of THREE CASES for

REACTOR BUILDING
88
2.0
PARTICIPATION
FUNCTION
G1
FREQUENCY 2
0.670 I 0 ! 0.108'O 3
FREQUENCY
0.2601' 0.2071'
PERIOD (SEC)
0.242
DAMPING RATIO
&
JJg
TN|2I
0.0799
G6
* 4
.' 1
. \x
G4
.
-
-i
16
G/2
IO
FIG. 7
10
1ST MODE PARTICIPATION FUNC TION of CASES for ENTIRE STRUCTURE
20
PARTICIPATION
FUNCTION
G1
FREQUENCY2
O.27610* 052010'
FREQUENCY
0.527icf 0A9310'
PERIOD (SEC)
DAMPING RATIO
FIG. 8
0.9
0.0934
2ND MODE PARTICIPATION FUNC TION of CASE 3 for ENTIRE STRUCTURE
- 89
fl
PARTICIPATION
FREQUENCY"?
O.W5*tf
0.395'0
FUEQUENC
O.7MIO'
OJ6T>O
G-I
FUNCTION
PERIOD ISEO
MMPi.0
1
O.OaS
ATiC
0 036?
6 ...
+
I
I
"o
G2
I,
>
.'
G5
^^V
03
FIG. 9
3RD MODE PARTICIPATION FUNC TION of CASE 3 for ENTIRE STRUCTURE
B1
MAXIMUM
G1
DISPLA CEMENT
R0CKING>5VAY
/ /
///
///
'//
I"
'i 3
' /
i /// /
I ///
CENTRO 1940 NS
TAFT 195? EW
f // /
DESIGN
MAX.ACC.300GAL
// /
li !
VI /
/
2.
3.
4.Cm
DISPLACEMENT
FIG. 10
RESPONSE VALUE (1) of CASE 3 for REACTOR BUILDING
90
B l G1
A
M XIMUM OVERTURNING MOMENT
El CENTRO 1940 NS
. lAFI 19S2 EW
DESIGN
MAX. AC C . O G .
10 10" t m
MOMENT
FIG.
11
RESPONSE VALUE
Bl Gl
(2)
of
C ASE
3 for
REAC TOR
BUILDING
A
M XIMUM OVERTURNING MOMENT
TAFT 1952 EW
MAX.ACC 300GA L
30 ' IO6 im
MOMENT
FIG.
12
RESPONSE VALUE (3) of CASE 2 and CASE 3 for REACTOR BUILDING
- 91 -
DISCUSSION
T. H. LEE, U. S. A.
In the s o i l - s t r u c t u r e interaction, the radiation damping is a function o the ex-
citing frequency. In your Table III, constant values were given for radiation damping. Are
these values the average values over the frequency range or the maximum values ?
J. JIDO, Japan
Table III in my paper shows the damping ratios to this p a r t i c u l a r s t r u c t u r e in
Japan. I suppose the other s t r u c t u r e s would have different damping ratios from this example.
K. UCHIDA, Japan
You use the absolute values as the expression of displacements in your paper. I
consider it better to make the displacement projected on the r e a l axis. What do you think of
my consideration ?
.
J. JIDO, Japan
I am studying whether it is better to consider the actual r e s p o n s e displacement

as the real part of complex displacement or as the absolute displacement.
L. ESTEVA, Mexico
You mentioned that you wanted to take into account hysteretic damping in t e r m s
not of an equivalent viscous damping, but rather in t e r m s of a h y s t e r e t i c , variable stiffness

analysis. However, your equations do not show that stiffness is actually v a r i e d from step to
step. How did you t r e a t the hysteretic damping then ?
J. JIDO, Japan
I can t r e a t the hysteretic damping by expressing the stiffness of s t r u c t u r e by

complex number ( K R + iKj).
Because by expressing the stiffness of s t r u c t u r e by a complex number, the damping ratios
become constant to each mode and as a result this damping can be considered as hysteretic
damping.
K 2/5
ASEISMIC DESIGN OF NUCLEAR REACTOR
BUILDING
STRESS ANALYSIS AND STIFFNESS EVALUATION

OF THE ENTIRE
BUILDING
BY THE FINITE ELEMENT METHOD

Y. TSUSHIMA, Y. HAYAM1ZU, K. NISHIYAMA,
Takenaka Komuten Co. Ltd.,
Technical Research Laboratory, Tokyo,
Japan
ABSTRACT
The purpose of this paper is to evaluate the spatial characteristics of

stress and stiffness of a nuclear reactor building having a complex wall
arrangement, a normal tendency of nuclear reactor buildings, by the finite
element method.
At first, the method of analysis is explained. Using this method,
interaction effect between wall and shell, axial effect of wall at right
angle to direction of lateral force and then stress distribution of each
wall and shell can be obtained.
Subsequently, an analysis example is shown in fairly great detail.
This example is of an actual nuclear reactor building and analysis is performed in order to design this building. Stiffness evaluation has been
performed by using normally the bending-shear theory. Finally, results of
stiffness evaluation obtained by using the method explained in this paper
are compared with that of the bending-shear theory.
1.
INTRODUCTION
Generally, the structural components of nuclear reactor buildings are

walls, shells and slabs which have some irregular openings and also are
forced to be of a complex arrangement spatially. Accordingly, in order to
obtain accurate stress distribution and perform stiffness evaluation of the
entire building, a structural analytical model is required, and that model
should be as near as possible to the actual condition of a nuclear reactor
bullding.
Up to now, many engineers have used the bending-shear theory for the
purpose mentioned above. The bending-shear theory is brief and simple, and
is an effective method of obtaining approximate stress distribution and
stiffness of a building structure. However, It is impossible to evaluate
and to obtain the following effects and values of which when the structure
has a complex wall arrangement:
(1) interaction effect between wall and shell
(2) axial effect of wall at right angle to direction of lateral force
- 94 (3) share ratios of each wall and shell for entire external force
() stress distribution of each wall and shell
In order to solve these problems, In this paper it Is attempted to
analyze stress distribution and stiffness evaluation of the entire structure
by using the finite element method and its application. This analysis method
can evaluate the spatial characteristics and this is the new method of the
authors.
In this method, eacli wall and shell is treated individually as an
assembly of flat elements, and Is reduced to a small order stiffness matrix
which has the vector of nodal displacements on floor level and on vertical
edge surface at some intervals. Subsequently, these small order stiffness
matrices are superposed considering the actual condition of the entire
structure. This superposed stiffness matrix represents a stiffness matrix
of the entire structure. Giving appropriate forces to the entire stiffness
matrix, each nodal displacement mentioned above is computed by Gauss Reduc
tion or other method. Stress analysis of each wall and shell can be com
puted from these nodal displacements.
2.
METHOD OF ANALYSIS
In this section, the above method is described in detail. complex

wall arrangement such as in a nuclear reactor building must be analyzed
essentially as a three-dimensional problem. Although the finite element
method is a good method for evaluating the stress of wall and shell, it Is
necessary to solve very large order simultaneous equations, because con
siderable time and core volume is required. Sometimes it is impossible to
solve the entire structure as a three-dimensional problem In the present
condition. Therefore, it is needed to divide the entire structure into
single walls or shells which are effective for aseismic design and possible
to compute.
One wall is treated as an assembly of triangular and rectangular flat
elements which have two nodal displacements at each nodal point. While, by
assuming that the behavior of a continuously curved surface can be ade
quately represented by the behavior of a surface built up of small, flat,
elements, a shell is treated as an assembly of trapezoidal, flat, elements
which have three nodal displacements and two rotations at each nodal point.
Since until now plane stress and plate bending problems by the finite
element method have been introduced in many reports and are well known,
hence it will not be necessary to explain them in this paper.
Now, the method mentioned above is explained by the analytical model
which is a structure composed of four walls and one shell as shown in Fir.
1(a). This model Is divided into the individual structural components as
shown in Fig. K b ) . At this time, some nodal points are selected properly
on the vertical edge surface of each structural component at some intervals,
in which the vertical nodal displacements of the structural components agree
with each other. These structural components are analyzed individually by
95
using the finite element method and are reduced to the smaller order stiff
ness matrices with the nodal displacements, 2 ^ 5 in lateral direction "u"
and, 6 . 13 In vertical direction "v" is shown in Fig. 1 (C).
Let [KA], [ K B ] , [KC ], [KD] and [] be the five stiffness matrices of
each structural component, and [K A ] = [K E ], [K B ] = [K D ] by assuming the
symmetry of structural model on the vertical center plane at right angle to
u direction. Then the stiffness matrix of the entire structural model
becomes as follows:
[K S ] = [KA] + [KB] + [KC ] + [KD] + [KE]
(1)
In which [Kg] is the stiffness matrix of the entire structure. The relation
ships between the external forces and the displacements become respectively
[[] + []](> = [ K A E ] ( V A ) >
= [[KB] + [KD]]
BD
11'
BD
21'
BD
31'
VA
12
C
'21'
C
22
[KC]
V
BD
K
12' 13
BD
BD
K
22' 2 3
BD BD
K
32' 33
Tf B D
K
'
u
V
(3)
'1
c
11'
(?)
In which, () and (u) represent the vectors of the lateral external forces
and displacements. All lateral nodal displacements on each floor level are
assumed having the same value in this structural model. Also, (q^), (qg)
and (v^), (vg) represent the vectors of the vertical forces and nodal dis
placements on the vertical edge surface of each structural component. These
vectors can be written respectively as follows:
^6
(pl
iqA>
q7
q8
q
.-n
(qR>
q12
^ >
vg
-13
N
3
' {VA'
un
(u) =
11
11
"io"
,(vBi =.
vil
V12
v
13
10
()
To obtain the stiffness matrix of the entire structure [Kg], substituting Eqs. (2), (3) and () into Eq. (1), the relationship between the external
forces and displacements becomes as follows:
C'A > =
^B
BD
C
K
ll + 11
BD
21
BD
K
+ K
31
21
,BD
12
BD
32
BD
C
13
12
BD
BD
K
33
+ K C
22
(6)
- 96 and the stiffness matrix of Eq. (6) represents the stiffness matrix [Kg] of
Eq. (1).
Hence, when the external forces of Eq. (6) are known, the lateral displacements of each floor (u) and the vertical nodal displacements ( v ) , (Vg)
can be decided solving this equation, and the stress distribution of structural components can be computed easily by using these solved displacements.
It is needless to say, in the stress analysis, these displacements are given
at the specific nodal points 2*13 as the boundary condition.
Then, multiplying the stiffness matrices of the structural components
by these displacements mentioned above the results obtained are the external
forces acting on these structural components. By computing the ratios of
these acting external forces to the total external forces acting on the
entire structure, those ratios are the share ratios of lateral external force
when the deformation of floor slab is ignored.
Next, if (q^l and (qg)are null vectors, eliminating the vertical nodal
displacements fv^} and (vg), the stiffness matrix of Eq. (6) is reduced to
the stiffness matrix concerned with (p) and (u), i.e. it becomes as follows:
(p) = [KR](ul
(7)
where [Kp] Is the stiffness matrix reduced into the order of vector (u).
Supposing (p) is expressible in the inertia forces which will be generated
In the structure during a earthquake motion, that is ...
fp) = - [M]{ii)
(8)
where [M] = the mass matrix and (ii) = the vector of the absolute acceleration on each floor level, respectively. Substituting Eq. (8) into Eq. (7),
the following equation can be obtained:
[M](l + [KR]{u) - <o)
(9)
This Eq. (9) is the equation of motion without damping, and thus the matrix
[Kp] Is the stiffness matrix which is required to perform dynamic response
analysis.
The method explained here still has some problems remaining which are
described below.
1.
On the vertical edge surface between wall and shell, the compatibility of the vertical nodal displacements Is satisfied, but
the compatibility of the lateral nodal displacements and the
rotation components are Ignored.
2.
The wall at right angles to the direction of lateral force affects axial forces only, but bending effect resulting from wall
thickness is ignored since its effect Is very negligible.
3.
Since core volume of computer limits the nodal points number which
can be established on the vertical edge surface of wall and shell,
the specific nodal points, at which the vertical nodal displacements agree with each other, are defined properly at some intervals
so as to be able to expect effect for axial forces.
- 97
Now the blocK diagram showing the computational procedure of this
method is shown in Fig. 2.
3.
EXAMPLE OF ANALYSIS
3-1.
Outline of Building
This model is of a nuclear reactor building of BWR type having power
generating capacity of 500 MW In Japan.
The plan of ground floor and the section A- in X direction is shown in
Fig. 3(a) and (b). This building has four stories having 15.5 m total
height above ground level, and two stories having 16.70 m total depth below
ground level and it is supported by a stiff and deep shale layer. This
building is completely square in plan at the lower part having a length on
one side of 63.00 m, while the upper part is also considerably symmetrical
and thus the influence of tortion may be'negligible. The structural compo
nents of the building are made almost all of reinforced concrete except for
the steel roof truss. Major items of specification of concrete are as
follows :
Compressive strength at weeks
Young's modulus of elasticity
Polsson's ratio
F c = 225 kg/crrr
E = 2.1 15 kg/cm^
= 0.1667
The dynamic characteristics of this nuclear reactor building are

described in Reference [VIII].
3-2.
Modellzatlon of Building
At first, the actual nuclear reactor building Is replaced by a structur
al analytical model, and the thicknesses of the structural components are
assumed. In preliminary design, the approximate stiffness of the structural
model is computed by using the bending-shear theory. The earthquake response
analysis is performed by using this approximate stiffness. Then, the thick
nesses of structural components are changed considering shear forces and
deformations obtained from the results of the earthquake response analysis.
Again, the approximate stiffness of the structural model is computed by using
the bending-shear theory by changing the thickness of structural components.
This process is performed repeatedly until satisfactory results are obtained.
When results of response analysis become fairly adequate for design
purposes, the stress analyses of structural components are examined by using
the method explained in this paper as final static analysis, and if results
of stress analyses are satisfactory, detail member designs for the entire
structure can be started.
In order to perform the above static stress analysis, external forces
are needed, and generally, these are the building weight itself and vertical
and lateral inertia forces which will be generated in the structure during
earthquake motion. These external forces are used in this paper, and the
lateral external forces only are shown in Table 1. These forces are the maxi
mum values of Inertia forces at each floor level.
- 98 The spatial arrangement of walls and shell which will be effective as

resistant structural components for external forces is shown in Fig. . The
structural components (walls and shell) numbers are 32 for the X direction
and 29 for the direction. The thicknesses and the divided mesh condition
of' a typical wall and shell In these components are shown in Fig. 5 (a) and
(b). These figures show the number of the floors.at the left hand and the
number of the selected nodal points to satisfy the vertical nodal displace
ments on the edge surface of structural components. AI30, a "t" at the
right hand and inside of these figures shows thickness (mm). Of course,
in these structural components, the most numbers of nodal points, are in the
flask type shell as shown In Fig. 5(b), and it is needed to solve the about
3000 order simultaneous equations for a half part of this shell.
After superposing all the stiffness matrices of the structural components
reduced to the matrix order of the lateral displacements and the selected
vertical nodal displacements, the stiffness matrix of the entire structure
considering spatial characteristics is obtained, and In this case the Btiff
ness matrix is of the order of I80 180 square. Of these ordere, 13 are the
lateral displacements and the remaining orders are the vertical nodal dis
placements, at the selected nodal points.
32.
Results of Analysis
In this section, the results of analysis are shown. The results of
W2, WC, WH, WS1 structural components are shown mainly In Fig. , because of
limitations of space.
a)
Share Ratio of Lateral External Forces
The share ratios of the lateral external forces In both directions are
shown in Fig. 6. From this figure, it can be found that major parts of the
lateral external forces are acting on the main walls at the outside and near
the outside. On the other hand, the share ratios of the shell at the middle
part is comparatively of small values except at floor levels 6 and 8. This
is because the widths of main walls become narrow nearby these floor levels.
b)
Deformation Condition
The deformations of W2, WC and WH are shown in Fig. 7 (a), (b) and (c),
respectively.
Fig. 7 (a) shows the deformation of W2 which is parallel to the direc
tion of lateral forces.. The deformation condition Is ..almost linear and the
top displacement is 2.1 cm to the base. Shown by the ratio of this dis
placement to the total height of the structure, R = 0.33 x 10"3.
Fig. 7 (b) and (c) show the deformations of WC and WH which are at
right angles to the direction of lateral force. These walls subject the
concentrated tensions and compressions for the vertical direction at the
left and right outsides of each wall. Therefore, the deformations of out
side and middle parts differ appreciably. Also, the difference of outside
and middle parts shows a spatial characteristic obtained by using this
method, and this characteristic cannot be evaluated by using the bending-
- 99 shear theory.
c)
Stress Distribution of Walls
The principal stresses and reactive forces distributions of W2, WC and
WH are shown in Fig. 8 (a), (b) and (c), respectively.
Fig. 8 (a) shows the principal stress and reactive force distribution
which are parallel to the direction of lateral force. In this figure, it
will be seen very clearly that the tensions flow from the right side of the
top to the left side of the base and compressions.flow from the left side of
the top to the right side of the base. It will be seen also that the stress
values near middle stories are larger than those near top and bottom stories
since this wall has a narrow width above the vicinity of middle stories. The
stress distribution around the opening is considerably disturbed and the
values are large in comparison with those of other parts. On the other hand,
the reactive force distribution under the base slab shows a nearly triangular
distribution except the reactive force of the compressive edge. This reactive force of the compressive edge becomes very large because, of consideration of a condition completely fixed under the base slab. This distribution
will become more smooth when actual deformation of soil is considered.
Fig. 8 (b) and (c) show the principal stress and.reactive force distributions of WC and WH, these walls being at right angles to.the direction of
lateral force. In these figures, it will be readily.seen that WC is pulled
up to the top side by the concentrated tensions at both vertical edges and
WH is pushed down to the base side by the concentrated compressions at both
vertical edges. Namely, WC shows a principal stress, distribution like a
simple beam of large depth supported at both edges, and WH shows a principal
stress distribution like a column of large width and thin depth subjected to
axial compression excepting the upper.part. .These stress distributions cannot be evaluated by using the bending-shear theory.
d)
Stress Distribution of Shell

The principal stress distribution of WS1 is shown in Fig. 9. The shell
is divided Into two parts at the center on the vertical sectional plane which
is parallel to the direction of lateral force, and analysis of the shell is
performed individually for the two parts. Subsequently, the lateral displacements and the selected vertical nodal displacements.of these two parts
coincide with each other. Accordingly, the torsion effect is ignored in
analysis of the shell.
In this figure, it will be seen that the major part of the shell is
subjected to compression stresses, and the stress values are small in comparison with those of W2 because the share ratio of the lateral external forces
are comparatively small and the shell Is thicker than the wall. Also, the
stress distribution is considerably disturbed around the opening.
1.
COMPARISON OF STIFFNESSES
In preliminary design, the bending-shear theory may be the effective

method to evaluate the stiffness of the entire structure and the analysis
- 100 can be performed In a short time. But in the bending-shear theory, the
entire structure is replaced by a single bar which is subjected to bending
and shear stresses, and walls at right angles to the direction of lateral
force, subwalls and a flask type shell which has the characteristics itself
are treated in the same way as the normal resisting walls. Therefore, the
results of analysis using the bending-shear theory may tend to overvalue
stiffness of the entire structure for the actual condition.
While the method explained In this paper has some unsolved problems re
maining at present, by using this method, stiffness of the entire structure
can be obtained considering the spatial characteristics of wall arrangement.
These stiffnesses are compared with the periods and the participation
functions obtained by performing eigenvalue analyses which employ the same
weight distribution. This weight distribution is shown in Table II. The
eieenvalue analysis Is performed by using the Jacobi's method.
The results of analysis are shown in Table II and Fig. 10 for two stiff
nesses obtained by using the bending-shear theory and the method explained in
this paper. Table II shows both periods from the first order to the seventh
order for the X and Y directions. Fig. 10 shows both participation functions
from the first order to the fourth order for the X and Y directions.
In these results, It can be seen that the periods and the participation
functions show the values and the distributions very near to each other in
the Y direction, and the difference between those periods is about 8 per
cent for the first order. It can be seen that in the X direction, the
periods and the participation functions are considerably different from each
other and the difference of those periods is about 16 percent for the first
order and also the shapes of the participation functions are not in good
agreement as in the Y direction.
Summarizing these results, If results for stiffness obtained by using
the method explained in this paper are more near the actual condition than
those for stiffness obtained by using the bending-shear theory, it can be
said that the bending-shear theory is brief and simple and is an effective
method of obtaining an approximate stiffness of the entire structure In a
short time, but as seen in the results of the X direction, the character
istics themselves of walls at right angles to the direction of lateral force,
subwalls and a shell must be given thorough consideration.
5.
CONCLUSIONS
To evaluate the spatial effects of wall arrangements which are normal

for nuclear reactor buildings, the structural model must be replaced as near
as possible to the actual condition and analyzed. For this purpose, the new
method explained in this paper gives more satisfactory results than the bend
ing-shear theory.
That is to say, by using this method, the detail stress distribution
each structural component can be obtained considerably near to the actual
condition as shown in Figs. 6, 7, 8 and 9, and the difference with the
- 101 bending-shear theory In stiffness evaluation is apparent as shown in Fig. 10

and Table II.
From the results analyzed only for the specific structural model treated
in this paper, which may not necessarily represent the general tendency, it
can be seen that the use of this method makes it highly possible to carry out
a reasonable structural design.
ACKNOWLEDGMENT
The authors wish to thank Dr. H. Tajimi, Professor of Ninon University,
Mr. F. Horle, Chief Research Officer of Odaka Laboratory, and Mr. I. Funahashl,
Research Manager of Takenaka .Technical. Research Laboratory, for their helpful
discussions and suggestions, and .the cooperation of structural engineers
engaged in the design of the nuclear reactor building mentioned in this paper.
REFERENCES
[I]
[II]
[III]
[IV]
[V]
H.C. Martin: An Introduction to Matrix Methods of Structural

Analysis. McGraw-Hill, I966
O.C. Zlenklewicz and G.S. Holister, editors: Stress Analysis,
John Wiley & Sons, I965
O.C. Zlenklewicz and Y.K. Chenng: The Finite Element Method In
Structural and Continuum Mechanics,McGraw-Hill, 1967
Ivar Holand and Kolbein Bell, editors: Finite Element Methods In
Stress Analysis, TAPIR-TRYKK, I969
3. Tlmoshenko and J.N. Goodier:
Theory of Elasticity. McGraw-Hill,
I95I
[VI]
S. Tlmoshenko: Strength of Materials. D. Van Nostrana, 1955
[VII] Lecture Notes, Course of Matrix Methods of Structural Analysis by
Computer. Society of Steel Construction of Japan (JSSC), I968
[VIII] J. Jido and Y. Tsushima: Aseismic Design of Structures with
Nuclear Reactors, Technical Reports, The 1st International
Conference on "STRUCTURAL MECHANICS IN REACTOR TECHNOLOGY"
Berlin, Germany, September 1971
102
(BEGIN )
CFNJT)
( C O c TOTAL T O 01 WAIL m
FIG.. 1
ANALYTI
C AL MODEL for
EXPLANATION and COMPUTING PROC ESS
FIO. 2
S'IL LL
BLOCK DIAGRAM of SYSTEM
'1X0
1.1000
:8
d'
J H.T1{
I i I ix I IKX> I i r o I rm I i m I mi I m I
tXP
{ a ) PLAN ol GROUNO FLOOR
FIG. 3
(b)
A A SECTION
PLAN of GROUND FLOOR and TYPICAL SECTION
103
W2
^WH
/[,WC
YJWSljl/J/l
d>"
~
DIRECTION
| \ ^ l Jwf\
of LATERALEXTERNAL
FORCE
1T|i\
J\ \
fl Juiz]
CD
CD
FIO.
SPATIAL ARRANGEMENT of WALL AND SHELL
I M Ol
(a) W2
FIG. 5
(b)
KM 112
IXWIU
WS1
THICKNESS and DIVIDED MESH CONDITION of TYPICAL WALL and
SHELL
104
Ifi.
17
1!
s3
sb
15(1
ISS
ai
51S
(5J)
Ili
223
5?n
ss
12.6
iti
S E
> ?
21f,
A' -K
FIG. 6
22.'
SHARE RATIOS of
LATERAL EXTERNAL FORCES
(PERCENT)
a^zas
^_
--
H1
f
I
^l
ci>
I a | i l M | p n ] n iirl w j
<s
(a)
<MF
3 D
i '
e <s>
W2
(b)
WC
D
D
j i c j '/. J
WW !TOJTOJTWlJTll|m.
S 4 )
(5 ST SI
(C)
WH
FIG. 7
DEFORMATION C ONDITION of
WALLS, W2, WC and WH
105
FIG. 8 (ni
C C
PRINCIPAL STRESS, and REACTIVE FORCE.DISTRIBUTION of W2
<s>
50
COtfcgtarf)
ENSON
'
COMPRESSION
ns
WJ
wi
I ICO
FIG. 8 (b)
I IPO
7.B0
I IPO
1 IPO
I im
inm
PRIN
C IPAL STRESS and REACTIVE PORCR DISTRIBUTION of WC
- 106
so
I 3
FIG. 8 ( c )
I rao 1 n 1 Vff I ? I
aasa I
PRINCIPAL STRESS and REACTIVE FORCE DISTRIBUTION of WH
FIG. 9
tggiÂwT)
SD
igtxa*wfi
PRINCIPAL STRESS. DISTRIBUTION of SHELL, WS1
107
\N
vS.
7ni
t,h
fi
\ /
(ist
li
M 1 i '1
I /
102010
10
20
30
XDIRECTION
Y DIRECTION
By Evaluation of B onding Shear Theory

By Evaluation of MMhod in This Paptr
FIG. 10
COMPARISON of PARTICIPATION FUNCTIONS in X.and Y DIRECTIONS
Floor Level
Lateral External
Forces
( t)
13 '
12
11
10
9
5
7
6
5
4
3
2
fABLE
TABLE I
FLOOR LEVEL..
2109.
2417.
9447.
10120.
2773.
8705.
5 565.
16004.
3006.
17236.
7131.
5 3 79.
COMP
A RISON of. PERIODS.In X and ."f DIRECTIONS. and WEIGHT DISTRIBUTION
XDIRECTION
Y DIRECTION
Order
CASE1 sec. CASE2 sec.
1
2
3
4
5
6
7
LATERAL EXTERNAL
FORCES ACTING, on EACH
0.161
0.087
0.061
0.192
0.106
0.069
0.049
0.035
0.031
0.029
CASE 1 sec CASE2 sec.

0.165
0.085
0.061
0.04 2
0.035
0.031
Floor
Level
0.1 7 S
0.091
13
Weight
(t)
1104.2
12
1509.4
11
6601.1
10
104 05.3
14672.1
CASE1 By Evaluation of Bending Shear Theory
21508.4
CA5E2 By Evaluation of Method This Paper
35053.1
'
0.044
0.033
0.031
0.025
0.02 5
0.063
0.044
O.C 34
0.031
0.034
OH
DISCUSSION
H. WOLFEL, Germany
How do you calculate the dynamic loads caused by earthquake ?
Y. HAYAM1ZU, Japan
Generally, we obtain the dynamic loads by the following method:
1. We calculate the dynamic loads by the standard code in Japan.

2. By using an analytical model evaluated by the bending-shear theory and the other method,
we perform the dynamic r e s p o n s e analysis, and from the result of this response analysis,
we obtain the m a x i m u m shear forces at each floor level. We a s s u m e that these shear forces
a r e the dynamic loads.
K 2/6
ASEISMIC DESIGN
FOR JAPAN EXPERIMENTAL FAST REACTOR (JOYO)
. AKINO, M. KATO,
ABSTRACT
This paper explains the aseismic design of Japan Experimental Fast Reactor (50 MWt)
called "JEFR" or Japanese nickname "JOYO" which is being constructed at Oarai site in
I bararli Prefecture, along the shore of the Pacific Ocean.
Even though the aseismic design of JOYO Is being progressed now in detail, fundamental
design requirements were fixed and some interesting design activities have been continued. This
paper introduces those matters, including decision of the design earthquake, idea of plant lay
out, explanation of dynamic analyses of main items, design modification of safety rods, recipe
for making flood response spectra, problem of sodium coolant piping and experiment of graph
ite shielding structure.
1.
INTRODUCTION
PNC (Power Reactor and Nuclear Fuel Development Corporation) has intended to experi
ence and solve various problems of technology, through the design, construction and operation
of LMFBR (Liquid Metal Fast Breader Reactor) in Japan, which aims to develop prototype
fast reactor and large scale commercial fast power reactors in the future. In the nature of the
case, the first reactor plants involve structural and mechanical problems which have never been
appeared in the designs and constructions of light water reactor plants under construction in
Japan, and moreover, most of those peculiar questions relate with the aseismic design.
As is well known, Japan is located in a zone of high seismicity in the world, and very severe
inquiry has been made of the aseismic design for every nuclear facility to avoid public disaster
due to probable earthquakes. On the other hand, PNC is one of new quasi-governmental organi
zations and its staff has little experience in the aseismic design. Therefore, JAPC (The Japan
Atomic Power Co.) has been entrusted, as a consultant, to assist PNC relating to all of the
aseismic design of JOYO with the contract since two years before.
Basic philosophy and method of the aseismic design for JOYO are almost the same with
those for large scale commercial light water reactor plants in Japan. "Technical Guidelines for
Aseismic Design of Nuclear Power Plants (written in Japanese)", published by Japan Electric
- 110 Association in April, 1970, and was edited by a special committee, explains the above philosophy and method, and a member of the committee will introduce this Technical Guidelines at
the Conference. Therefore, this paper does not touch upon such general philosophy, criteria
and method regarding the aseismic design of nuclear facilities, and it refers only to special topics
on the aseismic design appeared in the project of JOYO.
2.
DESIGN EARTHQUAKE
2.1. Special Site Condition
In the case of nuclear power plants which are being constructed or planned in Japan,
sound rock layers for bearing heavy reactor building sufficiently to withstand strong earthquakes are searched in the course of site selections. In the case of JOYO, even though it does
not generate electric power, its size and weight, structural complexity, construction cost and
safety requirements are comparable with those of commercial nuclear power plants, and a
subsoil profile of JOYO's site shows very deep sand layers up to 162 m below the ground
surface. However, since the bottom of reactor building foundation was located 32 m deep
from the ground surface, this project presented us with a new problem how the design earthquake be selected considering an effect of very thick sand stratum between the bottom of
building and the base rock.
2.2. Maximum Acceleration at Base Rock
First of all, an evaluation of the expected value of the maximum acceleration as the
design earthquake at the base rock in Oarai area was made as in the case of other nuclear
power plants. Magnitudes, focuses and epicentral distances of 13 earthquakes that occurred in
the past 1100 years were investigated, and using these data and also referring to the materials
prepared by Kanai [1 ) and Seed ( 2 ] , the values of accelerations for the maximum earthquake
at the base rock were estimated. Kanai's formula gave the maximum constant velocity 3.2
cm/sec, and this value can be converted into the acceleration 100 gals as the maximum probable figure, and Seed's figure represented less than 50 gals. On the other hand, Kawasumi (3)
presented a seismic probability map assuming that the whole land of Japan is covered by a
standard uniform soil condition. This map shows 200 gals on the ground surface at Oarai
area, and it is said that this value could be reduced to one half to one third, and therefore,
70 - 100 gals acceleration at the base rock would be an expectant value. While, an actual
observation of natural small earthquakes in a period of about one year was carried out at the
site, and its records indicated the above reduction factor being one quarter to one fifth.
Therefore, we concluded from the above information that as the maximum acceleration
of the design earthquake, 100 gals or 3.2 cm/sec in velocity could be selected at the base rock
(162 m below the ground surface) in a conservative sense.
2.3. Observation of Small Earthquakes

JAERI (Japan Atomic Energy Research Institute) carried out interesting observations
of natural small earthquakes by installing seismographs in and on the ground. Many instrumentations were arranged (a) around JPDR (Japan Power Demonstration Reactor) which
is borne by shale and (b) around JMTR (Japan Material Test Reactor) which is borne by sand
stratum. The former is located at Tokai 19 km North of Oarai and the latter is some 400 m
away from JOYO at Oarai. These observations presented useful data in comparison with each
other, and the conclusions on Oarai site are as follows:
Ill
(1)
Deep sand stratum of 22 m below the ground surface can be regarded as a vibratory base
as well as the base rock of shale, because both records represented a constant velocity
spectrum of ground movements as shown in Fig. 1.
(2)
Predominant periods of the ground are 0.15,0.5 and 1.1 sec on the surface, and 0.5 and
1.1 sec at the elevation of 22 m below the surface as shown in Fig. 2. Those periods mean
that 0.15 sec is the natural period of the over burden and 1.1 sec and 0.5 sec can be regard
ed as the natural periods of whole sand strata corresponding to the first and second mode
vibrations, respectively.
Therefore, 22 m layer can be chosen as the vibratory base instead of 165 m layer, but
to perform conservative calculation the latter elevation was defined as the vibratory base.
2.4. Amplification of Sand Stratum
In order to evaluate the amount of amplification of ground movement due to the exist
ence of deep sand strata, a theoretical calculation and actual observation were carried out.
In the theoretical calculation by means of the theory for multilayer reflections, the
following matters were considered:
(1)
(2)
Kanai's report [4] was referred to,

The base rock located 165 m below the surface was regarded as the vibratory base,
(3) ' Input vibration at the vibratory base was sinusoidal motions having parametric various,
frequencies to make a frequency response sectrum,
(4)
Coefficient of viscosity was calculated referring to the formula in the paper by Kanai
[] and using density and coefficient of rigidity of each stratum,
(5)
Case A or to account the amplification effect from the vibratory base towards the
bottom of an equivalent building, which has 0.2 sec natural period and was located 22 m
below the surface at its bottom, three subsoil strata were taken, and Case or to account
that towards the open surface, seven strata were taken, and
(6)
Required physical numbers of all strata were given from soil information obtained at
the site.
The conclusions of the calculations were shown in Fig. 3 and its summary is as follows:
(1)
For Case B. 5.6 times amplification was evaluated at 1.1 1.2 sec. and 7 times amplifi
cation at 0.3 0.4 sec.

(2)
For Case A. 4.8 times amplification was evaluated at 1.1 1.2 sec. and contrary to the
above, only 1.6 times amplification at 0.3 sec.

The above tendency appeared in the experimental records. Fourier velocity spectra of
four recorded earthquakes at elevations of 0.18 m (on the surface) and 22.18 m (under
ground) were averaged and plotted in Fig. 2. The amplifications for longer periods were
approximately the same, but for shorter periods larger amplification occurred just on the
surface.
Referring to the above theoretical calculation and experimental evidence, in order to
decide the value of the maximum acceleration to apply the bottom of the reactor building
foundation, we made such judgement, that a selection of 1.5 times amplification factor to
multiply the value of 100 gals at the base rock for shorter periods was adequate, and that the
acceleration acting at the bottom of the reactor building should be small for longer periods,
because the acceleration at the base rock reduced owing to a velocity constant tendency,
though an amplification factor being large as shown in Fig. 3.
- 112 2.5. Selection of Earthquake Waves

As apparent from the previous investigations, two different tendencies should be taken
into consideration for selecting the design earthquake waves, the one has the spectral peak at
a range of 0.3 - 0.4 sec periods and the other peak at 1.0 - 1.1 sec. Locking for many records
obtained by Strong Motion Accelograms and after examining them, the following two waves
were selected and their values of the maximum accelerations were decided respectively to
normalize them for the purpose of designing JOYO:
El Centra NS, 1940, Maximum Acceleration = 150 gals
Akita Record EW, 1964 (obtained at Niigata Earthquake on building of Akita Prefec
tura! Government), Maximum Acceleration 100 gals
By the way, Fig. 4 represents the response spectra for 5% of critical damping of the
above two design earthquakes.
3.
PLANT LAYOUT
An original conceptual layout of buildings indicated that the reactor building together
with the containment vessel was one individual structure, and several other buildings, in which
many A class facilities were supported, were arranged around the reactor building. However,
the bottom of the reactor building foundation is located at -32 m below the ground surface
since an elevation of the operating floor has to be coincided with the ground surface for con
venience of transporting the spent fuel cask car. In such a layout, not only the reactor building
but other buildings should be A class, and many trenches and cable tunnels should be arranged
between the reactor building and others on loose back fill, and it would be difficult to make
accurate aseismic designs for many buildings, trenches and tunnels.
Therefore, in order to avoid the above risky and troublesome designs, one large rigid and
strong building consisting of the reactor building at the center and auxiliary building having
monolithic basement was proposed as shown in Figs. 5A and 5B. The intention of designing
such a large building was to house every A and class facility in it, except the main secondary
sodium cooling system and its large air coolers.
4.
DYNAMIC ANALYSES OF MAIN ITEMS

4.1. Models for Response Calculations
For different purposes of the response analyses, two analytical models were selected.
The first model consisted of the reactor building, containment vessel and auxiliary building,
as shown in Fig. 6A, to obtain the response accelerations, moments, shearing forces, etc. for
those structures, and the second model consisted of the reactor building, auxiliary building,
reactor vessel and vessel internals as 'shown in Fig. 6B, to obtain the response accelerations,
moments, deformations, stresses, etc. for the vessel internals. Owing to a limitation of com
puter capacity for working out time history calculations, the containment vessel was omitted
in the latter model.
4.2. Spring Constants of Soil
As mentioned in 2.5., two earthquakes or El Centra and Akita waves were selected for
the design, however, Akita wave does not affect the above two models, because a range of
natural periods of the first mode of the both models is 0.25 - 0.35 sec and an amplification
factor due to Akita wave in this range is very low. Therefore, just only El Centro wave is
meaningful for the response calculation of the above two models. It is said that plural input
waves should be applied to the response analysis for the design, however, we found out such
other appropriate scheme that setting a certain range for spring constants of soil which pro-
- 113 duce rocking and swaying vibration modes can generate some amplifications in a broad range
instead of selecting plural input waves.
There are several references presented by Tajimi 16), Timoshenko et al. | 7 ] , Toriumi
[81 etc. to account the spring constants, but those formulas gave different results, namely,
It can be said that working out the spring constants is an uncertain problem, therefore setting
a certain range for the spring constant is an advisable technique.
The calculated numbers of and Ks obtained from the above formulas and averages
of these numbers were regarded as corresponding to the case of the hardest soil condition,
and one half of the above averaged numbers were regarded as corresponding to the other case
of the softest soil condition. Duplicate response calculations applying to El Centra Earth
quake wave for both soil conditions were performed, and the designs of all A class items have
been required to satisfy the both cases.
4.3. Calculated Results
Calculated results of the response analyses were shown in Fig. 6A for the first model
and in Fig. 6B for the second model. With respect to the first model, the dynamic response
analysis gave an insignificant result for the design of the buildings and containment vessel,
compared with the distribution of respondent acceleration and the static requirement which
was defined as the seismic coefficient represented by the step-wise full lines in Fig. 6A. How
ever, many computed outputs for the first model were used as inputs for calculating Floor
Response Curves which will be explained in 6. With respect to the second model, the calcu
lated results were valuable for the hardware designs of the reactor vessel and its internals.
4.4. Application of Finite Element Method
In the above analyses it was assumed that the input earthquake excitation took place at
the bottom of the main building foundation, even though the building hid deeply into the
ground and thick soil surrounded the side-walls of the building. A member of ACRS (Advi
sory Committee of Reactor Safety) asked that it should be clarified what effect due to input
excitation taking place along the side walls, especially near the ground surface, might exist
for the response of the main building. This question belonged to an academic problem, and
we could not reply straightforwardly, however, we performed dynamic response calculation
by means of the finite element method referring to the paper by Tsushima et al. [ 9 ] , for the
main building model surrounded by a large amount of soil as shown in Fig. 7. In order to
investigate the effect pointed out on the above, two comparable models were selected assum
ing the two dimentional problem. In the first model the side wall of the building touched
the soil, and in the second model certain clearance between the side wall and soil was ideally
prepared. The input excitation was El Centra Earthquake normalizing 100 gals acting along
the bottom line of the soil model, and physical numbers obtained from soil data were used
for '*, soil strata.
Looking at the figures of calculated result, for instance regarding the maximum accelera
tion shown in Fig. 7, much difference appeared between two models in the upper portion of
the soil, but in the portion of the building any appreciable deference did not appear, and
rather less amplification at the building roof of the first model than the second model was
observed. Then the method and results of the usual dynamic analyses mentioned in the above
4.1 through 4.3 were accepted.
114 -
FUEL ASSEMBLIES A N D C O N T R O L RODS

5 . 1 . Capability of Scramming
The most important problem in the aseismic design of J O Y O was whether or not the
safety rods can be inserted into the core when a destructive earthquake excites the reactor
building. Dynamic behaviors of key items (see Figs. 8 A and 8B) relating to this problem will
be as follows:
(1 )
The reactor vessel moves together w i t h the concrete structure,
(2)
The core barrel and core cover structure, which are vertical cantilevers, move independ-
ently and some relative displacement between the top of barrel and the b o t t o m of cover
structure occur in sodium coolant, and
(3)
The hexagonal fuel and blanket assemblies lean on the core barrel owing to the existence
of clearances between the assemblies, and this amount of deformations is critical.

5.2. Design M o d i f i c a t i o n
in the original design, there were no pads along the outer surface of hexagonal assem
blies considering bowing deformation, charging and discharging, and an accumulation of 3.2
mm clearance between each assembly intimated much deformation of the assembly columns.
In order t o estimate the amount of the above deformation, numerous theoretical calculations
were worked out but it turned out a failure since it was a non-linear problem accompanying
a chattering vibration.
Therefore, a design conception was reconsidered such that the dimension of every hardware should be appropriate to assure the scramming even if the m a x i m u m relative deformat i o n , which can be defined f r o m a possible geometrical relation and not from an analyses,
takes place. According to this design conception the following three design modifications
were performed:
(1)
3.2 mm clearances between all assemblies were reduced to 1.2 mm by means of new
provision, namely additions of 6 pads at the top of assemblies along the hexagonal out-
side edges,
(2)
In the original design 6 control rods were identical, but in the current design the clear-
ances between the rods and their lower guide tubes are 1 m m for 2 control rods and 5 m m
for 4 safety rods t o give more assured insertion of the safety rods than the control rods,
and
(3)
94 mm inside diameter of the upper guide tubes was enlarged to 144 m m giving head to
an anticipated large relative deformation.
FLOOR RESPONSE SPECTRA

For the designs of vital items, the time history response analyses were made as mentioned
in 4.3., however, for the designs of such equipment as piping, etc. classified as As and A classes,
a simple and convenient analytical method using the floor response spectra now being general
zed in Japan is also applied. In the case of JOYO, t w o earthquake waves having an isolate char
acteristic were selected and the range for the spring constants of soil was considered as mentioned in 2.5 and 4.2 respectively. Therefore, a special treatment should be carried out for
making the floor response spectra.
115
As an example, the floor response spectra for the 4 t h floor which correspond t o an eleva
t i o n suspending the reactor vessel and for 1 % of critical damping are shown in Fig. 9. Computer
calculation drew the curves A for the hardest soil and for the softest, and the t h i r d C was the
artificial design curve. In making the curve C the following!, were taken i n t o consideration:
(1)
Hill I covers the elastic vibration mode of the reactor building which appears in the curve
(2)
1 covers the rocking vibration mode which is affected mainly by changing the spring
B,
constants of soil,
(3)
III was drawn judging f r o m the response due t o the A kita wave,
(4)
The left f o o t corresponds t o the m a x i m u m response acceleration of the building at the
15)
The right f o o t corresponds t o the maximum response displacement of the building due t o
same elevation, and
El Centra wave.
7.
S O D I U M C O O L A N T PIPING
LMFBR piping design needs a peculiar deliberation owing t o its high temperature, t h i n
pipe thickness, doublewalled primary system and aseismic supports. First of all, a w i n d i n g pip
ing arrangement was made t o reduce thermal expansion stress as low as possible for the main
primary cooling system as shown in Fig. 10, and a trial of the response calculation as the aseis
mic design is being w o r k e d out by fixing or replacing the seismic supports, or hydraulic snub
bers. The final winding and locations of supports will be decided by comparing the stresses due
to the thermal expansion, thermal shock, seismic vibration, pressure and weight, and this study
is now being carried out. A n experiment t o evaluate a rigidity of the doublewalled pipes was
made, and a development of new type hydraulic snubber including irradiation and deterioration
tests has been performed. A s the applicable code USA S .31.7.0., Nuclear Power Piping and
its Case Interpretations were referred t o in the design of the piping systems.
8.
GRAPHITE SHIELDING STRUCTURE

In order t o protect the concrete shielding structure, the graphite shielding structure will
be piled up between the reactor vessel and the safety vessel as shown in Fig. 8A . A n original
conception of the graphite structure was a group of columntype stacks, but this conception
was abandoned because no design of an adequate restraint withstanding the earthquakes could
be f o u n d . Contrary t o the original conception, a new proposal was the pile up of sideways gra
phite blocks connected w i t h the safety vessel by means of outermost hole drilled into each
block and vertical steel rod inserted into the hole. A
certain clearance between the reactor
vessel and graphite structure w i l l be kept not t o allow any contamination t o the vessel. How
ever, the design of the graphite structure has not yet been finalized, and its model test using a
vibration table will be carried out in May, 1 9 7 1 , t o evaluate stress d i s t r i b u t i o n , factor of impact,
size of key, etc., and Fig. 11 represents the portions of the reduced model.
- 116 9.
CONCLUSION
Construction w o r k of the containment vessel is being carried out and engineering detail
design of many components are also proceeding now. Some outcomes which have been solved
or concluded up t o this date in the preparation and design stages are mentioned before. We have
a responsibility for finalizing the design and construction of J O Y O , and we do not know what
new bothersome problems in the aseismic design may arise in future, but we should find appropriate solutions on all such cases t o the best of our knowledge.
10.
ACKNOWLEDGEMENT
Thanks are due t o the Power Reactor and Nuclear Fuel Development Corporation for giv-
ing us participation in the project and for permitting our presentation of the paper in the
conference.
Appreciation is expressed t o engineers and research members of the contractor and venders
for their assistances and cooperations in the analyses and calculations and to Mr. T. Uchida for
his kindful submittal of useful data for earthquake observations at Tokai and Oarai performed
by Japan A t o m i c Energy Research Institute.
REFERENCES
(11
K. Kanai et al.: "Expectancy of the m a x i m u m velocity amplitude of earthquake motions

at bed r o c k " , Bull. Earthq. Res. Inst., Univ. of T o k y o , V o l . 4 6 , 1968.
[2]
H. B. Seed, I. M. Idriss and F. W. Kiefer:
"Characteristics of rock motions during earth-
quakes", Earthq. Eng. Res. Center, Univ. of Cal., Report No. E ERC 68-5, Sep. 1968.
[3]
H. Kawasumi:
"Measures of earthquake danger and expectancy of maximum intensity
throughout Japan as inferred f r o m seismic activity in historical times". Bull. Earthq. Res.
Inst., Univ. of T o k y o , 29, No. 3, 1 9 5 1 .
[41
K. Kanai et al.:
"Some features of strong underground earthquake motions computed
f r o m observed surface records",
Bull. Earthq. Res. Inst., Univ. of T o k y o . X L V I , 1968,
Part 3.
[5]
K. Kanai et al.: " R e l a t i o n between the amplitude of earthquake motions and the nature
of surface layer. I l l " , Bull. Earthq. Res. Inst., Univ. of T o k y o , 3 1 , No. 4, 1953.
(61
H. T a j i m i :
"Basic theories on aseismic design of structures".
Rep. of Inst, of Industrial
Science, Univ. of T o k y o , V o l . 8, No. 4 1959 (in Japanese).

(71
181
S. P. Timoshenko and J . N. Goodier: " T h e o r y of elasticity", 2nd Ed., 1959.

I. T o r i u m i :
" V i b r a t i o n s in foundations of machine",
Technology Rep. of Osaka Univ.,
No. 5, 1955.
[91
Y. Tsushima et al.:
"Earthquake response analysis of soil-structure system", Trans, of
Architectural Inst, of Japan. Summ, of Technical paper, Sep., 1970. (in Japanese).
117
JUT A * (.1.
Ulk
Fig. 1. Re ponse acclration spectra of earthquake
> observed simultaneously around JMTR and JPDR.

Olla
Illa
I
/.
/1' '~>\ I'
y>
.s ~"
Fid 2. Fourier velocity spectra o observed earthquake waves

MODEL OF
SOIL PROFILE
CASK A
A
C SE g
Ti
prnioD'SFO
Fig. 3. Theoretical velocity amplification spectra of ground strata.

DISPLA CEMENT'"
EL s r n i i s, it.n
' i'.,. fc1*.
Fig. 4. Response spectra of design earthquake waves.
118
n i n i , AHM n*
IMU I M . SNTIM HUI
Ml I
nm.
ARKA M ta S K I < >M AH
(OUf.lM. SVSTKM l IH M
Kig 5
Fig
(?)
I ;;.v"
>Ofr,v,M'.r
\\)
V e r t i c a l section ni main building
! l'Un ..I main bui hl.HL, al ( . 1 .
Jim
1 19
tACTOa III II Ms,,

UIIU AL T B LI LDI NC
VESSEL
IEL
INTIKALS
SOCL
ASE or sorrrsT so,

s mixes or sou.
II
: 33 JJ\Jj
Fig 6A. Results of earthquake response analysis for the f i r s t model, or for buildings
and containment vessel.
,.
hi p.
Fig 6H
Results of earthquake response analysis for the second model, or for reactor
vessel and its internals.
mo
MH MAIN B ULDING
SS SIHROLNDED SOIL
1 M I I.AL
.'
1W
(40
71
Mil
ILL
J I M
HRST MOUKI
vw
f"'
IM
4w
III!
70
IM
WO
rt
...
;o
ss
tOND MOOI
Fig. 7. Responded acceleration for two models of building and surrounded soil
by means of finite element method.
- 120
PRIMAR 1
I A M P I I *
m .KT
Fig. 8A. Vertical section of reactor structure.
Fig 8B. Plan of reactor
121 -
^DISPLACEMENT ;
" PERIOD U M I
Fig. 9. Typical floor response spectra for reactor buildin
Fig. 10. Analytical model of primary sodium coolant pipe.
-J2
i'-; '-JL
Fig. 11. Model of graphite structure for vibration test.
- 122 -
DISCUSSION
C. B. SMITH, U. S.A.
In the vibration tests of graphite shielding:
a) How was the experimental structure tested ?

b) Did your measurements include observations of "impact" between individual graphite
blocks ?
c) In your opinion is "impact" likely to be a significant factor after the graphite has undergone
radiation damage ?
"
J^
K. AKINO, Japan
a) It is the largest shaking table in the world, its owner is Desaster Prevention Center,
Japan Government, at Tsukuba in Ibargi Prefecture. The maximum loading capacity is
500 metric tons, and its control is carried out by displacement of either sinusoidal or random vibrations.
b) We provided accelerometers and U-gauges or pick-up to observe relative displacement
of each block. Those records indicate "impact", for instance, 10 g in acceleration.
c) Now, the other group is performing irradiation tests of the graphite material. After receiving our and their data, we should have discussion about that point you asked.
K 2/7
BERECHNUNG DER ERDBEBENSCHWINCUNCEN VON STRUKTUREN

MIT DER FINITE-ELEMENT-METHODE MECHANISCHE
MODELLE VON KERNKRAFTWERKEN MIT EINBAUTEN
K. MARGUERRE, M. SCHALK, H. WLFEL,
Institut fr Mechanik,
Technische Hochschule Darmstadt, Darmstadt, Germany
Prof. K. KLOTTER zum 70 Geburststag gewidmet
ABSTRACT
For the analysis of seismic vibrations, complex structures are usually idealized by lumped parameter models. With the finite element method however,
natural frequencies as well as forced vibrations can be calculated with higher
accuracy. Here deflections are used as generalized coordinates. The elements
used are beams with bending flexibility in two planes under consideration of
shear deformation and rotational inertia. Torsion, normal strain, initial
stress and elastic foundation may be taken into account. The element stiffness- and mass-matrices are found using the principle of virtual displacements. For the calculation of forced vibrations the method can be combined
with the response spectrum modal analysis as well as with the time history
modal analysis.
The choice of the mechanical model of the structure is of high influence on
the quality of the seismic analysis. The basic considerations for selecting
the mechanical model of buildings and equipment are discussed and models are
demonstrated for several nuclear power plant structures. Special attention
is given to the interaction between building and equipment.
Erdbeben veranlassen XK'./-Gebudc und ihre Einbauten zu SchuinE'.m^cn,

genau v:i Druckstfte, Kaschlnenunwuchten, o. .. Das Schwingun^sproblem zerfallt in drei 'dnterprobleme: 1. Erregung, 2, Abbilaun
der Strukturer, auf berechenoare mechanische -".oeiie, 3< Recht-nvcrfcihren. Wir wolJcn zuerst das Hechenverfahren allgemein errtern,
dann unter den IJcsondsr-lniiten des Erdueber.;";, und dabei kur?, auf <Jic
ErreRUH c-ir.ne'ncn. Als Drittes behandeln wir die Kodc-ilabLildunjj,
gioie!: nit blick auf die Erribebenerregung und auf die besonderen
Kral fcwsr-ks-Strukturen,
124
1.
Rechenverfahren allgemein
Am geeignetsten fr die Schwingungsberechnung von kompliz.1 ert^n
Sti'ukturen ist das Verfahren der finiten Elemente, a) weil sich wa
ndt Strukturen praktisch beliebiger Topologie erfassen lassen, b)
weil bei guten Ansitzen die Genauigkeit bei gleichen Aufwand we...n;
lieh iijher ist als z, B. bei der Abbildung auf ein System nit dis
kreten 'lassen (lumped parameter model). Als "Koordinaten" knen
Verschiebung oder Kraftgren gewhlt werden; wir verwenden hier
Verschiebungen und Drehungen, kurz: Verrckungen, Die Matrixsch.. ;1:
weise ist fr ein computerorientlertes Rechenverfahren besonders
zweckmig.
i AuGyl.n_^r_Schwlngungs g_le i chun

Der Zusammenhang zwischen dem GesantVektor <fj,(t) der glob.'.j
einander unabhngigen) Verrckungen einer schwingenden Struk t u r .n>
aen zugehrigen globalen Krften pJO<t) wird hergestellt dure 1; Cc
dynamische Steifigkeitsmatrlx 5,a(t) :
Sv,U)-arlU)
U)
pp(t) .
Trennt man Bekannte und Unbekannte

9 |
%*'
9.
unbekannte Verrckuriger.
bekannte
Verrckungen
bekannte Kr.''t o ,
unbekannte Kr til t ? .
so s p a l t e t sich die Matrixgleichung auf:
s.
"s ~ sii
Im Sonderfall der reinen Krafterregung (z. B. Drucksto) gilt irit
9.= o
S q ,
d')
Im Sonderfall der reinen Verschiebungserregung (.. IJ. Erdbeben) gilt

mit = O :
S-q ""S.,. .
d")
Die Lrdbebenerregung unterscheidet sich also von der Krafterregung

nicht prinzipiell; nan mu nur die erregende Kraft p(t) aur; cj mit
Hilfe einer dynamischen Stelfigkeitr.inatrlx S ( t ) bestimmen.
Die globale dynamische Stelfigkeitsmatrix S(t) gewinnt man aus den
dynamischen Steifigkeltsmatrizen S^t) der Elemente, Man formulier!
zunchst die kinematische Kompatibillttsbedingung, d. h, den Zu
125
sammenhang zwischen den lokalen und den globalen Koordinaten, mit
Hilfe einer
Inzidenzmatrix I ;
dann zeigt sich, da L - transponiert auch fr die Erfllung der
statischen Kcmpatlblllttsbedingungen sorgt, mit dem Ergebnis [ S ] :
S /S/L .
(2)
E,l,rin_tm_afcr;lêri
Die Steifigkeitsmatrizen S^ der Elemente stellen den Zusammenhang

her zwischen den RandVerrckungen und den Randkrften :
V" r'rr
Bei kontinuierlichen Elementen sind die Glieder s.. dieser Matrizen
1. a, transzendente Funktionen. Da die s,,. nicht nur unbequem sind,
sondern berdies in die Lsung der Gl (1) numerische Schwierigkelten
hineintragen, ist es zweckmig Nherungsauodrtlcke zu benutzen. Dar,
geschieht ber das Prinzip der virtuellen Verrckungen mit Hilfe von
Ritzanstzen h,(x) fr die rtlichen Verrckungen. Die s ^ ergeben sich dann als Integrale vom Typ Jh.(x)h. (x) dx und spalten
sich in zwei Terme auf:
S; = c, m i K a P
wobei beide Matrizen voll besetzt sind und die Matrix [cjic)= C bei
geeigneten Anstzen identisch 1st mit der statischen Steifigkeits
matrix. (Eine notwendige Bedingung fr die Brauchbarkeit der Matri
zen C und M 1st, da multipliziert mit einer Starrkrperverrfickung
C das Produkt Kull, M die GesamtMasse bzw. Drehmasse liefern
mu.) Aus der Gleichung (1) wird
M-cj + C q =
mit
(3)
Die Ritzanstze reduzieren das Schwingungsproblem mit unendlich vie

len Freiheitsgraden auf das Problem (3), das II Freiheitsgrade hat,
wobei die Zahl im Allgemeinen gleich sein wird der Zahl der Ver
riickungsKoordinaten q .
Die vollbesetzte Massenmatrix ist ein ganz wecntlicher Vorzug des
skizzierten Verfahrens es 1st wenig sinnvoll, die statische Stei
figkeit einer komplexen Struktur mit einem groen Aufwand an Matri
zcniiberlegungcn zu bestimmen, die ebenso wichtige "asr.enrnatrix aber
durch Abbildung des Gebudes auf einen j'unk'.massenSchv.'lnger f_.'<:iz
rrob nur anzunhern.
126
i. 3 osung_der_Schwiigungsg^leichung
Der Weg zur Lsung der Gleichung (3) fhrt ber die Eigenwerte und
Eigenvektoren des schwingenden Gebildes. Bei p = 0 geht (3) mit den
Ansatz <j(t) = -e
ber in das allgemeine Eigenwertproblem
(C ) = 0
C O
fr dessen Lsung heute ausgezeichnete numerische Verfahren zur Ver

fgung stehen; man findet die Eigenwerte <A und die zugehrigen
Eigenvektoren At .
Fr p(t) i O
baut man nun die Unbekannten <J(t) auf aus den 1.1
genvektoren. Ist
die Katrix der Eigenvektoren, so ersetzt
man cj durch einen Vektor vermge der Transformation
<Jf(t) = -tjU)
=ZfnJ<) .
(5)
Dank der O r t h o g o n a l i t t s r e l a t i o n e n
frC-^
;/1 O
jar L*K
('I')
entkoppelt sich das Gleichungssystem (3); wenn man von links mit
multipliziert, erhlt man
wobei sich die Hauptglleder der Diagonalmatrizen M
und C
au3
bestimmen. Die Auflsung der entkoppelten Gleichungen
v. r^ * ^(). = pi
[:=,]
(')
1st Tr jede beliebige Erregerfunktion p(t) mglich. Hat man (t)

gefunden, so liefert (5) den rtlichen Ausschlagsvektor <J = q ( 1 ) .
iii J^,EPf_ung_
Ist die Dmpfung nicht zu gro (wir bei mechanischen Schwingern mei
stens: < 105! der "kritischen Dmpfung"), so kann man le Eigenvektoren
berechnen, als ob die Dmpfung Null wire. Aber natrllcn werden die
Amplituden der erzwungenen Schwingungen v o r allem in Resonanznhe
von der Dmpfung erheblich beeinflut. Diesen Einflu schtzt mor.
ab, inden man ber die Annahme hinaus, da die Dmpfung ~q sei, noch
annimmt, da Dmpfungs und Steifigkeltst'atrlx einander proportional
127
selen. Dann entkoppelt die Operation (5) auch den Danpfungsanteil,
und man erhlt fr n
eine Gleichung vom Typus
>riu.-rjL * rLL-jL
+ Ctty_ = p i l i ) ,
(7)
die man mit Hilfe des Duhamelschen Integrals
.(t_T)
103t, das die Bewegung aus der Ruhe herau3 (Anfangsbedingungen

y(0) = li (0) = 0 ) beschreibt.
Erdbeben
Wie modifizieren sich die allgemeinen berlegungen in dem besonderen
Fall der (Erdbeben)Fupunkterregung? Ausgangspunkt 1st die Gl (1"),
Schreiben wir fr die an allen Fupunkten gleiche Erregung (statt
q ) , so lautet sie
M-q + cf9 M.H Cf.H,
(9)
wobei die Matrizen M . , C?. einspaltig sind. Der statische Anteil

der Erregung fllt heraus (s. die Bemerkung vor Gl (3)), wenn nan
statt der Absolutverrckungen o die Relatlvverrckungen y be
stimmt :
yi<) = q(t) - eH(t)

Darin ist
der sog. Boolesche Vektor, der 1 ist, wenn sich qj| ,
und 0, wenn sich qLi bewegt. Man erhlt
M y + Cy = Me + M.)z pit)
(9')
und wenn man eine Zeitlsung y(t) sucht (time history modal analy
sis) luft die Rechnung vile vorher geschildert. Fr die Absolutbe
schlennigungen (t) (in der Erdbebenliteratur meist ),' (t) ge
nannt), ergibt sich insbesoruere
cj(t) = E f c K i ^ j t ) * ;pt) + ez(t)
(io)
Kelativverschiebungen y (t) und Absolutbeschleunigungen q (t)

sind die beiden interessierenden Gren: Die Relativverschiebungen
sind konstruktiv wichtig, weil sie darber Auskunft geben, ob z, 3,
Trennfugen ausreichen, und die Absolutbeschleunigungen wird man in
vielen ''allen brauchen, wenn man die Beanspruchungen berechnen will.
Denn die SchnittKrfte wird man "direkt" nur bei solchen Strukturen
ausrechnen, bei denen Modell und Wirklichkeit einander sehr nahe
i'o.,,f>ri. Tt dio '.triikt.iir komolexer. so weicht das Schwingungcmodell
128
von der Wirklichkeit im Detail zu sehr ab, und dann geht man zwecK
mig einen Umweg: man rechnet aus q durch Multiplikation mit den
rtlichen Massen iriy eine "dynamische Belastung" aus und bestimmt
dann statisch die zugehrigen Schnittkrfte mit Hilfe eines Modells,
das an den interessierenden Punkten die Wirklichkeit genauer be
schreibt.
Soweit die "time history analysis". Nun aber weist beim Erdbeben die
erregende Funktion p(t) sc schnelle Vorzeichenwechsel auf, da man,
um q(t) numerisch zu bestimmen,zu einer ungemein feinen Untertei
lung der tAchse gezwungen 1st, d. h. zu einem erheblichen Rechen
zeit Aufwand. Man beschrnkt sich daher meistens darauf, die Ant
wort der Struktur auf ein Spektrum zu bestimmen (response spectrum
modal analysis), das a) fr ein bestimmtes Erdbebengebiet allgemein
gltiger 1st als^rgend eine'gemessene p(t)Kurve und b) sehr viel
weniger Aufwand erfordert.
Da das SpektrumVerfahren aus der Erdbebenliteratur gelufig ist,
wollen wir nur eine Bemerkung machen zu der Frage der Resultiercndcn
bildung. Was man ausrechnet ist fr Jede Eigenform der Beitrag r>,
+ )
zur Relativverschiebung y bzw. zur Absolutbeschleunigung

'y; "as
man nicht kennt ist der Phasenwinkel. Addiert man nur. an Jeder Stelle
einfach die Betrge, so berschtzt man die Resultierende gewaltig,
denn es ist unmglich, da die Maxima der einzelnen Elgonformen ber
all zugleich eintreten. Der Wirklichkeit nher (wenn auch nicht not
wendig von der sicheren Seite) kommt man, wenn man die Eigenforinen
"vorzeichengerecht" addiert, d, h. so wie sie sich ergeben, wenn ,:
an irgend einer Stelle Fhasenglclchhelt annimmt (dae Ergebnis ist
unabhngig von der Wahl dieser Stelle).
Ein bei Erdbebenschwingungen viel verwendeter Gedanke ist der eine]
quadratischen Mittelwertbildung, der wahrscheinlich besser irt als
die BetragsAddition. Er sollte aber nur In einer (modifizierter.) Fori'
verwendet v/erden, die zwei Forderungen erfllt, die man an drn Mittel
wert vernnftigerweise stellen mu: da er die Bodonbeschlcunigun;
liefert, beim elastischen Gebilde an Fupunkt, beim starren berall.
Das letzte Wort ist ber die Auswertung der SpektrumErgebnisse noch
nicht gesprocher'.
3. MoJellahbj lern,
Nur die einfachsten Strukturen lassen sich "direkt" auf .'; cliv.j i.i:i:; . .1
berechnen, normalerweise ist man auf eine Modellabbildung ':'i,',cv.icr o.i.
und von deren Gte h fin,; t die Brauchbarkeit der Ergebnisse cut v.crci
dend ab. Daher zu dieser Krage einige hinweise grundstzlicher Avi ,
ergnzt durch Erfahrungen an beispielen. ^Leider knnen wir sii d.u: L
129
Stelle nicht eingehen auf die Wechselwirkung GebudeBoden, weil fr
die Standorte in unserem Raum bisher zu wenig verlssliche Aussagen
ber die BodenKennwerte vorliegen.]
3.1 Grundstzliches zur_Modellabbildung
Die Wahl des Modells hngt ab von dem, was man sucht. Handelt es
sich um lokale Zerstrungswirkungen (Flugzeugabsturz), so ist das
Schwingungsverhalten an dieser Stelle Wellenausbreitung wichtig.
Anders beim Erdbeben; dort sind gefhrlich (rufen Resonanzvergre
rungen hervor) nur Frequenzen unter 10, notfalls 15 Hz, also sind
nur die untersten Elgenfrequer.zen der KKWGebude wichtig, d. h.
+) vgl. den Fupunkterregten Einmassenschwinger, dessen Gleichung
lautet ra + c y = o
das Modell mu diese Frequenzen liefern. Die Feinstruktur einzelner
Teile zu bercksichtigen wrde einen ungerechtfertigten Aufwand be
deuten und kann berdies numerisch gefhrlich sein. Beim Explosions
sto kommt es auf die Ausdehnung an: handelt es sich um einen Rohr
bruch, der nur benachbarte Teile in Mitleidenschaft zieht, so sind
nur diese Teile zu betrachten, trifft der Sto von auen auf groe
Teile der Wandungen, so ist das globale Verhalten wichtig wieder
die untersten Eigenfrequenzen der ffesamtstruktur.
3_._2_KKWGebude
a) Ist das Gebude ausreichend mit ber Hhe und Breite durchgehen
den Wnden ausgesteift (dies ist z. B. bei einem im Grundri ge
schlossenen Querschnitt meistens der Fall), so kann man es auf
ein Balkenmodell abbilden, das bei Unsymmetrie dann gekoppelte
Biege und Torsionsschwlngungen ausfhrt. Sind die Steifigkeiten
aufgelst, so ist ein Balkenmodell,selbst fr die Berechnung der
untersten Eigenfrequenzen,nicht zulssig. Als Modell mu man
dann eine (wenn auch vereinfachte) rumliche Struktur verwenden.
b) Symmetrieeigenschaften erlauben das mechanische Modell zu verein
fachen.
) Hat das Gebude eine Symmetrieebene, so sind die Schwingungen
in dieser Ebene unabhngig von den Schwingungen senkrecht dazu,
die ihrerseits Jedoch mit den Drehschwingungen um die Hochachse
gekoppelt sind, Ist der Schwinger ein Balken nach a ) , so findet
in der Symmetrieebene eine reine Biegeschwingung statt, Ist
eine Baikenabbildung nicht zulssig, so bringt die Symmetrie
eigenschaft eine wesentliche Erleichterung insbesondere dann,
wenn zustzlich eine ausreichende Zahl durchgehender Decken
vorhanden ist: es gengt in Richtung der Symmetrieachse ein
ebenes Modell zu betraenten.
130
f) Bei Doppelsymmetrie sind beide Schwlngungsrichtungen voneinan
der und von der Drehung des Gesaratgebudes um die Hochachse
unabhngig. Beim Balken trennen sich die beiden Biegungen und
die Torsion,
Sind eine oder mehrere dieser Voraussetzungen nur nherung3weise
erfllt, so liefert die Betrachtung des ebenen (oder eindimen
sionalen) Modells nur ein vielfach allerdings ausreichendes
Nherungsergebnis.
Oft auch sind bei einem Gebude die Voraussetzungen fr eine Ver
einfachung der Struktur nicht in allen Gebudeteilen gleichermaen
erfllt. Es kann dann notwendig werden, das Gebude oder einzelne
Teile fr die verschiedenen Schwingungsrichtungen auf Jeweils andere
Modelle abzubilden.
3_3_KKWEinbauten
Gebude und Einbauten fhren gekoppelte Schwingungen aus, mssen
also als Gesamtsystem betrachtet werden. Sind Jedoch die Massen
der Einbauten klein gegen die der Decken, auf denen sie stehen, 30
knnen die Einbauten fr die Berechnung der GebudeSchwingungen
weggelassen werden. Die Schwingungen dieser Einbauten lassen sich
Jedoch nur aus dem Gesamtsystem ermitteln, entweder indem man von
vornherein koppelt, oder mit guter Nherung indem man die "Aus
gnge" des Gebudes an den Stellen, wo die Einbauten sitzen, als
"Eingnge" fr diese betrachtet. Bei der Spektrummethede ist dieses
letzte Verfahren Jedoch nicht ohne weiteres verwendbar.
3._311_ durchgefhrten Berechnungen
Flg. 1 zeigt Grundri und Schnitt des Reaktorgebudes eines Druckwasserreaktors (PWR). Da es sich um einen geschlossenen Querschnitt
handelt, der auerdem mit durchgehenden Wnden und Decken versehen
ist, lnt sich das Gebude fr horizontale Erregung auf einen balken
abbilden, bei dem Jedoch unbedingt Schubnachgiebigkeit und Drehtrhelt bercksichtigt werden mssen. Bedingt durch die ausreichende
Symmetrie treten nur gewhnliche Biegeschwingungen auf; bei dem Gebude eines PWR mit etwa 1000 MW Leistung liegt die unterste Eif.enfrequenz bei ~10 Hz.
Fig. 2 zeigt das rumliche Schwingungsmodell der ber Rohrleitungen
miteinander gekoppelten Einbauten eines PWR. Die Befestigungen der
Einbauten an das Gebude sind auf Federn abgebildet. Da die Massen
von Reaktordruckbehlter und den vier Dampferzeugern D. - D, nicht
vernachlssigbar sind gegenber den Gebudemassen, mu das gekoppelte System Gebude-Einbauten berechnet werden.
- 131 Flg. 3 zeigt das Reaktorgebude (und gleichzeitig das Schwingungsmodell) eines Siedewasserreaktors (BWR). Auch hier ist wieder die
Abbildung auf einen Balken mglich; da es sich Jedoch um einen unsymmetrischen Querschnitt handelt, mssen die gekoppelten BiegeTorsionsschwingungen berechnet werden. Fig. zeigt zwei der rumlichen Eigenformen eines BWR von etwa 800 MW. Die tiefste Eigenfrequenz liegt bei 3,6 Hz, Fig. ) und d) zeigen den Verschlebungsund Beschleunigungsverlauf dieses Gebudes fr das auf lm/sec normierte und geglttete Beschleunigungsspektrum des El Centro Bebens
von 190 mit 7 Dmpfung.
In Flg. 5 1st der Schnitt durch einen Siedewasserreaktor (800 MW)
gezeigt und in Flg. 6 das zugehrige Schwingungsmodell mit den zu
den 3 tiefsten Biegefrequenzen gehrigen Eigenformen, ferner in Flg.7
der Verschiebungs- und Beschleunigungsverlauf. Das Schwingungsmodell
1st hier eine aus mehreren Balken zusammengesetzte Struktur.
In einem Maschinenhaus sind die Steifigkeiten i. A. aufgelst, Scheiben sind kaum vorhanden, so da als Schwingungsmodelle nur rumliche
(oder mehrere ebene) Modelle in Betracht kommen. Fig. 8 zeigt das
ebene Schwingungsmodell eines Maschinenhauses. Die Kopplung des
(schweren) Hauptturbosatzes mit dem Gebude mu bercksichtigt werden;
sie wird besonders deutlich in der zweiten Eigenform (Fig. 9 ) . Fig. IC
zeigt den Verschiebungs- und Beschleunigungsverlauf.
. Zusammenfassung
Die Arbelt berichtet ber Schwingungsrechnungen fr erdbebenerregte Kernkraftwerksgebude. Zunchst wird das Rechenverfahren dargestellt: Gewinnung der Schwingungsgleichung mit Hilfe der Methode der
finlten Elemente, speziell gekoppelt mit Ritzanstzen fr die Verrckungen innerhalb der Elemente, Lsung ber di'e Eigenvektoren,
insbes. fr das Erdbebenproblem.
Da man Schwingungsrechnungen nie fr wirkliche Gebude sondern nur
fr Modelle durchfhren kann, ist eine gute Abbildung ein wesentlicher Teil der Aufgabe, Nach einigen allgemeinen Hinweisen wird
das Abbildungsverfahren an charakteristischen Beispielen gezeigt.
- 132
Literatur
[lj
Ahorner, L., H. Murawski und G. Schneider

Die Verbreitung von schadensverursachenden Erdbeben auf
dem Gebiet der Bundesrepublik Deutschland.
Zeitschrift fr Geophysik, Band 36, 1970
[2]
Hansen, R.J.
Seismic Design for Nuclear Power Plants.
M.., Press, Cambridge, Mass. und London 1970
[3]
Hurty, W.C. und M.F. Rubinstein

Dynamics of Structures.
Prentice-Hall, Nev/ Jersey 196
[k]
International Association for Earthquake Engineering

Earthquake resistant regulations: a world-list.
Tokio 1963
5]
Marguerre, K. und H.-T. Woernle

On the approximation of continuous systems by systems
with a finite number of degrees of freedom,
IUTAM Symposium, Paris 19U5
[6]
Schwarz, Rutishauser und Stiefel

Numerik symmetrischer Matrizen,
Teubner Verlags-Gesellschaft, Stuttgart 1968
[7]
US Atomic Energy Commission.

TID 702: Nuclear reactors and earthquakes
[8]
Wiegel, R.L.
Earthquake Engineering
Prentice-Hall Inc., Englewood Cliffs, N.J. 1970
133
fe
irkr
ur id
hi
w;.tvj..v.^y^.vî'.'rr.irv.^y'.'T^.^. .;::.:..v :~^~?^: ; ' , . . v .'.. A J . . < ^ . r . A
a) Schnitt
b) Grundriss
Fig.1
Reaktorgebude (PWR)
134 -
v^w
rumliches
Schwingungsmodell
Fig 2
System der
Einbauten (PWR)
Schwingungsmodell
fr Horizontalschwlngungcn
Fig. 3
Reaktorgebude (BWR)
- 135 -
Jm
f2 = 5,5Hz
d) 2. Eigenform
Fig.4
= 16 Hz
'b) 4. Eigenform
c) Verschiebungen
Reaktorgebude (BWR)
Fig. 5
Siedewasserreaktor ( BWR )
d) Beschleunigungen
- 136 -
oCr-
-c
-c
0- )
W
f, =3,6 Hz
3=11Hz
SCH
) Schwingungsmode
Fig. 6
b) 1. Eigenform
c) 2. Eigenform
d) 3. Eigenform
Siedewasserreaktor ( BWR)
"\0
(
(
c-
-5
1'!H
o) Schwingungsmodell
Fig. 7
b) Verschiebungen
Siedewasserreaktor (BWR)
c) Beschleunigungen
13 7
Schwingur.gsmodell fr Horizontalschwingungen
Fig. 8
Maschinenhaus
rritrimmammaiuBrmnamme
2. Eigenschwingungsform i f2 = 2,4Hz
Rg. 9
Maschinenhaus
138
/^/^W/;^y/AV/A7/^/>v/;i&/'
0 12
) Verformungen
Fig. 10
Maschinenhaus
3 4 rn/s2
b) Beschleunigungen
139
DISCUSSION
Q
Y. HAYAMIZU, Japan
1. By which method do you analyze the eigenvalue of the m a t r i x ?

2. Up to how many degrees of freedom can you analyze the eigenvalue of the m a t r i x
fi^
H. WOLFEL, Germany
1. We use the Householder QR algorithm.

2. Our examples had about 50 degrees of freedom.
. OMATSUZAWA, Japan
Could you tell us the natural frequency of the turbine foundation ? And how much
is the maximum relative displacement to the turbine building ?
. WOLFEL, Germany
a) The natural frequency of the turbine foundation was about 3 cps.

b) Relative displacement between turbine and building was about 20 mm (for the top of the
building 40 mm).
Q
A
Q
O. HENSELEIT, Germany
Did you consider plastic deformation and the influence of soil ?
H. WOLFEL, Germany
No, we didn't but we considered damping with 1% of critical.
U. HOLZLOHNER, Germany
Haben Sie auch die " s t e p - b y - s t e p integration" versucht ?
H. WOLFEL, Germany
Nein, wir haben bisher ausschliesslich die "modal analysis" angewendet.
K 3/1
STATISTICAL
ON
TREATMENT
REACTOR
OF
BUILDINGS
A. AMIN,
SEISMIC
AND
LOADINC
EQUIPMENTS
A.H.-S. ANG,

University of Illinois, Urbana, Illinois, U.S.A.
ABSTRACT
A statistical approach for the seismic analysis and design of nuclear reactor
buildings and facilities is proposed.
The concepts are derived through random vibration
theory; a comparison of specific results obtained on this basis with those obtained for
a normalized set of recorded accelerograms attests to the validity of the proposed
statistical
approach.
142
1.
INTRODUC TION
Random vibration concepts have been used
in seismic analysts since the early
fifties; see for example Goodman, Rosenblueth, and Newma rk [1] .

the squareroot of the sum of squares of the maximum modal
The procedure of taking
responses, for example, s
based on certain approximations derived from the theory of random vibration.

to reasonable approximations for regular and symmetrical
This
leads
buildings for which only the
lateral deflections and the story drifts are the quantities of primary
interest.
However,
in the seismic analysis of nuclear reactor facilities, the systems are quite complex,
and response quantities other than displacements are also of significance.
Moreover, in
certain soil dynamic evaluations, the number of exceedances beyond a high stress level
is also of interest.
Because of the complexity of the systems involved
in a reactor
facility, a direct random vibration analysis, therefore, appears to be potentially useful

and desirable.
In this paper the available random vibration concepts that are
of practical
significance for reactor facilities are summarized, and specific numerical
results
obtained therefrom are compared with those from a normalized set of recorded accelerograms.
This comparison demonstrates
the validity of using a direct random vibration approach in
seismic ana lys i s.

The implementation of the procedure requires a description of the spectral
density for the design earthquake.
At present this information can be analytically
derived from the postulated average response spectrum, which

purposes of design.
specifying a response level corresponding

quake.
is presumably specified for
The dispersion of the maximum response is taken into account by

to a probability of exceedance during the earth
This probability serves as a consistent basis for specifying the response level
for design; this provides an alternative to the current practice of using an amplified
design spectrum.
2.
2.I
PREDIC TION OF EXTREME STRUC TURAL
RESPONSE
Approximate Distribution
It is wellknown
that the exact distribution of the extremes of the random
system response X(t) within an earthquake duration t
is not available.
number of studies, including the works of Amin, Tsao, and Ang

Kameda
[3] have shown that when high response levels of design
However, a
[2] and Goto and

interest are
considered,
it i s reasonab le to ignore the statisti ca 1 dependence of the up and downcrossngs of

the response levels = +_ b, as shown in Fig. 1.
If it is assumed that the response
process has zero mean and the up and downcrossings of level + b occur
in accordance
with a Poisson process, the distribution function of the random var i ab 1 e
can be d e s c r i b e d
m a x (t) I
(1)
. 0<t<t .
d
as,
F v ( b , t .)
X
d
m
Pr
< b) = I ( t .) = exp
m
e d
[2
'
0
v,
b
( t)dt ]
(2)
143
n which v . ( t ) is the rate of upcrossing at level = b and is obtained from the j o i n t

b
density function of X and X at time t as follows
v
b(t)
*fx X <b'*,t)dx
(3i
Assuming that the earthquake response s a Gaussian Process, the design response
level b corresponding to an exceedance probability p (t.) can be evaluated numerically
from the above equations.
Also if the response is assumed to be a stationary process,
the evaluation of b can be obtained in closed form as follows:

(k)
b =
where,
t J '
MO
[2 n ( V )Yn
<5>
q = n(l }
e
e
In these equations
and
(6)
are the standard deviations of X(t) and X(t). The
evaluation of these quantities is described subsequently.

2.2
Range of Validity of Stationary Response Assumption

Recorded earthquake accelerograms are clearly not stationary except perhaps
in the st rongphase duration.

the earthquake magnitude.
The duration of strong shaking, in turn, nereases w th
In addition to accelerogram nonstationarity, additional
evolutionary trends are also introduced into the response process because of zero
initial condi t ions.
A quantitative study of the influence of nonstatonarty was recently reported
by Ami and Gungor [*t]. The findings of this study can be stated as fol lows : For a
singledegree offreedom (SDF) system having 2 percent damping and a period less than 2
seconds, no more than 15 percent error is introduced if the response process s t reated
as a stationary process.
This conclusion applies to a magnitude 7 earthquake which is
comparable to the El Centro record of May 18, I9A0.
For larger magnitude earthquakes,
for which the duration of strong shaking is longer, smaller error will be introduced
by ignoring nonstationarity.
larger values of damping.
The same is true for systems with lower periods and/or
Since the fundamental periods of most reactor facilities are
less than 1 second and most structural dampings are taken to be larger than 2 percent,
the above discussion indicates that treating the earthquake response of power plants as
a stationary random process is reasonable.
The error involved is, of course, on the
conservative side.
2.3
Response Statistics
The use of eqs. (k) and (5) requires the variances of X(t) and X(t).
Evaluation
of these statistics for linearly elastic system is treated in many textbooks on random
144
vibration.
Pertinent equations of relevance to the subsequent presentation can be
summarized as follows:
G.(oj) = u 2 G ()
(7)
G (lo)du)
(8)
? =
G-()du
(9)
which G () is the power spectral density of the response X(t) which is related to the
input earthquake spectral density G..(iij) according to eq. (10)
y
G
()
where
()
"()
X
G..(D)
y
(10)
is the compi ex frequency response function and denotes the compi ex conjugate.
In seismic analysis, the method of modal analysis is usually used and a small
constant modal damping is assumed.
On this assumption, an arbitrary response X(t) can
be wri tten as
X(t) =
I B x k q k (t)
k=l
(II)
where = number of active modes, , =* a time invariant quantity depending on the shape
of mode k and the definition of the response X, and
^'k +
in which , = modal frequency (sec
"k ^k
'
y(t)
) ; and y ground acceleration.
response analysis of eqs. (11) and (12) (y = e
A steadystate
) , yields
() I H r 1
kl
(,2)
(l3>
. ()
. () = 1 ( !^ ) 2
i 2 ()
()
Using this information in eq. (8) we obtain
a\=
Similarly, eq. (9) yields
()0-
" ' k
(,5)
145
Bxkv<YV \ i
(,6)
1 n which
G., ()
tn
do
(17)
do
(18)
zk(o)z(o)
-O
and,
I 2 G. ()
co
kt
zk(o)z(o)
-00
For most power spectral density functions, G(), developed to describe earth
quakes, the Integrals In eqs. (17) and (18) can be readily evaluated by the method of
residues and using complex arithmetic features available In most computers.
Finally,
the double summation In eqs. (15) and (16) may sometimes be approximated, respectively,
by
n
L 0
.. 2 '
k1 o k
"kk
and
<20>
4\\*
k=l
.
k
When the frequencies of the active modes of a particular response are wel 1-separated,
eqs. (19) and (20) are reasonable approximations.
This Is usually true, for example,
for the lateral deflections of a cantllevered structure with regular distribution of

mass and stiffness.
However, eqs. (19) and (20) are poor approximations when floor
accelerations In a beam-like structure is considered.
For this reason, eqs. (15) and
(16) ought to be used when treating the response of a complex system.

In the following results, when eqs. (15) and (16) are used in eqs. () and (5),
the resulting value of b will be identified as b
, whereas the corresponding value of
b obtained with eqs. (19) and (20) will be referred to as b .

3
3-I
COMPARISONS WITH NORMALIZED ACCELEROGRAMS
Description of Excitations
Two horizontal components of the following four records obtained in the West
coast of the U.S. were used in this study:
El Centro (I2/30/3A and 5/8/0),
When k - i and if most of the contribution to the integral in eq. (17) comes from the
values of near . , a well-known approximation is .. ..(. )/2 ^. However,
in earthquake response analysis, frequency conditions exist for which this approximation
produces noticeable inaccuracy.
- 146 Olympia
(/3/9),
area under
and T a f t
(7/21/52).
The r e c o r d s w e r e n o r m a l i z e d t o have t h e same
t h e undamped p s e u d o - v e l o c i t y
The f o l l o w i n g e x p r e s s i o n
is
curve
(V - ) f o r - 0 . 1
to 2.5
used f o r
the s p e c t r a l
density of
second.
the
ground
acceleration,
G_(u)
1
G.(uj)
'
G,(u)G2(o)
2(^-)2
I Q).
-*
,
i n w h i c h , = 15-5 sec
I
shows
3-2
ui.
, 0 . - 0.6*42, 15*7 sec

1
2
t h e median s p e c t r a ,
comparison w i t h
(30)
, S
.5, obtained using eq.
(23)
.0052 f t
(21)
in eq.
the 8 n o r m a l i z e d
sec
-*
(K)
Figure 2
and
Us
records.
Sys terns Cons i de r e d

Three systems,
a single-story
stiffness
in
structure
2;
shown I n F i g . 3 as s t r u c t u r e s
Structure
the d i r e c t i o n
freedom.
this
perpendicular
Structure
structure,
structure
cribed
close
3 this
In Ref.
therefore,
t o each o t h e r ;
ratio
is
2 , and 3 a r e used I n
1.05
to
the ground m o t i o n ;
involves
10 d e g r e e s o f
2 and 3 a r e s e l e c t e d
for
structure
More d e t a i l s
2 the
pertaining
stories,
2 Is
mass and
there are,
5 identical
the
Structure
e between the c e n t e r s o f
3 i s composed o f
The p a r a m e t e r s o f s t r u c t u r e s
a c t i v e modes are
I,
1 i s a 10 d e g r e e - o f - f r e e d o m s h e a r beam.
shear s t r u c t u r e w i t h e c c e n t r i c i t y
two d e g r e e s o f
therefore,
each
resembling
freedom.
such t h a t
ratio
T-/T.
the periods o f
I105;
to these s t r u c t u r e s
the
for
are
des
[*l].
Comparison of Results
Columns 2 ,
response v a l u e s ,
integration of
accelerograms.
ground, y
rotation.
about
-1
the corresponding average s p e c t r a o f
comparative studies.
33
(22)
0.650 + 2.2 (H) 2 + 1.63 ()*

,
,
7
22
2 2
[ (L) ] + 2.2 (<-)
2
2
22 ( )
= S
2 <^) 2
[1 ( )
.
(21)
Table
I summarize,
respectively,
the e q u a t i o n s o f motion o f s t r u c t u r e s
Generally,
the a b s o l u t e
u is
5"percent
the d i s p l a c e m e n t
displacement
The s e c o n d h i g h e s t
1/8. A
smaller
3 and ^ o f
t h e a v e r a g e , and t h e s e c o n d h i g h e s t
value
of
range o f
I,
range o f
from a
2 and 3 u s i n g
o f a mass c e n t e r
t h e mass c e n t e r ,
the
stepbystep
the 8 n o r m a l i z e d
relative
and 0 d e n o t e s
to
the
the
floor
r e p o r t e d w o u l d have an exceedance p r o b a b i l i t y
damping was used f o r a l l
damping, a l a r g e r
the
value obtained
t h e modes o f
response values
than those
vibration;
reported
of
therefore,
in Table
for
I would
be e x p e c t e d .
The a b o v e r e s u l t s
random v i b r a t i o n m e t h o d .
are
compared w i t h
The l a t t e r
the
r e s u l t s are
response q u a n t i t i e s
b ' , b
'
b'
obtained with
and b . , w h i c h
are
the
14V
respectively, the median and the response level corresponding to a 1035 exceedance
probability.
These are presented In Columns 5 through 8 In Table I, and are given
in terms of the average and second highest values of Columns 3 and , respectively.
The results summarized In Table I indicate that the approximation of
eq. (19) may be poor even for canti levered beam-like structures; see for example the
response y
in Column 5, versus the corresponding result In Column 6 obtained using
eq. (15) for the variance.
The errors caused by the approximation of eq. (19) become
more pronounced for structures 2 and 3, as evidenced in Table I; these are structures
having modal frequencies close to each other.
Since this property is not uncommon in the
higher mode responses of complex systems, the evaluation of the variances through
eqs. (15) and (16) appears necessary when considering complex structures.
Columns 6 and 8 of Table I demonstrate the validity of the proposed random
vibration approach.
Specifically, these show that the assumptions of a Gaussian response
and Poisson occurrence of level-crossings produce a reliable means for estimating maximum
earthquake responses In MDF-systerns.
.
RESPONSE OF SECONDARY SYSTEMS
. I Generation of Floor Response Spectra

In current practice it is often necessary to specify floor spectra for the
design of light equipment items mounted on the primary structure.
When a deterministic
analysis Is used, this Information may be generated by making a time-history analysis

of the primary system using several accelerograms.
For each input accelerogram, the
floor motions are determined and used to evaluate floor spectra.
Final floor spectra
are specified by smoothing the floor spectra which are generated from various accelero
grams.
Clearly this Is an expensive and time-consuming process; for this reason, it is
impractical to study the effects of possible changes In the primary system frequencies
on the floor spectra.
Since the primary system Is a sharply resonant system, the effects
of changes in its frequencies on the floor spectra Is far more important than the signifi
cance to the deformationaI response of the primary system itself; minor changes in the
frequencies of the primary system can produce a significant effect on the response of
equipment mounted on the primary system.
A random vibration approach can be used to generate floor spectra rapidly
and, therefore, the influence of changes in the primary system frequencies on floor
spectra can be evaluated by making a sensitivity analysis of the results.
This Involves
the use of eqs. () and (5), in which X(t) and X(t), in this case, refer to the response
of the equipment as shown In Fig. 4.
The required equations for evaluating a
is given
below.
If the inertia and damping coupling between the equipment and the primary
system In Flg. Is ignored, and damped normal coordinates are used to consider, approxi
mately, the effects of damping, the equipment response X(t) in Fig. 4 can be written as
X(t) - I V> k (J)x k (0

k-1
(24)
148
+ 2
+ 2
2Vk

Vk
+
"k %
+ 2
(25)
1 1A
" *(t)
u \
<26>
where .. , y k = modal frequency and p a r t i c i p a t i o n f a c t o r , r e s p e c t i v e l y , f o r the primary

mode k; <l>Aj) " amplitude o f mode k at f l o o r j ; and ,
= damping values for the
primary and secondary systems, r e s p e c t i v e l y , and = frequency of the secondary system.

The complex frequency response o f the equipment, determined from eqs. (24)
through (26),
is
() = V k ( j )
a (ul)
k=l
[<V2(<)zlk(t)r'
(27)
In which z k is given by eq. (14) and

a () 1 + 2i B, ( )
k
2(,
= , - (^)2
(28)
u>|k
i 2,2(>-)
(29)
Using eq. (27) in eq. (10) and using eq. (8) yields the variance of X(t) as
= vI vI
VA(J)(J)
k A
k<> a fo> c yM
dJ
30)
|2() | 1 ( () ()
The integral in eq. (30) can be readily evaluated by residues; the tedious algebra can
be avoided by using complex arithmetic routines available in most computers.
When the primary system consists of a single mass
.2
I t s discussed in Ref.
[| + /.g2 (^-)2]G..()
Y
' 11
.
2 ~ 2^ ' I lz| j ^
[5] that because of the t a i l behavior of \z |
and | . . |
for
for l i g h t l y damped systems, c e r t a i n bounds e x i s t for the f l o o r response spectrum.

S p e c i f i c a l l y , when /.. <^ 1/3, the spectrum approaches that of a SDFsystem w i t h
damping . and d i r e c t l y subjected to the ground motion y; whereas when _/.. > 3. the
pseudoacceleration of the f l o o r spectrum approaches the pseudoacceleration of the
primary system.
This behavior is i l l u s t r a t e d in F i g . 5 for the NS component of the
May 18, I9*t0 El Centro accelerogram, w i t h f . .

*t .2
= ../2 = 5.8 eps.
Random V i b r a t i o n of MDFEqui pment

The f l o o r response spectra may be used w i t h c u r r e n t l y a v a i l a b l e rules for
combining modal maxima to analyze MDF equipments mounted on a s i n g l e f l o o r of a reactor
149
bullding.
However, the floorspectra cannot be directly used for equipments, such as
piping systems, that are connected to several

piping systems is described
floors.
A response spectrum approach for
in Ref. [5]; t is mentioned that a random vibration approach
to the analysis of piping systems would be preferable.
Such an approach should also
Improve the accuracy of results for MDF systems resting on a single floor.
No conceptual
difficulty exists for the application of random vibration principles to MDF secondary
systems along the lines proposed herein.
However, additional studies are needed to
simplify the fourfold summation
in calculating all the coupled terms
expression for
involved
(two summations
In the
for the building modes and two for the equipment).
This i tern Is currently under study.

5.
CONCLUSIONS
The fundamental periods and damping o f most r e a c t o r
facilities
the seismic responses o f these s t r u c t u r e s can be t r e a t e d as a s t a t i o n a r y

Practically
f e a s i b l e procedures f o r the a n a l y s i s . o f
are a v a i l a b l e from random v i b r a t i o n t h e o r y .
reactor f a c i l i t i e s
The use o f
set o f
recorded accelerograms.
This shows t h a t
used to o b t a i n r e l i a b l e estimates o f seismic

for
the a n a l y s i s o f equipment responses.
the proposed s t a t i s t i c a l
results
approach can be
the use o f s p e c i f i c
the allowance f o r
maximum response can be made more s y s t e m a t i c a l l y

of
leads to
i n t e g r a t i o n o f a normalized
loading on the main s t r u c t u r e as w e l l as
This avoids
and lengthy s t e p b y s t e p i n t e g r a t i o n ; a l s o ,
random process.
and equipments
random v i b r a t i o n
that are In agreement w i t h those obtained from the d i r e c t
are such t h a t
accelerograms
the d i s p e r s i o n i n the
than by means o f a simple
amplification
the average response s p e c t r a .

6.
The study described herein
probabilistic aspects of structural
ACKNOWLED GMENT
is part of a continuing research program on the
safety and design, being carried out in the
Department of Civil Engineering, University of Illinois at Urbana,

program is currently supported by the National
grant GKI812X.
The numerical
Science Foundation
results reported
research
in the comparative studies were
obtained by Mr. I. Gungor as a part of his doctoral dissertation.

are gratefully
Illinois, USA. The

through
His contributions
acknowledged.
REFERENCES
[1]
GOOD MAN, L. E., ROSENBLUETH, E. and NEWMARK, . ., "Aseismic D esign of Firmly

Founded Elastic Structures," Trans. ASCE, Vol. 120 (1955).
[2]
AMIN, M., TSAO, H.S. and ANG, . .S., "Maximum Response Statistics of Simple
Linear Systems to Earthquakes of Prescribed Intensity," Proc. ASCEEMD Specialty
Conf., Purdue University, November I969.
[3]
GOTO, H. and KAMEDA, H., "Statistical Inference of the Future Earthquake Ground
Motion," Proc. 4WCEE, Santiago, Chile, Vol. 1 (I969).
[4]
AMIN, M. and GUNGOR, I., "Random Vibration in Seismic AnalysisAn Evaluation,"

Presented at ASCE National Structural Meeting, Baltimore, Maryland 1971.
[5]
AMIN, M, HALL, W. J., NEWMARK, N. M., and KASSAWARA, R. P., "Earthquake

Response of Multiply Connected Light Secondary Systems by Spectrum Methods,"
Proc. of ASME
First Congress on Pressure Vessel and Piping, San Francisco, Ca.
May I97I
150
TABLE I - COMPARISON OF R E S U L T S :
OF ACCELEROGRAMS
Response
Quanti ties
RANDOM VIBRATION VS. INTEGRATION
(B - .05 In all modes)
Values from Records

Range
(2)
(1)
Ave.
(3)
2nd
Highest
(4)
(a) Structure 1 ,T
Uj 10
(u ? - u 6 ) IO1*
.50(30)
AVE.
(5)
(6)
1.01
1.00
(30)
b
.io
2nd H Ighest
(7)
(8)
- 2 Sec
291-477
383
467
208-409
283
322
.95
98-158
.99
98
95
.99
.99
1.00
.94
97
.9
122
144
2.19-4.08
3.O7
3.49
.7
.98
.90
.99
4.10-6.64
5.10
6.07
98
95
97
93
77
93
11
( u 1 0 - u 3 ) IO
(b) Structure 2 ,T, - 0.66 Sec

L
u 10
IO
y
IO
432-822
565
659
76
92
469-935
670
906
1.59
1.15
4.41-8.22
5.70
6.68
487-937
692
901
.75
91
1.38
1.00
75
.91
1.55
1.12
1.39
1.01
(c) Structure 3 ,T, - 2 Sec

Uj IO
425-809
600
669
.79
1.04
.85
1.13
||
I72O-277O
207O
2250
.75
1.00
.89
1.12
240-397
271
318
3k
.72
95
120-214
155
173
1.16
2.69
1.26
I.78-3.78
2.74
3.49
72
98
.66
.90
3.06-5.07
3.6O
4.14
.70
91
.72
542-778
702
771
2.28
1.08
2.45
LIS
8OI-I52O
1000
II90
2.04
.98
2.01
.98
u IO
(u 5 - u A ) IO1*
, IO
h
, IO
5 IO
72
2.48
151
x(t)
FIG. I
0.8
FIRST PASSAGE PROBLEM
0.7
+ +
8
2
o
+ Average of 8 Normalized Accelerograms

Median Spectra, 6g(<u) =6|() G2(ui)
Median Spectra, G-() = G.(ui)
I
1.0
2.0
3.0
Period, T, sec.
FIG. 2
COMPARISON OF MEDIAN PSEUDO VELOCITY SPECTRA TO AVERAGE

SPECTRA OF NORMALIZED ACCELEROGRAMS -- =0.05
0.625 m 1
1.05 k
0.682 m
0,738 m
0.795 m
1.10k
1.15k
1.20k
0.852 m '
1.25k
0,909 m
1.30 k
0,966 m
1.35k
1,022m l
1,40 k
1.079 m '
1.45 k
1,136 m '
1.50k
agi
2EZZZ3ZZEJ
ttii
2
CMi
KTe L
viii/r/ritimtiim
Pion
k V7
I
t^w
k
Hx
'c.M. ,. x, Cflj
kx
yû + yg
r
m, J
taammmazax
Typical Pion
i=5
=4
=3
=2
FIG. 3
m, J
m, J
ve
! Vi
=l
Structure 2
l!
kiVe
*i 'W'
Struc fure !
y = Jj+yg
Structure 3
STRUCTURES USED IN THE COMPA RITIVE STUDIES
m, J
m, J
m, J
153
iiitiiiiiiiizzm
^P~*
Equipmi
ipment
rmitiY/ittiirm Floor j ofPrimary System
nir!?
FIG. 4
lOOOi
SDF EQUIPMENT MOUNTED ON A PRIMARY

SYSTEM
0,1 0.2 0.40.6 I
4 6 10 20 4060100
Frequency L, cps
FIG. 5
RESPONSE SPECTRUM FOR STRUCTURE 2 AND ITS

LIMITING VALUES ,= 0.02, 2 =0.005, EL CENTRO
154 -
DISCUSSION
N. N. KULKARNI, India
In case of p r e s s u r e tube type of r e a c t o r s fuelling machines also operate during
operation for on-loading fuelling. If an earthquake occurs during operation "MCA" can result.
It is n e c e s s a r y to establish a c r i t e r i o n for the design of fuelling machines. Can the authors
present some s t a t i s t i c a l data for such a c a s e ?
J. M. DOYLE, U. S. A.
It s e e m s that if you consider the fuelling machines as an equipment item, the
methods outlined in our paper could be used to obtain the floor motion at the location of the
machine. Therefore, the information you want would need to be calculated for each individual
c a s e . It would depend, of c o u r s e , on the design basis earthquake, and the dynamic properties
of the p r i m a r y s t r u c t u r e .
K 3/2
THE RESPONSE SPECTRUM ANALYSIS

TO AN ARTIFICIAL EARTHQUAKE
WITH TWO GROUND PREDOMINANT PERIODS
H. SATO,
ABSTRACT
Analysis of the response spectrum of structure system simulated by one-degree-of-freedom

to an artificial earthquake; stationary random vibration with two ground predominant period,
is made. This is investigated as the extensive study for the case of single predominant period. Then for the estimation of maximum of the artificial earthquake and response wave form
the probability density function of extreme is made use of. According to this analysis both
spectra show good coincidence about their shape which could not be seen in the analysis by
single ground predominant period. For the case that the structure system has nonlinearity of
elasto-plastic characteristic the response spectrum of velocity and displacement in terms of
standard deviation is obtained. However general characteristic of the spectrum is same as
that for single ground predominant period, the results of the analysis show better agreement
with the spectrum to actual earthquake record than those by single ground predominant period
did. The expectation and the variance of the spectrum are obtained by assuming probability
density function for realization of system parameters as ground predominant period, natural
period and damping ratio of structure model. Considering the existence of two ground predominant period close each other suggests that seismic force does not vary irrespective of the
fluctuation of the system parameter.
1. INTRODUCTION
The response spectrum of one-degree-of-freedom system simulating structure shows a characteristic of dynamic response of the system to an earthquake record and this also implies
characteristics of the earthquake wave form itself. It is given by Housner et al. [1] that
the general characteristic of the response spectrum for a number of earthquake is described
as that the velocity spectrum keeps constant for the natural period of the system higher
enough and the acceleration spectrum has at least one peak and decreases as the period becomes large. Further study by Tajimi [2] has made it obvious that the period where the peak
stands corresponds to predominant period contained in earthquake wave form.
There are several merits that artificial or simulated earthquake is made use of for the
response analysis in place of actual earthquake record. Statistical approach to predict response spectrum can be made analytically by simulated earthquake as well a? it is obtained
experimentally. And it also makes it possible to distinguish effect of reciprocal relation of
- 156 the system parameters as ground predominant period, natural period and damping ratio of structure to the spectrum, shich is easily masked for the spectrum to actual earthquake record because of its complexity.
Response spectrum is originally plotted by taking maximum of time history of the response. As for the acceleration response it also can be represented by taking the ratio of
maximum between input earthquake and the response, response spectrum of accleration amplification factor. According to the analysis that the maximum is in proportion to the standard
deviation and random vibration corresponding to earthquake has the characteristic of band
limitted white noise filtered through one-degree-of-freedom system the natural period of
which is equal to predominant period of ground, the response spectrum obtained through the
simulation agrees well by covering a number of spectra to earthquake records in sense of an
envelope [2]. The author has made an investigation that he applied the probability density
function of extreme by Rice [3] in order to find maximum, that is, where the function is
small enough was assumed the maximum [4]. The results were similar to the case that the standard deviation was adopted for the maximum.
These analyses were all performed by the simulated earthquake with sinle predominant
period in spite that the spectrum to earthquake has several peaks. Then this paper studies
the statistical analysis to simulated earthquake with two ground predominant period which is
the simplest case for multi-predominant period. The results are compared with those actual
earthquake records. The study is extensively made for the case that structure system has
elasto-plastic nonlinearity. For both cases the approach made in this paper gives a better
agreement even for the shape of the spectrum in addition to its magitude [5), [b].
If once the response spectrum statistically given is considered reliable by comparing
with that of earthquake record, seismic force applied to structure can be approximately predicted by being given the system parameters. However, these parameters realized after construction have naturally stochastic characteristics in estimating them at stage of design
beforehand. The mean and the variance based on the analytically estimated response spectrum
can be evaluated by assuming probability density function for the parameters. The analysis is
made for the spectrum with two ground predominant period, too, and the results are compared
with those for single predominant period.
However it has not been studied in this paper, whether ground predominant period is one
or more than :>ne is expected to make difference as for results of the analysis taking noiistationarity of earthquake and the response of two-degree-of-freedom system simulating building-machine structure system, into consideration.
2. FORMULATION OF BASIC EQUATION FOR LINF.AR SYSTEM
The earthquake acceleration is assumed stationary random vibration with gaussian probability density function. It is made evident by Kanai [7] that the ground can be simulated by
one-degree-of-freedom system basically for the earthquake with single predominant period. The
power spectrum at base of the ground is supposed to be hand limitted white. The maximum of
the simulated earthquake and the response of structure model simulated by onc-degrce-of-freedom system to the earthquake is obtained from the probability density function of extreme,
that is, where the function reaches small enough is supposed to be maximum.
If two predominant periods are contained in the simulated earthquake, the transfer function of the ground model can be represented as follows,
- 157
2u)j2hj2So \
2ujjihjis*Lij!
(D
H,(s)=
s1*2uixhiSMu\
sôo^h^s+ijl
If =0 in e q . ( l ) , i t i s equal t o t h a t of s i n g l e predominant p e r i o d system on which a number of

s t u d i e s have been c a r r i e d o u t . , i = 2ir/Tj j , 1I1J2 = 2TT/TJ2, hj and h j 2 a r e t h e predominant
circu
l a r frequency and t h e e q u i v a l e n t damping r a t i o of t h e ground model. Tj and T 3 2 a r e t h e ground

predominant
period.
The p r o b a b i l i t y d e n s i t y f u n c t i o n of extreme p(y) f o r a time f u n c t i o n a l random p r o c e s s

I ( t ) with Gaussian d i s t r i b u t i o n i s given as follows by Rice [ 3 ] ,
/UU-li
Io I
I2
_ _ exp{
y 2 } * 3 y exp(
/ 7 "
2r.I0I1.-I2)
2/7L.
1
p(y) =
y2
l2
){lerf
y} (2)
/2(Io U12)
where y = I ( t ) / / T 0
and
(3)
2
IT|f|llts)| k
i2=Ii5i7j|sH(s)| kdM,
^ ^ ^ ( s l l ' k
to
(4)
H(s)=H,(s)Hi(s)
(5)
ll(s)=H1,(s)H>(s)lij(s)
(6)
For the random v i b r a t i o n c o r r e s p o n d i n g t o e a r t h q u a k e and f o r t h a t of r e s p o n s e of s t r u c t u r e

system t o i t , H(s) i s given as e q . ( 5 ) and e q . ( 6 ) . These a r e s u b s t i t u t e d i n t o e q . ( 4 ) and e q . ( 2 )
in o r d e r t o o b t a i n p(yj , where
2cubhfcsKi)2
llt(s)=
(7)
s 2 +2u) b h k s+j 2
Hf(s) =
C8)
(1*,5)2
2*)2
Integral is performed by making use of residue. It is given as formula by Newton et al. [8].
Hk(s) and Ht(s) are the transfer function of structure model and that of band limitting filter.
Hhere k is a constant related with constant power spectrum at the base of the ground, and
i[*2 are time constant of high pass and low pass filter characteristics, tok and hb are natural
circular frequency and damping ratio of structure model. The maximum is represented by a point
where p(y) is small enough, in this paper p(y)=0.01 is adopted.
3. EXAMPLES OF RF.SP0NSE SPECTRUM BY THE STATISTICAL APPROACH
Fig.l fa) shows an example of response spectra for acceleration amplification factor by
the statistical approach. Ti=0.2s and Tj2=0.Ss are adopted by taking comparison of the spec
trum with that of actual earthquake record into consideration. Equivalent damping ratio of the
ground model h3i=0.4 and h^20.3 are used. The former is the value recommended as a standard
for the case single predominant period and the latter is an example for computation. =0.015s
and t|J2 = 3.0s are equal to break point frequency fi = 10.6Hz and f2=0.531Hz respectively. As damp
ing ratio of structure model hi =0.07 is utilized.
The response spectrum for =0 has simple single peak shape. The magnitude of the factor
at Ti=.2s shows maximum and the magnitude of the factor at the peak scarecely varies if the
predominant period moves as long as the period is single. The factor at Tji=0.2s decreases as
becomes large. This implies that merging the component of long period makes the acceleration
amplification factor decrease in comparison with the case of the original single predominant
period. As the result for certain value of the spectrum has two peaks at the both predomi
- 158 nant periods of the ground.

Even if T|2 becomes longer than that of the examples, the tendency does not change. Fig.l
(b) shows another example of the combination of two predominant periods. In this case the
longer predominant period exists at five times as much as the short one, however, the sensiti
vity decreaseing the amplification factor at original short predominant period and increasing
that at another long predominant period is much smaller than the case aforementioned.
These results make it obvious that the maximum value of a response spectrum occurs when
the natural period of the structure model coincides with the predominant period of the simu
lated earthquake containing single component, in other words even if the natural period of the
structure is equal to either of multi predominant period of the ground, the amplification fac
tor is not larger than that for single predominant period.
Fig.2 (a) and (b) ahow the displacement response spectra by the statistical computation.
The parameters used for these correspond to those in Fig.l (a) and (b) respectively. These
figures explain that the appearance of longer predominant period simply increases the dis
placement response in longer period than longer predominant period. This phenomenon is really
observed about the response spectrum for violent earthquake as Niigata {June 16, 1964).
In Fig.3 the analytical results and those by actual earthquake records such as El Centro
(NS, May 18, 1940) and Taft (NS, July 21, 19S2) are compared. =0.9 are taken for the analysis
because of the better agreement with the spectrum by the actual records. Although there is
still discrepancy between both computations, the shape of the spectrum shows good agreement.
The analytical response spectrum obtained for single predominant period shows same amount of
amplification factor as those for the actual record and covers those as an envelope, however
it never gives better agreement about the shape as is seen in Fig.3 for the spectrum by El
Centro and Taft [4].
Fig.4 is the comparison of the displacement spectrum to earthquake records and that by
the analytical computation. This also shows that this simulation provides better agreement
than that with the single predominant period of the ground. The magnitude of the displacement
spectrum at T J 2 depends on that of .
In Fig.5 the power spectrum of El Centro and that of the analytical ground model are com
pared, is set as 1.0. Frequency where the spectrum reaches the maximum is made equal for
both. The actual record has more dominant power than the analytical model around low predomi
nant frequency. For the higher predominant frequency the difference should be said quite large.
It is worthwhile noticing that in spite of the discrepancy such correspondence as the magni
tude and the shape of the response spectrum in Fig.3 and Fig.4 can be obtained.
4. FORMULATION OF RESPONSE OF ELASTO-PLASTIC NONLINEAR SYSTEM
Response spectrum of elasto-plastic system to the stationary artificial earthquake was
studied as for the displacement and the velocity. For the estimation the standard deviation
was used as the measure of the maximum. The method of equivalent linearization by Sawaragi [9]
was made use of for the analysis of the nonlinearity. The discussion made it clear that gener
al characteristic found in response spectrum could be explained by the analysis, however there
remained some differences as for the precise shape of spectrum [4].
Basic equation for the system will be summarized below at first. Ihe model of the vibra
tion system is represented as Fig.6. The system has solid friction characteristic at the top
of spring. The displacement and restoring force show a special hysteresis curve as Fig.7. It
- 159 is t he simplest expression of t he e l a s t o - p l a s t i c deformat ion syst em. The equat ion of mot ion
for t he syst em can be writ t en as
mx=-cx-f-ma(t)
f=ky, i=y: f<|F|
(9)
f=F(xy)/|xy|: f>|P|
where m: mass, c: damping consta nt of the structure model, k: spring constant of the structure
model, x: relative displacement of mass to the ground, y: relative displacement of top of
spring to the ground, F: yield force and a(t): the ground a ccelera tion. The system can be re
presented by a block diagram shown in Fig.8. La pla ce tra nsform of input to nonlinear element
Z(s) can be given as follows,
Z(s)=
(H(s))
(10)
KS +(iu2+2),hkiOsn (2(dkhb+ic)
eq.(l) is used as H(s) for the case of two predominant period of ground. u b and h h are natu
ral circular frequency and damping ratio structure model for linear behaviour, and is equi
valent linea r gain for the nonlinear element. and X in Fig.8 mean relative velocity and dis
placement respectively. This block diagram shows that displacement of the system is obtained
as output of the open loop through an integral. This suggests that response of displacement is
originally unsta ble. Really the displacement response to actual earthquake record is generally
has permenent set.
As the equivalent linea r gain is used as
<=/27T(A//7)
(11)
where A is saturation va lue of nonlinear element and corresponds to F in Fig.7. The variance
of the input to the nonlinear element I, is given as
<*='
Z(s)|2d)
(12)
Z(s) is represented such a function of as is seen in eq.(10). Since < is defined for the
input level to the nonlinear element like eq.(ll), eq.(ll)and eq.(12) are combined and solved.
It is necessary to give another condition for the solution, that is,
=/7
(13)
I,îjjH(s)| 2 dL,
(14)
6 is an index which provides the ratio of the yield seismicity of structure model to the
standard deviation of earthquake acceleration. Since Laplace transform of the velocity and
displacement response is given as
U(s) = (l^|)Z(s)
(15)
(16)
The variance of these are obtained by same type of equation as eq. (14).
[[jU(s)|2dui
(17)
,=^|(5)|2
(18)
As soon as is obtained by soving eq.(ll) and eq.(12), I and I, are acquired by substituting
this into eq.(17) and eq.(18). Although these formulation were already given except Hj(s), the
summary is described for convenience of discussions.
160
5. EXAMPLE OF CALCULATED RESPONSE SPECTRUM FOR THE NONLINEAR SYSTEM
Fig.9 is example of displacement response spectrum. = coincides with linear response.
The parameters shown are same with those found in case of rponse spectrum of linear accelera
tion amplification factor. In this figure predominant periods exist in short period, so that
details of difference hardly seen. This will be shown later. General characteristics appearing
according to nonlinearity do not change in comparison with the spectrum for single predomiant
period. These characteristics are that the stronger the nonlinearity is, that is, the smaller
is, the larger the displacement response is in short period. For example =0.3 implies yield
seismicity of 0.09 for the case of the maximum input acceleration 0.3g. As increases the re
sponse spectrum with nonlinear characteristic approaches the linear response spectrum. 0.9
is the parameter which made good agreement with acceleration response spectrum to actual re
cords in linear system.
Next the velocity response spectrum is payed attention to. Some tendency as in single pre
dominant period, which in short period the spectrum for nonlinear system becomes larger than
that for linear system and in. longer period this characteristic reverses, is found. Fig.10
shows an example of velocity response spectrum using same parameters as Fig.9. Taking that ve
locity response does not show permanent excursion as displacement response into consideration,
the response spectrum for the nonlinear system depicts such shape that the spectrum for linear
system is suppressed. Although slope depends on magnitude of , the spectrum for nonlinear
system increases almost linearly as the period gets large. General characteristics for nonli
near system aforementioned are similar to the case of single predominant period. The existence
of the predominant period 0.2s mades the spectrum for linear system increase around the period
and region where the spectrum for nonlinear system stands larger than that for linear system.
These suggest that the velocity response spectrum is appropriate to describe the response
characteristic of nonlinear system.
Fig.11 shows comparison of velocity response spectrum of various cases. As the spectrum
for actual record that by El Centro is taken. As for the linear system the spectrum by the
artificial earthquake with predominant period 0.2s agrees well in short period with that by
El Centro. On the other hand the spectrum for the predominant period of 0.5s may be closer to
the real one than that for 0.2s in long period, the tendency in short period is quite differ
ent. On the contrary the spectrum with two predominant period is considered to be equal in short
period. In longer period there exist some differences quantitatively, but the tendency explains
that of the spectrum by real earthquake. The differences can be made small by fitting system
parameter . The spectrum by the artificial and actual earthquake for nonlinear system coin
cides remarkably both quantitatively and qualitatively taking some differences of into con
sideration.
Fig.12 shows ' an example of velocity power spectra of response of structure model with
nonlinear characteristic. System parameters of ground model are same with the various cases
aforementioned. As the natural period of structure model Tb=0.2s, that is, fk=5Hz is taken.
For the linear system dominant power occupies around 5Hz. As diminishes, this component
looses its power and gradually the component of 2Hz, which corresponds with that of Tt2=0.5s,
gains the power. =2.0 which means the system is very close to linear only decreases the power
of 5Hz and that of 2Hz does not show particular change. This implies that the existence of
small has same effect as increase of damping in structure model. This also affects the response
- 161 behaviour of machine structure system if it is appended to building structure system.

It is made obvious that although there is some differences as for the response spectrum
between the analysis by the artificial and actual earthquake, the characteristic of the spec
trum for the latter can be explained applying the analysis of equi linearization for the non
linear element.
6. EXPECTATION AND VARIANCE OF THE ACCELERATION AMPLIFICATION FACTOR
The analysis by the statistical computation makes it possible to predict seismic force
applied to structure system during earthquake by knowing the system parameters as ground pre
dominant period, natural period and damping ratio of structure system [10]. However these pa
rameters are given as design value, the realization of these usually differs from the estima
tion. According to observation the predominant period appearing in earthquake recorded at an
observatory point moves around as is seen in Fig.12 after Kanai [11]. This can be said that
appearance of predominant period possesses a probabilistic characteristic. Prediction of na
tural period and damping ratio also have same sort of probabilistic characteristic as the
ground predominant period. Then if probability density function is fitted for realization of
the system parameters, the response spectrum given by the statistical analysis as Fig.l can be
represented as that having its expectation and variance.
This is given by the equation for the case that only ground predominant period is assumed
probabilistic as follows,
E[A]=
2=
''
ft"
(!,/,)(,),
{A(Tk/T.)}2p(T})dTj-{E[A]}2
(19)
(20)
V
where E [A]: the expectation of the acceleration amplification factor, A(T_i/T.): the response
spectrum of acceleration amplification factor obtained by the statistical analysis, p(T,):
assumed probability density function factor and : standard deviation of amplification fac
tor. The expression of eq.(19) and eq.(20) mean that T b is fixed through the integral.
As aforementioned generally the other parameters also can be probabilistic variable. If
Tk and hk are treated as the variable, eq.(19 and eq.(20) are written as tripple integral. In
addition to these as the earthquake wave form can be considered a sample from a statistical
population, the amplification facto itself fluctuates around a mean according to Shibata et
al. [12]. If this probabilistic characteristic is estimated by other analysis, the effect also
will be able to be combimed with that of these parameters. Fig.14 shows a result of estimated
mean and 3 width for the case that in estimation of ground predominant period and natural pe
riod of structure normal probability density function is assumed. Standard deviation of the
normal distribution are taken for two abscissas. The damping ratio of structure system remains
definite. This makes it evident that the amplification factor is maximum and corresponds with
the value of the original response spectrum by neglecting probabilistic characteristic. As the
standard deviation increases, mean value of the amplification factor decreases and 3 width
around the mean is made wide. Adding another probabilistic characteristic to either of the
first parameter naturally introduces same tendency.
For the case that damping ratio is also taken having probability density distribution
Monte Carlo method is applied to performing the integral. According to Fig.14 the mean and the
standard deviation are given as
- 162 E[A]=3.40
=0.163
(21)
for log-normal probability density function with ov =o T b =0.10. Taking the damping ratio as a
stochastic variable,
E[A] = 3.08
= 0.346
are provided for 0^=0.004 and same
(22)
and o T b of normal distribution. Since the integral is
carried out by Simpson method for eq.(21) and by Monte Carlo method for eq.(22), direct com
parison is difficult. However the general tendency that E[A] diminishes and increases does
not vary.
Although eq.(21) and eq.(22) are obtained for h b =0.07 and the spectrum of the single pre
dominant period, Fig.IS shows an effect of two ground predominant period. In Fig. 15 only the
natural period of structure system is provided the probabilistic characteristic, and the nat
ural period is varied keeping ratio of the standard deviation to the natural period constant.
Fig.15 (a) is as for two ground predominant period and Fig.15 (b) is as for single predominant
period. The behaviour of mean and 30* width show that they keep constant for the former and
they are almost same tendency as the originally estimated spectrum for the latter. Zigzag
curve depends on using Monte Carlo method. However, the result means that once the ground pre
dominant period appears more than one at close period each other, the predicted amplification
factor should be constant irrespective of the change of natural period. This is considered im
portant from practical viewpoint in estimating seismic forces.
7. CONCLUSIONS AND ALNOWLEDGEMENT
It is assumed that the earthquake is simulated by stationary random vibration with
Gaussian distribution, at the base of ground it has band limitted white spectrum. The maximum
of the simulated earthquake is given by the point where the probability density function
reaches small enough. These were already studied by the author as for single ground predomi
nant period. In this paper the research that the ground characteristic is represented by com
bination of one-degree-of-freedom system, that is, two ground predominant period is extensive
ly made. This makes it possible that not only the magnitude of the acceleration amplification
factor, but also the shape of the response spectrum coincides with those by the actual earth
quake record. The analysis introduces better agreement in displacement response spectrum, too.
Response spectrum for structure model of one-degree-of-freedom system with nonlinear
characteristic simulating elasto-plastic deformation is also obtained to the artificial earth
quake with two ground predominant period by the statistical analysis. The method of equi line
arization is utilized for the treatment of nonlinear characteristic. Velocity and displacement
response spectrum explain well the general characteristic found in the spectrum to the actual
records qualitatively. As for the velocity spectrum the coincidence is seen also quantitative
ly by adjusting the comparison of the system parameter.
The mean and the variance are given in predicting the acceleration amplification factor
taking the probabilistic characteristic of the system parameters into consideration. The ex
istence of two ground predominant period causes a tendency of the mean and 3 confidence in
terval do not vary in spite of the change of estimated period of the system. This suggests
that it is generally difficult to economize design seismicity in practice of dynamic aseismic
design.
Author is very grateful to Professors S. Fujii, A. Watari and II. Shibata at University
of Tokyo for their fruitful discussions. He also expresses his cordial thanks to Messrs. K.
Suzuki, M. Komazaki and M. Ohori for their helpful arrangement performing the computation.
63
REFERENCES
[1] HOUSNER, G;!., MARTEL, R.R.and ALFORD, J.L., "Spectrum analysis of strong motion earth
quakes". Bull. Seism. Soc. Am., 432, 19534.
[2]* TAJIMI, H., "Basic theories on aseismic design of structures'.', Rep. Inst. Ind. Sci.,
Univ. of Tokyo, 84, 19593.
[3]
RICE, S.O., "Mathmatical theory of random noise", Bell Syst. Tech. J., 23, 1944, 24,
194S.
[4]* SATO, H., "A study on aseismic design of machine structure", Rep. Inst. Ind. Sci., Univ.
of Tokyo, 151, 196511.
[5]
SATO, H., "Response of structure system to a model earthquake motion with two ground pre
dominant periods", J. Inst. Ind. Sci., Univ. of Tokyo, 2111, 196911.
[6]* SATO, H., "Response of nonlinear structure system to a model eathquake motion with two
ground predominant period", Proc. JSME, 70017, 197010.
[7]
KANAI, K., "Semiempirical formula for the seismic characteristic of ground", Bull. ERI,
Univ. of Tokyo, 35, 19571.
[8] NEWTON Jr., G.C., et al.. Analytical design of linear feedback controls, John Wiley, 1957.
[9]
SAWARAGI, Y., "A survey on statistical study of nonlinear control systems". Trans Fac
of Eng., Univ. of Kyoto, 14, 19589.
[10]* SATO, H., "A study on confidence limits of characteristics of response spectrum". Proc.
3rd Japan Earthq. Eng. Symp., 197011.
[11]* KANAI, K. and OTSUKI, T., Aseismic design, Corona, 1962.
[12]
SHIBATA, H. and SHIMIZU, N., "Some considerations on response analysis of piping systems
to multiinputs
Response estimation and its fluctuation", Proc, 3rd Japan Earthq.
Eng. Symp., 197011.
In JaDanese
164
0.6
0.2
0.8
0.4
1.2
/,
, 0.015
4.0'
ft =3.0
h,,'0 4
3.0
Tg,= 0.20.ec
hg; 0.3
Tg.0.5jec
hb0.07
2.0
1.0
0.2
0.4
0.6
0.8
1.0
T'b sec
1.2
cv <9
,0.013
0
4.0
Ti ' 3 0
ng: 0
3.0
/
/'
T a ,'0.25sec
h a:
\\
0.25
0 51)
0.11
IO
Tg/I.25ltc
"
h b '0.07
2.0
*"_"2^
^,
^"^
~"^"~"" ^" _'~"
1.0
Fig.l
0.2
0.4
0.6
1.0
1.2
1.4
1.6
, sec
(a)
(b)
Response spectrum of a c c e l e r a t i o n a m p l i f i c a t i o n
factor
165
0.01 cm/gol
5.0
X*
g/,' 015
'
10
hg, 0 4
3.0
^'
Tg, 0.20 sec
hg,o.3
T,,. 0.5sec
h'0.07
yS
/A'S.'
fy/s^s^
/ # X ^"^
gs^
1.0
/^
02
^ V _^*'*
y'sf^s*'
s*
S,K''s'"'.''
O 1.2
0.4
0.8
0.4
0.6
o.a
1.2 Ttsec
,/
0.01 cm/gol
00
.0.015
.3.0
8 0
hg,- 0.3
T,,.|.25ic
4.0
0 25
0.5
075
1.0
,
.*'*'
h,,-04
., 25e
60
'
'
///....""
h.'0.O7
.-
"
/ / / ^ .--"-"^
. .
______
^â^""
2.0
Fit
(a)
(b)
Response spectrum of d i s p l a c e m e n t
166
,/o_
,; 0.2 sec
1,, o.5sec
0.8~.
4.0
._.
,,0.2sec
Tg,= 0.4sec
=0.75~.
EL
CENTRO
3.0
2.0
1.0
0, 0.015 (,= 3.0

h,0.4
0.2
h,_0.3
0.4
h_=0.07
0.6
0.8
1.0
T65ec
Fig. 3
Comparison of the acceleration response spectra by the analysis with those by actual
earthquake records
/ /
/ / ^
V0.0I5
A'3.0
h,; 0.4
h,; 0.3
ht'0.C7
// ;
/ ;f
/' J'*'
/ /
0.2
;/\
L CtNTRO
+
TAFT
O VO.Ztic
T|_*0.5ttc
-,^
O-T.,-0.2MC
T,,*<Mttc
'0.75-1.0
0.4
0.6
0.8
1.0 TeSec
Fig.4
Comparison of the displacement response spectra by the analysis with those by actual
earthquake records
167
=1.0
^ =0.015
A =3.0
T=0.2s
h=0.A
Tn=0.5s
hj=0.3
A NA LYSIS
O
0
Fig.5
2.0
4.0
EL CENTRO
6.0
8.0 Hz
Comparison of power spectrum of the theoretical model with that of El Centro
'illuminili,
Fig.6
Scheme model of the nonlinear system
- 168
0 /
/
x
A
Fig.7
Scheme description of c h a r a c t e r i s t i c of the e l a s t o - p l a s t i c deformatie
+Y
Fig.8
s
^biy*-
______
Block diagram of the nonlinear vibration system
16!)
4 " f e w ^ 2 5
VO.3
T=0.5s
0.3
o 0.3
oo
0.6
0.2
0.1 f
-/
/ s f >
\J^^
j t ^ ' / ^ S j t '
Tb S
0.0
00
Fig.9
X P
1.0
2.0
3.0
4.0 5.0
Comparison of the displacement response spectra as for linear and nonlinear system
cm
^cm/s'
=0.9
=a5S
* * * \ ^ >
0.3
0.2
0.1
0.0
fi /~~"~"*
p z v ^
jV_/
0.3
0.6
0.0
Fig.10
o
a
'
2.0
o
s
1.0
2.0
3.0
UQ 5.0
Comparison of the velocity response spectra as for linear and nonlinear system
- 170
80.0
a CENTRO
--
LINEAR
- NONLINEAR
(0 3 g)
h.=0 07
40.0
'/,.--''
*0,9 . i=02s , v=ass
- - fl=oe, - - fl.1.5
~ 4 - . = 0 . l,,=0.2s , --
- =0 . V=0.5s . fl.oo
20.0
0.0
Fig.11
0.0
0.4
0.2
0.6
Tes
1.0
0.8
Comparison of the velocity response spectrum of the nonlinear system by the analysis
with that by actual earthquake record
Linear
0 ~~
Fig.12
10
Hi
A n a l y t i c a l power s p e c t r a of r e s p o n s e of the s t r u c t u r e model
171
iVlO
8
6
4
2
0O
Fig.13
0.2
0.4
0.6
Tg
0.8
sec
A relation between the ground predominant period in earthquake observed at a place

and its occurrence frequency (after Kanai fll])
S*~
IIA)
.. ,
,/g 0 ()3>
- '
L<rtsy '"
40
*"^*
"*^ ^^"
~'_^^*
<^* ____ ~ * ^*
^ *..
3.0
"
Tf d 1.0 SflC
TM'LOsfC
2.0
'
. "*
'"'
,0.4
h. 0.07
Q.\^
;"
~~~
j ^
^ ff.sec
OA^^
0.4
i*sec
0.2^
to
0.__^
^-
0.2
Fig.14
0.3
Mean and 30 width of the acceleration amplification factor (single ground predomi
nant period)
172
V 02 VO. 5
s
0.2
0.4
0.6
hg0.4 h"007,
(1
0.5
Fig.15
0.8
1.0
1.2
1.4
(R/T 0 2
(4)+3<
10
ly 0.4 ly 0.3
0.9 10.07
E (A)
EAh3t>E(A)3.
1.5
()<
2.0
25
VTJ
(a)
(b)
Mean and 3 width along a response spectrum of acceleration amplification factor
- 173 DISCUSSION
__
Ch. CHEN, U.S.A.
In one of the slides the relationship between ground predominant periods and the
occurence frequency is shown. Does this relationship apply to all ground conditions, i r r e
spective of competent rock or soft soil ?
H. SATO, Japan
No, it is a datum measured at an observatory. I wish I could have this kind of

data for various conditions from the engineer's view point.
_
G. SCHNEIDER, Germany
What is your opinion about Prof. Kanai' method to find out natural or predominant
periods of microseismic signals depending on geologic environment and to use these results
for predicting predominant periods in earthquake signals ? Since most reactors are or
will be installed on more or less thick layers of sediments this would be a good method to
determine the amplification of low impedence surface layers.
H. SATO, Japan
I am intending to find out the ground predominant period and also the probabilistic
characteristic appearing the period. It would be helpful for us to estimate the response of
structure by making use of the results of the statistical approach.
. WOLFEL, Germany
Is there any possibility to use the response spectrum of a nonlinear one degree
of freedom system to calculate nonlinear systems of several degrees of freedom ?

H. SATO, Japan
Yee, there is. It may become more complicated and we may have to spend more
computer time. It is possible to extend this method for the case of multi-degree-of-freedom
systems.
U ESTEVA, Mexico
The problem of taking into account not only two, but many more ground periods
has been studied previously by assuming continuous systems to represent the ground layers.
Some of these studies have limited themselves to obtain amplification functions for the Fourier
spectra or for the power spectra of the motions, under the assumption of stationarity, but
174
others have introduced c o r r e c t i o n s in o r d e r to account for finite duration and non-stationarity.

For the case you studied, what was the advantage of working with a t w o - m a s s system ?
How did you obtain the p r o p e r t i e s of the t w o - m a s s s y s t e m from those of the continuous s y s t e m ? How do you obtain amplification functions for response s p e c t r a r a t h e r than for F o u r i e r
spectra ?
H. SATO, Japan
I can take the advantage by this rather simple analysis to make the effect of
another ground predominant period to the response s p e c t r u m obvious. Once we know the
effect, we can compose the multi-degree-of-freedom s y s t e m as far as the number of degrees
is considered. I made use of the probability density function of extremes by S. O. RICE. The
maximum is r e p r e s e n t e d by the point where the probability density function becomes small
enough. This is made for the random process corresponding to the earthquake motion and
the response to it. The ratios for both estimations a r e taken as the amplification factor.
K 3/3
THE INFLUENCE OF SEISMIC PULSE TIME

ON STRUCTURE-FOUNDATION INTERACTION
R.J. SCAVUZZO,
Department of Mechanical Engineering,
Rensselaer Polytechnic Institute, Hartford Graduate Center, East Windsor Hill,
Connecticut,
R.R. LITTLE,
Department of Mechanical Engineering,
The University of Toledo, Toledo, Ohio, U.S.A.
ABSTRACT
The influence of the duration of seismic ground motions on inertia forces

of a power plant is investigated considering ground-structure interaction
effects. This study is based on the ground accelerations measured during the
1940 El Centro earthquake and the 1952 Taft earthquake. Results of the
analysis indicate that ground motions from five to ten seconds in length are
sufficient to evaluate seismic forces on heavy structural components.
1.
INTRODUCTION
The response motion of stiff heavy structures to seismic excitation is

changed from free-field earthquake motions because of structural inertia
forces. The magnitude of this change depends upon the ground stiffness, the
modal characteristics of the structure, and the relationship between the
natural frequencies of the structure and frequency characteristics of the
seismic input wave motion (see Scavuzzo, Raftopoulos and Bailey [1-4]).
It is the main purpose of this study to determine the influence of the
duration of the seismic ground motion on the shock loads in a nuclear power
plant. In addition, ground-structure interaction effects are also studied
for two different ground motions. Significant ground motions often last for
sixty seconds. However, practical limitations of the computer such as core
storage and cost make ground-structure interaction analyses based on seismic
motion of this length difficult using either finite element analyses such as
those done by Chiapetta or Isenberg [5], [6], or the analytical solutions
referenced above. As a result, this study was undertaken in order to
investigate the effects of time on seismic shock loads of large building
structures.
The two earthquake motions used for this study are the north-south
component of the 1940 El Centro earthquake and the N21E component of the 1952
Taft earthquake. Parametric studies are based on the first fifteen seconds
of the El Centro earthquake and the first thirty seconds of the Taft earthquake .
176
The structural model used in this study is a simplified

of a nuclear power plant previously employed.
representation
This dynamic model consists
of three masses, a base mass, a containment vessel mass and an internal

structure mass
(Figure 1 ) .
Fixedbase frequencies of the containment vessel
mass and internal structure mass are approximately 4.0 cps and 5.0 cps,
respectively.
The natural frequency of the containment vessel is close to
measured values of the fundamental mode of the
EGCR by Matthiesen and Smith,
[7].
NOMENCLATURE
a
Dilatation
Area of the structure base
() wave velocity
Shear
Half the base width
Young's modulus
F (t)
Lateral force at the base of a structure
(S) wave velocity
f (t)
Surface shear stress when
Base mass
|x| < c
Effective mass of the j ^ n mode
j
u(x,y,t)
Lateral displacement in the halfspace

(xdirection)
u(t)
Lateral displacement of the center of the base
up(t)
Freefield lateral displacement at the center

of the base
v(x,y,t)
Vertical displacement in the halfspace

(ydirection) .
Natural circular frequency of the j

VE
Lame's constant
Shear modulus y
Poisson's Ratio
Ground density
Rayleigh wave velocity

3
mode
(1+v) (l2v)
(1+v)
THEORY
This study is based on the analysis presented in References
[3].
[2] and
In this formulation of the soilstructure interaction phenomenon,
the ground is represented by a twodimensional elastic half space.
The
governing equations for dynamic elasticity problems in two dimensions are:
(+) (
3x3y
) + uV2u
(1)
___j
3t2
(+) (
3x3y
JLV) +
3y
72
177
Making use of the coordinate system introduced in Figure 2, the boundary

conditions used in this solution are:
o 0
,y0
ff(t)
|x|<c,
|x|c.
(2)

=
=0
y
xy
y =
By employing Laplace and Fourier transforms the solution for the displace
ment at the origin caused by a shear force which varies arbitrarily with
time in the interval c < < c is obtained (eq. (3)). In order to simplify
the inversion of these transforms Poisson's ratio is made equal to 1/4.
For this case, a = /3 b where a and b are the and S wave velocities,
respectively.
u(0,0,t)
\
2ttc i
>o
f(0
Im g ( - ^ )
(3)
where Im g ( ) is the imaginary part of g ( ) .
The function g(T) is defined
by eq. (4/.
5(T)
/T
(4)
2 2
cicj - T ) + T 2 / j - T 2 / l - T 2 1
The imaginary p a r t of g(T) can be evaluated a s :

< <
/3
Im g(T) =
3T(1
_ 2)/2
< < 1
2(T 2
-)(.'
32- 1
-X.T2
(2 - )
+ 2 / 2 / 2 1
>
2(2 | ) (2 -2=) (2
-%)
(5)
178
It should be noted that for t < , there is no contribution from the
second integral in eq. (3). In the interval < t < the function,
bt
c
a
b
Im g ( ) has one form and for t > c- the term has a second form. A singularity
occurs when t = where v is the Rayleigh wave velocity. However, the
second integral of eq. (3) is bounded in the Cauchy sense.
Interaction Equations
For an N-mass structure subjected to an arbitrary lateral foundation
acceleration, the horizontal loads at the foundation can be expressed in
the following form using the notation of O'Hara and Cunniff [8],
5" 8
F(t)
(l) sin u.(t-t)dT + M (t)

J
o
(6)
where the positive direction of F(t) is assumed to be the same as the shear
stress f(t) which acts on the surface in the interval - c < < c, M. are
the effective modal masses and u(t) is the lateral displacement at the
center of the foundation located at the origin. The lateral displacement,
u(t), equals the sum of the displacement due to the inertial shear forces
(given by eq. (3)) and the free-field displacement, u (t).
u(t)
(OdT - Ji f(c
2 cu
Ira g(p) ddi + u (t)
(7)
" Jo Jo
Jo
Fit)
By assuming that f(t) = - and by taking the first time derivative of
eq. (7), f(t) can be eliminated by substitution from eq. (6) to obtain the
following interaction equation in terms of the acceleration at the center
of the foundation.
1r V j JT
2TTCM
()
sin
(tt)dt
(0
Im g ^ 1 )
Jo -
+
^[uUOImt
EM IVO
aln
di
. ( t - i - O d ' . d t
(8)
179 -
In eq. (8) the properties of the structure are defined in terms of the
base mass M , the effective modal masses M., the circular natural frequencies,
,, the base half width c. The properties of the ground are specified by
the shear modulus, u , and the shear wave velocity b. Because Poisson's
ratio has been made equal to 1/4 and a value of 100 lb/ft was used for the
weight density of the soil in all calculations, the stiffness of the ground
is completely specified by the shear wave velocity. Values of 500 ft/sec.,
1000 ft/sec, 2000-ft/sec. and 4000 ft/sec. were employed in these parametric
studies.
4.
DISCUSSION
The solutions to twenty-eight problems were studied in the evaluation

of the effects of time on soil-structure interaction. The input parameters
specified in these problems are tabulated on Table I. In addition, the
acceleration response ratio, which is defined as the acceleration response
determined from the foundation acceleration, (t), (eq. (8)), divided by
the acceleration response of the input acceleration, (t), at the fixedbase frequencies of the system, are also listed on Table I. F or Cases 1
through 24 there are two modes of vibration ; there is one mode of vibration
for Cases 25 through 28. In these latter four problems, the effects of the
internal structure dynamic mass were not considered so that thirty seconds of
input motion could be studied.
Two different earthquake motions were used as input motions to eq. (8).
In Cases 1 through 12, the first fifteen seconds of the 1940 El Centro,
California, earthquake measured in the - direction was employed. The
first fifteen seconds of the 1952 Taft, California, earthquake recorded in
the N21E direction was used as input in Cases 13 through 24; the first
thirty seconds of the Taft earthquake was employed in the last four cases.
Plots of these input accelerations are presented on Figures 3, 4, and 5.
Typical output accelerations, corresponding to the grafted inputs are plotted
on Figures 6, 7 and 8 for a soil shear wave velocity of 10DO ft/sec.
Reductions of the peaks in the input acceleration caused by structurefoundation interaction can be observed on these graphs.
Assuming no structural damping, the acceleration response spectrum in
units of g's is determined from the integral
S(u>,t) =
() sin u>(t-r)di
(9)
180
where is either the input acceleration, (t), or the foundation acceler
ation, (t). The spectrum response is the maximum value of S(UJ, t) , for any
time, t. Spectrum responses of the two mode problems (C ases 1 to 24) are
presented on Tables II and III. The time, t, at which the integral of eq.
(9) is a maximum is also listed on these tables. By comparing the results of
problems which are similar except for the length of the input motion, the
effect of time on foundationstructure interaction can be evaluated.
Results of the studies on the El Centro input are the easiest to evaluate.
Peak accelerations of this motion of 0.25 g's or more occur between 1.7
seconds and 4.89 seconds. After five seconds, the magnitude of input accel
eration motion decreases. Peak values of the spectrum response integral
based on this input motion varies from 4.95 seconds to 10.42 seconds and
depends upon the specified response frequency, . On Figures 9 and 10, the
integral, S(ii),t), is plotted as a function of time for the El Centro earth
quake motion. The magnitude of S(ui,t) is large for the fifteen seconds
considered because there is no damping associated with the integral of eq.
(9). However, the peak acceleration responses of the calculated foundation
motion occurs between 2.2 and 3.24 seconds. On Figures 11 and 12, S(io,t)
is plotted using the calculated foundation motion, for the two fixedbase
frequencies and a ground stiffness specified by a shear wave velocity of
1000 ft/sec. In addition to the reduced magnitude of the response acceler
ation, damping of the motion can be observed.
Energy in each mode of vibration in the structure is lost by radiation
of waves into the ground. As a result of this energy loss in each mode, the
integral S(ui,t) is damped at the modal frequencies and, therefore, will not
increase in magnitude with time. An estimate of the equivalent structure
damping associated with this energy loss is between 5% and 10% of the critical
value. As a result, peak accelerations and, therefore seismic forces of the
structure occur during the first five seconds of input motion which is the
violent portion of the earthquake motion.
The input motion of the Taft earthquake shown on Figure 4 is much more
uniform in acceleration magnitude over the first fifteen seconds than the
El Centro earthquake. As a result, the calculated foundation accelerations
plotted on Figure 7 are more constant in magnitude even though the peaks of
the input motion are reduced. As seen in Table I, the acceleration response
ratio tends to be reduced as the duration of the motion is increased. This
change occurs because the spectrum response of the freefield motion
increases more with time than the spectrum response of the foundation motion.
The response integral S(io,t) is plotted on Figures 13 and 14 at the first
mode frequency for the input motion and output motion, respectively. As
seen on these graphs and Table III, the maximum values of the spectrum
response of both the input and output motions occur at approximately ten
seconds for most cases. The only significant deviation occurs in the second
mode with the stiffest soil considered (b = 4000 ft/sec). Again there is no
- 181 tendency for the integral of the foundation motion, S(u>,t), to increase with
time because of radiation damping. As a result, if the portion of the Taft
earthquake between three and eleven seconds are used as input, the maximum
seismic forces acting on the structure can be evaluated considering inter
action effects.
Using the one-mode two-mass system described above, input motions from
five seconds to thirty seconds in duration are considered (Cases 25 to 28).
As seen on Table I, the acceleration response ratio is unaltered after the
first ten seconds of motion are considered. The peak response occurs at
9.9 seconds which is similar to the two-mode model. The increase in the
acceleration response ratio associated with these cases is caused by the
reduction in total weight of the idealized structure.
In the preceding discussion, the spectrum response at the fixed-base
frequencies of the idealized structure was considered. These frequencies
were considered because structure inertia loads are determined at these
values. However, the response of light weight structures without structural
damping attached to the base mass, , can.be determined from the spectrum
curve at other frequencies. By examining the changes in the spectrum
response curves with time, it was found that at most frequencies the curve did
not vary significantly with time. However, at a few values there were large
variations with time (a typical response curve is presented on Figure 15).
Seismic loads on light weight subcomponents have not been thoroughly investi
gated in this study. However, preliminary results indicate that longer
duration motions are needed to determine maximum seismic forces than for the
heavy-structural components.
5.
CONCLUSIONS
The results of this study show that ground structure-interaction effects

are significant for all soil stiffnesses studied. Furthermore, the ratio of
the output spectrum response to the frcc-ficld input spectrum response at the
fixed-base frequency varied from 0.18 to 0.19 for the softest soil studied
and varied from 0.65 to 1.1 for the hardest soil studied. For the stiffest
soil, this reduction ratio depended upon the input motion.
By comparing the response spectra for 5 sec. , 10 sec. , and 15 sec. of
motion, it was concluded that interaction effects of nuclear power plants can
be accurately evaluated using earthquake motions from 5 sec. to 8 sec. in
duration. Peak values of the spectrum response occurred during the most
violent portion of the earthquake motion. Damping of the structure associated
with radiation of stress waves into the foundation prevented amplification of
the response spectrum with time at the fixed-base frequencies of the structure.
Light weight equipment structures attached to the base of the structure
were not thoroughly investigated in this study. However, preliminary results
indicate that longer duration of seismic input motions are needed to determine
182
maximum inertia forces than for the heavy structural components.
frequencies, the acceleration responses increased with time.
6.
At some
AC KNOWLEDGEMENT
The authors are grateful to Dr. Paul C. Jennings of the California

Institute of Technology for providing the earthquake motions used as input in
this study and to the United States Atomic Energy Commission for supporting
the work presented in this paper through USAEC Contract AT(401)3822.
BIBLIOGRAPHY
[1]
S
C AVUZZO, R. J., BAILEY, J. L. AND RAFTOPOULOS, D.D. "Lateral Structure
Interaction with Seismic Waves", Journal of Applied Mechanics, Vol. 38,
No. 1, p. 125, 1971
[2]
S
C AVUZZO, R. J., BAILEY, J. L. AND RAFTOPOULOS, D. D. Lateral Structure
Foundation Interaction of Nuclear Power Plants During Earthquake
Loading, USAEC Contract No. AT(401)3822, Report 2, Research Founda
tion, The University of Toledo, August, 1969.
[3]
S
C AVUZZO, R. J., RAFTOPOULOS, D. D. AND BAILEY, J. L., Lateral
StructureFoundation Interaction of Nuclear Power Plants with Large
Base Masses. USAEC Contract No. AT(401)3822, Research Foundation,
The University of Toledo, Report No. 3, September, 1969.
[4]
S
C AVUZZO, R. J.., StructureFoundation Interaction of Nuclear Power
Plants Phase I Final Report, ORO38227, December, 1970
[5]
C HIAPETTA, R.,
Ground Motion.
April, 1970.
Effect of SoilStructure Interaction on the Seismic

ORO38224 (IITRI J6157) , IIT Research Institute,
[6]
ISENBERG, J., Interaction Between Soil and Nuclear Reactor Foundations

During Earthquakes. ORO38225 (AJA R69151200) September, 1970*
[7]
MATTHIESEN, R. B. AND SMITH, C. B. Forced Vibration Tests of the

Experimental GasCooled Reactor (EGC R) UC LA Report 694, August, 1969
[8]
O'HARA, G. J. AND CUNNIFF, P. F.

NRL Report 6002, November, 196 3.
Elements of Normal Mode Theory.
*AbabianJacobsen Associates, Los Angeles, California
TABLE I
TABULATION OF PROBLEMS STUDIED
Case
No.
Structure Properties
M
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
2.4
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.475
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
.310
0
0
0
0
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.06
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
15400
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
500
500
500
1000
1000
1000
2000
2000
2000
4000
4000
4000
500
500
500
1000
1000
1000
2000
2000
2000
4000
4000
4000
1000
1000
1000
1000
1940
1940
1940
1940
1940
1940
1940
1940
1940
1940
1940
1940
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
1952
NS EL Centro
NS EL Centro
NS EL Centro
NS EL Centro
NS E L Centro
NS EL Centro
NS EL Centro
NS EL Centro
NS EL Centro
NS EL Centro
NS EL Centro
NS EL Centro
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
N21E Taft
Duration of
Seismic Input
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
First
A base area of
10
Poisson's
containmentvessel mass, l b / s e c / f t .
10
internalstructure
10
2
,
Free Field Earthquake

Input, ' (t)
5
10
15
5
10
15
5
10
15
5
10
15
5
10
15
5
10
15
5
10
15
5
10
15
5
10
15
30
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
sec.
ccc.
sec.
Ac ce le ration Response
Ratio*
.201,
.188,
.188,
.339,
.317,
.317,
.588,
.550,
.550,
1.091,
1.020,
1.020,
.279,
.137,
.177,
.406,
.260,
.260,
.651
.404,
.404
.949,
.707,
.707,
.200
.182
.169
.274
.249
.231
.406
.369
.342
.794
.723
.670
.125
.126
.156
.200
.188
.251
.354
.305
.364
.691
.534
.648
495
288
288
288
base mass l b s e c / f t .
"1
Ground Proper ties
mass, l b s e c / f t .
fixedbase
containmentvessel
frequency,
cps
fixedbase
internalstructure
frequency,
cps
structureft.
ratio
3
ground density, lb/ft.
b soil shear wave velocity, ft/sec.
Ratio of base motion acceleration spectrum response to freefield acceleration spectrum response at the
fixedbase structure freqquencies (f.,f_).
184
TABLE II
COMPARISON OF ACCELERATION SPEC TRA FOR THE
TWO MODE MODEL SUBJECT TO
THE EL CENTRO EARTHQUAKE MOTION
DURATION
'n
cps
SOIL CONDITION
OF
5 SEC
INPUT
MOTION
10 S E C
15 S E C
FreeField
b = 900
b = 1000
b = 2000
b = 4000
Motion
ft/ s e c
ft/ i c e
ft/ s e c
ft/ s e c
4. 06
4.06
4.06
4. 06
4.06
1.
0.
0.
0.
1.
54
31
52
90
68
@
@
@
@
@
5.00
2.20
2.19
2 . 58
2. 56
.ec
ec
ec
ec
ec
1.65
0.31
0. 52
0. 90
1. 68
@
@
@
@
@
9. 67
2.20
2 . 19
2 . 58
2. 56
ec
ec
ec
ec
aec
1.65
0. 31
0.52
0. 90
1.68
*?
@
@
FreeField
b = 500
b = 1000
b = 2000
b = 4000
Motion
ft/ s e c
ft/ aec
ft/ s e c
ft/ e c
4. 95
4 . 95
4.95
4. 95
4. 95
1. 86
0. 37
0. 51
0. 76
1.48
(8
@
@
@
@
4. 95
2.72
2.72
3.24
3.24
ec
aec
ec
ec
aec
2.05
0. 37
0. 51
0. 76
1.48
@ 6.98
@ 2. 72
@ 2. 72
(S> 3. 24
@ 3. 24
ec
ec
ec
aec
aec
2.21
0. 37
0. 51
0. 76
1.48
t? 1 0 . 4 2 a r c
it 2 . 72 a e c
@ 2 . 72 a e c
@ 3 . 2 4 aec
@ 3 . 2 4 aec
9.67
2.20
2 . 19
2.
. 56
aec
aec
aec
^
ec
TABLE III
COMPARISON OF ACCELERATION SPEC TRA FOR THE
TWO MODE MODEL SUBJECT TO THE TAFT
EARTHQUAKE MOTION
DURATION
OF
epa
FreeField
b = 500
b = 1000
b = 2000
b = 4000
Motion
ft/aec
ft/aec
ft/eec
ft/eec
4.
4.
4.
4.
4.
06
06
06
06
06
0.
0.
0.
0.
0.
45
13
18
29
43
FreeField
b = 500
b 1000
b = 2000
b = 4000
Motion
ft/aec
ft/aec
ft/aec
ft/eec
4.
4.
4.
4.
4.
95
95
95
95
95
0.
0.
0.
0.
0.
86 <S' 4. 96Bee
1 I (S 3. 74 Bee
17 @ 3. 74 flee
30 @ 4 . 2 7 s e c
59 <a 4. 97 flee
@ 3. 78 aec
@ 3. 81 ec
<& 3. 81 e c
@ 3. 81 ec
@ 3. 79 s e c
INPUT
MOTION
10 S E C
5 SEC
DESCRIPTION
SOIL
li
SEC
0.92
0. 13
0. 24
0.37
0. 65
<S> 9. 90
@ 3. 81
(<i 9. 91
@ 9.90
' 6. 66
ec
ec
ec
ec
ec
0.92
0. 16
0.24
0. 37
0.65
(3 9. 9 0 e c
@ 1 4 . 2 8 ec
@ 9. 91 e c
@ *. 90 e c
@ 6 . 6 6 ec
1.
0.
0.
0.
0.
<8 6. 67
@ 8. 49
(?> 8. 49
(q> 9. 11
(S 6. 70
ec
ec
ec
aec
aec
1. 16
0. 18
0. 29
0.42
0. 75
@
@
@
Iff
@
16
15
22
35
62
6.67
10. 37
10. 37
10. 36
14. 48
aec
ace
ec
aec
aec
- 185 -
Containment Vessel
Internal Structure
Base
(a) Nuclear power plant

1
m - 475,000 lb-sec 2
1
ft
k - 0.31 IO9 l b / f t
m - 310,000 lb-sec 2
2
~ft
k - 0.30 IO9 l b / f t
2
H = 2,400,000 lb-sec 2
0
ft
Base Area = 15,400 ft 2
(b) Two-mode nuclear power plant model
Fig. 1.
The n u c l e a r power p l a n t shown i n F i g u r e l a h a s b e e n i d e a l i z e d
a s a two-mode t h r e e - m a s s model a s i n d i c a t e d i n F i g u r e l b . The m a s s e s
II , m. , and m. r e p r e s e n t t h e b a s e m a s s , c o n t a i n m e n t v e s s e l m a s s , and
i n t e r n a l s t r u c t u r e mass, r e s p e c t i v e l y .
186
Coordinate system used in the solution to the Lamb problem.

Fig.
A shear stress, f(t), which acts between -c varies arbitrarily with
time. The remaining portion of the surface is stress free. Normal
stresses over the entire surface are zero.
Fig. 3
North-south ground acceleration record of
the El Centro, California earthquake of May 18, 1940.
187
, , lil llllil 1
Aluiu iP ||mrt 1 h iffl Ili |J
iff 11 ir p i l m
li I 'Il 1
Fig. 4
N21E component of the ground acceleration
recorded at Taft, California July 21, 1952.
dJ/lli
Triff
11 ii
IIUAIILIAikai!
J 1 1 Ui I ila* Wjf
Bf
| PI
1! 1 1\f
'
'
rf
TIKT. SECONDS
Fig. 5
The first thirty seconds of the N21E component of the
ground acceleration recorded at Taft, California July 21, 1952.
188
Fig. 6
The computed output acceleration response (t) versus time for
the two-mode nuclear power plant model with a ground having the properties
of b=1000 ft/sec, p=100 lb/ft and v=l/4, subjected to the El Centro N-S
input acceleration (t) shown in Fig. 3
Fig. 7
The computed output acceleration ii(t) versus time for the twomode nuclear power plant model with a ground having the properties of
b=1000 ft/sec, p=100 lb/ft and v=l/4, subjected to the N21E component of
the Taft input acceleration (t) shown in Fig. 4.
189
"'
Ili
.1
kUiLJJUJIL rti'llljM/
iE
_*HWIfi |1 Imff flyjijv
n.ûti/lIMJl
fi 1'
''
SECONDS
Fig. 8 The computed output acceleration response (t) versus time for the
one-mode two-mass nuclear power plant model (using the base mass of
2
2
2,400,000 lb-sec /ft and the containment vessel mass of 475,000 lb-sec /ft)
J
with a ground having the properties of b=1000 ft/sec, p=100 lb/ft and
v=l/4, subjected to the N21E component of the Taft input acceleration
(t) shown in Fig. 5.
15
lb
TIME - SKCOWDS
Fig. 9
The absolute value of the spectrum response integral,
(eq. (9)), is plotted as a function of time for the El Centro
ground motion and a specified frequency of 4.06 cps.
190 -
ui
>,
SS
\
14
Ji
>
siika
lI
10
1fr
' il
Piff
.UI
6P
>
>
1*
Ib
TUO. itCOKOs
Fig. 10
(eq. (9)), is plotted as a function of time for the El Centro
ground motion and a specified frequency of 4.95 cps.
TIHEStCnHDS
Fig. 11
(eq. (9)), is plotted as a function of time for the calculated
foundation motion presented on Fig. 6 and a specified frequency
of 4.06 cps.
191
J! IP PI IIII '1
Ij A fili H
? 'WIM'
il
I
Fig. 12
The absolute value of the spectrum response integral, (eq. (9)),
is plotted as a function of time for the calculated foundation motion
graphed on Fig. 6 and a specified frequency of 4.95 cps.
12
1}
It
i>
it
TIHE StCUHDS
Fig. 13
The absolute value of the spectrum integral, (eq. (9)), is
plotted as a function of time for the Taft earthquake and a specified
frequency of 4.06 cps.
192
li
-AV ,
i!Il
ill
M/fl iilt IIIi
uil
io
i:
,u
TIrC-SECONDS
Fig. 14
The absolute value of the spectrum response integral, (eq. {9)
is plotted as a function of time for the calculated foundation motion
graphed on Fig. 7 and a specified frequency of 4.06 cps.
h
M
lift h
IKFUT ACC.
USPOKSE
SPECnUM
FLW\i
iv
Mwv
'r
\j V^\y
mQUENCT, CPS
Fig. 15
The input and output acceleration response spectrums for the
two-mode model based on the first fifteen seconds of the Taft input
acceleration (Fig. 4) and the first fifteen seconds of the output
acceleration (Fig. 7 ) . The shear wave velocity was 1000 ft/sec and the
fixed-base frequencies were 4.06 cps and 4.9 5 cps.
193 DISCUSSION
H. SHIBATA, Japan
1. We generally consider the effect of horizontal motion. But the effect of rotational motion
of ground surface should be stronger than that of the horizontal motion.
2. The notches in Fig. 15 might be very significant for the design. But to estimate the eigenfrequencies in an accuracy within 10% is very difficult, so the view point of the margin of
safety, the effect of peaks should be also considered in average. Then applying the effect of
such reduction for the design may be limited. How do you consider the case of applying this
effect to the design ?
yfi^
1. In the analysis which is presented, rotational motion was not considered. However, in
Ref. (4) the effect of rotation was considered, ground motions w e r e based on the finite element
analysis of Isenberg (Ref. (6)). These results showed that a c c e l e r a t i o n s associated with rock
ing a r e of the same o r d e r of magnitude as the l a t e r a l motions. F u r t h e r m o r e , accelerations
of the containment vessel m a s s caused by rotation tended to be out-of-phase with the l a t e r a l
motion making the final value very small.
2. Reductions in the spectrum response curve caused by interaction effects tend to shift with
the fixed base frequency of the s t r u c t u r e . Thus, the exact value of the s t r u c t u r e frequency
is not as important as it first appears. P a r a m e t r i c studies a r e presented in Ref. (3).
N.M. NEWMARK, U.S.A.

Would your conclusion regarding interaction be changed if the actual net m a s s
of the s t r u c t u r e , taking into account the weight of the excavation removed, w e r e considered ?
R. J. SCAVUZZO, U. S. A.
My conclusion concerning interaction effects would not be altered if this weight
were considered. This statement is based on two studies.
F i r s t , the studies reported in Refa(2) and (3) were made neglecting the large base m a s s of
the s t r u c t u r e . Interaction effects were s i m i l a r to those obtained with the base m a s s (Ref. (3)).
Also, finite element studies were conducted in which the containment s t r u c t u r e was embedded
in the soil. For this case interaction effects were still found to be significant but, of c o u r s e ,
altered from surface studies.
T. II. LEE, U. S. .
Would your conclusion on the duration of earthquake motion be changed if the
most violent portion of the ground motion occurs at 10 seconds later ?
- 194
R.J.
SCAVUZZO,
U.S.A.
No my conclusion would not change. The reason for this statement is that the
first portion of the earthquake which does not have high accelerations would not affect peak
structural loads significantly which will occur during the violent portion of the ground motion.
Thus, the first 10 seconds of motion could be neglected.
Q
1.
M. B E N D E R , U. S. A.
Would you suggest quantitative criteria for differentiating between light and heavy mass
structures for the purpose of analysis and suggest the most appropriate approach to give a
conservative design for intermediate mass elements of the structure ?
2.
With respect to uncertainties in the soil stiffness characteristics and variabilities in
structural response attributable to unknown physical properties, how would you use your
analysis methods as a design tool ?
^
1.
R . J . SCAVUZZO,
U.S.A.
Parametric studies have not been conducted that would lead to an accurate answer to
your question. However, as a guess I would suggest that a mass could be considered light
if it weighted less than 10% of the heavy mass.
2.
I would suggest that parametric curves be developed for the system under consideration.
Both the soil stiffness and structural properties could be varied.
A. HADJIAN, U. S. A.
Since the response spectrum is calculated at specific values of oscillator fre-
quency, the dips and peaks of response spectra should not be looked as absolute numbers
since any other choice of oscillator frequency would change the dips and peaks dramatically.
R. J. SCAVUZZO, U. S. A.
No,
this is not the case. The reason for this statement is that the dip in the
spectrum is caused by the vibration of the structural mass. Thus, if the fixed-base frequency of the structure were increased by 1/2 cps, the dip in the spectrum response curve would
also increase by 1/2 cps. Results of a parametric study based on the El Centro earthquake
are presented in Ref. (3).
K 3/4
ANALYSIS OF SOIL-STRUCTURE INTERACTION EFFECTS

UNDER SEISMIC EXCITATION
C.J. COSTANTINO,
School of Engineering,
The City College of the City University of New York, New York, U.S.A.
ABSTRACT
This report describes a numerical technique to treat the complete
dynamic soil-structure interaction problem. The structure embedded within
the free-field soil system is represented by its rigid body and elastic freefree modes, while the soil is treated by the finite element method including
nonlinear material properties. The application of the developed computer
program, termed the SLAM Code for identification, to a particular problem
incorporates the ability for treating potential separation and sliding between the soil and the structure. The results obtained for a particular
structural response problem indicate that the rocking motion induced in the
facility from a seismic input magnifies the motion of the structure at points
away from the center of gravity. This magnification increases the severity
of the structural motion by an order of magnitude.
1.
INTRODUCTION
The analysis of soil-structure interaction effects has been the

subject of extensive investigations for many years. The analytic approaches
to these problems generally fall into two categories, namely, exact solutions
to a rather restricted class of problem, such as that of Baron [ll, and
approximate solutions to a more general problem of interest, such as that of
Miller and Costantino [2]. The analytic solutions, although complete, generally suffer from the fact that only relatively simple problems can be
treated. Restrictions as to material constitutive laws and structural geometries must be imposed to make the solutions tractable. On the other hand,
the approximate solutions, although able to treat a wider class of problems,
suffer from the difficulty involved in estimating the degree of approximation in the solution.
In an attempt to improve this situation, a large computer program
was developed which has the capability for treating the complete soilstructure interaction problem including all of the primary aspects of the
response. The solution then becomes "exact", only, however, in the sense
that numerical approaches approximate continuous problems. The basic
- 196 configuration of interest is shown in Figure 1 and consists of a general

structure embedded within a soil/rock foundation made up of an arbitrary
number of material layers, each layer possessing its own, generally nonlinear, constitutive law. To this system, the seismic motion history is
applied in the form of either displacement, velocity or acceleration motion
records. This immediately brings to the fore an important aspect of the
problem, namely, how should the input motions be applied to the system.
Seismic motion records are generally surface records and in fact what is
needed are records with depth at a given site. This problem, however, is
currently unanswerable and is considered as external to the soil-structure
interaction problem of interest herein.
The wave propagation from the input location into the free-field
soil system can be treated by finite element methods of analysis (see
Costantino [3], [4] ) including the effects of nonlinear properties of the
soil. The computer code developed for this problem is termed the SLAM Code
for identification. The finite element approach has been taken in this
development to allow the user a general flexibility in treating problems of
rather complex geometry (material layering, structural inclusions, complex
boundaries, etc.).
To treat the interaction of the free-field soil/rock with the (say,
concrete) structure, the finite element mesh can be continued through the
structure wall. Thus, no other considerations need be made in the developed code (save for possible separation and sliding effects which will be
discussed later). However, this approach has serious drawbacks. Many more
elements would be required to treat the structure in this manner increasing
computer running times. Of more importance, however, is the following fact.
The stiffest material encountered in a typical problem is usually the structural material. In addition, the smallest sized elements in the problem
occur through the thin wall of the structure. This combination leads to
extremely high frequencies in the mesh, in turn leading to extremely small
time steps in the required numerical integration procedure. This in fact
occurs because the mesh in the structural wall is able to transmit the highfrequency through-the-thickness waves which would develop.
This refinement in the solution is generally unwarranted except for
some special situations. In fact, the usual structural representation in a
soil-structure interaction analysis consists of the rigid body modes of the
structure together with its lower free-free elastic modes of vibration.
This approach has been taken in the analysis presented herein.
To summarize the approach to the dynamic soil-structure interaction
problem then, the following aspects are included.
a) The free-field is represented by a finite element mesh and treats the
wave propagation problem through a layered soil/rock site, each material
in the free-field being allowed to have its own nonlinear constitutive
- 197 -
relationship.
b)
The structure is represented by its rigid body modes together with its
lower free-free elastic modes.
c)
Potential separation and sliding between the structure and the free-field
can be treated by means of a special element (of zero thickness) placed
between the structure and the soil.
It should be mentioned that if nonlinear behavior of the structure

is to be considered, then the use of the finite element mesh through the
structure must be used, as previously discussed.
2.
FREE-FIELD ANALYSIS
To treat the free-field wave propagation problem, the soil/rock

material is divided into small elements, these elements being connected to
each other at their vertices. The types of elements used depend upon the
particular problem of interest. For three dimensional problems, tetrahe
drons, cubes, etc. can be used. For the two dimensional problem for which
numerical results are presented herein, rectangular and triangular elements
are used for the mesh formation. The data that is developed is the motion
history (displacements, velocities, accelerations, stresses, etc. ) at these
node points or vertices as a function of time. This method of mesh forma
tion is essentially a physical one, as opposed to the more abstract approach
of finite difference methods.
In the analysis, any material constitutive law can be used, pro
vided of course that it can be suitably described for inclusion in the code.
The specific soil/rock models used will be described in a later paragraph.
The computational procedure starts from some time at which the complete
solution is known; that is, the displacements, velocities and accelerations
of all the nodes are specified, as well as the entire stress and strain his
tory up to and including this time. Typically this time is the zero or
initial time, although it need not be. The problem then is to determine
these same variables at the following instant of time, suitably taking into
account the nonlinearities introduced by the material properties.
A typical interior node point of a two-dimensional mesh is shown in
Fig. 2, this node being connected to its surrounding nodes through the inter
connecting elements. The equations of motion for this node can be written
symbolically as
**'<
-- * - ? f l
R
H M - _ - 2 F
(1)
198
where M, is the total nodal mass composed of the mass contributions from
each adjacent element, ( F,^ F W M ) are the horizontal and vertical forces
R
K
applied to the nodes (if any) and ( F., futt ) are the node resisting forces
developed by the distortions of the surrounding elements, the summation
being taken over all of the surrounding elements. C learly, a displacement
field causing only rigid body motions of the elements will develop no
resisting forces at the nodes. The details for computing the node resist
ing forces from the element distortion are presented by Costantino \_4~\.
Combining the equations for all the nodes, a set of second order
equations are developed for the entire mesh which can be written symboli
cally as
i^\x
4 K x
FA
F"
(2)
where M is a diagonal mass matrix, is a displacement vector con

sisting of the horizontal and vertical displacements of the nodes, is
the usual banded system elastic stiffness matrix and F is the vector of
applied nodal forces. F
is a vector of corrective forces which account for
the nonlinearities in the material stressstrain properties, or deviations
from the elastic case. These nonlinear correction forces are computed at
each time step for each element surrounding a given node.
The numerical integration scheme used to treat the wave propagation
phenomena follows directly from eq. (2). At a given time, t, when the
nodal displacements and velocities are known (together with the previous
histories), the nodal accelerations are computed by determining the applied
loads (determined from any applied surface pressures), and the nonlinear
correction forces (by knowing the current displacements and the previous
element displacement and stress histories). Knowing the accelerations at
this time, the displacements of the nodes can be determined at the following
time, t + A t , by a suitable numerical integration scheme. Knowing the new
displacements, the cycle can be started again by determining the accelera
tions from eg. (2) at time t + t, etc. In this manner, the solution is
marched out in time to obtain the complete history of the motion of all the
nodes. Examples of various solutions obtained in this manner are presented
by Costantino [ 4~\.
3.
MATERIAL CONSTITUTIVE LAWS
The computer program developed to treat this probien contains a

catalogue of material stressstrain laws which can be added to with little
difficulty without changing the operation of the code. Each material occur
ring in a particular problem can then be allowed to have any of the material
properties available in the catalogue. The current catalogue allows the
specification of the following stressstrain relations:
199
a)
b)
c)
d)
e)
elastic material, either isotropic or anisotropic,

linear compressible fluid,
elastic plastic material satisfying the Mises yield criterion with arbi
trary strain hardening effects,
elastic plastic material satisfying the CoulombMohr yield criterion,
a nonlinear material law which contains a stiffening effect under hydro
static pressure as well as a plastic dissipation under deviatoric
strains to account for compaction effects in soils.
The last three of this list are the only nonlinear laws currently
available in the code, and have been included in an attempt to at least
crudely approximate some known responses of soil/rock materials. C uite
apparently, none of these models are completely adequate but until further
advances in the state of the art occur, only such approximations are avail
able for applications to earth media.
4.
SOILSTRU
C TURE
INTERAC TION
The treatment of the interaction between the structure and soil

begins with the assumption of continuity, that is, the nodes at the soil
structure interface are assumed to be attached to the structure and move
with it. Separation and sliding between the soil and the structure is
accounted for separately and will be discussed in a following paragraph.
Let the displacements of the nodes attached to the structure be defined by
the vector X^ (a subset of the freefield displacement vector, X, defined
in eq. (2). If there are 1> such nodes and if two dimensional (plane) motion
is considered, the components of X, are then
[**] ' ^"*.,"", .w,
(3)
.^."M
where (i.,ur) are the horizontal and vertical displacements of the nodes.
The equations of motion of the structure are defined by its modal
equations
** % + K ^ -- Q *
(4)
where y is the mode vector of the structural degrees of freedom and con
siits of the rigid body coordinates plus an arbitrary number of freefree
modal coordinates. M is a diagonal mass matrix consisting of the modal
masses, K is a diagonal stiffness matrix and Q is the vector of applied
structural modal loadings. The displacements of the structure at the loca
tions of the attached nodes are obtained by superposing the modal vectors, or
X
where
t "
1*
(5)
is a matrix composed of the structural mode shapes.
From eq. (2), the equations of motion of the nodes attached to the
structure are
200
M/x t + F*
(6)
where is the vector of interaction forces developed between the nodes

and the structure. With these interaction forces, the corresponding modal
loads applied to the structure are then
Q s FTP
(7)
where the superscript indicates the transpose of the matrix. Substituting

eq. (6) and (7) into eq. (1), the equations of motion for the structure
become
V,
rK
(8)
where M_ is a nondiagonal mass matrix including the inertial coupling

between the structure and the freefield, and is defined by
Mj M
(9)
From this point on, the solution to the interaction problem pro
cedes in a similar manner to the freefield problem. At a particular instant
of time, the displacements of all the freefield nodes and structural mode
displacements are known. The accelerations of all the nodes (except the
attached nodes) are computed as before from eq. (2). The resisting force
vector, fv, of eq. (6) is determined during this computation. The modal
accelerations of the structure are then computed from eq. (8). The dis
placements of the freefield nodes and the modal displacements are then
determined at the following instant of time by the integration algorithm.
The displacements of the attached nodes are then computed from eq. (b). The
solution is then marched out in time as before.
As an example of a solution to the interaction problem, a simply
constrained rod problem was investigated with a rigid mass located at a
point in the middle of the rod. A step pulse pressure of unit magnitude is
applied to the end at time zero. The rod was represented by a series of
uniform rectangular elements with all the nodes constrained to move later
ally. The results for several runs are shown in Fig. 3. The solid curve
of this figure indicates the stress variation at a particular instant of
time along the length of the rod with no rigid inclusion. As may be noted,
the stress solution "smears" out the sharp shock front since the finite
solution technique cannot treat discontinuities. This induces the oscilla
tions about the true solution which tend to decay away from the front. The
comparison for displacements (see Costantino _4 ] ) indicates minor
201
differences only near the stress front. In all such computations, displacement calculations show good correlation with available solutions while
stress calculations contain these typical oscillations. This is true also
for nonlinear material problems.
The other curves of Fig. 3 indicate the computed solution at the
same time but with a rigid mass included in the mesh. The first curve is
for the case where the mass of the inclusion equals that of a rectangular
element while the second curve is for the case where the mass of the inclusion is ten times that of an element.
5.
SEPARATION BETWEEN SOIL AND STRUCTURE
In treating this separation problem, it is desirable to use a technique which does not deviate from the method of analysis outlined above. To
accomplish this objective, a new finite element model was developed. For
the two dimensional problem (planar motion), a rectangular element is used
which has a finite dimension in one direction and a zero dimension in the
normal direction. The properties of this element are determined by using
the limit process to the properties of the finite size rectangular element
normally used. Thus, of the four nodes comprising this element, two at one
end have the same coordinates and two have the same coordinates at the other
end. This element is situated between the soil and the structure so that
the side of finite length lies on the interface. Two nodes are then located
on and attached to the structure while two nodes are attached to the freefield nodes. When the nodes separate, a zero interaction force is computed
while in compression the full interaction force is transmitted. Shear
transfer and sliding motions are treated in a similar manner with the same
element. The details of this element model are to be presented in a forthcoming paper.
6.
RESULTS FOR AN INFRACTION PROBLEM
Using the developed program, the response of a structure subjected

to seismic inputs was determined. The structure, shown in Fig. 4, is a
relatively long building with cross-sectional dimensions as indicated. A
two dimensional analysis was thus undertaken. The foundation consists of a
two hundred foot soil layer resting atop a rock material. For this particular case, the soil was modeled as an elastic material with the properties as
shown, while the wave transmission through the rock was neglected due to its
high stiffness as compared to the soil.
The finite element mesh used for this problem is shown in Fig. 5,
and consists of a rectangular mesh, the total length of which is 4500 feet.
The seismic input applied to the left most mesh boundary is the horizontal
motion history obtained from the 1957 San Francisco Golden Gate earthquake.
As an indication of the transmission characteristics of the mesh, a problem
was first run without the structure. The results are shown in Fig. 6 in
- 202 which the input displacement motion is compared with the horizontal motion
in the middle of the mesh (2250 feet from the left most boundary). As can
be noted, the total motion response at the downstream location is a replica
of the input motion up until a time of approximately 2.1 seconds. Beyond
this time, the motion response is modified due to the reflections trans
mitted from the downstream or right most boundary of the mesh. It is quite
clear that the mesh must be long enough so that within the response time of
interest no fictitious boundary reflections will be encountered. Nontransmitting or "quiet" boundary considerations can be used to help reduce such
reflection effects but these will not be considered herein.
Shock spectra for both the input motion and for the downstream
motion were computed and are shown in Fig. 7. As can be noted, the spectra
for the computed motions lies below that for the input at all frequencies.
At the low frequency ranges (below 10 cps) the differences are due to the
shorter record length of the computed motions as well as the boundary
reflections. At the high frequency end of the spectra, the differences are
due to the characteristics of the mesh used. An approximate relationship
for the highest frequencies in a given mesh is
* 7
(10)
where c is the dilatational wave speed and A is the shortest element

dimension used in the mesh. For the elements used in this problem, the
highest frequency of the mesh is approximately 15 cps. The mesh therefore
cannot transmit frequencies beyond this cutoff frequency. This is clear
from Fig. 7.
After this initial free-field problem was run, a second problem was
solved with the same free-field mesh as previously used, but with the struc
ture shown in Fig. 5 attached to the mesh. The total length of the run was
for a real time of 3 seconds, as before. Part of the displacement motion
generated is shown in Fig. and is compared with the free-field motion
obtained from the first run. As may be noted, the structural response lags
the free-field motion, as expected, with significant rocking generated from
this predominantly horizontal motion.
Horizontal shock spectra for the upstream, center and downstream
points on the soil-structure interface are shown in Fig. 9 and again are
compared to that of the free-field (no structure) problem. As can be noted,
the rocking motion significantly increases the horizontal spectra at both the
upstream and downstream locations. Similar results for the induced vertical
motion of the structure are shown in Fig. 10, with the same conclusions
drawn.
7.
SHOCK SPECTRA INTENSITY FACTORS

To better indicate the magnification of the induced rocking motion,
- 203 shock spectra intensity factors were computed for the horizontal and verti
cal spectra of the upstream, center and downstream points on the soilstructure interface. These intensity factors were defined by Miller and
Costantino \,2] as
2>" J l i . U f
\'
("
v ' f, I * U T
(ID
] \ ^ 1
where Sp = displacement intensity factor, Sv = velocity intensity factor, and

5 = acceleration intensity factor, f is the frequency of the linear oscil
lator (cps), and is the corresponding period. The parameter ,_| i s
the value of the peak displacement of the shock spectra associated with a
given frequency, )Z
I the corresponding peak pseudo-velocity and \
|
the corresponding peak acceleration. These factors are measures of the in
tensity of the motion at the low, mid and high frequency ranges. For the
relatively short record lengths used in this problem, the displacement in
tensity factors are questionable.
The results of this calculation are shown in Table 1. As can be
seen, for horizontal motion, the factors for point A (directly under the
center of gravity of the structure) are approximately the same as that of
the free-field (with no embedded structure). However, at both the upstream
and downstream locations on the structure, the intensity factors are in
creased by an order of magnitude. For vertical motions, the intensity fac
tors are even further increased due to the rocking motion as well as the
induced vertical motion of the structure. Such increases in the factors
indicate the magnification of the motions seen by the structure and have
severe implications on equipment stored within the structure.
e.
SUMMARY
A numerical technique has been described which treats the complete

soil-structure interaction problem due to seismic motions. The approach is
quite general in its application and includes the effects of structural
flexibilities on the coupled motion as well as nonlinear behavior of the
soil materials.
For the particular structure analyzed, the intensity of the motion
sustained by the structure (as measured by the shock spectra intensity fac
tors) is increased by an order of magnitude indicating that soil-structure
interaction effects play a significant role in the seismic design of equip
ment mounted within the structure. It should be especially noted that the
usual assumption of a one-to-one correspondence between support point
motion and free-field motion may not be a conservative design criteria.
Rocking motions induced in the structure may easily result in a more severe
environment for points away from the e.g.
- 204 ACKNOWLEDGEMENT
The development of the SLAM Code was supported in part by the
National Science Foundation under Grant No. GK 3214 with The City College
of New York.
REFERENCES
1
BARON, M.L., "The response of a cylindrical shell to a transverse shock wave", Proc, of the Second U.S. National Congress
of Appi. Mechanics, ASME (1955).
MILLER, C.A. and COSTANTINO, C.J., "Structure-foundation

interaction of a nuclear power plant with a seismic disturbance;1
Nuclear Eng, and Design, 14, 332 (1970).
COSTANTINO, C.J., "Finite element approach to stress wave

problems", J. Eng. Mech. Division, ASCE, 93, 153 (1967).
COSTANTINO, C.J., "Two dimensional wave propagation through

nonlinear media", Journal of Computational Physics, 4, 147
(1969)
TABLE I
SHOCK SPECTRA INTENSITY FACTORS
LOCATION ON
SOIL-STRUCTURE
INTERFACE
( i t * Fig. I t )
FREE-FIELD
(NO STRUCTURE)
UPSTREAM POINT
(POINT B)
CENTER POINT
(POINT A)
DOWNSTREAM POINT
(POINT C)
(la-cycli/ic.)
HORIZONTAL
VERTICAL
(In./crol)
HORIZONTAL
VERTICAL
( In./ayola-aao.)
HORIZONTAL VERTICAL
2.466
0.S68
7.10
1.366
0.321
0.0762
31.604
12.361
146.84
60.767
3.761
1.6166
i.eze
3 5 97
4.06
24.SI7
0.210
0.666
16.167
13.4 4 7
137.10
62. 3TT
3.267
1.666
- 205
SEISMIC
INPUT
MOTIONS
Fig. 1
General
Configuration
Fig. 2
Typical Interior Node, N, and Surrounding Nodes, S
206
UJI 1 1 1 1 1 1 1 1 1 1 1 M m 1 1 1 I 1 I'
"
Rectangular
/
Element Mesh
V Rigid Inclusion
No
Inclusion
Mass Ratio I
Moss Ratio = 10
Analytic Solution at a given tin
10
20
DISTANCE
Fig.
ALONG
R e s p o n s e of
ROD
C onstrained
Rod w i t h R i g i d
Inclusion
200
W= 120 k /In
I= 2.2l05k,ec
140'
e.g.
Structure
70
200'
Soil L a y e r ' w = 100 pcf

C.~ 3 0 0 0 f ps
a
wmnmm^r
imrmm
Rock
Fig.
Soil
Layer
Structure
System
Investigated
- 207 -
4500'
"
CG.
200
^w
Input Motion
Histories
tyr
77 Wr
W7 wT
W7~ "*
jm I tm
Roller Supports
Along Base
Fig. 5
ih
imi
Tffr
__Reetangulor
Elements ( 5 0 * 5 0 ' )
Element Mesh u s e d f o r S o i l - S t r u c t u r e
System
1.0
Free-Fiel dInput
.
bl
< .
a.
^N
z^-.
co
a
.4o
O
I
f Computed Downstreor
Motion
.2
/
I
2.0
1.0
TIME (SEC.)
Fig. 6
F reeF ield Mesh Displacement Histories
3.0
208
IO
loo
niEouENCYicrt)
Fig. 7
Shock Spectra for Free-Field Horizontal Motion (No Structura)
F r - Field Motion Mllh

no Structure
LS
2 0
TIME (SEC.)
Fig. 8
Horizontal Displacement Histories with Surface Structure
20 9
\\
B A
Ksy
\.
^~
.5
/ ?
" W
1 / V.
SVPoint , Horizontal
Point A r U '
Ho I D tol
\
-Point C, H o r i z o n t a l
% \
.^^FrtFIld\
v \ Horizontal
10
N,
^v
100
FREQUENCY (CPS)
Fig. 9
Shock Spectra for Horizontal Motion with Surface Structure
fi
ll \\
VB A c /
//' \\
\
s~*S\
^"^
/
\ * y\
iP N l\
:
;
."'
N\^_Point
/ \ \ \
N
C,Vortical
^\^--Polnl
B.Vortleol
> 1.0
'
"
\ \
Point A . _ V ^ \
Vortical
\.y
\
'. v
\
Ff.Fltld, _ j , \
Vertical
\ \ .
.
\
\
\
FREOUINCYI)
Fig. 10
Shoe* Spectra for Vertical Motion with Surface Structure
K 3/5
SOIL-FOUNDATION INTERACTION
OF REACTOR STRUCTURES SUBJECT TO SEISMIC EXCITATION
T.H. LEE, D.A. WESLEY,
Gulf General Atomic, San Diego, California, U.S.A.
ABSTRACT
A theoretical Investigation has been conducted to study the soil-structure dynamic interaction effects on the seismic response of reactor structures.
The analysis was made by consid-
ering a linear, damped, unsymmetric, three-dimensional flexible structure coupled with an

elastic, homogeneous, Isotropic half-space.
The model is capable of handling mult I-componen t
ground excitation input and accounting for additional ground effects, such as the through-soil
coupling between translation and rocking motion.
Numerical results for some example problems dealing with a prestressed concrete reactor
vessel (PCRV) are presented in terms of both transient and steady-state responses. The
seismic behavior of reactor structures was found to differ considerably from that exhibited
by ordinary buildings.
Significant interaction effects were observed for various cases which
have not been previously analyzed.

I.
INTRODUCTION
The dynamic interrelationship between an elastic structure and its foundation is known to
have a significant influence on the structural response during earthquakes.
These interaction
effects are more pronounced for heavy structures, particularly when founded on a relatively
soft medium, and it has become necessary to incorporate the soil-structure interaction in the
seismic design considerations of today's large nuclear power stations.
The interaction phenomenon was initially studied by approximating the dynamic properties
of the soil with discrete springs and masses.
More rigorous solutions have recently been
developed by coupling the structure with an elastic half-space.
Among these recent contrib-
utions, one may cite the work of Parmelee [1,2], Luco [3], and Scavuzzo, et al. [A], The use
of an elastic half-space in an interaction problem has been shown to yield results which are
in good agreement with the measured data from Duke, et al. [5].
Parmelee coupled the elastic half-space with a mathematical model representing a singlestory building, and considered the translation and rocking degrees of freedom of the system.
Luco studied the interaction of soil with an infinitely long shear wall.
He assumed that the
ground excitation is such that the movement of soil is parallel to the direction of the shear
wall and hence, the motion of the system is purely trans latlonal.
Scavuzzo, Bailey, and
Raftopoulos obtained the time-history of the seismic response of nuclear reactors on an AEC-
- 212 sponsored s t u d y .
They c o n s d e r e d
and a c o n v e n t i o n a l
structure
the dynamic
N-mass s t r u c t u r e
can be c o n s i d e r e d
to e x i s t
in
their
p a p e r on t h e h i g h e r mode a s p e c t s and d e r i v e d
N-mass s t r u c t u r e w i t h
lateral
emphasis was t o d e m o n s t r a t e
interrelationship
in a l a t e r a l
translation
trans l a t I o n a l
formulat on.
utilizing
the f o r m u l a t i o n
an e l a s t i c
i d e a l i z e d so t h a t
the
techniques
nuclear
power
the problems
the e q u a t i o n s
for
have,
s o t rop i c
n a d d i t i o n
siderations,
a more g e n e r a !
to
half-space.
its
modal
time h i s t o r y .
Lagrange's
equations,
synthesis
time-history
In o r d e r
nume r i c a l
results of
p r e s e n t e d here
for
2.1
Derivation of
Governing
In t h e de r i v a t i on o f
studies.
The
a p p e a r as s u b s c r i p t s
Consider
the
indicai
0,
is
the o r i g i n o f
angular
Let
t h e s y s t e m a re
nvesti-
highly
manner.
Hence,
structures
such as
is
Additional
have been
displacement
column
transient
matrix
c o m p o n e n t s , each h a v i n g a
e q u a t i o n s were
formulated
t h e s y s t e m were o b t a i n e d
the v a l i d i t y
con-
included
of
the a n a l y s i s ,
simplified
models
and s t e a d y - s t a t e
from
the
through a
the
general
for
obtaining
responses
are
ves se 1 (PCRV).
the L a g r a n g a n
and summat i o n c o n v e n t i o n a r e used t h r o u g h o u t
is p r e f e r a b l e
in order
used a r e
shown
in F i g .
(=1,2,3)
the c e n t e r
energy o f
f o r m u l a t i o n was a d o p t e d
notation
superstructure.
ve l o c i t y v e c t o r P. r e l a t i v e
for
frame
is
to
The x .
of
t o enhance
identification
1 whe re
is a f i x e d
frame
t h e X.
the c l a r i t y .
this
The
indices
reference
s y s t e m and
the
s assumed t o be a r i g i d
i s assumed t o be f i x e d
The x.
in
for
pi.r p o s e s .
reference
the basemedium c o n t a c t
d e s i g n a t e d as u .
the p a s t
two C a r t e s i a n c o o r d n a t e
The base mat
area.
frame a l s o
in
5 lab
t h e base and
The v e l o c i t y
vector
rotates
an
with
frame.
the s t r u c t u r e b a s e
s y s t e m be d e s i g n a t e d as T w h i c h
= + TS
and
elastic,
Laws a p p r o a c h w h i c h has been used i n
The X.
t h e moving frame
the k i n e t i c
freedom.
and r o c k i n g ,
rotational
t o a number o f
Both
considering
c o u p l e d w i t h an
degrees o f
interaction
a moving c o o r d i n a t e s y s t e m .
located at
structure
Newton's
model
into
where
t h e use o f
In t h e s e
restricted
an
primary
s i m u l a t e d by a d i s c re t e s y s t e m w h i c h can
and t h r e e
reactor
and t h e s u p e r s c r i p t s
which supports a f l e x i b l e
origin,
through
t h e p r o b l e m i s p r e s e n t e d by
rigid-body
to demonstrate
notation
systems have been i n t r o d u c e d .
its
response o f
Tajimi's
Equations
t h e o re t i c a l
frame
is
the g o v e r n i n g e q u a t i o n s ,
application of
s e c t i o n e x c e p t when m a t r i x
to
responses o f
some example p r o b l e m s .
THEORETICAL FORMULATION
of
in a rather
d e f i n e d by a f r e e - f i e l d
specialized
a pres tressed concrete
2.
to d i r e c t
is
The t h r e e - d i m e n s i o n a l
and the
method.
six
translational
d e r i v e d were s u b s e q u e n t l y
(i = 1,2,3)
freedom.
complex unsymmetric
c o u p l i n g between t r a n s l a t i o n
equations
x.
of
flexible
The s t r u c t u r e
The s e i s m i c e x c i t a t i o n
prescribed
interaction
treated
approach
coordinates,
such as t h r o u g h - s o i l
n the a n a l y s i s .
contrast
steady-state
r e s u i t s we re p r e s e n t e d .
analysis
the
d i s c u s s e d Parme l e e ' s
an N-mass b u i l d i n g model
three-dimensional
wh i eh i s a 1 lo^/ed t o have t h r e e
Fourier
general
in
plants.
damped, unsymmet r i e ,
homogeneous,
for
for
half-space
No damping
[6]
t h e geomet ry and m o t i o n o f
n e c e s s a r i l y were
are not adequate
In the p r e s e n t w o r k ,
a linear,
half-space,
Tajimi
and r o c k i n g d e g r e e s o f
a moda 1 t r a n s f o r m t on t e c h n i q u e , and no n u m e r i c a l
gations
be tween on e l a s t i c
motion.
is
decomposed
(I)
denote, respectively, the kinetic energy of the base and the kinetic energy of
the remainder of the structure.
213
For a rigid base, it has been shown n [7] that
_B
1 B O O ___ 1 _B n n A
_
= = m u. u. + = . , . . + m e. ..u.l.r.
2
11
,. . .
. n _
_ ,JJ , k = 1 , 2 , 3 /
i j k i j k
,0v
(2;
'
where u. and , are, respectively, the components of uand , e.., is the permutation symbol[8],
'
J
R
1 JK
m s the total mass of the base, r. are components of the position vector of the mass center
B
of the base mat, and I., are the inertia tensors of the base with respect
to the x. coord i nate
p
IJ
system defined as
* =fBp
(6
ijrkrk
r r )dT
(3)
i j
where s
vector of
Let
the mass d e n s i t y ,
6..
1
J
is
a particle with origin

the r e m a i n d e r o f
The k i n e t i c
energy
the Kronecker d e l t a ,
a t 0 , and t h e
the s t r u c t u r e
may t h e n be e x p r e s s e d as a f i n i t e
where m
is a diagonal
placement
vector
Let
{ul
inertia
re la t i ve t o
designate
f r a m e , and l e t
matrix of
the
d e f i n e d as a 6X1 column
U,n
rigidbody
{u}
position
p
the base,
be r e p r e s e n t e d by a lumpedmass
sum c o n s i s t i n g o f
.
model.
t h e f o l l o w i n g terms
()
masses, and { u $ }
i s an Ncomponent
dis
the d i s c r e t e
vector of
structural
masses
relative
the s u p e r s t r u c t u r e b a s e
to
s y s t e m be
matrix
a r e components o f
displacements
t h e volume o f
the
frame.
vector of
,., .j
following kinematic
t h e components o f
= 1.2...N)
the d i s c r e t e
displacement
{UR}
where 0.
are
f i xed r e f e r e n c e
the d i s p l a c e m e n t
the
is over
(superstructure)
TS = i m n =
the x .
integral
r,
small
relation
rotation
relates
in a l i n e a r
=/!}
{u>}
(;
= 1,2,3)
vector of
to
the x .
the r i g i d b o d y
(5)
coordinate
frame and . s . .
displacements
{ U ^ } and t h e
The
relative
transformation
uS = A
R +
nr r

(n=l,2,...N)
(6)
(p=l,2,...L)
(r=l,2,...6)
where A
and a r e
here
relative
is
modal
data a s s o c i a t e d w i t h
base n a t u r a l
Let
let
rectangular
{u}
trans formation matrices.
to a r i g i d b o d y
frequencies
{u}
reference
are u s u a l l y
and mode
position of
The d i s p l a c e m e n t
the system.
known and t h e y a r e o f t e n
referred
in
terms o f
the
{u}
introduced
t o as
the
the " f i x e d
shapes."
the number o f modes be assumed t o be t h e same as t h e number o f

be e x p r e s s e d
vector
In s e i s m i c a n a l y s e s ,
fixedbase
= q (t)
ps^s
mode shapes o f
components
the d i s c r e t e
(p,s = 1 , 2 , . . . L )
in
structure
{u},
and
as
(7)
where q (t) are the generalized coordinates and 'C

is the modal matrix of the discrete struc
s
ps
ture. Substitution of Eq. (6) into (k) in conjunction with Eq. (7) yields
TS
T =
1 MS RR ___ 1 u

___ M SB,R
M U U + x / i q q + M U q
2 pr r
2 ps s
rs r s
/ox
loj
214
where
pr
PS
in
(9a)
Up nr
Ob)
tn Iq nt qp ts
n5B _ m.D
U,n
A.B
y
i n i r np ps
I.2....N)
(p.r
1,2,...6)
(q.p.s.t
The s t r a i n energy o f a lumpedmass s t r u c t u r e
(9c)
1,2,...L)
may be w r i t t e n as
U i k u u
2 qs q 5
or,
i n terms o f
the g e n e r a l i z e d c o o r d i n a t e s ,
(10)
as
0,)
ptVt
where
KS
pt
Equations
(2),
for s u b s t i t u t i o n
( 8 ) , and ( I I )
into
In the p r e s e n t
external
surface
contact
forces a t
will
the L a g r a n g e ' s
formulation,
k Y t
qs qp s t
be used i n c o n j u n c t i o n w i t h
the v i r t u a l
work
expression
equations.
the f o r c e s e x e r t e d on the base by the medium a r e
f o r c e s on the s t r u c t u r e b a s e
the s t r u c t u r e m e d i u m
system.
The v i r t u a l
i n t e r f a c e may be w r i t t e n as
t r e a t e d as
work a s s o c i a t e d w i t h
the
[7]
6WM OySU*
(12)
M
where 0_ are components
the partitioned
of the generalized forces associated with the interface stresses and

M
form of Q vector is consistent with Eq. (5), namely:
>- : :::!:!!
n which
P.
y_T.2dA
and ., are the components
and
M, j f . e ^ r . r ^ d A
of contact stress and the integrals are
(It)
over the entire
contact
area .
Using
the principle of superposition,

uj \
where u
is
the p r e s c r i b e d
ground d i s p l a c e m e n t
vector
freefield
let (U") be written as

+ "l
(r = 1,2,.. .6)
ground d i s p l a c e m e n t
due to the presence o f
v e c t o r and
the s t r u c t u r e .
oped h e r e , both i u ^ } and f u * ) a r e a l l o w e d to have s i x components;

rotations.
Each element
Treating q
following
and u
in {uG}
can have an Independent
as c o o r d i n a t e v a r i a b l e s
two systems o f e q u a t i o n s are
obtained:
(15)
the
additional
model
three t r a n s l a t i o n s
timehistory
and a p p l y i n g
is
In the g e n e r a l
devel
and
three
input.
the L a g r a n g e ' s e q u a t i o n s ,
the
215
S
SB'"R
+ C q + K q H
U
nm m
nm m
nr r
q
nmm
MB R + MS R + M S B q
QM
pr r
pr r
pn
p
(16a)
(n.m 1,2,...L)
(16b)
(p,r 1,2,...6)
where H
can be put
n partitioned form as
HSt
with all
the submatrices being 3X3.
(17)
Q
The elements in I
have already been defined in Eq.(3) and
MBB,.mB
U
'J
In Eq.
(16a), C
nm
(18)
B B
uBC
M., e. .. r.m
k
jk j
is known as the g e n e r a l i z e d damoinq m a t r i x o f
may be d e r i v e d from the d i s s i p a t i o n
do*
(19)
the s u p e r s t r u c t u r e ,
and
The prime denotes
the
term in the L a g r a n g e ' s e q u a t i o n s .
it
transDose o f a m a t r i x .
2.2
Solution of
the Dynamic
The s o l u t i o n
made t r a c t a b l e by f i r s t
then o b t a i n i n g
Its
that
tations
be t r a n s f o r m e d
solution
the complex frequency
found by I n v e r t i n g
e q u a t i o n s g i v e n by Eqs.
d e t e r m i n i n g the s t e a d y s t a t e
timehistory
require
o b t a i n e d as
Equations
to the system o f d i f f e r e n t i a l
into
its
(16a)
transforms.
and (16b) was
the i n t e r a c t i o n system and
through a F o u r i e r s y n t h e s i s method.
responses o f
The
techniques
the system be d e t e r m i n e d and the s e i s m i c
f r e q u e n c y domain.
the response
responses o f
The time h i s t o r i e s
of
The s t e a d y s t a t e s o l u t i o n o f
exci
the response were

the problem
then
is
follows:
For harmonic
response,
/\
q ( t ) = qM e
^n
n
'WE
iiR
nR
U = 0 e
r
r
'tut
fnn\
(20)
where is the circular frequency of the excitation, the bar over a vector designates the com
plex amplitude, and
i /\
when it is not used as a subscript.
Solving Eq. (16a) for the q
results in
' M 5 B 'G R
where U
20
and T " 1
i s the
nm
inverse of
()
The two sys terns o f e q u a t i o n s

t u t i n g Eq.
(21)
i n t o Eq.
(l6b).
nm
(l6a)
This
(21)
the complex m a t r i x ( )
nm
mr
and ( l 6 b )
JC
nm
(22)
nm
may be combined i n t o a s i n g l e s e t by s u b s t i
gives
(MB
+ MS ) U R + M S B D M SB ' R
\ pr
pr/
r
pn nm mr
r
where the g e n e r a l i z e d
KS
d e f i n e d by
KM u 1 = 0
pr r
(23)
f o r c e s have been r e p l a c e d by
p C ( l " ) G
{M
216
and the matrix D
nm
is defined as
D
= 2"]
nm
nm
The response o f
the s t r u c t u r e - b a s e system is
(25)
thus coupled w i t h
the displacements o f
the
elastic
half-space.
In E q .
( 2 * 0 , the m a t r i x
()
I t s elements are complex f u n c t i o n s
is
t h e dynamic s t i f f n e s s
matrix of
taken from the s o l u t i o n s o f
the h a l f - s p a c e
medium.
t h e dynamic r e s p o n s e o f
elastic
s e m i - i n f i n i te solids
p a r t o f an e l e m e n t
wave
(radiation
under h a r m o n i c s u r f a c e l o a d i n g o r h a r m o n i c i n c i d e n t w a v e s . The i m a g i n a r y
M
i n m a t r i x a c c o u n t s f o r t h e e n e r g y d i s s i p a t i o n due t o r a d i a t i o n o f t h e
damping).
In v i e w o f E q . ( 1 5 ) , E q . (23)
L2*1
\
reduces
pr
to
+ KM ) u 1 = - 2 DG
pry
pr
(26)
where
H1 - HB + MS M S B D MSB'
pr
pr
pr
pn nm mr
Equation (26) represents a system of six algebraic equations with complex coefficients.
r
When the freefield excitation components, , are prescribed, Eq. (26) can be solved for the
amplitudes and phase angles of u . The base steadystate motion can then be determined by the
second equation in Eq. (20) with
UR = DG + D1
r
r
r
Using Eq.
vector
( 2 6 ) , one can show t h a t
by t h e f o l l o w i n g
{ } is
(28)
related
to the f r e e f i e l d
acceleration
amplitude
relation
{01}
[()]-1[]{}
(29)
where
( ] = 2 [ M l ] + [ ( ) ] .
(30)
In accordance with Eq. (28), the base acceleration amplitudes can be expressed n terms of
the ground excitations by the equation
(R) [H R (| U )]{5C )
(3D
where
[/()] FU 2 [E (i )]"' [Ml] .
(32)
The amplitudes of the absolute accelerations of the structural masses can be put in the same
form as Eq. (31).
This gives
s ) = [H S (w)]( G )
(33)
where
[H S (i)] = ([A ] + [ B l M l D H M S e f W o u , ) ]
R
The H
(3
S
and H
tudes o f
matrices.
matrices
relate
the i n p u t e x c i t a t i o n s
the s t e a d y - s t a t e
and a r e o f t e n
amplitudes of
referred
the output q u a n t i t i e s
to as t h e complex f r e q u e n c y
t o the a m p l i
response
- 217 With the exception of certain physical

earthquake disturbance
transform.
phenomena such as faulting, the time history of an
is sufficiently well-behaved
to guarantee the existence of a Fourier
Let
{G(u)}=/"{G(t)} ."'dt
(35)
be the Fourier transform column matrix of the ground accelerations so that the response transforms of the absolute accelerations of the structural masses are
|u S ( u )} - [H 5 (U)]{ G (U>)}
(36)
Upon inverting Eq. (36), the time history of the structural
response is obtained as
{5(t)( = I r[HS(lM)]{G(u)} e1"'*,

2.3
Through-Soil
(37)
Coupling Effects
The medium stiffness matrix K M

the foundation flexibility.
in the interaction equations represents the influence of
The elements on its diagonal are the most significant ones since
the off-diagonal elements account for the stiffness coupling effects through the medium.
The
dynamic stiffness properties of a semi-infinite elastic solid under forced vibrations have been
determined by many
investigators
[9~1**] based on a partial bonding theory.
In the partial
bond-
ing conditions, it is assumed that no shear stresses will be induced for vertical or rocking
motion and, conversely, the normal stresses vanish for lateral
translational
motion.
These
assumptions enable us to find the medium stiffness parameters for each type of motion separately and to analyze the interaction problem by considering a diagonal
medium stiffness matrix.
The complete bonding condition has recently been investigated by Oien
[15] who studied
the harmonic wave diffraction problem in the steady-state motion of a rigid strip bonded to
the surface of an elastic half-space.
In his analysis, it has been shown that the horizontal
translation and rocking of the rigid strip is coupled

ing condition
numerical
is assumed.
through the medium when
the complete bond-
Using a different method, Karasudhi, et al. [16] also estimated
values of such coupling effects.
When the through-soil
coupling
terms are
the
included,
M
the resulting K
quency
matrix becomes nondiagonal.
The ability to readily adjust the complex fre-
response matrix of the system so as to account for these through-soil
coupling effects
is one of the important advantages of the present approach.

3.
NUMERICAL
COMPUTATIONS
As examples
illustrating
the applicability of the analysts, numerical
results were
obtained
for the seismic response of an 1100 MW(e) PCRV based on several simplified
models.
The parameters
representing
results readily available
lumped-mass
the foundation medium properties are based on the research
in the literature.
The models used may have either circular or
square bases although a comparison between a circular base with a square base having the same
area has been shown to exhibit good agreement
The accuracy of the computer results
for the case of a vertical
has
results were obtained both by applying Cramer's

subroutine.
been verified
to give the same answers.
load [13].
The steady-state
rule and by using a complex matrix
The two computing methods were found
pendent verification of the steady-state
vibrating
in several ways.
inversion
As another inde-
results, the data for a typical conventional
building
were also used as input constants for the single-mode model and the response curves are plotted
- 218 In F l g . 2 .
These curves a r e
i n c l o s e agreement w i t h those g i v e n e a r l i e r by Parmelee
c o m p u t a t i o n work a s s o c i a t e d w i t h
Fourier
transform algorithm.
the responses computed f o r

Using conventional
characteristics
of
V e r i f i c a t i o n of
time-history
the PCRV were o b t a i n e d .
motion
f r e q u e n c i e s , mode s h a p e s , and modal
base o f
i n Model I
freedom r e l a t i v e
( F i g . 3a)
is
the model
to the x .
Model I I
(Fig.
3b)
the s y s t e m .
The e a r t h q u a k e
t r a n s l a t i o n , and the
about the X 3 - a x i s , which
The r o t a t i o n s o f
the lumped
When the r o c k i n g degree o f
motion combined w i t h v e r t i c a l
modes a s s o c i a t e d w i t h
mass I s assumed t o be a t t a c h e d
freedom o f
The two d i s p l a c e m e n t
the top mass a r e e l a s t i c a ! l y
coupled.
freedom, u
vibrate
The
and *
to the top mass as shown in F i g . 3 b .
the system w i l l
a nuclear
3c)
r e p r e s e n t s an i d e a l i z e d
reactor structure.
in a coupled
to the main mass, m j .

the r i g i d - b o d y
ments and one t r a n s l a t i o n a l
the s y s t e m .
displacement
The appendage mass is
mass a n d , f u r t h e r m o r e ,
mass.
The e c c e n t r i c
location of
.
k.I
the f o u r modes o f
the
include
is not
the
Relative
so t h a t o n l y
The geometry has
lateral-rocklng-vertlcal
ground movement.
simulating
movement
three
rotational
to
displace
t r a n s l a t i o n s o f main mass
elevation
In l i n e w i t h
rigidly
freedom r e l a t i v e
four s t r u c t u r a l
the appendage mass thus g i v e s
the s t r u c t u r e such t h a t a ground t r a n s l a t i o n a l

all
These
located at a d i f f e r e n t
the mass c e n t e r o f m
structure
is a g a i n assumed to be a t t a c h e d
(North-South).
the o t h e r d i r e c t i o n s were assumed n e g l i g i b l e
components
The appendage
lumped-mass model
The main mass I s assumed t o have f o u r degrees o f
reference position of
are considered.
three-dimensional
The appendage mass, m ,
(or
Is s u p p o r t e d by a " b e a m - t y p e " member
type o f motion under the e x c i t a t i o n produced by a combined h o r i z o n t a l - v e r t i c a l

(Fig.
the base
[].
The s e i s m i c motion was assumed to
ground movement.
the two degrees o f
rigidly
been a r r a n g e d I n such a way t h a t
input.
The top mass, m i .
the t r a n s l a t i o n and r o t a t i o n o f
Model I I I
is
structural
r e p r e s e n t s an i d e a l i z e d unsymmetric system w i t h an e c c e n t r i c mass
can have a p r e s c r i b e d phase r e l a t i o n .
has two n a t u r a l
lateral
t o the system s t u d i e d by S c a v u z z o , e t a l .
appendage) w i t h mul 11 component ground d i s p l a c e m e n t

be a h o r i z o n t a l
(rocking)
the ground m o t i o n .
frame were i g n o r e d .
reduces
is
r e p r e s e n t e d by N masses i n s e r i e s , each h a v i n g
i s a l l o w e d t o t r a n s l a t e and r o t a t e
i s o m i t t e d . Model I
to
Some e x p l a n a t i o n
the models used i n the example p r o b l e m s .
t o the r i g i d - b o d y d i s p l a c e m e n t s o f
t o the d i r e c t i o n o f
masses r e l a t i v e
so t h a t
frequencies.
I n t h i s model was assumed t o have o n l y a s i n g l e component,
perpendicular
fast
r e s u l t s was a l s o made by comparing
The terms "two-mode PCRV" as used h e r e r e f e r s
n e c e s s a r y c o n c e r n i n g the geometry and c o o r d i n a t e s o f

The s u p e r s t r u c t u r e
The
r i g i d ground cases w i t h the v a l u e s g i v e n by o t h e r computer programs.
methods, the f i x e d - b a s e n a t u r a l
the i d e a l i z e d PCRV s t r u c t u r e h a v i n g two f i x e d - b a s e n a t u r a l
one d e g r e e o f
[I].
the F o u r i e r s y n t h e s i s method was done w i t h the a i d o f the
modes
from t h a t o f
the p r i n c i p a l
rise
In
(fixed-base)
the main
axis of
the main
to c o u p l i n g e f f e c t s
through
i n the N o r t h - S o u t h d i r e c t i o n w i l l
excite
structures.
RESULTS AND DISCUSSIONS

Steady-State
Response
The s t e a d y - s t a t e
amplitude
ratios of
results are presented
i n the form o f
f r e q u e n c y - r e s p o n s e curves w i t h
the d i s p l a c e m e n t s c o n s i d e r e d to be o f p r i m a r y
interest.
were not p l o t t e d .
S i n c e the complex a m p l i t u d e s a r e used in the e q u a t i o n s ,
that
I n the response diagrams a r e
the o r d l n a t e s
p l e x ampi I t u d e s .
the r a t i o s o f
the
The phase angles

it
is
understood
the a b s o l u t e v a l u e s o f
two com
219
The ground f l e x i b i l i t y is represented by the parameter V
which Is the shear wave veloc
ity of the halfspace medium defined as
Vs /Tp"
(38)
where and are the modulus of r i g i d i t y and the mass density of the halfspace medium,
respectively.
The condition
is the limiting case where the ground Is r i g i d .
The
results obtained with r i g i d ground represent the responses of the system without considering
the Interaction e f f e c t s .
For threedimensional analysis, the model Is assumed to have a c i r c u l a r base with radius
r .
the steadystate results supplied by B ycroft [11,12] for a threedimensional e l a s t i c
halfspace were u t i l i z e d .
1.5.
B ycroft's results are valid within the frequency range of iiir 0 /V
<
However, the extension of a halfspace analysis to higher frequency factors has subse
quently been made by Olen [15] and by Awojobi [ 1 7 ] .

Figures 4 and 5 present the response amplitudes of the 1100 MW(e) PCRV simulated by
Model I . B oth the singlemode (softmounted) and twomode (hardmounted) cases were considered.
These results clearly Indicate that a large nuclear power station w i l l behave d i f f e r e n t l y from
the conventional highrise buildings during earthquakes.
The Interaction effects tend to
reduce the response of a low, heavy structure due to the presence of r e l a t i v e l y large amounts
of radiation damping.
The response determined for a system simulated by Model I I , where a twocomponent ground
Input was used. Is given In F i g . 6.
The f r e e f i e l d displacement vector was assumed to have
two translational components; a l a t e r a l component

In Fig. 6 were computed with
2/3 of that for u .
being In phase with u
and a v e r t i c a l component .
and the amplitude of
The curves
was taken to be
I t Is interesting to note that as the medium shear wave velocity decreases,
the response at the second peak Is being reduced more rapidly than that at the f i r s t resonance.
I t has also been shown by the present authors [7] that the response curves can be affected by
a change In the phase relationship between u and .
9
9
The Influence of structureground interaction upon the response of higher modes has been
a subject of some International controversy.
Parmelee [1] and Luco [3] have both shown that
the higher mode response w i l l be affected appreciably by the Interaction phenomenon while
Tajimi
[6] claimed that most investigators in Japan believe that the e f f e c t Is n e g l i g i b l e .
Since the results of the present Investigation have again shown a rather pronounced Influence
on the higher mode response, I t may be desirable to reexamine the opinion of the analysts in
Japan on this particular issue.
The model for the ground medium in Model I I was subsequently replaced by a twodimensional
halfspace In order to study the Influence of throughsoil coupling between horizontal
tion and rocking.
transla
In this study, the halfspace response data given by Olen [15] and Karasudhi,
et a l . [16] were u t i l i z e d and the steadystate results for the case with V 1500 FPS are
plotted In Fig. 7 against those obtained without consideration of coupling.
Curves computed
for other values of V show the same changes, d i f f e r i n g primarily In their magnitude.
The
comparison Indicates that the throughsoli coupling effects not only tend to modify the natural
frequencies of the system as expected, but also to reduce the structural response.
220
^.2
TimeHistory
Response
The r e s p o n s e o f an i n t e r a c t i o n
s y s t e m u n d e r an a r b i t r a r y
f o r an e x a m p l e p r o b l e m b a s e d on Model
tation
i n t h e X3 d i r e c t i o n .
(May 19^0) was used as
structural
modes.
four
of
freefield
the comparison o f
excitation
in
the
total
results
Many p r e v i o u s
by t h e
This
al.
analyzed
time h i s t o r y
acceleration
ground f l e x i b i l i t y
inertia
t h e X3
to
for
(NorthSouth)
than
of
values.
For
this
appreciably
effects
if
interaction
f r o m the
the
foundation
freefield
t h e base
the s t r u c t u r e
effects
In the p r e s e n t
the s t r u c t u r a l
direction)
example p r o b l e m ,
l e a d s on t h e appendage mass as
For a r o c k
the
for
maximums
accelerations
is
undamped
{see,
[18]).
the
(n NorthSouth
effects.
the maximum v a l u e s
i s m e n t i o n e d because
interaction
in Scavuzzo, e t
investigators
the e x a c t
translational
the
motion
shown i n F i g . 8
i n X3 d i r e c t i o n w i t h
i s on s o i l ,
interaction
through a response spectrum a p p r o a c h .
by e x a m i n i n g
all
displacement
threedimensional
{ul(t)}are
base a c c e l e r a t i o n s
reactor
do n o t d i f f e r
the s t r u c t u r e .
can be s grv f i c a n t l y a m p l i f i e d
structure
t h e base
a r e seen t o have much l a r g e r m a g n i t u d e s
When t h e
t e n d t o be r e d u c e d by t h e
is p r e s e n t
f o r example,
the
input.
t h e peak base a c c e l e r a t i o n s
when damping
exci
California
.05 was used f o r
medium,
(Vg = 1000 FPS). A s e x p e c t e d ,
the X_axis
determined
t h e El C e n t r o ,
damping f a c t o r o f
t h e s i x components o f
i n p u t was
translational
components.
base a c c e l e r a t i o n
medium,
Time h i s t o r i e s
and r o c k i n g a b o u t
F i g u r e 9 shows
applied
responses.
of
on t h e
response o f
investigation,
Figure
this
10 p r e s e n t s
t h e appendage mass f o r
the
interaction
the g r o u n d medium becomes
the
is
effects
super
studied
the
a range
tend to
total
of
reduce
softer.
CONCLUSICN
The a p p r o a c h d e s c r i bed i n
space dynamic s o l u t i o n s
methods.
A lthough
s o l u t i o n of
here
t h e NS component o f
t r e a t e d as a d e f o r m a b l e
t h e s t r u c t u r e on f i r m s o i l
translation
the o t h e r
5.
to a ground
has s i x n o n z e r o c o m p o n e n t s , and t h e s y s t e m r e s p o n d s w i t h
the case w i t h
the
of
i n p u t and a modal
When t h e g r o u n d i s
seismic e x c i t a t i o n .
base
timehistory
I I I w h i c h was s u b j e c t e d
The t i m e h i s t o r y
earthquake
vector
is a p p l i c a b l e
sys terns
response o f
demonstrating
coup l e d ,
t h r e e d i mensional
both
response
the
are
indicated.
are
o b t a i n e d as compared
Further
elastic
the present
of
problems
and m a g n i t u d e s
variations
to r e s u i t s
for
for
low,
for
are
of
d e v e l o p e d by
the general
interaction
the s u p e r s t r u c t u r e
the u s e f u l n e s s
comb i n i nq t h e e l a s t i c
structures
investigation
power s t a t i o n ,
structuremedium
interaction
frequences
for
p r e s e n t e d h e r e can be used i n
the modal e q u a t i o n s
results
sugges t s a method o f
in
a nuclear
to a wide c l a s s of
the e q u a t i o n s
in which
Numerical
report
t h e p r i m a r y emphasis
dynamic
modifications,
this
to computer programs
are not
presented.
both higher
the
f o r m u la t i o n deve l o p e d
rob l e m s .
With
o f more
approprate
involved
decoupled.
synthesis
Significant
as we I I
massive s t r u c t u r e s
more c o n v e n t i o n a l
has been p l a c e d upon
the a n a l y s i s
the F o u r i e r
half
finiteelement
method
solving
of
as fundamenta 1 f r e q u e n c i e s
such as n u c l e a r
buildings.
for
modifications
power
stations
221
REFERENCES
[1]
PARMELEE, R., "Building-Foundation Interaction Effects," Journal of the Engineering

Mechanics Division, ASCE, EM2, April 1967, pp. 131-151.
[2]
PARMELEE, R., PERELMAN, D., LEE, S., KEER, L., "Seismic Response of Structure-Foundation
Systems," Journal of the Engineering Mechanics Division, ASCE, Vol. 9, No. EM6,
December I960, pp. 1295-1315.
[3]
LUCO, J., "Dynamic Interaction of a Shear Wall with the Soil," Journal of the Engineering
Mechanics Division, ASCE, EM2, April 1969, pp. 333-36.
[]
SCAVUZZO, R., BAILEY, J., RAFTOPOULOS, D., "Lateral Structural-Foundation Interaction of

Nuclear Power Plants with Large Base Masses," USAEC Contract No. AT-(0-1)-3822, Tech.
Report No. 3., September 1969.
[5]
DUKE, C. M., et al., "Strong Earthquake Motion and Site Conditions: Hollywood," UCLA
Department of Engineering, June 1969.
[6]
TAJIMI, H., Discussion of Ref. 2., Journal of the Engineering Mechanics Division, ASCE,
EM6, December 1967, pp. 29-298.
[7]
LEE, T., WESLEY, D., "Soil-Structure Dynamic Interaction Effects on the Seismic Response
of an Arbitrary Three-Dimens ional Structure," Gulf General Atomic Report GA-10^37,
March 1971.
[8]
SOKOLNIKOFF, I., Tensor Analysis, Theory and Applications to Geometry and Mechanics of
Conti nua, Second Edi t ion, John Wley & Sons, I ne., New York, 1964.
[9]
REISSNER, E., SAGOCI, H., "Forced Torsional Oscillations of an Elastic HalfSpace,"

Journal of Applied Physics, Vol. 15, 9, pp. 652-662.
[10]
SUNG, T., "Vibration in Semi-1nfi ni te Solids Due to Periodic Surface Loading," ASTM
Sympos i urn on Dynamic Testing of Soils, ASTM Special Technical Publication, No. 156,
1953, PP. 35-63.
[II]
ARNOLD, R., BYCROFT, G., WARBUTRON, G., "Forced Vibration of a Body on an Infinite Elas
tic Solid," Journal of Applied Mechanics," Vol. 22, Trans. ASME.,Vol.77, 1955,.391-00.
[12]
BYCROFT, G., "Forced Vibrations of a Rigid Circular Plate on a Semi -1nfi ni te Elastic
Space and on an Elastic Stratum," Philosophical Transactions, Royal Society of London,
Vol. 28, -98, January 1956.
[13]
THOMSON, W., K0B0RI, T., "Dynamic Compliance of Rectangular Foundations on an Elastic

Half-Space," Journal of Applied Mechanics, Vol. 30, Series E, No. h, December 1963.
[]
AWOJOBI, ., GROOTENHUIS, P., "Vibration of Rigid Bodies on Semi-Infinite Elastic Media,"

Proceedings of the Royal Society, London, England, Vol. 287, Series A, 1965, pp. 27-63.
[15]
OIEN, M., "Steady Motion of a Rigid Strip Bonded to an Elastic Half-Space," to appear
in Journal of Applied Mechanics, Paper No. 70-WA/APM-56.
[16]
KARASUDHI, P., KEER, L., LEE, S., "Vibratory Motion of a Body on an Elastic Half-Plane,"
Journal of Applied Mechanics, Vol. 35, December 1968, pp. 697-705.
[17]
AWOJOBI, ., "Approximate Solution of High-Frequency Factor Vibrations of Rigid Bodies

on Elastic Media," Journal of Applied Mechanics, Vol. 38, Series E, No. I, March 1971,
pp. II t 117-
[18]
SCAVUZZO, R., BAILEY, J., RAFTOPOULOS, D., "Lateral Structure teracti on with Seismic
Waves," Journal of Applied Mechanics, Vol. 38, Series E, March 1971, pp. 125-3.
222 -
Flg. 1.
Coordinate systems for theoretical model.
223
Apli tudet
SingleMode
COUVENTI B U LD HC
Modal Cvrplitg Factor ( .05
Poisson's R.tlo . 0
TRAS LITIOKBPCIIIIO
Amplitude of Top nasi
(a)
3.0
It
liti, frequency
10.0 .
(b)
_____
_4i
Hocking ___Mt___
>
(cj
Uet. fr
Total Displacement of B ase Mas

3.0
tatt.
*. cps
frequency
Fig, 2. Response amplitudes for buildings.
(a) Model I
Conventional N-Mass Building
(b) Model II
Unsymmetric System
(c) Model III

Idealized Three-Dimensional
Four-Mode Structure
225
SINGLEMODE PCU
Modal Oanping Factor
.0.
Poisson's Ralo
0
S l r u c i . i r p f r r i i u r n c y
-
Vr - 200
fa)
I i . de (.1 Top M.n,
TRANSLATIONROCKING
(b)
E*ct.
OC i r>g Anpl i LuO
frequency
le)
Base Mass Tran l a i
05
E.ct. frequency
Fig. . Response amplitudes for soft-mounted PCRV.
226
1*|ontockIng
i.l
l u Hjf,. . . ' . . .
r-fl
po?
i\:f\
ui
'><!i\
;/\
:ooo
'.'30
_._;>.
/
''
.<_>'
*" ^ S
" " " i
/
/
Fig. 5.
Response amplitudes for hardmounted PCRV.
227 -
(A)
TOTAL VERTICAL DISPLACEMENT OF APPENDAGE MASS
15
(B)
TOTAL HORIZONTAL DISPLACEMENT OF APPENDAGE MASS
IN PHASE WITH u
(v
- I u )
3 g'
FT/5EC
D -
EXCITATION FREQUENCY (H )
Fig.
6.
Response a m p l i t u d e s
ground m o t i o n .
of
u n s y m m e t r i c PCRV s u b j e c t e d
t o combined
lateral-vertical
- 228
V s - 1500 FT/SEC
THROUGH-SO IL COUPLING
NEGLECTED
MODAL DAMPING
FACTOR FOR
STRUCTURE
. = 0.02
THROUGH-SOIL COUPLING
INCLUDED
't
EXCITATION FREQ UENCY (H )
Fig.
7.
Influence o f
s tructure.
through-soil
c o u p l i n g on the s t e a d y - s t a t e seismic
response of
reactor
*iO~3
4
'^P \ v _v A /
^ V
^
East-West Translation
/v/
VU
NorthSouth Rocking
X CT
8-
<C7
4-
8h
---'-/\/
- O-
U-8 -
Vertical Translation
Torsion
xO
8
1
r-
_ 4
O
4
North-South Translation
TIME .SECOND
j,
EastWest Rocking
"S
3
4
TIME, SECOND
- 230
MAX. 0.3509
Fig. 9.
Comparison between base a c c e l e r a t i o n and f r e e - f i e l d
excitation.
- 231 -
vs= Fps
OB
0.4
0
-0.4
-0.8
j -
/
J W
/
1/ V
/ \ / ^"^
~. 0.908g
VV =2000 FPS
S
MAX. 0.604 g
V s = 1000 FPS
TIMF., SECOND
F l g . IO . I n t e r a c t i o n e f f e c t s on seismic response of appendage mass.
- 23:
DISCUSSION
A. HADJIAN, U. S. .
Contrary to Fig. 5 and m o r e in line with Tajimi's thinking, Bielek's (1971) Ph. D.
Thesis at Caltech shows that the 2nd mode of a two m a s s model of a containment s t r u c t u r e
does not show any variation of the frequency as compared to a fixed based model. This was
p a r a m e t r i c a l l y shown to be true for a wide variation of soil p r o p e r t i e s .
T. H. LEE, U. S. A.
The interaction effects on higher modes, in my opinion, will depend on the system
p a r a m e t e r s , at this stage I don't know what c a s e Bielek has studied and what approach he has
used. I will be glad to have a copy of his t h e s i s for further studies.
H. SATO, Japan
The transfer function used in the equation of F o u r i e r inversion would have i n t e r
action t e r m between each ground component. I understand that these can be theoretically
estimated.
Did you m e a s u r e the interaction t e r m which is supposed to appear in transfer function of
multi-degree-of-freedom system ?
What sort of method of calculation did you use to obtain the F o u r i e r spectrum for the analysis ?
T. H. LEE, U. S. A.
The m e a s u r e d data for the Hollywood Storage Building, Los Angeles (Kern Country
Earthquake, 1952) a r e for the horizontal translation only. To compare with these m e a s u r e d
data, I only considered the single horizontal displacement of ground motion in my computer
analysis. I agree with you on the point that the actual building response may be affected by
a multiple-component ground motion during that earthquake. However, I did not make any
attempt to incorporate other components in that particular comparison study.
K. UCHIDA, Japan
I think the influence of soil interaction of higher mode may depend upon both the
type of vibration model and the soil stiffness used in the model. For example in case of
complicated vibration model like nuclear power plants composed of various vibration elements,
the natural modes with respect to soil motion may often be computed in the higher modes as
you mentioned in your paper. So in these c a s e s the higher modes may be important. However
in case of the high r i s e buildings, tall chimneys like cantilever type, I consider only the first
few modes may be important for the response calculation.
What do you think of my opinion ?
233
T. H. LEE, U. S. A.
Prof. Tajimi's opinion was on the analysis of a cantilever-type structure. It is

quite possible that for this particular type of system, interaction effects on the higher modes
are truly negligible. The consideration of the first few modes for analysis of high-rise build
ings appears to be justifiable.
K. AKINO, Japan
With respect to flexible structures, especially tall buildings, we have experimen
tal data in Japan for their natural periods. If we suppose that the first period is unity, the
second is approximately 1/3 and the third is approximately 1/5, and those numbers correspond
to natural periods of shear mode vibration modes of a cantilever beam. Therefore, it can be
said that vibration modes of tall buildings are independent of the soil interaction. How do you
think whether Japanese have to reconsider the influence of structure-ground interaction
upon the response of higher modes, as you pointed out in your paper, (the second paragraph
from the bottom in page 9)? However, we agree with your opinion for rigid structures such
as the reactor buildings. (This question is turned to me from Dr. Tajimi).
T. H. LEE, U. S. A.
The interaction effects depend, in a complicated way, on the parameters in the

system. Naturally, they will be influenced by the inertia property of the structure. The ex
perimental data in Japan are for cantilever-type systems whose m a s s e s are relatively small
in magnitude as compared to those used in a reactor model. I must emphasize here that the
interaction effects on higher modes can be significant in unsymmetric three-dimensional
Systems. Where the response of structural m a s s e s depends upon the rotations as well as
translations of the system. The mode shapes of such unsymmetric systems differ appreciably
from those used by Prof. Tajimi in his paper.
As I remember, I have never given any opinion for rigid structures because the model de
veloped at Gulf General Atomic is for flexible structures. This must be a misunderstanding.
K 3/6
DYNAMIC CALCULATIONS USING A FRAMEWORK ANALOGY

TO PREDICT THE SEISMIC RESPONSE
OF A NUCLEAR REACTOR
D.A. JOBSON,
United Kingdom Atomic Energy Authority,
Reactor Group, Risley, Warrington, United Kingdom
ABSTRACT
The feasibility of predicting the response of a nuclear reactor system to a spectrally

defined earthquake is established in the context of a particular example.
The associated
dynamic calculations were carried out on FRAMES, which is a UXAEA Reactor Group program for
the vibration analysis of skeletal structures.
Although the pipework and framed structures
could be represented quite directly in this way, further work was needed to simulate other
components, such as the reactor vessel and core. The use of grid-framework methods in this
connection is described, as is their application to the modelling of the ground strata supporting the reactor.
The complexity of the combined system led to the use of physical par-
titioning, and full advantage was taken of a number of rationalisations which this technique
suggested.
The analysis demonstrated that soil/structure interaction introduced significant extra
degrees of freedom.
The associated modal shapes involved simultaneous sideways movement and
rocking of the raft, at frequencies which depended primarily on the values used for the soil
moduli, thus emphasising the need for reliable data on the latter.
1.
INTRODUCTION
The work described was aimed primarily at an assessment of seismic effects on a nuclear
power station, with particular reference to the loads imposed on the core.
Extensive use was
made, in this assessment, of a general-purpose computer program (Jobson and Litherland [1])
which enables vibrating shapes and frequencies to be computed for skeletal structures. The
need for a modal analysis arose from the form in which the earthquake data were presented.
The latter were given in spectral terms (see Hudson [2]) and took the form of Housner-type
response curves, normalised to a ground acceleration of 7% of g.
In the present study, attention was focussed on those features of the system which could
influence the seismic response of the graphite core stack (see Figs 1 and 2) and these were
judged to be:
(i)
(ii)
Soil/structure interaction.
Response of the diagrid and its support system, including the main vessel, ducts
and heat exchangers.
236
(iii)
Response of the core, including the support plates and the steel restraint
structure.
2.
SOIL/STRU
C TURE MODES
The ground strata consisted of silty sand and clay, resting on soft rock.
was a thin layer of sand and then clay to a very great depth;
Beneath this
the reactor raft was founded
on the rock.
The application of spectral analysis first requires that the relevant natural frequencies
and modal shapes of the system be found.
If soil/structure interaction is ignored, there is
no ambiguity as to what is meant by the "system".
A relatively flexible building founded on
solid bed rock extends only as far as the latter. The above method is obviously inadequate
for situations where the raft and the biological shield are both massive and stiff, whereas
the ground is relatively deformable. The boundaries of the "stress bulb" on which the reactor
sits were based on the deadweight stresses Induced in the ground.
These limits were chosen
as the line beyond which the vertical stress felt by the soil was less than 10 of the mean
vertical stress under the foundation and indicated that most of the enclosed soil volume was
clay.
The above boundaries served as a conceptual convenience to limit the extent of the
system for which modal analysis was required, see Figs'3 and ,,
They were supported in each
reference plane by attaching horizontal and vertical springs to its boundary nodes. The
spring stiffnesses were derived from finite element models of the ground lying beyond the
boundaries considered.
The inertia of the enclosed ground was modelled by a lumping procedure
and the stiffnesses of each ground stratum were represented by equivalent elastic frameworks,
(see Jobson [3]).
clay.
The most important of the soil parameters was the Young's modulus for the
The considerable uncertainties in ascribing an effective value to the latter makes
that used (4300 p.s.i.) a provisional figure which is subject to review.
It was derived from
fast undrained triaxial compression tests, after making allowances for cyclic loading to a
peak strain of 01%.
The completed work showed that this was not significantly exceeded.
The inertias of the foundation raft and the structures on it were similarly represented by
lumped masses and the associated rigid body movements were found to be significantly larger
than those induced by distortion of the concrete for the modal shapes plotted in Figs 5 and
6.
Only relatively few of the many modal shapes found had any significant effect on lateral
movement of the reactor.
Strictly speaking, a complete computation of the modal shape at
each of these frequencies required a modelling of the steel structures on the raft, together
with the core.
It was argued however that the dynamic effect of the latter in the 01 Hz
range would have been negligible so far as the system considered was concerned.
It followed
that the modal shape at these frequencies could be "completed" by inferring the diagrid and
core movements subsequently.
in the 01 Hz range.
The amplitudes of the raft followed from the spectral curves
The associated spectral displacements (u0) apply directly only to the
motion of a single mass on a spring, see for example Biggs [].
More generally each needs
multiplying by the appropriate participation factor (). For an earthquake oriented in the
and y directions respectively, takes the form:
237
"*
+
V^x*
2) '
=
_
_________________
mi"- 22 i
SV m(<
Px + V
2 +
2\
'
The summation is taken over all the r masses, each of which has component displacements
, tri and
respectively in that mode, relative to the reference axes Ox, Oy and Oz.
Neglect of the relatively small masses of the steel and graphite features of the system in
this calculation involved negligible error.
The maximum displacements of the raft were
obtained from u 0 and the relevant , together with the modal amplitude (q_) of the component
considered.
The spectral amplitudes (u0 _) of the raft so obtained are given in Table I.
The tabulated values are spectral and not absolute displacements.
Interpreted physically,
they are the peak displacements of the raft, as seen from moving ground that is substantially
uninfluenced by the presence of the reactor.
As the methods described above are novel, they
were compared with a conventional analysis of soil-structure interaction, using the method of
Whitman [5]. Although there was a remarkably good agreement of frequencies, the modal shapes,
see Figs 5 and 6, are much more complicated than could be derived from the Whitman analysis
alone, and hence the responses are different.
3.
DIAGRID, VESSEL AND DUCT SYSTEM

The 2A support legs for the spherical steel vessel are equally pitched round a circle
which is approximately 4-5 ft in diameter.
Their ends are so cradled that no bending is
induced in them by radial expansion of the sphere. An equivalent system of six uniform
pillars was devised, having the appropriate component stiffnesses. Similar simplifications
v/ere made for the pillars supporting the diagrid which was modelled structurally by reference
to a smeared analysis.
The enclosing pressure vessel was simulated by a lattice structure, based on an equa
torial ring and a parallel set of six horizontal small circles, together with a perpendicular
set of six meridians.
They were rigidly jointed at their intersections and a curvilinear
rectangular element of the shell was notionally associated with each of these nodes, see
Fig. 7.
By suitably lumping the membrane, bending and torsional stiffnesses of each element
so defined, an equivalent skeletal grid-framework (Hrennikoff [6] and Jobson [7,8]) was
derived.
Corresponding values were similarly determined for the members forming the equi
lateral lattice in plan round each of the poles and these were further adjusted to take
account of the fairing lying within the bottom of the vessel.
The inertia of the vessel, as
well as that of the diagrid and support system, was represented by a set of lumped masses,
distributed as symmetrically as possible.
Associated problems of core storage and machine
time required that the number of inertial degrees of freedom in each sub-system be limited
to 78.
It was therefore decided to neglect the obliquity of the side-ducts in plan and
define equivalent top and bottom ducts, disposed on each side of the vessel, along the major
axis of the foundation raft.
The restrictions on flexibilities were far less severe and each
length of heat exchanger vessel and duct was represented by an appropriate prismatic member.
The component stiffnesses of each bellows was assessed and combined with those of the tongues
lying within them.
The guides which prevent lateral motion of the ducting presented no
difficulties since the program allowed imposition of appropriate component restraints.

The lowest natural frequency of this combined system was found at 199 Hz and the next
critical was found at 302 Hz.
Neither of these modes induced any substantial diagrid
- 238 movement however, such as that seen at the next critical (3*39 Hz, see Fig. 8). Although
there were uncertainties about the behaviour of the heat exchangers, subsequent computer runs
showed that the core movements were not in general sensitive to their response.
Only an
unlikely synchronism between the natural frequency of a heat exchanger and one of the modes
of the main system could substantially modify the top duct movement and its interaction with
the main modes of the vessel.
The dynamic modelling and characteristics of the core, together
with its associated restraint structure will be considered in the next section.
4.
GRAPHITE CORE STACK AND RESTRAINT STRUCTURE

Figure 9 shows the graphite core construction in which, at the assembly stage, each
brick is separated in plan from its neighbour by a small gap (exaggerated for clarity in the
figure), the individual columns of bricks being stabilized by keys to those surrounding it.
A layered model was used to simulate this structure, each node of the latter being
supported vertically by columns having a stiffness derived from tilting tests on a model core
stack.
The octagonal bricks are arrayed on a square pitch with clearance between each, and
an interlinking system of mutually perpendicular keys.
The further interstitial bricks,
which are loosely fitted between them and are similarly interlinked by sliding keys, thus
give rise to what is equivalent to a cross-braced lattice pattern.
Reference has already
been made to the studies which have been made of lattice systems of this kind.
terms, if = [,
tice and =
{e , ty,
xv}
In smeared
} is a column vector of stress components in the plane of the lat

similarly denotes the corresponding engineering strains as con
ventionally defined, then to a first order:

C e
(2)
For each graphite layer it was found that:
C =
Cl I
Cl 2
C21
C22
Csi
C32
Cl 3
-G
C23
-G
C33
Ci)
which is unusual in that it corresponds to an isotropic lattice that has been so designed as
not to resist all-round tension.
Its elastic constant (G) was inferred from tests on keyed
connections and the flexural characteristics of each lattice member were so chosen as to
reproduce C_.
The enclosing restraint structure is plated towards its base and is diagonally braced
towards the top.
A rationalised system having 12 sides was devised and the section constants
for the plated portion were so chosen as to reproduce the elastic constants of the material
it replaced, again by the use of grid-framework methods.
Advantage was taken of symmetry and
only one half of the core and the restraint structure was modelled.
The mass was distributed
over four layers and the response to harmonic excitation at the level of the diagrid obtained
for the frequency range 0-6 Hz at intervals of 025 Hz.
The otherwise "free-free" system was
located positively by dummy springs which were sufficiently light as to introduce negligible
errors above 01 Hz.
These runs demonstrated that, in the range 0-1 Hz, rigid body displace
ments were much larger than any distortion of the core shape in response to either unit force
- 239 or to unit moment.
It was concluded that in this frequency range, the interactive effect, or
receptance of the core on the diagrid would be substantially the same as that for a rigid
body.
An equivalent dynamical system was thus devised, consisting of two lumped masses, lying
on the centre-line of the core.

The forced responses further showed that there was negligible straining of the crosssection of the core due to excitation by alternating forces and moments at the base of the
core in the range 0-5 Hz. Core distortion could therefore be represented in this frequency
range simply by flexing of its centre-line and led to the use of a flexible "dumb-bell"
for computations in this range.
Such a simple model could not adequately represent the core
at higher frequencies, where its behaviour became increasingly complex.
In the 3-5 Hz range
it was difficult to decide whether a rudimentary modelling of the core, with a fairly complete
representation of the vessel and associated circuits, would yield better results than a
sophisticated modelling of the supported core, with a crude modelling of the vessel/duct and
heat exchanger system.
It was judged that the latter would certainly be more appropriate
above 5 Hz and it was finally decided to analyse the 3-5 Hz range by a combination of both
models.
5.
RESPONSE OF CORE/VESSEL COMPLEX

Modelling of the vessel/duct and heat exchanger system, together with the definition of
a rigid dumb-bell to represent the core enabled the modal profiles of the complete system to
be inferred at each of the soil/structure natural frequencies given in Table I.
The asso
ciated core accelerations were found from the known spectral movements of the foundation raft
about its virtual pivot position for each mode, and it was found that deformation of the sup
port pillars increased the core movements, see Table II. The natural frequencies and modal
shapes of the vessel/duct system were also found for the above model in the 1-5 Hz range using
the flexible dumb-bell representation of the core. The presence of the latter reduced the
frequency at which substantial diagrid movement was felt from 339 Hz down to 295 Hz.
Further frequencies were found, again involving diagrid movement parallel to the longitudinal
axis of the raft at 3*61 Hz and 4-3 Hz. Only one significant mode was found in the 1-5 Hz
range for diagrid oscillation parallel to the transverse axis, which was attributed to the
additional lateral restraints on the ducts.
Ine corresponding frequency was 3*77 Hz and in
all of these cases the diagrid movement was substantially horizontal.
The associated profiles
of the core stack at each natural frequency were deduced from its narmonic responses to
excitation of the base.
It is possible to repeat the above process by computing the actual receptances of the
distorting core at each natural frequency and correcting the receptances of the flexible
dumb-bell in a subsequent recalculation.
There are, however, indications to suggest that the
modes of the gas circuit are not sensitive to the receptance of the core in this frequency
range and the modal shapes of the vessel/core complex were obtained without iteration.
They
were used to compute the corresponding participation factors and nence the spectral accelera
tions, see Table III. These results are the same as would have been obtained had soil/struc
ture interaction effects been ignored completely and this is a point whicn will be recon
sidered subsequently.
240
In an alternative representation, particularly aimed at computation of the higher modes
of the core (above 5 Hz), the core and restraint structure were modelled in full detail.
This allowed only a crude simulation of the vessel and attached ducts. Lumped masses were
used to represent each of the latter, based on the total mass of the vessel, plus a nominal
allowance for the attached ducts.
It followed that such a model would give the same fundamental frequency in each of the
two principal planes, and this came out at 3*09 Hz. It corresponds to 295 Hz (longitudinal)
and 377 Hz (transverse) obtained from the previous model. A further limitation was that no
other frequencies were found in this range to correspond to the different ways in which the
ducts could participate in core motion.
Further significant core modes were found at 5*9 Hz and 133 Hz, see Fig. 10(a), (b) and
(c). C omputation of the modal participation factors, again ignoring any reactive effect on
the foundations, led to the results given in Table IV. Negligible distortion of the cross
section of the core was found for the frequency range covered by Tables II, III and IV. This
was attributed not only to the effectiveness of the lattice layers in preserving their shape,
when account is also taken of the stability of the graphite columns and of the restraint
structure, but also to the absence of any excitation of the 'breathing' modes of the core
found at b'U, 9 and 12 Hz (see Fig. 11(a), (b) and (c)).
6.
REA
C TIVE MOTION DUE TO SOIL/STRUCTURE INTERAC TION
The analysis presented so far suggests that soil/structure interaction had two effects:
(i) It introduced additional lowfrequency modes which did not interact strongly with
those of the reactor structure because of the low value assumed for the soil
modulus.
(ii) By obtaining the structural modes of the system with the implied assumption that
the foundation and the soil did not participate in the modal motions, these fre
quencies and amplitudes came out the same as if soil/structure interaction effects
had been ignored entirely.
It is the second assumption that leads to results which are in conflict with current
views about soil/structure effects. The latter are known not to be wholly deleterious,
particularly for heavy stiff structures [9]. This can be illustrated by reference to a very
simple model in which the mass mi representing the foundation is much larger than the
"structural" portion of the system. The spring constants are such that the natural frequency
of the large mass is very much lower than that of the smaller one. With the above provisos,
the frequencies of the coupled system are virtually the same as those of each mass on its own
spring i.e. u_ = /(ki/mi) and m_ = /(k2/m2). The modal participation factor for the first of
these two frequencies is virtually:
'a
_____ _ miXi
lX* m,X,*
_1_
X,
since at this frequency Xi = X and mi 2. The corresponding spectral displacements of

mi and 1TI2 are thus substantially equal at the lower (soil/structure) frequency:
i,\
'
241
di = dj = - u. = u
a
a
(5)
The modal participations for the higher (structural) frequency require however that
account be taken not only of the relatively large movement of the small structural mass m 2 at
this frequency, but also the relatively small movement of the much larger foundation mass mi.
An approximate analysis shows Xi to be very nearly, XjiT^/mi. It follows that the modal
participation factor at the higher frequency is virtually zero since IhX = 0. Physically it
implies that the higher mode is such that the masscentre of the combined system remains
virtually stationary. These conditions appear to be well satisfied for the reactor considered
and require that the participation factors for the structural modes be reduced by taking into
account the reactive effect on the foundation. Using m, and , to denote the 'structural
and "foundation" masses respectively, the modified modal participation factor corresponding
to is:
+
_
~ /2 + 2
'
2
'
Since the effect on the denominator of is virtually negligible. It follows

that the Reactive Factor (r) to be applied to is approximately:
'
3 +
" ~ "
Zm^
= 1 +
ap"
^7'
Similarly:
'
r = -?-
ry
1 + =
antpv
(8)
The second term in each expression comes out negative, as a result of the reactive shears and
moments exerted by the structure on the foundation. The values of r have been found for each
of the structural modes in this way and their effect is to reduce all predicted accelerations
for the high frequency modes (above 1 Hz) to which they apply. It should be noted that they
have no relevance to the soil/structure interaction modes. The diminished accelerations on
the core support are given in Table V, and may be compared with the undiminished values given
in Tables II, III and IV.
7.
SDMMARY AND CONCLUSIONS
A seismic analysis of the vital components of a nuclear pov/er station has been described
in which the primary aim was to assess the loads imposed on the core restraint structure. It
followed that the motion of any components which interacted with that of the core needed to
be included. These were judged to be the core diagrid, its supporting pillars, together with
those for the vessel and the vessel itself. It was found that allowance also needed to be
made for the attached ducts, together with the heat exchangers to which they are connected.
Seismic excitation of such systems must essentially be limited to displacements 'which can be
accommodated by a substantially recoverable elastic response and the object of the study was
to provide data to enable this to be checked.
242
Extensiv? use was made of FRAMES, which is a powerful generalpurpose program for the
dynamic analysis of skeletal structures.
This was deployed in conjunction with gridframework
methods, which enable elastic continua to be represented by equivalent lattices.

Modal analyses were carried out by means of an eigenvalue/eigenvector subroutine, which
vas based on the use of a matrix deflation method that determined each frequency in turn from
the lowest value upwards.
Dynamic interaction between subsystems was accounted for by
noddling the dynamic flexibilities of the attached system at each interface.

Deformable ground implied that the natural modes of the raft/structure needed to include
that part of the subsoil which participated significantly in the motion.
In assessing the
latter account also had to be taken of the fact that the ground properties depended on the
associated bearing pressure.
Definition of the boundary for modal analysis is to some extent
arbitrary but is net judged to be critical.
The various layers of ground within the support
ing stress bulb were represented by an equivalent elastic framework and its distributed mar.'
was lumped on the nodes of the latter.
The boundary of the framework war anchored by equiva
lent springs having stiffnesses based on a finite element modelling o
the surrounding ground.
Subsequent analysis of the forced response of the structural features mounted on the raft
showed that the sideways movement was amplified so far as the core was concerned by associated
deformation of its support system.
One further significant effect resulted from including soil/structure interaction.
Vibration of the structural features at each relevant natural frequency gave rise to reactivo
forces and moments on the foundation and ground.
These, in turn, caused small displacement.
j{ the latter 'which were, nevertheless, significant in that the masses involved were rela
tively large. C onsequently the participation factor for each of these modes due to excitt'UMI
from the surrounding ground was reduced.
As a result, the spectral accelerations of the
higher frequency structural modes were less than would have been predi'.tod had soil/structure
interaction effects been ignored.
8.
C
A KNOWLEDGEMENTS
The author wishes to acknowledge the contributions made by his colleagues to the work
described, and to thank the Managing Director of the Reactor Group, , for permission to
publish this paper.
243 -
REFE REICHS
[1] JOlSOll, D. A. and LITHERLAND, J. R., "vibration analysis by computer: a user's guide tc
programs for the natural and forced oscillations of skeletal structures , TRG Report
1919(R), (1969)
[2] HUDSON, D. E., 'Response spectrum techniques in engineering seismology , Proceedings c;"
1956 World Conference. Earthquake Engineering, Earthquake Engineering Research Institut'-,
1956
[3] JOBSON, D. ., "Lattice analogies for plane elastic problems". TRG Report 1339(R),
Part 2, (1966)
[4] BIGGS, J. VI., "introduction to structural dynamics". McGraw Hill, New York, (1969)
[5] WHITMAN, R. V., "Seismic design for nuclear power plants". MIT Press, (1970)
[6J
HRENNIK OFF, A,, "Sulution of problems of elasticity by the framewerk method", J. App.
Medi.. ASME, pp A169-175, December 191
[7J
JOBSON, D. ., "Grid Analogies for the elastic bending of plates". TRG Report 13<VO(R),
Part 2, (1967)
[8] JOBSON, D. ., "The representation of elastic solids by space lattices". TRG Report
V J U ( R ) , (1967)
[9] "Nuclear reactors and earthquakes". USAEC Division of Technical Information, 0-72,
(1963).
- 244 TABLE I
RESPONSE OF FOUNDATION
Maximum spectral
displacement of top
of slab/(in.)
Frequency
/(Hz)
Mode
Maximum
rotation
/(rdn)
Longitudinal
direction
2
0-46
1.75
0-001C
050
015
0-0006
076
0-10
0-88
0-01
0-97
0-01
0.41
098
0-0011
057
0-27
0-0001
0-88
005
LOA
0-03
0-0001
Transverse
direction
TABLE II
LOW FREQUENCY RESPONSE OF CORE
Frequency
/(Hz)
Spectral displacement
of core/(in.)
Lateral acceleration
of core/(g)
Base
Top
Base
Top
Longitudinal
axis
0-461
2-033
2-406
0.0441
0-0522
Transverse
axis
0-409
0-571
1-272
0-248
1-666
0.223
0-0217
0-0083
0-0284
0-0074
24 5
TABLE
III
INTERMEDIATE FREQUENC Y RANGE UNDIMINISHED RESPONSE
Lateral acceleration of core/(g)

Mode
Frequency/(Hz)
base
top
Longitudinal axis
295
361
403
00118
00015
00058
00688
00102
00162
Transverse axis
377
00012
00065
TABLE IV
HIGH FREQUENC Y RANGE UNDIMINISHED RESPONSE
Lateral acceleration of core/(g)

Mode
Frequency/(Hz)
Longitudinal and
transverse
590
1330
base
top
0050
0043
0052
00026
TABLE V
DIMINISHED SPEC TRAL RESPONSE
Lateral acceler tion of core/(g)

Frequency/(Hz)
r(= '/)
base
top
00441
00002
0
00003
0
0002
00522
00012
0
00007
0
00001
00441
00522
00217
00083
00002
00032
00008
00284
00074
00001
00033
00001
00235
00296
0050
0060
Longitudinal axis
0461
2953
3606
4025
590
1330
10
0017
0006
0043
0001
0040
Root, square value (L)

Transverse axis
10
1 0
0190
0064
0018
0409
0571
3771
590
1330
Root square value (T)
2
Resultant /(L' + T )
246
GOAPHiTt
PRE S S U R E
CORE
VESSF L
SECONDARY
BIOLOGICA
SHIELD
HFAT
EXCHANGERS
EXPANSION
OUTLET
DUCT
JO.NTS
PRIMARY
BIOLOGICAL
SHIELD
J.
DIAGRID
GRID
HEAT
CORE
BIOLOGICAL
DUCT
t EET
PRESSURE
INLET
SHIELD
FIGURE 2
PADS
The- Reactor B u i l d i n g n o v a t i o n
EXCHANGERS
PRIMARY
INLET
SUPPORT
FI GURT 1
GRAPHITE
The Reactor Building Man
VESSEL
DUCTS
24 7
I.fyITrTKAL
FIGURI 3
SECTION
L o n g i t u d i n a l S e c t i o n of S u p p o r t i n g Ground
24 8
" V -" 5P,
<^ Q^T
FIGURE -* C ross Section of Supporting Ground
^r *^
94
Sb
(kA
ritr f^.
/
~i
\
\
trr
u Ih
"\
iyjf2QE.2_
0_K>J_ ^.._
:-
LOUGITUDIU&.L
5E
C TIOU
LOUGITUDIrJ^L
Longitudinal Modes due to SoilStructure Interaction
2.
ECTiOJ
l i .
.'.
1
'

IM
11
1
1
<J
/
bl
i'
4
73
/ (
(74
ta
bl
bS
%t
52,
.^
4o
4)
12.
*7
32
'9
20
24
1
Sn
ti
39
I
<L4
3^
Moot i
MtSc.
IO
11
MODE. Z.
* V
TS W
^b
2
CB.OSS
CB.OSS
5E
C TIQU
FIGURE 6
T r a n s v e r s e Modes due to S o i l S t r u c t u r e
Interaction
5E
C TOU
251 -
FIGURE 7
Lattice Representation of Sphere
to
CJI
IO
FIGURE 8
.Vcdal Shape c f Vessel/Duct S y s t e - a t 3-39 Hz
253
FIGURE 9
- Graphite Core Lattice
"/
M
" *
,ui
ri
\ \
'\Vs.V?\
/ , ',,
's
r ^ T ^
10
(a) 3 C") Hi
(c) 13 3 Hi
iIG'JRE 10
A s y r r m e t r i c C o r e Modes
/
IO
t*
1
1
.o
'
II
<
4ft
it
14
fel
tl
/
/
v 1\
"X\\
<-.
-* V '
\^sS^^ "
1 1
V
\
1I
1)
11
sV.
" ,"'
_" ' _
,.
'X
I
(o) 6 17 Hi
(c) 142 Hi
FIGURE 11
Symmetric Core Modes
256
DISCUSSION
K. AKINO, Japan
We c a r r i e d out vibration test for the graphite shielding s t r u c t u r e , and lest results
show us that the graphite pile-up s t r u c t u r e does not have eigenvalues.

In your paper, you proposed the tramework analogy including the graphite core structure nul
you stated natural period of the s t r u c t u r e to be t> cps. Do you have any experimental evidence
concerning that figure ?
.
W. T. LAW TON, U. K.
The graphite core was simulated by a 3-dimensional a r r a y ot damped masses

and springs; the stiffnesses in the horizontal plane were calculated to reflect the freedom to
particular horizontal movements permitted by the keyed construction, while the vertical
springs were determined from experimental m e a s u r e m e n t of the lateral stiffness of a column
of graphite bricks.
Measurements have been made of the response of a section of the cure when excited on
vibrating table and clear evidence of eigenvalues observed. Checking of these natural frequen
cies against calculation is not yet complete.
K 3/7
PARAMETRIC ANALYSIS OF SOIL-STRUCTURE INTERACTION

FOR A REACTOR BUILDING
R.V. WHITMAN, J.T. CHRISTIAN, J.M. BIGGS,
Massachusetts Institute of Technology, Cambridge, Massachusetts, U.S.A.
ABSTRACT
A reinforced concrete reactor building to be located over deep soil

deposits is analyzed to determine the effects of varying flexibility of
the soil and variable damping. The soil structure interaction is
shown to be significant, the stiffest foundation giving the most critical results for all responses except those of the base itself. The
interaction of the first three structural modes and the soil is complicated, but in general the first mode dominates the outer structure, and
the second and third modes dominate the inner structures. When damping
is allowed to vary among the components of the system, the results
indicate that the conventional approach of constant damping in all modes
is conservative.
INTRODUCTION-PHYSICAL PROBLEM
The reactor building discussed here is to be located on the flood
plain of a major river. Beneath the structure the soils are predominantly
sands and gravels, with occasional thin strata of clay. Because of the
siting conditions, soil-structure interaction had a particularly important
effect upon the dynamic response of the building during possible earthquakes, and hence a detailed study was made of the interaction effects.
The results of the study were used as a basis for choosing the dynamic forces eventually used for the design of this building. They also may be
used to guide future analysis of other similar buildings.
The structure itself (see Fig. 1) includes an exterior cylindrical
building, a primary containment or drywell consisting of a shell that is
cylindrical in the lower portion and conical in the upper, and the reactor
vessel, which is supported on a pedestal within the primary containment.
The roof of the exterior building is a steel framed dome. The primary
containment is completely independent of the outer building, except at the
base mat. The vessel and the pedestal are connected to the containment only
- 258 by stabilizer springs and by a drywell floor seal, which is treated as a

spring in the analysis. The entire edifice is supported on a circular base
mat. The structure is of reinforced concrete, except for the reactor vessel
and its biological shield.
The ground response spectrum for the design basis earthquake is shown
in Fig. 2. This is based on a peak ground acceleration of 0.2 g. Only
horizontal ground motions are considered in this paper.
MATHEMATICAL MODEL
The building together with its contents is treated as one integrated
dynamic system. The dynamic model is composed of 60 nodes, at which mass
is lumped, and these are connected by structural segments and springs as
illustrated in Fig. 3.
The soil and foundation is accounted for by translation and rocking
springs at node 60. The rotation of all horizontal planes is assumed to be
identical because vertical deformations in the walls of the exterior building and the primary containment are negligible. The circular base mat is
assumed infinitely rigid; the mass of the mat and an effective mass of soil
are lumped at node 59. The model was made axially symmetric about the
vertical axis, and hence no rotation about this axis occurs.
Because some of the masses are rigidly connected, there are 54 independent nodes. Thus, there are 5 5 degrees of freedom, including the rotation of the entire system. The modal equations for the system were developed and solved by Rayleigh's method as described by Biggs (1964) and Pahl
(1970) . The root mean square sum of up to eight modal contributions was
used to evaluate the response of the structure and foundation system.
SOIL SPRING PARAMETERS
From the theory summarized by Whitman (1970a, 1970b), a possible extreme
range of values for the rocking and shear spring stiffnesses was selected
as shown in Table 1. In the analyses three foundation stiffnesses were considered: hard, medium and soft. If different damping ratios were to be
assigned to different portions of the system, a damping ratio of 5% would be
associated with the rocking spring and 25% with the horizontal spring. A
comparable ratio for the structure alone would be 4%. If a single damping
ratio were to be used for the entire system, 7% was recommended. This latter
ratio was in fact used for the first analysis reported below.
BEHAVIOR FOR RIGID FOUNDATION
The response of the structure for the case of a totally rigid foundation shows its behavior without effects of soil-structure interaction. Such
an analysis helrjs to isolate those parts of the subsequent analyses that are
- 259 due to foundation effects alone. An analysis was carried out for the case
of a rigid foundation by setting both foundation springs to large values.
The frequencies of the first five modes of the structure on a rigid
foundation are listed in Table 2. Fig. 4 shows the shapes of the first
three modes.
The first mode involves the outer cylinder only. The second and third
modes involve response of all the structure except the cylinder. In the
second mode all parts of the structure are in phase, but the reactor internals (mass 14) are out of phase in the third mode.
When the damping ratio is 7* in all modes, the responses tabulated in
Table 3 occur. It is clear that the most important modes are the first and
second. The cylinder responds most strongly in the first mode, and the rest
of the structure responds in the second and third modes. This is predictable from the mode shapes.
The spectral accelerations and participation factors are also shown in
Table 2. Because of the shape of the response spectrum, the first three
modes have the same spectral accelerations of 0.38 g. The higher modes
have decreasing participation factors and decreasing spectral accelerations.
They would, therefore, be expected to contribute less to the response than do
the first three modes.
It can be concluded that the response of the structure alone is characterized primarily by the first three modes, the first two being most important. The first mode dominates the response of the cylinder; the second
and third dominate the response of the rest of the structure. All three
have identical spectral accelerations.
EFFECTS OF FOUNDATION FLEXIBILITY
The effects of foundation flexibility can best be understood by first
considering the response of a perfectly rigid structure resting on the
flexible foundation. The total weight of the structural system, including
9
soil, is 194,000 Kips, and the rotational inertia about the base is 2.48x10
Kip-ft. These figures together with the spring constants of Table 2 lead
to frequencies of Table 4. The distribution of energy between rocking and
swaying shows that the first mode involves primarily rocking and the second primarily swaying for the whole range of possible foundation conditions.
The distinction is sharpest for the soft foundation.
All of the frequencies
of the structure on a rigid
increases, the frequency f,
of soil-structure exclusive
as the foundation stiffens.
of Table 4 are lower than that of the second mode

foundation, but, as the foundation stiffness
approaches that of the second mode. Thus, effects
of the cylinder could be expected to increase
A second observation from these results is that the fundamental frequency of the structure alone is very close to that of the swaying spring
alone for a hard foundation. Increased interaction could be expected in
260
that case.
A prediction of the fundamental frequency of the combined system can be
made by the DunkerleySouthwell approximation (Jacobsen and Ayre, 1958):
f >- f
RF

f
RS
where f and fR_ are the fundamental frequencies of the cases with a rigid
foundation and with a rigid structure. Application of this rule to the re
sults of Table 4 gives values of 1.18, 0.92, and 0.706 Hz for the cases of
hard, medium, and soft foundations, respectively.
RESPONSE OF COMBINED STRUCTURE AND FOUNDATION
The response of the combined structure and soil system with 7% damping
in each mode was calculated with the computer model with the results in
dicated in Tables 5 through 9. In addition to the runs summarized in
these Tables, computer runs were made in which the stiffness of the drywell
structure and of the floor spring were changed. The foundation stiffness
was again varied from hard to medium to soft. The effect of foundation
stiffness on response was substantially just as in the basic cases that
are described here in detail, although there was some change in the patterns
of modal dominance.
Examination of the information in Table 5 shows the following points:
1. The response of the first mode and its frequency depend primarily
on the foundation springs. The participation factor of the first
mode is nearly independent of foundation type. However, the change
in fundamental frequency of the first mode as foundation stiffness
decreases causes a decrease in spectral acceleration. This occurs
because of the shape of the response spectrum in this range of
frequencies. The frequencies predicted by Dunkerley's rule are
close to those of the first mode.
2.
Foundation stiffness has a marked effect on the frequencies of

mode 2. The frequency of the fourth mode is virtually independent
of all changes used in the analysis. This suggests that this mode
primarily involves internals, as is the case.
3. Because of the location of the range of frequencies for the second,
third, and fourth modes on the response spectrum, the spectral
acceleration of all these modes is a constant 0.38 g., as for the
first three modes of the rigid foundation case.
4. A decrease of foundation stiffness drastically decreases the part
icipation of modes 2, 3, and 4.
Figures 5,6, and 7 illustrate the first four modes of computer runs 4,
5, and 6. C omparison shows that mode 1 is essentially the same for all
three foundations stiffnesses and involves swaying, rocking, and some struc
- 261 turai deformation becomes less important. Modes 3 and 4 are primarily structural modes, mode 4 being almost entirely internal. Thus, modes 3 and 4
of the combined case look very much like modes 2 and 3 of the rigid foundation case.
A further understanding of the soil-structure interaction is obtained
by examining the proportion of energy distributed among structural deformation, swaying, and rocking. This is tabulated for the first two modes in
Table 6, which shows that as the foundation becomes softer there is less
energy in the structural deformation. In the first mode the decreased
structural energy comes from an increase in rocking energy. In the second
mode it comes from both rocking and swaying but primarily from swaying. In
all cases there is a significant contribution from both soil and structure,
but the structural part of the response decreases with decreasing foundation
stiffness.
The effects of all these factors on accelerations at various points
in the structure are seen in Table 7. The pattern of modal domination is as
expected from the previous paragraphs. The behavior of the cylinder and
roof is dominated by the first mode, and accelerations decrease for the
softer foundations. The foundation accelerations increase with a softer
foundation because of the increased importance of swaying. For the response
of shield and vessel a stiffer foundation gives greater acceleration.
The forces in the connectors and the shears and moments in the same
parts of the structure as were examined for the case of the rigid foundation
are listed in Tables 8 and 9. The variations involve a sometimes complicated interaction of modes, but two points appear clearly:
1. In all cases the worst conditions occur when the foundation is hard.
2. The outer cylinder, roof, and base respond in the first mode. The
rest of the structure responds most critically in the second mode.
DETERMINATION OF VARIABLE MODAL DAMPING
While in the past it has been common practice in such analyses to assume
that damping is the same in all modes (in this case 7%), intuitively it
makes more sense to vary the amount of damping in different modes. As has
been seen, the degree of soil participation varies greatly between modes.
Modes in which horizontal distortions in the soil are significant would
be expected to have much larger damping.
The investigation of the effects of variable damping is based on the
concept that the energy absorbed by damping in each portion of the system
is proportional to the damping coefficient multiplied by the total modal
energy contained in that portion.
This is equivalent to taking the modal
damping coefficient as an average of the coefficients in the various parts
weighted on the basis of the modal strain energies in the separate parts.
This may be expressed mathematically as
- 262 D = - i - .
where D represents
E represents
E .represents
D. represents
Dni. D.
i
damping in the n'th mode

the total strain energy in the n'th mode
the strain energy in part i for the n'th mode
the damping for part i.
In the present case, the system consists of three parts:

1. The superstructure including the building, the containment, and
the reactor (D. = 4%)
2. The soil rocking spring (D, = 5*)
3. The soil swaying spring (D3 = 25*)
To compute the modal strain energy in the three parts it is convenient
first to compute the total kinetic energy in the entire system, which must
also be the total strain energy in the mode, E . The strain energy in the
two soil springs is easily computed from their modal displacements. The
energy in the structure is then the total minus that in the two springs.
The results of this calculation are shown in Table 10. To illustrate
the calculation, in mode 1 of run 4, the percentage of total energy contained
in each part is as listed in Table 6. The weighted damping coefficient is
then
.236 X 4 + .563 X 5 + .201 X 25 - 8.8
This value would be used to compute response in this mode.
All modes higher than the third mode have nearly 4.0% damping because
they are largely structural modes. The weighted damping in mode 1 is rel
atively unaffected by changes in foundation stiffness. Because of the
shift in energy content, mode 2 increases its damping and mode 3 decreases
its damping as the foundation becomes softer. These results are consistent
with Tables 5 and 6.
Since with constant damping the major response occurred in modes 1, 2,
and 3 and since the damping in these three modes will now be increased, the
overall response should decrease. Response which is dominated by the first
mode should change least, since the damping in this mode increases only
slightly. On the other hand, there may be considerable change in response
which is dominated by the second mode. Some change in patterns of modal
domination can also be expected.
THE EFFECT OF VARIABLE MODAL DAMPING
Varying the modal damping does not affect participation factors or
frequencies of modes. It will affect the modal spectral acceleration, and
hence the acceleration and forces in the structure.
Table 11 shows the acceleration response for the three cases. It
should be compared with Tables 5 and 7.
The pattern of modal participation and dominance is more complicated
- 263 than it was in the case of uniform 7% damping, but the same general trends
are still evident. All accelerations are smaller than they were for uniform
7% damping. With the uniform damping of 7% the acceleration of the foundation mat is greater than the peak ground acceleration, whereas with variable
damping it is, in most cases, approximately equal to the ground acceleration.
The latter result, which seems more reasonable (Biggs and Whitman, 1970),
is the result of larger damping in the second mode.
The forces and moments at selected points are tabulated in Tables 8 and
9. Again the pattern of modal dominance is more complicated, but the same
general trends are observed as in the case of uniform damping. The magnitudes of all forces and moments tabulated are reduced by variable damping.
The decreases are least for the outer structure (dominated by the first mode)
and greatest for points whose response is dominated by the second mode.
CONCLUSIONS - COUPLED FOUNDATION AND STRUCTURE
The analysis of the combined system of soil and structure leads to the
following conclusions applicable to this specific case:
1. For all responses except base acceleration and displacement,
the cases with the hard foundation are most critical.
2. The response of the outer structure is dominated by the first mode
involving structural deformation, swaying, and rocking, the last
being the most important.
3. The response of the rest of the structure depends on first, second,
and third modes, but the response is largest when the second mode
dominates. In those cases swaying and structural deformation are
most important.
4. The decreased response in the first mode with decreased foundation
stiffness reflects the smaller spectral acceleration of this mode
for the smaller fundamental frequencies, even though the modal
participation factor is unchanged.
5. The response of the second and third modes is affected considerably by the foundation stiffness, even though the spectral accelerations for those modes is unaffected. This occurs because of the
interaction between swaying of the foundation and the first two modes
of the structure alone. The interaction is strongest with the hard
foundation, and hence the response of these modes is greatest in
this case.
6. The fourth mode involves largely internal response, but its participation is greatly increased by hard foundation conditions because of
the increased importance of swaying.
Therefore, the hard foundation should be used to obtain the most critical
stresses at all points and to obtain the acceleration everywhere except on or
very near the mat. The soft foundation should be used to obtain displacements everywhere and the accelerations on or near the mat.
264
CONCLUSIONS - GENERAL
The study illustrates how the interaction of structure and soil may
affect the response of a reactor building. It also shows how a detailed
examination of modal response can reveal patterns in the soil-structure
interaction.
The response of the model with weighted damping is significantly less
than that with uniform damping, especially those portions where the response
is strongly affected by foundation swaying. This is because a large portion
of the energy of the dominant first and second modes is in the soil rocking
and swaying springs. Results obtained using nominal uniform damping in all
modes may be conservative for the internal portions of a reactor building and
for equipment mounted on the foundation and internal structure.
REFERENCES
BIGGS, J.M. (1964). Introduction to Structural Dynamics, New York,
McGraw-Hill, p. 105.
BIGGS, J.M., WHITMAN, R.V. (1970). "Soil Structure Interaction in Nuclear
Power Plants", Proceedings, Third Japanese Symposium on Earthquakes Engineering, Tokyo, pp. 89-95.
JACOBSEN, L.S., AYRE, R.S. (1958).

McGraw-Hill, pp. 113-115.
Engineering Vibrations, New York,
PAHL, P.J. (1970). "Modal Response of Containment Structures", Seismic

Design of Nuclear Power Plants, R.J. Hansen, ed., Cambridge, M.I.T. Press,
pp. 344-400.
WHITMAN, R.V. (1970a). "Evaluation of Soil Properties for Site Evaluation

and Dynamic Analysis of Nuclear Power Plants", Seismic Design of Nuclear
Power Plants, R.J. Hansen, ed., Cambridge, M.I.T. Press, pp. 270-305.
WHITMAN, R.V. (1970b). "Soil Structure Interaction", Seismic Design of

Nuclear Power Plants, R.J. Hansen, ed., Cambridge, M.I.T. Press, pp. 245-269.
- 265 -
TABLE 1
FOUNDATION SPRING CONSTANTS
FOUNDATION TYPE
SPRING CONSTANTS
SWAYING (kg)
ROCKING (kR)
HARD
1.2 X 10 6 K/ft.
7.50 X 10 9 K-ft/Radian
MEDIUM
0.8 X 10 6 K/ft.
SOFT
0.5 X 10 6 K/ft.
TABLE 2
MODAL RESPONSE OF RIGID FOUNDATION CASE
MODE
FREQUENCY
PARTICIPATION
FACTOR
SPECTRAL
ACCELERATION
2.33 Hz
1.49
.38g
4.25
5.01
.38
4.87
-4.06
.38
7.70
-1.57
.33
9.11
- .69
.29
- 266
TABLE 3
RESPONSE OF RIGID FOUNDATION CASE
LOCATION
ACCELERATIONS
ACCELI:RATION
0
0
0
0
FOUNDATION
ROOF (PT.43)
TOP OF SHIELD (PT.22)
BOTTOM OF VESSEL (PT.9)
(b)
LOCATION
BASE (PT.59)
BOTTOM OF SUPPORT (PT.21)
BOTTOM OF SHIELD (PT.26)
BOTTOM OF SKIRT (PT.8)
BOTTOM OF DRYWELL(PT.42)
BOTTOM OF CYLINDER (PT.58)
(c)
X
X
X
X
X
X
io2 K
1
3
2
2
2
io2
10 2
10 2
10
10 2
20g
60g
39g
31g
SHEARS AND MOMENTS

1XJM.
SHEAR
MODE
306
3.97
3.63
8.62
58.1
299
DOMINANT
MODE
1
2
2
MOMENT
438.
3.29
2.70
1.70
53.5
397
X
X
X
X
X
X
10 4 K-ft
io4 K-ft
io4 K--ft
10 K-ft
io4 K- ft
in4 K-ft
DOM.
MODE
1
2
3
2
2
1
FORCES IN CONNECTION
FOI(CES
SPRING
VESSEL ON SHIELD (3-22)
SHIELD ON DRYWELL (22-29)
5.37 X 10
134.9
X 10
DOM.
MODE
TABLE 4
RESPONSE OF RIGID STRUCTURE CASE
FOUNDATION
TYPE
UNCOUPLED
FREQUENCIES (Hz)
ROCKING SWAYING
COUPLED
FREQUENCIES (Hz)
f
l
2
ENERGY DISTRIBUTIONS
MODE 1
MODE 2
ROCKING
SWAYING
C
RO KING
SWAYING
HARD
1.57
2.25
1.36
3.82
71%
29%
28%
72%
MEDIUM
1.11
1.84
1.00
3.02
81%
19%
21%
79%
SOFT
0.81
1.45
0.74
2.14
81%
19%
19%
81%
Note:
Uncoupled frequencies refer to one degree of freedom systems for rocking only and swaying only.
Coupled frequencies refer to system with two degrees of freedom where rocking and swaying occur
simultaneously.
TABLE 5
RUN NO.
FOUNDATION
TYPE
DESCRIPTION OF SOILSTRUCTURE INTERA CTION RUNS

MODAL RESPONSE 7% DA MPING
FREQUENCY (Hz)
A
SPECTR L A CCELERA TION
A
P RTICIP
A TION FA CTOR
1
2
3
4
1
2
3
4
1
2
3
HARD
1.24
3.46
4.11
4.66
.303
.38
.38
.38
1.67
1.79
2.31
.72
MEDIUM
0.96
2.86
3.88
4.64
.234
.38
.38
.38
1.62
.79
.39
.02
SOFT
0.74
2.25
3.81
4.62
.179
.38
.38
.38
1.59
.82
. 0 6 . 0 8
268
TABLE 6
ENERGY RATIOS
RUN
MODE
NO.
NO.
ENERGY RATIOS
STRUCTURE
SWAYING
ROCKING
23.6%
30.4%
20.1%
63.5%
56. 3%
6. 1%
13.5%
5.8%
17.5%
79.3%
69.0%
14.9%
7.7%
1.8%
16.0%
82.4%
76.3%
15.8%
TABLE 7
ACCELERATIONS
RUN
NO.
FOUNDATION
DOMINANT
MODES
NOTE:
ROOF
(PT.43)
LOCATION
TOP OF SHIELD
(PT.22)
BOTTOM OF VESSEL
( PT . 9 )
.253g
.580g
,449g
,373g
.303
.463
.244
.238
.314
.370
.176
.183
SEE NOTE
Mode 1 dominates except in run 4, where mode 2 dominates.
RUN
NO.
SHEAR
BASE
(PT.59)
MOMENT
TABLE 8
FORCES AND MOMENTS
BOTTOM OF
SUPPORT (PT.21)
SHEAR
MOMENT
4,,07xl04K
4.85xl06K-l
6,,2xl02K
3.92xl04K-l
3. 40
"
3.70
6. 0
2, 91
"
2.80
5 .7
Dominant
Modes
1
RUN NO.
BOTTOM OF
SKIRT (PT.8)
SHEAR
MOMENT
7.75xl02K
1.28xl04K-l
3.49
0.52
2.34
0.35
Dominant
Modes
2,1
2,1
BOTTOM OF
SHIELD (PT.8)
SHEAR
MOMENT
2 .83xl02K
2.74xl04K-l
2.70
1,,37
1.48
2.15
0. 9 8
1.07
BOTTOM OF
CYLINDER (PT. 58)
SHEAR
MOMENT
BOTTOM OF
DRYWELL (PT.42)
SHEAR
MOMENT
10.5xl03K
2,1
7.84xl05K-l
2.89xl04K
3.76xl06K-l
7.4
4.88
2.14
2.75
6.0
3.65
1.61
2.03
1,2
270 -
RUN NO.
TABLE 9
FORCES IN CONNECTIONS
SPRING
VESSEL TO SHIELD (3-22)
SHIELD TO CYLINDER (22-29)
4.74 X 10' K
2.21
"
7.08
"
1.50
"
4.98
"
DOMINANT
MODES
13.6 X 10
1,2
1,2
TABLE 10
WEIGHTED MODAL DAMPING
RUN NO.
DAMPING IN MODE:
5
4.1
4.0%
4.0%
4.6%
4.0
4.1
4.0
4.0
4.4
4.0
4.1
4.0
4.0
4.2
8.8%
17.4%
6.8%
4.1%
8.4
20.8
4.3
8.1
21.5
4.0
TABLE 13
FORCES IN CONNECTIONS FOR WEIGHTED DAMPING
SPRING
RUN NO.
VESSEL TO SHIELD (3-22)
SHIELD TO CYLINDER (22-24)
4.01 X 10
2.06
"
6.28
"
1.37
"
4.37
"
DOMINANT
MODES
NOTE:
1,2,3
13.37 X 10
1,2,3
Mode 1 always dominates for soft and medium foundations;

modes 2 and 3 dominate for hard foundation.
TABLE 11
ACCELERATIONS FOR WEIGHTED DAMPING
RUN
NO.
MODAL SPECTRAL ACCELERATION ( g )

1
2
3
4
FOUNDATION
ACCELERATION
(g) AT:
TOP OF
SHIELD
(PT.22)
ROOF
(PT.43)
TOP OF
VESSEL
(PT. 9)
.279
.256
.391
.619
.179
.504
.382
.334
.221
.240
.604
.632
.195
.396
.229
.192
.170
.240
.628
.632
.200
.200
.310
.146
1,2,3 in 4
1 in 5 & 6
DOMINANT MODES
2 in 4
1 in 5 & 6
TABLE 12
FORCES AND MOMENTS FOR WEIGHTED DAMPING
RUN
NO.
BASE
(PT.59)
BOTTOM OF SUPPORT (PT.21)
SHEAR
MOMENT
SHEAR
MOMENT
BOTTOM OF SHIELD (PT.26)

SHEAR
MOMENT
3.61x10 K
4.47x10 K-l
4.6x10
3.10xl04K-l
2.36xl02K
2.24x10 K-l
2.91
"
3.47
"
4.2
"
2.16
1.16
1.25
2.37
"
2.65
"
3.8
"
1.68
0.78
0.88
Dominant
Modes
1
1,2
RUN NO.
BOTTOM OF SKIRT (PT.8)

SHEAR
Dominant
Modes
MOMENT
1,2
BOTTOM OF DRYWELL (PT.42)

SHEAR
MOMENT
1,2,3
1,2,3
BOTTOM OF CYLINDER (PT.58)

SHEAR
MOMENT
6.62x10
1.12x10 K-l
8.2x10
6.41xl05K-l
2.66xl04K
3.42x10 K-l
3.22
0.52
5.8
"
4.20
2.01
2.57
2.12
0.34
4.6
"
3.14
1.52
1.92
1,2,3
1,2,3
1,2
1,2
05
Figure I:
IDEALIZATION
OF
STRUCTURE
007
0.1
02
0.3
05
0.7
1.0
UNDAMPED
PERIODISK.)
Figure 2. GROUND RESPONSE SPECTRUM-DESIGN BASIS EARTHQUAKE
I r t q . " 4 . 2 5 c pi
f r #q : 2 3 3 C p i
I r t q . ' 4 . 6 7 cpt
.-I
MODE 3
Figure 3: THE
DYNAMIC
MODEL
Figure 4 : MODE
SHAPES-
RIGID FOUNDATION
CASE
275
Iraq.>3.46 cp
f r q. 4.66 cpt
Figure 5 : MODE
SHAPES RUN
276 -
frq. * 3 . 6 6
cpi
fftq. 4 . 6 4
[7
Figure
6 : MODE
SHAPES - RUN 5
cpi
277 -
(req. = 2 . 2 5
Cps
ff eq. - 4 . 6 2 Cp:
L7
Figure
7 : MODE
SHAPES - R U N
- 278 -
DISCUSSION
K. MAR.GUERRE, Germany
The soil is r e p r e s e n t e d by springs and damping. What about the m a s s of the soil ?
J. M. BIGGS. U. S. A.
An effective soil m a s s was added to the base mat - both for translation and rock
ing. The values used were those previously derived by Dr. Whitman.
G. KLEIN, Germany
In this country there is a strong discussion about damping. Are your figures:
Concrete s t r u c t u r e
4%
Soil swaying
25%
Soil rocking
5%
a conservative assumption ?
J. M. BIGGS, U. S. A.
I think these values a r e conservative. These were applied to the Design Basis
Earthquake. Lower values were used for the smaller Operating Basis Earthquake.
J. D. STEVENSON, U.S.A.
How do you determine the base motion acceleration for the rigid body model
when in the product
S a is
= 0
In the rigid foundation case the acceleration shown in the slide for the base mat
is m e r e l y the peak ground acceleration. It cannot be computed from the model analysis.
K. UCHIDA, Japan
1. How do you estimate the stiffness between the pedestal and the dry well (horizontal springs
seen at the middle part in your vibration model) ?
2. Are there strong interactions between them ?
A
This depends upon the nature of the seal. If, as in some c a s e s , it is a metal
bellows, the spring constant is computed from the properties of the corrugated plate. If the
- 279
seal is rubber, perhaps the spring constant should be taken as zero. In the case of a bellows
seal, the spring is very stiff and has a significant effect on the forces in the pedestal and
dry well.
D. LUNTOSCH, Germany
Isn't it a rather poor representation modelling a cylindrical and conical shell by
only one lumped m a s s for each ring ? What about the accuracy of the model ?
Ovalling of a cylindrical shell does not occur under earthquake loading because
it is a n t i - s y m m e t r i c a l . Local bending might be significant if large concentrated m a s s e s were
attached to the shell. However, this does not usually occur and therefore I think use of a
single node for the complete ring is satisfactory.
A. HADJIAN, U. S. A.
The analog of the soil s t r u c t u r e interaction not only includes m a s s and springs
but also a large damping (for lateral motions). Although your mat motions, using the energy
damping concept, are more valid than just using modal damping values, would not the high
damping values used for the 2nd mode reduce also the motion of the r e s t of the s t r u c t u r e .
Any good solution should place the damping where it really is.
The weighted modal damping values apply to the whole mode. Therefore, to the
extent that the upper structure responds in the second mode, the higher damping does affect
the upper structure.
K 4/1*
DEVELOPMENT OF ASEISMIC DESIGN OF PIPINGS

VESSELS AND EQUIPMENTS IN NUCLEAR FACILITIES
H. SHIBATA, A. WATARI, H. SATO, T. SHIGETA,
Institute of Industrial Science, University of Tokyo,
Tokyo,
A. OKUMURA,
Faculty of Science and Engineering, Waseda University,
S. FUJII, M. IGUCHI,
Faculty of Engineering, University of Tokyo,
Tokyo
This report involves the development of aseismic design procedures of pipings,

vessels and equipments in Japan. These mechanical structures show their various characteristics of vibration. Pressure boundaries, a containment vessel
and safety systems belong to such structures. The vital components of nuclear
power plants are classified to class "A" as the classification in case of
aseismic design in Japan. All components in class "A" are required to be designed in dynamic sense based on design basis earthquake, of which level is
decided in consideration of local seismisity.
For the dynamic design, the following processes are the most important:
i)
estimating natural frequencies and modes of the system,
ii)
estimating its damping characteristics,
iii) estimating the behavior of the system during strong earthquakes,
iv)
deciding the design criteria, especially allowable stresses to earthquake loadings.
The authors, as members of the sub-committee of "Aseismic Design of Nuclear
Facilities" in the Japan Society of Mechanical Engineers, have been developing
the fundamental techniques and programs from process i) to iii).
DYNAPS, the program for eigen-values and modes analysis of three-dimensional
piping systems, and SHELVIA, the program for those of shell-of-revolution
were developed. Both are designed for arbitrary configurations of pipings or
shells and use a transfer matrix method which was originated by one of the
authors.
The method of response analyses of a two-degrees-of-freedom system and a
multi-degrees-of-freedom system to two inputs are discussed based on results
by a random vibration theory and an analog computer technique.
Through several field experiments, the parameters of full-size power plants
were made clear. Also some results of the computation by DYNAPS and SHELVIA
were proven by both field experiments and model experiments.
Philosophy on deciding the allowable stress to seismic loads in Japan is also
discussed briefly.
282
DISCUSSION
-^
G. KLEIN, Germany
Do you use different damping factors for considering:
1. Design base earthquake ?

2. Maximum potential earthquake ?
H. SHIBATA, Japan
We design only for "design earthquake". But for the analysis of "hypothetical
earthquake" usually we use the same value. In some c a s e s , we use another approach to check
the margin of the safety, for example, elasto-plastic analysis.
K 4/2
SEISMIC DESIGN COEFFICIENTS OF EQUIPMENT

IN NUCLEAR POWER PLANTS
C.-W. LIN,
Westinghouse Electric Corporation,
Power Systems, Pittsburgh, Pennsylvania, U.S.A.
ABSTRACT
Seismic coefficients for the design of equipment in nuclear power plants have been
shown to be proportional to the single degree response spectrum.
Using the two degree
response spectra, constructed according to a numerical time integration study for the ElCentro 1940 N-S earthquake obtained by previous investigators, equipment seismic design
coefficients for other earthquakes can be easily computed using present theory.
1.
INTRODUCTION
Although eeismic analysis of a multi-story building has been much simplified since the
response spectra technique was introduced by Alford et al. [1], and the root-mean-square
combination of different modes was recommended by Clough [2], the design of a piece of equipment to resist an earthquake is still a difficult task for which a universal approach has
yet to be found.
There are, to the author's knowledge, at least two different approaches
being adopted in the United States. These are briefly stated as follows:
a)
When time history accelerogram of one or more actual earthquake records normalized to the
proper density or maximum ground acceleration is available, a direct integration will
generate the response spectra of equipment (see Blume, Keith [3], [4]).
b)
If the N-S component of the El-Centro 1940 earthquake can be used as the design earthquake, the results of the two degree of freedom response spectra constructed by Penzien
and Chopra [5], [6], may be readily used.
Wien design criteria is such that response spectra different from those of the El-Centro
site have to be used, one would be likely to choose a time history representative of the past
earthquake or an earthquake generated artificially by using a stationary, Gaussian distributed input function having a power spectral density which is filtered to represent the site
characteristics, and either go through a time integration study or repeat the same analysis
outlined by Penzien and Chopra [5], to obtain the required two degree of freedom spectra.
These analyses are not only costly, but also time consuming.
Very often, when proper specif-
ication has to be made even before preliminary design of a building is completed, the sophistication of these analyses does not seem to justify the loss of accuracy due to the use of
preliminary design data and the delay of schedule.
284
Using the same model, this paper illustrates how the results presented by Penzien and
Chopra [5], in obtaining two degree of freedom response spectra, can be easily converted to
useful information for different site conditions when either the local ground power spectral
density or the single degree response spectrum is known.
2.
DYNAMIC
RESPONSE OF A TWO DEGREE OF FREEDOM SYSTEM
Using the fact that the mass of equipment is usually small compared with the supporting
structure, the interaction effect will be more from the structure to the equipment than vice
versa.
C onsequently, Penzien and Chopra [5] suggest a separate two degree of freedom system
for each of the N normal modes of the building without the equipment in analyzing this inter
action effect.
Figure 1 shows this system, in which subscripts n and a Indicate the quanti
ties derived from the n th building mode without the equipment and from the equipment, re
spectively.
M is the mass, K the spring constant, C the damping factor, X the displacement
relative to the support, and U (t) is the support motion, with

N
N
U ft) ( E m . , / ra. 2 ) U (t) U (t)
i
in
i
in
11
8
" 8
iel
where
(1)
Is the dimensionless th building mode shape quantity for i th floor, with m
representing the floor level mass,

respectively,
damping ratio.
tion factor
and
the corresponding frequency and damping rutio,
is the frequency of the equipment by itself, and
is the corresponding
The use of the ground motion time history multiplied by the th participa
to represent the support motion indicates that when quantities from the
equipment are taken equal to zero, and the remaining system is equivalent to the generalized
building model with the th mode being excited alone.
The equations of motion can be derived from the mathematical model shown in Figure 1.
However, instead of working directly with chose equations, a combined single equation can
be obtained as the following:
(X X ) + 2 (6 + 1) ( ) + 2 ( + 1) (
a
aa
a
a
a
a
= + 2
)
a n
(2)
+ 2
where
= /
a
a n
(3)
Recall at this point that one of the basic assumptions made by Penzien and Chopra (5],
in using the mathematical model of Figure 1 is that the building entirely controls the
response of the equipment, while the effect from the equipment is little felt by the
building.
With this approximation in mind, one has :

X + 2 X + 2 - - n U
ng
Eq. (4) can be solved in terms of a convolution integral.
(4)
By differentiating the solution
twice with respect to t and neglecting the integration terms containing damping ratio, one
arrives at:
- 285 t
-
+aU
= o sino (t - )e
n
ng
nn
n
After rearranging eq. (4) by grouping
+ U
(t - )..
U ()d
g
(5)
to one side of the equation and then elimin
ating this using eq. (5), the resulting equation can be substituted into eq. (2) and then
solved for X - X . The net result is:
a
( - )
a n
et t
-** (t - )
- V 1 I sino* (t A)e a a
a
a o
(6)
( ) ddX
where:
*
a
3.
( + 1 ) 1 / 2 , *
a
*
a
( + 1 ) 1 / 2
a a
VARIANC E OF ( )
a
Since design earthquakes are usually very strong and have long duration, it can be de
duced from Caughey, Stumpf, and Bycroft [7], [8], that each of such earthquakes forms a
process which is ergodic, Gaussian distributed with zero mean.
From these observations, one
can take the temporal autocorrelation function of U as:

< U (t)U (t') >
g g
I G<o)coso(t t')do
where G(o) is the power spectral density.

senting a white noise process.
(7)
Very often this is assumed to be a constant repre
However, physically, such a process can never exist.
For a
large majority of strong earthquakes, a semiempirical formula suggested by Kanai [9] can
be used.
With the help of eq. (7), the variance of (X
X ) may be derived from eq. (6) as:
o2(t) < [Xa(t) X n (t)] 2 >

(-^>2
a
(8)
t t
o**(t + t' ')
I sino* (t ) sino* (t' ') a a
o o
*
sinn) ( - )sin<*>
( + * ')
(* ')
G(ii) cosi(r, r, ')dr,dr,'dXd\ "dio

Knowing the integrals involved in eq. (8) are convergent, one may reverse the order of
Integration.
When G(J) is known, the integration can be carried out either explicitly or
numerically, depending on how G(iii) is given.
Since
is considered to be small in most of
the structural analysis, when integrations with respect to and " have been carried out
by reversing the order of Integration for , ", and , the quantity l/|z(oj)j2 , where
|()| 2
[F,V + 2 2 2 (2 + 2 ) 2 + (2 2 ) 2 ]

(9)
- 286 as obtained in the process will have a sharp peak at o - . Hence, when making contour
integration with respect to o, the main contribution to the integral will come from the
region around - . Using the same analogy as Caughey and Stumpf [7], which originated
from the Laplace's method of evaluating integrals, eq. (8) may be very closely approximated
by:
o
t t
o 2 ( t ) G(o ) ( - W o 2 s i n o * ( t - ) s i n o * ( t ' - ' )
a
a
a
o o
-** ( t +
(10)
t ' - - ' )
<'
|()|2
C0Sw(X
"
- '
-e
[coSA * cosu) + cosuX *
[COSCDX
COSL
sinto ' cosLoX + simo * + e
[1 +
simo
( + ' )
- 2 + 2
( + ' ) + sinto ' ] } dio) dXdX'
o r , one may w r i t e :
02(t)
= G( n ) ( , o a , t , t n . a ,
a , )
(11)
, , , C , C , , ) represents the rest of the terms of the rieht hand

n
a
n
a
a
side of eq. (10) multiplied by G( ) . Thus, the maximum of the rootmeansquared relative
in which (
displacement of (X
X ) is given by:
D
max
= [G() ) ] 1 / 2 [f
( , , t, , , , ) ] 1 / 2
max
a
a n
(12)
For any given , , , and , eq. (12) shows that the rootmeansquared relative
n a n a
a
displacement of (X X ) is proportional to the square root value of the ground power
spectral density evaluated at the frequency . This implies that when analysis of the two
degree of freedom system of the present model has been carried out for any particular ground
power spectral density, results obtained can be converted immediately to yield rootmean
squared relative displacement of (X
X ) for any other ground power spectral density.
This is:
G* ( )
* ( , , , , , ) = ( n * \ ) 1 / 2
(, , , , , )
max
a
' a' a' '
G ()
max
a'
a
a'
(13)
where
r e p r e s e n t s the known v a l u e of t lie r o o t m e a n s q u a r e d r e l a t i v e d i s p l a c e m e n t c a l
v
max
culated based on the particular
ground power spectral density G(UJ ) , and
represents the
r
max
unknown rootmeansquared relative displacement corresponding to the given ground power
spectral density G ( ). In the case that the spectral acceleration response S ( , ) is
given rather than G ( ) a similar result as eq. (13) can be derived.
Using eq. (7) and the same analogy in obtaining eq. (12), it is possible to show that
the maximum probable response of a single degree of freedom system, which approximates the
287
single degree response spectrum S (in ,f, ), is in proportion to the square root of its power
spectrum density.
Therefore,
S* ( , )
* ( , , , , , ) " _a . " ,".
(, , , , , )
max
a' a
S ( , ) max
a
' a
a'
a

Eq. (14) has
(14)
a direct application when the two degree of freedom response spectra obtained
by Penzien and Chopra [5] for the ElCentro 1940 earthquake are used.
Using their notation
with:
C
( , , , , , ) (
an
* a' ^' s a' a
' a
) 2/
a g1max
(15)
representing the seismic coefficient based on ElCentro earthquake, eq. (14) can be written
( , , , , , )
an
n
a' n
a
a'
S* ( , )
'C
( , o . , . , )
(<"_
) an
(16)
where
. . . . .
,2
(, , ,
a*
,
a
)
a n g
(17)
NUMERICAL EXAMPLES
To illustrate how eq. (16) can be used to convert the results obtained by Penzien and
Chopra 15], corresponding to the ElCentro earthquake to results required for equipment
design based on other design earthquakes, two different sets of single degree spectral
accelerations, S ( , ) and S ( , ) representing the ElCentro earthquake and a typical
earthquake, respectively, are plotted in Figure 2.
Figure 3 shows the spectral ratio which
is defined as S*(io , ) / S ( , ) .
a
Knowing the spectral ratio, seismic coefficient C
for the typical earthquake can then
be computed using eq. (16). Figures 4 to 12 show these results.
Table 1 shows the parameters
involved in identifying each curve in these figures.

The following procedure for seismic design of equipment located at the I th floor of a
building is outlined briefly in order to present a selfcontend treatment.
a)
b)
c)
Plot spectral ratio S ( U ) , ) / S ( G J . ) against T for each in consideration.

v
an
n
a
Plot seismic coefficients C ( , , , , , ) for each combination of and

an n
n
n
a
a
a
for different mass ratio and equipment frequency against the building period .
C ompute dynamic characteristics of the building and the equipment such as , , ,
, , , etc.
d)
From the figure showing the pair of damping ratios

equipment, read in the seismic coefficients C
and
used for the building and
for each mode.
- 288 e)
Multiply each C
mode shape
by its proper building participation factor a
C.*
ia
f)
[ ( < C * ) 2 ] 1 / 2
in an
Design the equipment to resist a maximum horizontal seismic force of:

F
a
where W
5.
and the normalized
and take the root-mean-squared sum, i.e.,
C, *W
ia a
is the weight of the equipment.
DISCUSSION AND CONCLUDING REMARKS

For a given set of building and equipment parameters, such as , , , , and
a
&
, the present study shows that the ratio of the seismic coefficients for two different
design earthquakes is in proportion to the ratio of the spectral accelerations.
Using this
conclusion, together with the two degree response spectra obtained by Penzien and Chopra [5],
based on a numerical time integration study for El-Centro 1940 N-S earthquake, equipment
seismic design coefficients for other earthquakes can be easily obtained.
In contrast to the time consuming and high cost process of the numerical Integration
scheme, which in general requires the help of a large capacity electronic computer, the
semi-graphic procedure reported in the present analysis is simple to apply and easy to con
struct.
When results by Penzien and Chopra [5] are not available, the relationship established
within the text allows one to use the present typical spectral accelerations and seismic
coefficients in constructing the necessary seismic coefficient curves.
For damping ratios
and , and the building frequency not appearing in the prepared figures, extrapolan
a
tion will yield a close estimate to the true value.

The response spectra of the El-Centro 1940 earthquake shown in Figure 2 are actually
the design spectra averaged over four different earthquakes.
The use of these averaged
spectra in conjunction with Penzien and Chopra's [5] results undoubtedly provides some
safety margin over the design since these spectral values are lower than the true spectra of
the El-Centro 1940 earthquake.
Finally, it may be pointed out that the shaded area in Figures 4-15 indicates where one
would be had the peak of the ground response spectra been used in the design.
289 -
REFERENCES
[1] ALFORD, J. L., HOUSNER, G. W., and MARTEL, R. R., "Spectrum Analyses of Strong-Motion
Earthquakes," Earthquake Research Laboratory, California Institute of Technology,
August 1951.
[2]
CLOUGH, R. W. , "Earthquake Analysis by Response Spectrum Superposition," Bulletin of

the Seismologicai Society of America, Vol. 52, No. 3.
[3]
JOHN A. BLUME 6. ASSOCIATES, ENGINEERS, "Summary of Current Seismic Design Criteria for
Nuclear Facilities," San Francisco, California, September 1967.
[4]
KEITH, J., "Seismic Design of Critical Equipment in Nuclear Reactor Plants," John A.
Blume & Associates, Engineers, 1968.
[5]
PENZIEN, JOSEPH and CHOPRA, ANIL K., "Earthquake Response of Appendage on a MultiStory Building," Proceedings of the Third World Conference on Earthquake Engineering,
Vol. II, 1965, New Zealand, pp. 476-487.
[6]
SEXTON, H. JOSEPH, KEITH, EDWARD J., Discussion of the above paper, Proceedings of the
Third Wcrld Conference on Earthquake Engineering, Vol. II, 1965, New Zealand,
pp. 488-490.
[7]
CAUGHEY, . . , and STUMPF, "Transient Response of a Dynamic System Under Random

Excitation," J. Appi. Medi., December 1961, pp. 563-566
(8]
BYCROFT, G. N., "White Noise Representation of Earthquakes," ASCE, Vol. 86, No. EM2,
April 1960, pp. 1-16.
[9]
KANAI, ., "Semi-Empirical Formula for the Seismic Characteristics of the Ground,"

Bulletin of the Earthquake Research Institute, University of Tokyo, Tokyo, Japan,
Vol. 35, 1957.
290
APPENDIX II NOTATION
The following symbols are used in this paper:
C
damping factor of the equipment.
damping factor of the nth building mode.
spring constant of the equipment support.
spring constant of the generalized nth building mass.
mass of the equipment.
generalized mass of the nth building mode.
M
W
S . S *
a' a
m
G
equipment weight.
spectral accelerations,
mass of the building at ith floor.
"
power spectral density.
C , C *
an
an
seismic coefficients,
C*
ia
resultant seismic coefficient,
displacement of equipment relative to the support.
displacement of generalized nth building mode relative to the support.
ground movement.
= generalized nth building mode support movement.
equipment frequency.
nth mode building frequency.
equipment damping ratio,
nth building mode damping ratio.
0,
in
dimensionless nth building mode shape quantity for ith floor,
mass ratio of equipment to building,
frequency variable.
spectral density at bedrock.
parameters depending on local geology.
gravity
2/ , period of the equipment.

a
2/ , period of the building.
participation factor of nth building mode.
design seismic force for the equipment.
h ,
8
g
g
acceleration.
291 -
TABLE 1
PARAMETERS FOR TWO DEGREE
RESPONSE SPECTRA FIGS. 4-12
CURVE NO.
1
2
Up
0.20
0.002
0.010
0.025
0.40
0.002
0.010
0.025
0.60
0.002
0.010
0.025
10
0.80
0.002
11
0.010
12
0.025
13
1.00
0.002
14
0.010
15
0.025
Xnlt)
X0(l]
FIXED
REFERENCE
1
>JHy^l4,
MwSwwwuyMsWA&"^
Us
FIG. 1 TWO DEGREE OF FREEDOM SYSTEM
Sa
(n, i n ) , ELCENTRO
SO | . n | ( n ) , * 7 . l
1940 EA RTHQUA KE
TIMES
EMPIIFIED TYPICA L EA RTHQUA KE
/ *
'
/
/
//
" ^
.
0..
v.
//
v.
II
\o 0
\
\
1
1
20
2';
'
5;
//
102
2Crt
1 ^ ^ ^ ^ .
___J0^5
20';
0.2
0.4
0.6
O B
Tn
FIG.2 ACCELERATIO N SPECTRUM
1.0
293 16.0
14.0
12.0
So i m i , n)
Sa (un, n)
10.0
FIG.3
SPECTRAL RATIOS
Sa ("n. in) / S a (.. in)
VS. Tn
n = 1.0
= 0.02
0.00
1
-an
9'
/2
/ /
\
\ \
\ \
L M
li
VI
/'
/ \
y///
'////
0.2
SEISMIC
\!i
1 //
\ \v
12
v/77 -rrrr/ /
0.4
0.6
T
FIG.4
10
XM
\\
lil
\
'S \
/ ' ) \\ \M\
COEFFICIENTS
0.8
-1
/7
1.0
- 294
n 1.0
n 0.05
i 0 = 0.00
Can
1 t
1/
. \
/ / \Y
'
V" ^&
*\.
V/
y y7uL
///
0.4
0.2
0.6
/ / /
0.8
1.0
T
FIG.5
SEISMIC COEFFICIENTS
n '0
n = o.io
{ 0.00
an
2
7
4
1
S's
/ /\
//
S2^?^
0.2
/ / /
\ /
8
?>___2 _
7?^
0.4
T
FIG.6
^
0.6
***^*^^" J
w&
0.8
1.0
295
"n ' 10
7.
0.02
4
an
11\
',
f 0 0.02
Ie
W /
' w &f*&
1 / J x\v
6
' z^^fc^,"
0.2
'A..s
0.6
0.4
s
0.8
1.0
FIG.7
"n = 1.0
f = 0.02
fa " 0.05
an
g's
7.
I 1
lv
0.2
FIG.8
\
\
'9
/V
\\>iv^2
0.4
0.6
0.8
1.0
296
"n = ' 0 0
n = 0.05
= 0 5
7
4
1>
//
//y
' /
/r/
//yi
^^
SAX
^^^ S L
Wk
1
0.2
0.4
0.6
'
>'
'
0.8
>'
1.0
Tn
FIG.9
"r, ' 10
7.
1
' . 005
/ 8
/ ' ^\ \\
\ \
' \ \ C? ^ s \
\^-
12
//
ly'
' / / / .
0.2
0.4
0.6
In
FIG.10 SEISMIC
o ' 2
COEFFICIENTS
' / 2 ?1
08
I)
297
"n * 1.0
= o.io
o = 0.02
7
1
yi
10
y* ^ ^ i
Jl
*"
" 1 2 ^ * ^
^?4ie
0.2
0.4
0.6
0.8
1.0
Tn
FIG. 11 SEISMIC
COEFFICIENTS
"n 1.0
f = 0.10
f = 0.05
7 .8
S^fe^
JO
r11.12
y**"
""/^*^_
/ //,
0.2
0.4
0.6
Tn
FIG. 12 SEISMIC COEFFICIENTS
0.8
1.0
298
DISCUSSION
P. MITTERBACHER, Switzerland
Do you consider the case of a horizontal tube or tank with end-closure and not
completely filled with water under the influence of an earthquake ?

J . D . STEVENSON, U . S . A .
For design purposes the simplified method presented in TID 7024, chapter 6,
earthquake design of nuclear facilities is normally used. This procedure does consider slosh
effect.
H. WOLFEL, Germany
I think your method is very conservative. Did you compare your results with a
time history analysis and can you give us an estimation of the failure ?
J . D . STEVENSON. U.S.A.
No comparison with time history was performed. The method does tend to be
conservative in so far as it applies to single degree of freedom s y s t e m s . However, in systems
which might not respond as a single degree of freedom system some additional conservatism
should be required. In addition the technique is meant to be used as simple procedure to
develop design c r i t e r i a to the manufacturer. Avoiding to evaluate a design in detail requires
to be conservative to insure design adequacy.
H. SHIBATA, Japan
How do you define the m a s s ratio of equipment to a building ? In most c a s e s , the
m a s s of a building might be less than the total m a s s of it, especially in a coupling condition
of higher modes, if you use a set of two-degrees -of-freedom s y s t e m s .
J . D . STEVENSON, U.S.A.
Generally the m a s s of that portion of the s t r u c t u r e that will be excited by a
p a r t i c u l a r piece of equipment is much less than the total m a s s of the s t r u c t u r e . Actual points
of support have to be evaluated to estimate affected building m a s s . An o v c r - r s t i m a t i o n of
affected building m a s s is p r e f e r r e d since this will lead to more conservative results.
K 4/4
ASEISMIC DESIGN OF ASYMMETRIC STRUCTURES

AND THE EQUIPMENT CONTAINED
CH. CHEN,
ABSTRACT
The dynamic structural response Is a function of these basic assumptions regarding

seismic input, soil structure interaction, structural and dynamic properties of
buildings, etc.
A general reviev of the current assumptions is included.
The linear
dynamic analysis of a coupled stubby asymmetric building of a nuclear power plant

is presented as an example.
rigid in its own plane.
Each floor of the building is modeled as a diaphragm
Consequently three dynamic degrees of freedom are assigned
to each diaphragm, two translational in its own plane and one rotational about the
vertical axis.
The equipment installed in the building must be designed dynamically
to insure the safe shutdown in case of accident caused by the possible strong-motion
earthquake. Conventionally single "floor design spectrum" at each floor level is
used to design these equipment.
However due to the rotational effect, the structural
response varies from point to point on the same floor; also two "floor design spectra"
corresponding to two translational degrees of freedom of the diaphragm are obtained
for each directional input.
1.
INTRODUCTION
To insure public safety, the structures and the equipment installed in
a nuclear power plant in such countries as the U.S.A., Japan, etc. are required
to be designed dynamically to withstand a possible strong-motion earthquake.
The
theory of structural dynamics with deterministic or random input can be found in

books by Biggs [l], Hurty & Rubinstein [2], and Lin [3]. The aseismic design of
concrete structures was treated by Blume et. al. [U]. However, the calculated
structural response depends on assumptions regarding seismic input, soil structure
interaction, the structural dynamic properties of buildings, etc.
The designer
should make these assumptions as realistic as possible.

Due to the layout of the equipment and the interior shielding walls, most
auxiliary buildings and control buildings of nuclear power plants are asymmetric.
Here , asymmetry means the center of mass of the floor does not coincide with the
center of rigidity of the same floor; furthermore, these mass centers and rigidity
- 300 centers of different floors are not on the same vertical axes.
displacements and torsional displacement are coupled.
Hence the translational
The results indicate that
the coupled displacements have a significant effect on the equipment design.

2.
SEISMIC INPUT
The intensity of shaking and the frequency of occurrence of a possible
future earthquake at a nuclear power plant site should be first estimated by the
seismologist and the geologist.
Engineers interested in this topic can consult
to references [5] through [9]
The recommendation from the analysis of the local
geology and seismology will then be translated into some forcing function or input
at the base of the structure.
Due to the nondeterrainistic nature of earthquake
records, there are vorks dealing with random input [10 - l6].
Unfortunately, owing
to the scarcity of strong motion acceleregrams, the statistical properties of the

random motion can not be determined precisely [9> l1*]*
Hence there are efforts
devoted to the analysis of least favorable response [IT , l8]. But the disadvantage
of this approach is that the design may be too pressimistic to be practical even
for the least favorable response in the "local" sense only [17]. The other alternative,
though not necessarily a better way due to the uncertainty about the nature of a
future earthquake and the local soil influence, is to use the deterministic input
based on the past strong motion earthquake accelerograms [19]. These digitized
computer data cards are available from Environmental Science Service Administration,
U. S. Department of Commerce.
in the next section.
The influence of the local soil will be discussed
Since the calculated structural response due to time history
input depends very much on the damping and the dynamic properties of the structure
which are not easy to be determined precisely, and since the future ground motion
are difficult to predict, engineers have been using the technique based on design
spectrum which is a smoothed curve relating the damped maximum responses of single
degree of freedom oscillators due to certain deterministic input to their undamped
natural frequencies.
The generation of design spectrum is discussed in references
[20] through [23]. Sometimes the simulated earthquakes [2h] are also used as input.
These digitized computer data cards are available from Earthquake Engineering Research
Laboratory, California Institute of Technology.
The advantage of earthquake time
history analysis is that the modal responses can be combined easily on a time basis.
The disadvantage is that a large amount of computer time is required to obtain a
satisfactory solution, especially when a parametric study is involved.
The advantage
of design spectrum analysis is its simplicity in application and saving in computer

time.
The disadvantage is that the maximum modal responses are combined on some
assumed basis, i.e. the square root of the sum of squares.

3.
SOIL STRUCTURE INTERACTION AND SOIL AMPLIFICATION

Soil structure interaction has been treated by various approaches. Barkan
[25], Whitman & Richart [26], and Whitman [27] used equivalent damping value and
spring constant derived from a model consisting of a ri#id base resting on linear
elastic half space to handle the interaction.
This method has the advantage of
- 301 simplicity in its application.
However the different damping values between soil
and structure may cause difficulty in the analysis. An intuitive formula for combined
damping value was suggested by
Professor
Biggs as mentioned in reference [27].
Another pitfall of this method is that the basic assumption of no separation between
base and soil in the derivation of spring constants may sometimes be violated in
the analysis.
Other more sophisticated schemes such as the finite element method
or the lumped parameter method are sometimes used to model the soil.
Related works
in this area can be found in reference [28]. These methods can also he used to
predict the soil amplification or the influence of local soil properties, if the
soil model is extended to the bedrock.
But these methods again suffer from the
difference in damping values between soil and structure.
The linear elastic half
space theory was also applied to predict the influence of structure on the free
field motion [29, 30, 31] and to predict the structure response on flexible foundation
[32].
Generally speaking, the effect of interaction is small for structures built
on competent rock.
As to the soil amplification effect or the influence of local
soil properties, both the shear beam model and finite element method have been used
[28].
h.
STRUCTURAL PROPERTIES AND MODELLING

All structures must first be idealized into the equivalent mathematical
models to facilitate the analyses.
In modelling, one often intends to cover as
much as possible the essential characteristics of the stiffness and mass distribution.
On the other hand, one also hopes that the model will be simple enough for easy
interpretation of the results, or hopes to avoid the feeling of seeing the trees
but not the forest.
A compromise should usually be achieved between these two ends.
The mathematical models used in the past varies from shear buildings [33]
to cantilever-type buildings [3^]
There are also assumptions
of shear walls, floor diaphragms, etc.
concerning the behavior
One ought to choose the model which fits
closely the characteristics of the structure. It is always premature to predict

which assumption is more conservative before the dynamic analysis is performed,
because changing the stiffness and inertia properties has an effect on the eigenvalues
and eignvectors which in turn have an effect on the stresses and displacements.
Multistory structures have been analyzed by several different techniques
recently.
Clough et. al. [35] simulated the static behavier of a symmetric three
dimensional structure as a series of two dimensional frames connected by inextensible

links at each floor level.
They improved the analytical method by evaluating the
lateral stiffness of each frame with the tri-diagonalization procedure and superposing
these to obtain the overall building stiffness [36]
Kokinopoulos [37] treated
the multistory building of very flexible floors as a one dimensional catilever beam,
and the overall stiffness matrix was derived by Solving Simultaneously the threemoment equations.
Weaver et. al. [38, 39] followed the modified tridiagonal approach
to obtain an overall stiffness matrix for static and dynamic analyses of a general
three dimensional frame without shear wall. The floors are assumed to behave like
diaphragms.
Weaver & Oakberg [UO], and Manning [Ul] extended their works to include
- 302 the shear walls. Winokur and Glck lk2]
analyzed statically the asymmetric multistory
structures by combining the stiffness matrix of each individual stiffening element

into the overall stiffness matrix.
Glck [1*3] analyzed the asymmetric structure
by the "continuous method," and was discussed by Nynhoven ii Adams, Biswas & Tso
5.
EXAMPLE
The linear elastic dynamic analysis of the coupled auxiliary structures
of a nuclear power plant is presented here as an example.
The characteristics of
these structures are quite different from those conventional structures mentioned
before.
First of all, due to the requirement of biological shielding from the radio-
active emission caused by hypothetical accident, the vails and floors are exceedingly
thick. Secondly, as a result of the equipment layout, interior walls are discontinuous
from floor to floor.
Thirdly, several structures are coupled together.
To reveal the torsional effect about the vertical axis of an asymmetric

structure, the floor is assumed to behave like a diaphragm rigid in its own plane
and with three degrees of freedom, two translational, and one torsional [38, ^ i ,
k6".
Due to the Stubby proportion of the actual structure, the flexibility of the
wall along its strong axis is controlled by the shear deformation [U6]. Hence the
analytical method described by Ayre [7] can be applied here.
For Stubby structures
with foundation keyed into competent rock, it can be assumed to be fixed at the
ground.
The mathematical model is as shown in Fig. 1.

Because of asymmetry, the center of stiffness does not coincide with that
of the mass at each floor.
Furthermore, the centers of stiffness of the floors
do not fall on the Bame vertical axis; neither do the centers of gravity.
It should
be pointed out that the classic shear center of open section based on bending deformation
does not apply here.
To facilitate the analysis, a common coordinate system is established.
Then the stiffness matrix of each floor can be transformed as
k
oy
o
k
(1)
where k , k , and k. are the translational and torsional spring coefficients, and
x
y
(;)
T. is the transpose of T..
1
The subscript i refers to floor number, and R , R are

th
' y
the coordinates of the center of stiffness at i
floor.
The mass matrix can be
transformed similarly.
M.
T.
*:
'V
(3)
303 where m, m
moment of inertia respectively, and
are the lumped mass and the
T, is the transformation matrix defined
(
where C , C
are the coordinates of the center of gravity of the i
floor. The
transformed individual matrices can be combined into the overall matrix easily.
The equations of motion of the system are
[M]
{D}
[C]
{D}
[K]
{F}
{D}
(5)
where mass matrix [M] is composed of the subraatrices M, on the diagonal. [C] is
the damping matrix, [K] is the stiffness matrix and {D} is the relative displacement
vector.
The overhead dot indicates time derivative.
(F) = - I T ] '
[M2]
The force vector {F} i3 defined
(6)
{G}
where [] consists of submatrices
on the diagonal.
[M ] is composed of submatrices
M, on the diagonal, and {} is the ground acceleration vector.

The solution method of Weaver et. al. [39] will be followed here. We first
solve for the undamped natural frequencies and mode shapes from the free vibration
equations.
[M]
{D}
[K]
{D}
{0}
(7)
For the ease and efficiency of computer programming, the mass matrix is decomposed
into lower and upper triangular matrices,
[M] =
[U]
[U] =
[MC]'S
(8)
[U]
[T]
(9)
Substituting eq. (8) into eq. (7) and premultiplying both sides by ([U] ),-1
~ we have
([u]')"1 [u]' [u] (D} + ([]')"1 ] [] -1 [u] (D) = {0}
(10)
(Ut)
Cu J
[Kr]
Dr)
[0}
304
where
{Dj} =
[U]
{D}
(12)
and
([U]')"1
[Kj] =
[] [ U ] 1
(13)
Let the normalized eigenvectors of [ ] "be [V]. Premultiplying both sides of eg.
t
(il) by [V] and making use of the orthonormal property of the eigenvectors, we
have
tv]' (Dj.) + iv]* n y Iv] IV]' {Dj} = {0}
+ [
(15)
{0}
{D }
N =
(11,)
where
{DN) =
[ V ] ' {Dj} =
[ V ] ' [U] {D}
(16)
and
t \ j
= [V]' [Kj.] [v] = ( [ u ] _ 1 [ v ] ) ' [] t u ] " 1 [V]
[ ^] a r e t h e e i g e n v a l u e s of t h e system and [U]
(IT)
[V] a r e t h e e i g e n v e c t o r s of [K].
Assuming p r o p o r t i o n a l damping and t r a n s f o r m i n g e g . ( 5 ) i n t o t h e normal c o o r d i n a t e s

{D } , we have
where
damping.
{DN} + f c N J {5N) + fx N J {DN} = ( [ u r V ] ) ' {F}
(18)
[ \ j
(19)
2(5 [ \ j *
2 [ \ j
is the circular frequency of

The J
mode, and is the percentage of critical
component of eg. (10) becomes
j + 2 "j 5 j + a] Dj " " Y J 'i '<*>

Where
<2>
is the participation factors defined as

Yj = ([U]_1[V])j
[T]' [MC] {e}
(21)
305
and the ground acceleration vector G) is replaced by
{0} = G
o
and G
f ( t ) {e}
(22)
is the maximum ground acceleration, f(t) is the time function of ground motion,
and {e} is the earthquake directional vector.

If one wants to adopt time history of strong motion earthquake as input,
then eg. (20) can be used directly to solve for the modal response.
If one wants
to use the design spectrum method, then the solution of eg. (20) with zero displacement
and velocity as initial conditions and with small damping 12] is
D,
= ! i V G - ^ ( _ ) f(-r) Sin , (t-x)dx
(23)
i J
Defing the spectrum value as 120]
S () = [*0
f(T) e"uJ
(t_x)
Jo
(Sk)
Sin U(t-T)dx]max
the maximum modal response is
"
^-Sa'"j
(25)
The responses in the common coordinates are

(D) = [U]" 1 [V] {D,,}
(26)
Since the maximum response do not occur at the same time, the method of the square
root of the sum of squares is taken to combine the maximum modal responses.
results of the example are shown in Tables I to IV
Some
and the input design spectra
are shown in Fig. 2.

6.
EQUIPMENT DESIGN
For equipment with mass small compared with the mass of the structure,
it has generally been acceptable to use floor design spectra as input for aseismic
design.
The floor design spectra can be generated either with time history of ground
motion as input [6] or with ground design spectra and structural response as input
[1+8, 1*9, 50]. Were the torsional effect of an asymmetric structure not considered,
only one floor design spectrum would be obtained at each floor level.
However with
the inclusion of the torsional effect, two floor design spectra corresponding to
two translational degrees of freedom of the floor are obtained for each directional
input.
Generally speaking, the magnitude of the extra design spectrum depends on
the characteristics of the structure.
These spectra are shown in Fig. 3 and h>
and they should be applied simultaneously for the design of equipment.
306
ACKNOWLEDGMENT
Thanks axe due to Dr. G. J. Patterson, research engineer of Gilbert Associates,
Inc., for some stimulating discussion, and to Mr. M. Plica, project structural
engineer of the same company, for preparing the mat hematic model in the example.
307
REFERENCES
1.
Biggs, J.M. Introduction to Structural Dynamics McGraw Hill 196k.
2.
Hurty, W.C. , M.F. Rubinstein, Dynamics of Structures, Prentice Hall I96I1.
3.
Lin, . K. Probabilistic Theory of Structural Dynamics, McGraw Hill I967.
1*.
Blume, J.., N.M. Newmark, L.H. Corning, Design of Multistory Reinforced Concrete
Buildings for Earthquake Motions, Portland Cement Association 1961.
5.
U.S.A.E.
C . (Division of Technical Information), Nuclear Reactor and Earthquake,
TID702I, August 1963.
6.
U.S.A.E.
C . (Division of Technical Information), Summary of Current Seismic Design
Practice for Nuclear Reactor Facilities, TID25021, September 1967
7.
HouBner, G. Vf.. "Engineering Estimates of Ground Shaking and Maximum Earthquake

Magnitude", U^" World Earthquake Engineering Conference, Vol. I, Santiago, Chile
1969.
.
Lomenick, T.F. and NSIC Staff, Earthquake and Nuclear Power Plant Design, Nuclear
Safety Information Center, Oak Ridge National Laboratory, 0RNLNSIC28, July 1970.
9.
Housner, G.W. , "Strong Ground Motion," "Earthquake Engineering", Edited by R. L.

Wiegel, PrenticeHall, Inc. 1970.
10.
Eringen, A. C. "Response of Tall Buildings to Random Earthquakes", Proceedings of

3 r d U.S. National Congress of Appi. Mech., June 1958.
11.
Goldberg, J.E., J. L. Bogdanoff, D. R. Sharpe, "The Response of Simple Nonlinear

Systems to a Random Disturbance of the Earthquake Type". Bulletin of the
Seismologies! Society of America, Vol. 5^ , No. 1, February 196b.
12.
Penzien, J., "Applications of Random Vibration Theory", Earthquake Engineering,

Edited by R. L. Wiegel, PrenticeHall, 1970.
13
Geto, ., K. Toki, "Structural Response to Nonstationary Random Excitation",

U t h World Earthquake Engineering Conference, Vol. I., Santiago, Chile 1969.
lb.
Penzien, J., S. C. Lin, "Nondeterministic Analysis of Nonlinear Structures

Subjected to Earthquake Excitations", litn World Earthquake Engineering Conference,
Vol I, Santiago, Chile 1969.
15.
Kouskoulas, V. J., W. C. Hurty, "Probabilistic Description of the Response of

Linear Structures to Random Inputs", Journal of Applied Mechanics, September 1970.
16.
Ilarayana Iyengar, R. , K. T. Sundara Raja Iyengar, "Probabilistic Response Analyses

to Earthquakes", ASCE, Engineering Mechanics Division, June 1970.
17.
Drenick, R. F., "Modelfree Design of Aseismic Structures", ASCE National

Structural Engineering Meeting, Portland, Oregon, April 1970.
18.
Shinozuka, ., "Maximum Structural Response to Seismic Excitations", ASCE,

Engineering Mechanics Division, October 1970.
19.
Strong Motion Earthquake Accelerograms, Digitized and Plotted Data. Earthquake

Engineering Research Laboratory, California Institute of Technology, July 1969.
20.
Alford, J. L. , G. W. Housner, R. R. Martel, Spectrum Analyses of StrongMotion

Earthquakes, Earthquake Engineering Research Laboratory, California Institute of
Technology, Revised August 1961*.
308
21.
Newmark, N. M., W. J. Hall, "Seismic Design Criteria for Nuclear Reactor

Facilities," Fourth_ World Conference on Earthquake Engineering, Vol. II, Chile,
1969.
22.
C ornell, C.A. , "Design Seismic Input," Seismic Design for Nuclear Power Plants,
Edited by R. J. Hansen, The M.I.T. Press, 1970.
23.
Housner, G.W., "Design Spectrum," Earthquake Engineering, Edited by R. L. Wiegel,

PrenticeHall, 1970.
2U,
Jennings, P.C., G.W. Housner, N.C. Tsai, Simulated Earthquake Motions, Earthquake
Engineering Research Laboratory, California Institute of Technology, 1968.
25.
Barkan, D.D., Dynamics of Base and Foundations, Translated from Russian by L.

Drashenska, McGrawHill, I962.
26.
Whitman, R.V., F.E. Richart, "Design Procedures for Dynamically Loaded Foundations,"
ASCE, Soil Mechanics and Foundations Division, Nov. 1967
27.
Whitman, R.V. , "Soil Structure Interaction," Seismic Design for Nuclear Power
Plants, Edited by R.J. Hansen, The M.I.T. Press, 1970.
28.
Werner, S.D., A Study of Earthquake Input Motions for Seismic Design, prepare for
U.S.A.E.C. By Agbabian Jacobsen Associates, R691^925, June 1970.
29.
Parmelee, R.A. , "Building Foundation Interaction Effects," ASCE Engineering

Mechanics Division, April 1967
30.
Scavuzzo, R.J. , D.D. Raftopoulos, J. L. Bailey, Lateral Structure Foundation

Interaction of Nuclear Power Plants with Large Base Masses, U.S.A.E.C. T1D25391*,
September I969.
31.
C astellani, ., "Foundation Compliance Effects on Earthquake Response Spectra",

ASCE, Soil Mechanics and Foundations Division, July 1970.
32.
Parmelee, R.A., D.S. Perelman, S.L. Lee "Seismic Response of Muliple Story
Structures on Flexible Foundations", Bulletin of the Seismologicai Society of
America, Vol. 59, No. 3, June 1969.
33
Salvador!, M.G., "Earthquake Stresses in Shear Buildings", Transactions of ASCK,

Vol. 119, 1951*.
3I+ Blume, J.., "Structural Dynamics of CantileverType Buildings", Fourth World

Conference on Earthquake Engineering, Vol II, Chile 1969.
35
C lough, R.W., E.L. Wilson, I.P. King, "Large Capacity Multistory Frame Analysis
Program", ASCE, Structural Division, August I963.
36.
C lough, R.W., I.P. King, E.L. Wilson, "Structural Analysis of Multistory Buildings",
ASCE, Structural Division, June 196*4.
37 Kokinopous , F.E. , "Aseismic Dynamic Design of Multistory Systems", ASCK, .'itrm: turai
Division , June 1966.
38 . Weaver, W. , Jr. , M. F. Nelson , "Three Dimensional Analys is of Tier Building" , A: 1','K,
Structural Division , December 1966 .
39
Weaver, W., Jr., M.F. Nelson, T.A. Manning, "Dynamics of Tier Buildings", ASCK,
Kngineering Mechanics Division, December iy6ii.
1+0.
Weaver, W. , Jr. , R.G. Oakberg, "Analysis of Frames with hear Walls by Finite
Elements", Proceedings of Symposium on Application of Finite Kiemcrit^Method3 in
Civil Engineering, Vanderbilt University and ASCE, November l'j69.
309
Ui.
Manning, .., Jr., The Analysis of Tier Buildings with Shear Walls, Ph. D.
Dissertation, Stanford University, April 1970.
!2.
Winokur, ., J. Glck, "lateral Loads in Asymmetric Multistory Structure",

ASCE. Structural Division, March 1968.
3.
Glck, J., "Lateral-Load Analysis of Asymmetric Structures", ASCE, Structural

Division, February 1970.
*.
Wynhoven, J.H. , P.F. Adams, J.K. Biswas, W.K. Iso, Discussion of "Lateral-Load
Analysis of Asymmetric Multistory Structures", ASCE, Structural Division,
November 1970.
5.
Bergstrom, R.N. , S.L. Chu, R.J. Small, "Dynamic Analysis of Nuclear Power Plants
for Seismic Loading", Presented at the ASCE Annual Meeting, Chicago, Illinois,
October 1969.
U6.
Plica, ., C. Chen, Dynamic Analysis of the Auxiliary Structures, Gilbert

Associates, Inc., Reading, Pennsylvania, Report No. 17^8, January 1971.
hj.
Ayre, H.S., "Interconnection of Translational and Torsional Vibrations in

Buildings", Bulletin of the Seismic Society of America, 28, 1938.
h.
Biggs, J.M., J.M. Roesset, "Seismic Analysis of Equipment Mounted on a Massive

Structure", Seismic Design for Nuclear Power Plants, Edited by R. J. Hansen, The
M.I.T. Press, 1970.
U9.
Amin, M., R.P. Kassawara, W.J. Hall, "Response of MDF-Systems Using Spectra",
Submitted to Structural Division of ASCE, 1971, Publication Pending.
50.
Amin, M., W.J. Hall, N.M. Newmark, R.P. Kassawara, "Earthquake Response of
Multiply Connected Light Secondary Systems by Spectrum Methods", to be appeared
in Proceedings of ASME First National Congress of Pressure Vessel and Piping
Technology, San Francisco, May 1971.
- 310
TABLE I - MODAL FREQUENCIES
f ICENYA4.UtS
PREOuewciii-oiD/stc
FRfoueKiti-crs
naioossic
0.1*4047910
Ol
14.101
2.2*1
0.441
0 . 1 1 1*1 6 JO
Ol
17.44*
t.ao*
0.116
O.U4M
OS
24.446
1.476
O.SI
0.10*4**4*0
Ol
24.424
6.1)
0.116
0.41*111*10
Ol
10.110
6.111
0.20T
0.142**1 MD
0*
17.771
6.011
0.166
0.2TT0I161D
04
92.4)2
.t
.ll*
0.15147740
0*
94.020
4.31
0.101
0.4791410
4*.81
11.11*
0.040
0.9*79401*0
04
14.412
12.200
0.0i
0.IS470I71D
0*
2.11
14.711
0.061
0.097)0
0*
1.170
14.12*
0*0*1
0.10140*420
01
100.701
14.021
0.067
.*
01
101.00
17.201
0.011
0.1217*9*10
01
110.947
11.441
0.037
0.1*4400000
OS
114.017
1.49
0.036
0.1M44114O
OS
114.949
21.119
0.066
0.217792*40
Ol
147.949
11.4*4
0.0
0.220**70
03
192.941
14.247
0.O*l
0.212471030
01
194.091
21.114
0.0*0
0.274*49*10
03
149.124
14.174
O.Ol
0.10414*210
03
174.447
27.741
0.016
0.99*097090
03
1*4.4**
10.142
O.Ol
0.401474110
01
200.41*
1.141
0.01
0.41kl**4D
03
I0S.014
11.241
0.010
0.41499S91D
03
211.2**
9.461
C.Ol
0.9047270
OS
224.476
19.71
0.0]
0.12M40710
Ol
27*.*41
14.401
O.OJT
0.110770420
03
240.441
*.*
0.016
0.4*0101400
03
142.774
41.011
O.Ol*
0.071
0.744*11700
01
172.414
41.460
0.042141120
03
2*1.4*4
46.7*1
0.071
O.V1O979470
03
101.717
41.027
0.071
O.*l*7710
Ol
04.474
6.770
0.P71
.*1740770
03
114.101
10.021
0.070
0.101*441*0
04
121.6*6
11.09
O.Ol
0.1141049ID
06
11.0*
91.04
0.011
0.I211I214D
06
14*.012
93.114
O.Ol
0.1313 .7 760
0*
171.941
14.631
0.017
0.11*140770
06
144.430
41.173
0.01
0.1*0024170
4)1.20
44.1*0
O.Ol*
0.2407141*0
06
440.447
10.0*0
0.013
DYNAMIC RESPONSE IN MODE

MODE
OPKT
SHAPE
3.19717D-05
7.5615JD-06
9.28002D-07
9.576830-05
1.05763D-05
2.698390-06
1.772440-02
7.581860-0*
5.441440-04
9.70*720-05
1.08*370-05
2.732180-06
3.881310-06
3.*56610-06
9.*8686D-08
1.3027*0-05
9.087700-06
1.576930-07
2.*8837U-05
1.666620-05
1.871**0-07
*.150*70-05
*.18725D-05
6.080660-07
3.99*13D-06
*.7B625D-06
'5.108920-08
1.25662D-05
8.793690-06
2.102980-08
1.285*0D-05
1.2821SO-05
8.3933*0-08
3.888770-02
1.083560-01
7.662960-0*
4.23545D-02
1.387230-01
8.396810-04
4.565360-02
1.681*90-01
9.09755D-04
I FOR 0.27 G EARTHQUAKE AT THE ORIGIN OF THE GLOBAL COORDINATES
IN X QUAKE
CM,RAD
3.502*3D-05
8.28353D-06
1.016610-08
1.0*9120-0*
1.158610-05
2.956020-08
1.9*1670-02
-8.305760-0*
5.960980-06
1.06313D-0
1.187910-05
2.9930*0-08
*.251890-06
3.7866*0-06
1.03926D-09
l.*27l2D-05
9.95537D-06
1.727*90-09
2.72595D-05
1.8257*0-05
2.05012D-09
*.5*6750-05
*.5870*0-05
6.661230-09
*.375*80-06
5.2*3230-06
-5.59671D-10
1.376600-05
9.633300-06
-2.303770-10
1..00130-05
1.404570-05
-9.19*72D-10
-*.260060-02
1.187010-01
-8.39*600-06
-*.6398*D-02
1.519680-01
-9.198520-06
-5.001260-02
1.8*20*0-01
-9.966160-06
0.5322
SPECTRUM RESPONSE
PARTICIPATION FACTOR IN X OUAKE PARTICIPATION FACTOR IN Y OUAKE
OPNT
I N y auAKE
CM,RAO
9.03*600-0*
2.1367SD-0*
2.622360-07
2.706230-03
2.988660-0*
T.62512D-0T
5.008570-01
-2.1*2*90-02
1.537650-0
2.7*2370-03
3.06*230-0*
7.72061D-07
1.096790-0*
9.76T72O-05
2.680800-08
3.681280-0
2.568010-0
.56090-08
7.031660-0
.7095D-0
5.28B3D-08
1.172840-03
1.1832*0-03
1.718280-07
1.12866D-0
1.352500-0
-l.36BD-08
3.55096D-0
2.8930-0
-5.9*2620-09
3.63230D-0*
3.623110-0*
-2.37180D-0B
-1.09889D 00
3.061930 00
-2.16941D-0
- 1 . 1 9 6 8 6 0 00
3.9200*0 00
-2.37278D-0*
-1.29008D 0 0
.751570 00
-2.57079D-0
0.4181
10.7847
ACC
IN X OUAKE
N.RAO/SECSO
6.971500-05
1.6B82D-05
2.023530-06
2.088250-0*
2.30619D-05
5.88389 0-06
3.86484O-02
-1.65320-03
1.186520-03
2.1161*0-0
2.36450D-05
5.957570-06
.463290-06
7.537210-06
2.068630-07
2.8*06*0-05
1.98159D-05
3.*3BS2D-07
5.2590-05
3.63*090-05
.080720-07
9.050190-05
9.130380-05
1.325900-06
8.709280-06
1.043650-05
-1.114010-07
2.74008D-05
1.917*80-05
-.585590-08
2.802850-05
2.79576O-05
-1.830190-07
-8.7955D-02
2.362720-01
-1.670920-03
-9.235480-02
3.024880-01
-1.83090-03
-9.95487D-02
3.666530-01
-1.983740-03
ACC
IN r OUAKE
N.RAO/SECSQ
1.798310-03
.253160-0
5.2197*D-05
5.386680-03
5.948850-04
1.517760-0
9.969430-01
-.26*570-02
3.060650-02
5.*5862D-03
6.099280-04
1.536770-0
2.18312U-0
l.9*2*D-04
5.336070-06
7.327500-0
5.111560-04
8.86970-06
I.399630-03
1.374220-04
1.0526 30-05
2.3 14520-03
2.355200-03
3.420190-05
2.246580-04
2.69212U-0
-2.873620-06
7.068100-0
4.946 190-04
-1.18286D-06
7.230010-0
7.211720-0*
-4.72100D-06
- 2 . 1 8 7 3 2 0 00
6.09*690 00
-.310180-02
- 2 . 3 8 2 3 1 0 00
7.802730 00
-.72295D-02
- 2 . 5 6 7 8 8 0 00
9.457890 0 0
-5.117100-02
y-i
H
I
w
M
o
co
M
H
S
O
0
c
DYNAMIC RESPONSE IN MODE
2 FOR 0.27 G EARTHQUAKE AT THE ORIGIN OF THE GLOBAL COORD INATES
OPMT IN X QUAKE
CM,RAO
MOOE SHAPE
1.85667005
2.26575D06
2.73510008
4.60468005
4.03470D06
1.36048007
7.94497003
2.12613003
4.29429D05
4.71816D05
4.10144006
1.52240007
1.99599D06
9.4179D07
1.0911008
6.920T1D06
2.32523D06
2.65915008
1.36808D05
12361D06
1.10580D07
2.21312D05
8.06251006
1.39143007
1.70779006
6.B6426D07
9.8725BO09
5.91667006
1.47659D06
2.70582008
5.71051006
1.09908006
2.11576D08
2.34821002
2.14935002
3.76814D04
5.91429002
2.43537002
4.39135004
9.56988D02
2.70828002
4.99324O04
OPMT I N Y QUAKE
CM,RAD
ACC IN X OUAKE
M.RAO/SECSQ
ACC I N Y QUAKE
M.RAD/SECSQ
1.08012007
1.31811008
1.59116010
2.67879007
2.3*720008
7.91*63010
*.62201D05
1.23668005
2.*9822D07
2.7**81D07
2.38603008
8.85660010
1.16117D08
5.*7892D09
6.3*932011
*.02615008
1.35271D08
1.5*697010
7.9588*D08
2.39892008
6.*3300D10
1.287*9007
.69039008
8.09*69010
9.935UD09
3.99331009
5.7*341DU
3.4*20*008
8.59010009
3.B0075D04
4.63816D05
5.59897D09
9.*261D04
B.25935D05
2.78500008
1.62640001
4.35235002
8.79074006
9.65B43D04
8.39597D05
3.11646008
4.08594005
1.92792D05
2.23420D09
1.41672004
4.75992005
5.44349D09
2.80056004
8.44134005
2.26365D08
4.53043004
1.65046004
2.84836D08
3.49597D05
1.40517D05
2.02099009
1.21119004
3.02269005
5.53902009
1.16B98004
2.24990005
4.33111D09
4.80698001
4.39989D01
7.71367D05
1.210700 00
4.98538D01
8.98944005
1.959030 00
5.54406001
3.6752008
.23151D09
5.10808013
8.59970008
7.53521009
2.50B3D12
1.48380D05
3.97076006
8.02001010
8.8U63D08
7.65986009
2.84323012
3.72771009
1.75889D09
2.03832013
1.29251D0B
4.34260009
.96623013
2.55502008
7.70124009
2.06518012
4.13323D0B
1.50575008
2.59863D12
3.18946009
1.28197D09
1.B4380013
1.10500008
2.75767009
5.0533B013
1.06649008
2.05264D09
3.95138D13
4.38552005
4.01413005
7.03738009
1.1045500
4.54829005
8.20129009
1.78727D04
5.05798005
1.18393003
1.4*7800
1.7**07006
2.93622003
2.5727700*
8.6752*006
S.06619001
1.35575D01
2.73830003
3.00858003
2.6153300
9.70772D06
1.2727600
6.00544005
6.95950D07
.130600
l.8271O0
1.6956*006
8.72369D04
2.62946D04
7.05122D06
1.41122003
5.1*ll*D0*
8.87260006
1.0889900*
*.37707005
6.29536007
3.77283D04
9.*1561005
1.72539006
3.6*13600*
7.0083BD05
1.3*913D06
l.*97360 00
1.370560 00
2.40279002
3.77131D 00
1.552940 00
2.80019002
6.10234D 00
1.726960 00
1.02215D04
9.32537D09
3.18399D02
SPECTRUM RESPONSE
PARTICIPATION FACTOR IN
PARTICIPATION FACTOR IN
62*9
OUAKE
CUAKE
10.4121
0.0009
1.57*12D10
3.32211008
6.39392D09
1.23085D10
I.3660800*
1.25039D0*
2.19213D06
3.**066D0*
l.*167800*
2.55*69006
5.56731D0*
1.5755500*
2.90*83D06
313
TABLE IV SRSS OF ALL THE MODES
SASS OF THE DYNAMIC RESPONSE OF ALL 42 KODES FC 0.77 C EARTHQUAKE AT THE CENTERS OF HASSES
155
1
2
3.
A.
Y
6
l
8.
,
10
l
12
13
14
D PKT I H QUAKE
CM.RAD
1.
f
'.i
.43272E02
472*2E02
i 74103E06
2. 4 T M T I 0 2
2.96096E02
I.20494E09
l.93730C0I
.42016E01
; .99746E04
! 499MC02
.10M2E02
' . 14)117E09
.61293E03
3442*103
.I7184E0*
.971B9E02
.21434E02
.89436E0*
.09496C02
.3979BE03
. U I M H
.2B.24E02
.97B41E02
29342E03
.69732E03
.32361E03
41393E07
30230E02
.69I09E03
34I19E06
74424C02
34499E02
.42370E05
.3)4*78101
.31749101
59497104
67920E 00
.19282E01
32770E04
.49361E 00
.90463E01
.15207E04
OFHT I N V OUAKE
CM.RAD
..47310E-03
.75404-02
.430SOE-0
.19094E-02
E.9*997E-02
r.ei793E-06
.79691E01
.646201
I.O2369E04
F.93469E02
.51996E01
.791909
.4327203
3949202
.2986306
.94029E02
I.UI23E02
.984470*
.9S127E02
U171Z2E02
).87046E0A
I.T5116E02
I.414T6E02
.2780709
I.945A6E03
I.50792E03
.*875flf06
.7860302
1630202
.293280*
269702
.30753E01
.9877409
.92301E01
. 7 I 9 0 2 E 00
[.33790E04
..31634E02
.52034E 00
2339204
.0462SE01
.23833E 00
.01993E04
ACC IM I QUAKE
M.KAD/SfCSt)
ACC I N T OUAKE
H.*AO/SECSQ
1.82T23E 00
1.477&1E 00
0.13436E02
3.08433E 00
2 . 9 1 4 8 3 E 00
1.08582E01
5.40064E 00
3.T2683E 00
2.18778E01
7.91401E 00
9.81483E 00
9.47910E01
1.47108E 00
8.89290E01
3.35995E02
2.47846E 00
1.50T22E 00
6.89492E02
3.8940AE 00
1 . 0 7 9 7 4 F 00
9.B053&E02
4 . 74491E 00
2.33699E 00
1.47026E01
1.0175ftE 0 0
9.03612E0I
2.1633BE02
2.43747E 00
9.65603E01
4.9U49E02
9.20042E 00
2 . 9 3 3 3 6 E OO
2.31260101
3.44990E 00
2.32779E 00
1.0203IE01
3.942T1E 00
2.48914E 00
1.56869E01
7.84461E 00
1.31349E 00
2.336&0E01
8.T0&29E01
1.T9T10E 00
4.91624E02
1.43862E 00
2.99933E 00
T.6B968E02
3.90311E 00
6.32192E 00
1.22092E01
4.00987E 00
1.98430E 01
S4089E01
T.94132E01
1.80269E 00
3.647T3E02
1.87T40E 00
3.40743E 00
7.98498E02
2.77740E 00
3.36743E 00
1.08639E01
3.44427E 00
3 . 6 1 7 7 0 : 00
1.40448E01
8.36103E01
1.94644E 00
4.0-2
1 . 7 76 7 II
00
1.39977E 00
6.962B0E-02
9.17411E 00
6.97092E 00
3.300&0E-01
1.296BE 00
4.03112E 00
1.16914E-01
7.9DS83E-01
9.277S3E 00
1.72129E-01
S.313B3E-01
6 . 9 9 4 3 9 E 00
2.46788E-01
314 -
EL. 32.3 <
^^7<
^WW
EL. 17.0
>S
<. s&u^^St
^^=^f^
y>
J
-ORIGIN-FOR
DYNAMIC MODEL COORD.
EL. 4.0
FIGURE 1 -
DESIGNATE PO'NT SETS
MATHEMATICAL MODEL
315
FIGURE 2 - GROUND DESIGN SPECTRA
316
DI PHRAGM NO. 8
EQUIPA DA MPING 0 5%
\STRUCT URAL DA MPING 5/o
1
QUAKE X RESPONSE
\
0 27 G EA RTHQUA KE
~ ^
<
.01
.02
04
.06 08.10
.20
40
.60 80 1.0
PERIOD ISECONDI
FIGURE 3 EXAMPLE OF FLOOR DESIGN SPECTRUM
>
DIAPHRAGM rIO 8
EQUIPMENT DA MPING 0 5%
STRUCTURAL DA MPING 5/o
Y QUAKE Y RESPONSE
r\T7C
1 ,\u 1 I K .11 1
1
y
01
.02
04
.06 08 10
20
40
60 80 10
PERIOD ISECONDI
FIGURE - EXAMPLE OF FLOOR DESIGN SPECTRUM
- 317 DISCUSSION
P. C. RIZZO, U. S. A.
If you used design s p e c t r a approach for s t r u c t u r e , what method was used to
generate floor response s p e c t r a ? If Biggs' method is used, for what plants (foreign and
U. S. ) has it been used successfully ?
.
Ch. CHEN, U. S. A.
Response spectrum method (Biggs' method) was used to generate floor response
c u r v e s . In the U. S. this method was applied successfully in a nuclear power plant in Pennsylvania (I don't want to mention the name of the plant without our client's permission). This
method was also applied to a job in Japan,
H. SATO, Japan
I would like to ask a question about Figs. 3 and 4. It s e e m s to me that the same
sort of frequency curves which were given in these figures is too wide at the peak frequency
even if the value of damping ratio is taken into consideration.
How did you obtain these r e -
sults ? I would like to know the quantitative c h a r a c t e r i s t i c s .

Ch. CHEN, U. S. A.
These r e s u l t s were obtained from the envelope of a p a r a m e t r i c study. Because
the dynamic property of the concrete v a r i e s over a wide range of values.
J . D . STEVENSON, U . S . A .
The response s p e c t r u m curve you present is actually an envelope curve r a t h e r
than an actual response curve since only one of the p a r a m e t e r s your have considered can
actually exist. Is that c o r r e c t ?
A
Q
Ch. CHEN, U. S. A.
Yes, it is an envelope.
O. HENSELEIT, Germany
What a r e the n u m e r i c a l values of dynamic Young's modulus you used in your
calculation ?
Ch. CHEN, U. S. A.
We used the static value of 210,000 k g / c m
to a dynamic value of 300,000 k g / c m 2
- 318
and s o m e t i m e s higher value.
K. AKINO, Japan
I have two questions which a r e related. The first is, in the U. S. A. who provides
a conceptual design of s t r u c t u r a l layout; in your c a s e , Westinghouse or Gilbert ?

The second i s , if you a r e given the s t r u c t u r a l layout shown in Fig. 1 in your paper, you have
to c a r r y out unreliable complicated dynamic analysis including torsional vibration mode.
However, a s t r u c t u r a l engineer can provide a better balanced and m o r e stable structural
layout than you have now. Which is the better way ? Either you c a r r y out the analysis as you
do or s t r u c t u r a l engineers provide a m o r e adequate s t r u c t u r a l layout for the a s e i s m i c design ?
By the way, do you know an ideal proposal prepared by Kaiser Engineers which is axisymnictr i c a l c i r c u l a r r e a c t o r and auxiliary buildings.
.
Ch. CHEN, U. S.A.
The layout is decided by Gilbert. It is true that some of the complex analytical
methods can be avoided by different design layout, as using expansion joints to separate buildings, or making the building reasonably symmetric to avoid the coupling effect of torsional
modes and translational modes. But in this specific job, due to the limitation of site condition
and other reasons the coupling effect is almost not avoidable. As a matter of fact, on the
other jobs we do p r e p a r e the layout in such a way as to avoid the coupling effect.
K. SHIBATA, Japan
It s e e m s very dangerous to put the result of a p a r a m e t r i c study on the response
analysis curve. We should consider the margin of design in the view-point of the whole dynamic design, and not include it in your response curve or estimation of damping co-efficient
so on.
Ch. CHEN, U. S. A.
Yes, I agree the p a r a m e t r i c study should be applied in a broader sense instead
of on each individual case to avoid the conservatism on top of conservatism.
K 4/6
DYNAMIC ANALYSIS OF VITAL PIPING SYSTEMS SUBJECTED

TO SEISMIC MOTION
CH. CHEN,
ABSTRACT
The linear dynamic analysis of the three dimensional piping system of a nuclear
power plant is based on a lumped parameter model.
spectrum input are used and discussed.
Both time history input and design
Also both dynamic stiffness matrix and
flexibility matrix are employed in the eigenvalue problem and in calculating internal
stresses and support reactions.
The interaction of piping with the structure can be
treated either by floor design spectrum or by combining them into one model.
1,
INTRODUCTION
The early development of dynamic piping analysis was carried out by Naval
engineers [l]. To assure public safety, the designers of the nuclear power plants
In such countries as the U.S.A. and Japan have also been required to perform dynamic
analysis of piping subjected to seismic motion [2, 3, U].
Since the state-of-the-art
of the subject has constantly been improved, it is the purpose of this study to
summarize the improved analytical procedures which emphasize economy of computation
time and accuracy of responses.
The interaction of the piping system with the structure plays an important
role in the response of the system.
Several methods are available to treat this problem.
The conservatism of the result depends on how each method is properly applied instead
of on which method is applied.
?..
EQUATIONS OF MOTION AND SYSTEM RESPONSE

The equations of motion based on a lumped parameter model of a three dimen-
sional piping 3ystem subjected to seismic input may be written as:

[M] U ) + tC] (x) + [K] {x} = -[M] ty}
where the system is assumed to be linear elastic.
(1)
The symmetric, positive definite mass
matrix [M] is diagonal for lumped mass approach but non-diagonal for consistent mass
approach [5]. The positive definite stiffness matrix [K] and the positive senidefinite
320
damping matrix [C ] are both symmetric,
{x} is the relative displacement vector.
dots over the variables indicate time derivatives,

at the support.
The
{y} is the input acceleration vector
The piping response is affected by the mathematical model used, and
some modelling consideration vas discussed by Harrington and Vorus [6j.

Either the time history of the support acceleration or the floor design
spectrum can be used as input.
In view of the fact that hundreds of piping systems are
analyzed dynamically in a typical FWR plant, and that the time history analysis is time
consuming, it is more practical to use floor design spectrum as input.
Of course time
history analysis has the advantage of providing resposes as a function of time on

condition that the mathematical model is correct and that the assumed material properties
and time history input are exact.
If the time history is used as input, eq. (l) can be solved either by direct
time integration [7 , 93 or by superposition of normal modes. With either the time
history method by superposition of normal modes or the design spectrum method, the
eigenvalues and eigenvectors of the system have to be solved first.
They are obtained
by solving the free vibrational equations.
{} + []" 1 [] {} = {0)
] []
(2)
() + () = )
(3)
where [A] is the symmetric, positive definite flexibility matrix of the system.
Methods
of deriving stiffness matrix or flexibility matrix were discussed by Przemieniecki [10],

It is well known that the stiffness matrix requires much less time to derive.
Various
techniques of solving eqs. (2) and (3) are available [11, 12]. The eigenvalues of
[Ml
[] are the squares of the system circular frequencies, whereas those of [A] [M] are
the inverses of the squares of the system circular frequencies.

Rayleigh [13] showed that the sufficient condition for a damped system to
prossess classical normal modes is that the damping matrix is a linear combination of
mass matrix and the stiffness matrix.
Caughey [X-U ] pointed out that the necessary and
sufficient conditions for the existence of classical normal modes are that the damping
matrix be diagonalized by the same transformation that uncouples the undamped system.
So after obtaining the normalized eigenvectors [] of eqs. (2) or (3), we can apply the
orthogonal transformation
(x) = [] ()
t o eqs. ( l ) , and premultiply both sides of the equations by the transpose of [].
the uncoupled equations are
{} + ( + ] ) {rl} + [ \ ]
= - [ ] [M] {y}
(5)
Hence
321
where we made use of the orthonormal conditions

[] [] [] = [ Y ] ,
(6)
[] [] [] f ? ] ,
()
and the proportional damping relations

[C] [M] + [].
The i
(8)
component of eqs. (5) is

^ + (C + cui^) f\ + ^ 1 - 1 ) [M] {y}
where {.} 1 ie the transpose of the i
(9)
column of []. If we define the percentage of
critical damping as

* ""i
~~2^
(io)
then eq. (9) becomes

1 + 2u i n i + ^1 = { 1 ) [M] {y}
(11)
If time history of support acceleration is chosen as input, eq. (11) can be applied
directly to obtain the modal response as a function of time.
If design spectrum is
chosen as input, we can replace the support acceleration vector as

(y) y 0
where y
(d) f(t)
(12)
i s the maximum support a c c e l e r a t i o n , (d) i s t h e earthquake d i r e c t i o n v e c t o r and
f ( t ) i s the time function o f support a c c e l e r a t i o n .

have the same time f u n c t i o n .
[15,
16].
Here we assume t h a t a l l
supports
Supports of d i f f e r e n t time f u n c t i o n s can a l s o be handled
Furthermore d e f i n e the p a r t i c i p a t i o n f a c t o r , , as
ti
= ( ^ 1 [M] {d}
(13)
which is a measure of the extent to which the i
normal mode participates in
synthesizing the total loads on the system [17]. With the initial conditions
n(o) = () = o,
(lit)
the solution of eq. (11) is [17]
li
(t)
" ""2
e , u i ( t - T ) f(T) Sino^t-T) dT
(15)
322
ing.rJl
-
"'
where we assumed small damping,^1
* 1.
Let the acceleration response spectrum
value be [18, 19]
Sa(u) = y
o
[/e
o
^ ^
f (T)Simo(tt) di]
max.
(lo)
Then the maximum modal response is

. Sa
(ni)max = J ^
i
(IT)
l h e maximum d i s p l a c e m e n t s a t each d e g r e e of freedom in i
mode a r e
(x.)
= (.)
{.}
max
max
(13)
If one is interested in the maximum equivalent forces applied at each degree of freedom
the maximum absolute modal acceleration will be derived first.
(rl
i+Yiyof(t))max
Then the maximum forces i n i
= Y
It is [.'O]
iSa
(U,)
mode are
{F.} = Y.Sa[M] {.} .
(20)
This set of forces may also be obtained from maximum displacements and system stiffness
matrix,
CF.}
= (n.)
[] {.}.
L max
1
(21)
The internal stresses and support reactions for each mode are obtained by applying
statically either the set of displacements of eqs. (13) or the set of farces of eqs. (,'U)
or (21) to the system.
The most probable response of the system is the square root of
the sura of squares (SRSS) of all the contributing modes.
It hod been general practLce
in the past to apply the SRSS of the displacements or the equivalent forces to obtain
the internal stresses and support reactions.
By doing this, the computation time can be
saved, but it will not yield the most probable system response.
This is due to the fact
that the internal stresses and support reactions are not only a fune t iun ol" the magni
tudes but also the signs of the applied displacements or forces, and the s igns of the
SRSS of the displacements or forces are lost.
The SRSS method was first used for structural analysis [21]. It give
satisfactory results for most cases when the frequencies of the modes are we 11 separated.
But it happens quite often that some modes of a three dimensional piping system
have frequencies close to each other.
at same time.
The maximum response of these moles may occur
To accomodate this, the absolute sum should be applied to close
2'
frequency modes, and then take SRSS with the rest of the modes.
Since some modes are
insignificant in comparison with others, it is desirable to choose the contributing ones.

Following the derivation, we can see that the modal acceleration as defined in eq. (19)
is a natural basis for modal selection.
The final maximum stress at a point is then compared with the allowable one.
In case of overstress, perturbation technique can be applied to choose the design
changes [22]. A general approach without analysis is to put rigid restraints, e.g.
snubbers, at location of maximum deflection; hopefully this can drive the fundamental
frequency toward the higher frequency side of the peak area of the design spectrum.
Of
course this does not promise to be on economical redesign.
3.
C OMPARISON OF STIFFNESS AND FLEXIBILITY MATRIX METHOD

For stiffness matrix method, the overall stiffness matrix is obtained by
combining the individual branch stiffness matrices as done in displacement finite

element method [23]. This matrix includes elements corresponding to branch points which
are not assigned as mass points. These unwanted elements can be eliminated by condensa
tion scheme as follows.
Let the overall stiffness matrix be partitioned as
K..
11
K.,
lj
(22)
where the subscript i indicates elements corresponding to mass points.
The modified
matrix will be
& = [ *ii ] [ *u ] V ' t f j i 1
(23)
When the branch stiffness matrix is generated, each point is assigned with six degrees
of freedom, three translational and three rotational.
It is a general practice in
dynamic analysis to consider three translational degrees only [2lj],
Under this case we
can apply the condensation scheme again to obtain the dynamic stiffness matrix.
- [k^] a^-1
[] = aui
ik3i]
(2k)
This matrix [] is the one used in eqs. (2). These maximum displacements obtained in
eqs. (l) correspond to translational degrees of freedom of mass points only.
The
rotations of mass points are obtain as
(,) ,, 1 [ ;)
JJ
Ji
Let
(25)
- 324
The translations and rotations of branch points not assigned as mass points are obtained
V " -[Kjjrl
[R
ji ]
(2T)
The total displacement vector is

(x) =
# )
X
(28)
Applying these displacements {} to individual branch stiffness matrix, the internal

stresses and support reaction can be calculated accordingly.
If one wants to use flexibility matrix method, the flexibility matrix [A] in
eqs. (3) can be obtained either by taking the inverse of [K] in eq. (2k)
unit load method.
or by applying
When unit load method is used, we will solve the set of simultaneous
equations
[] (x) - IJ
(29)
as many times as there are number of dynamic degrees of freedom.
Since the only thing
which changes is the force vector on the right hand side of eqs. (29), we can
triangularize [ic] once and store it.
The displacement vectors due to different force
vectors will be obtained by back substitutions.
Various efficient methods [11, 25, 26]
are available for solving eqs. (29). After the maximum displacements in eqs. (l8) or
maximum forces in eqs. (20) are derived, one can obtain the internal stresses and the
support reactions either by the same way as described in stiffness matrix method or by
applying these forces to the system to solve for the overall displacements (x) or
multiplying these forces by those influence coefficients obtained in eq. (29)
Although
it is well known that the stiffness matrix method is faster in obtaining the internal
stresses and support reactions^
(21) in calculation.
Some programs still use forces defined in eqs. (20) or
The possible reasons are two.
Firstly engineers working on
dynamics are more familiar with the concept of equivalent static load method.
Another
reason is that engineers perform the dynamic piping analysis by modifying the static or
thermal piping stress program available, and the simplest way of modification 13 to use
unit load method.
k.
PIPING STRUCTURE INTERACTION

For piping with small mass in relation to the structure, it is a general
practice to use floor design spectrum as input.
The spectrum can be derived either by
time history method [2], or by design spectra method [15, l6, 27]. The conservatism of
the spectrum obtained depends on how each method is properly performed instead of which
method is used.
With the design spectrum method, special care should be exercised to
325 -
obtain the amplification curves and to combine the modal responses.
With the time
history method, special care should be exercised to obtain the proper time history and
to perform parametric study.
Due to the abrupt changes of the unsmoothed response
spectrum obtained from the actual strong motion earthquake records, the general trend
is to use simulated earthquake [28] as input such that the unsmoothed response spectrum
derived from it will simulate closely the design spectrum.
For primary coolant loop of a PWR plant, the mass is not small comparing with
the supporting structure.
The response will usually be overestimated if the floor
design spectrum is used as input.
Under this case, the loop and the structure can be
combined into one model and analyzed using ground design spectrum as input [29]. The
other alternative
is to perform component mode analysis [30].
ACKN0WLEGEMENT
Thanks are due to Dr. G. J. Patterson, research engineer of Gilbert
Associates, Inc., for stimulating discussion and to Mr. D. K. Croneberger, Chief
Structural Engineer of the same company, for helpful comments.
- 326 REFERENCES
1.
Crawford, L. , " P i p i n g under Dynamic L o a d i n g " , J ; _ of t h e Ameri cai; S o c i e t y ol" Naval

E n g i n e e r s , I n c . , May 1956.
2.
U...E.C. ( D i v i s i o n of T e c h n i c a l I n f o r m a t i o n ) "Summary of C u r r e n t Seismic Pes if:;

P r a c t i c e f o r Nuclear Reactor F a c i l i t i e s " , TID-25021, September 1967.
3.
B e r k o w i t z , L . , "Seismic A n a l y s i s of Primary P i p i n g System for Nuclear G e n e r a t i n g

S t a t i o n s " , Reactor and Fuel P r o c e s s i n g Technology, Vol. 1 2 , No. L , March 19o9
U.
L i n , C.W., "Seismic A n a l y s i s of P i p i n g System", Nuclear E n g i n e e r i n g and Design,

Vol. 1 1 , No. 2 , March 1970.
5.
A r c h e r , J . S . , " C o n s i s t e n t Mass Matrix f o r D i s t r i b u t e d Mass System", ASCK, S t r u c t u r a l

D i v i s i o n , August 19^3
6.
H a r r i n g t o n , R . L . , Vorus , W . S . , "Dynamic Shock A n a l y s i s of Shipboard Equipment",

p r e s e n t e d a t t h e Meeting of t h e Hampton Roads S e c t i o n of t h e S o c i e t y of Naval
A r c h i t e c t u r e s and Marine E n g i n e e r s , I 9 6 6 ,
Wilson, E . L . , R.W. Clough, "Dynamic Response by S t e p - b y - S t e p Matrix A n a l y s i s " ,

Symposium on t h e Use of Computerein C i v i l E n g i n e e r i n g , P o r t u g a l , 19o2.
8.
Chan, S . P . , H.L. Cox, W.A. B e n f i e l d , " T r a n s i e n t A n a l y s i s of Forced V i b r a t i o n s oV

Complex S t r u c t u r a l - M e c h a n i c a l Systems" , J Roy . A e r o . Soc . , 66 iy6,? .
9.
C a k i r o g l u , A . , Ozmen, G. , " N u m e r i c a l - I n t e g r a t i o n of F o r c e d - V i b r a t i o n
ASCE, E n g i n e e r i n g Mechanics D i v i s i o n J une I 9 6 0 .
Equations",
10.
P r z e m i e n i e c k i , J . S . , "Theory of Matrix S t r u c t u r a i
11.
R a l s t o n , . , "A Fir_s_t| Course^ i _ Numerical A n a l y s i s " , McGraw H i l l , lyb'p.
A n a l y s i s " , McGraw H i l l , I 9 6 8 .
12 .
Wilkinson , J .H. "The .Algebraic^ Eigenvalue Problem" , Oxford U n i v e r s i t y P r e s s , 196!; .
13.
Lord Rayleigh "Theory of Sound", Dover p u b l i c a t i o n s , N.Y. Vol. I ,
k.
Caughey, T.K. , " C l a s s i c a l Normal Modes in Damped L i n e a r Dynamic S y s t e m s " , J _._ o

Appi. Meen., June I 9 6 0 .
15.
Ainin, M., W.J. H a l l , N.M. Newmark, ond K.F. Kassawara, "Earthquake Response oi'
M u l t i p l y Connected Light Secondary Systems by Spectrum Methods", t o appear in
P r o c , ci* ASME F i r s t N a t i o n a l Congress of P r e s s u r e Vessel and Pi ping, Technology,
San F r a n c i s c o , Hay 1971
16.
Amin, M., R.P. Kassawara, W.J. H a l l , "Response of MDFSecondary Systems Using

S p e c t r a " , Submitted t o ASC E, S t r u c t u r a l D i v i s i o n , 1971 p u b l i s h p e n d i n g .
17.
H u r t y , W. , M.R. R u b i n s t e i n , "Dynamics of S t r u c t u r e s " , P r e n t i c e H a l l , J 96U.
13.
A l f o r d , J . L . , Housner, G.W. , M a r t e l , H.H., "Spectrum A n a l y s i s ol' Strong Motion

E a r t h q u a k e s " , Earthquake E n g i n e e r i n g Research L a b o r a t o r y , C a l i f o r n i a I n s t i t u t e of
Technology . Revised August I96I4.
19
U.S.A.E.C . ( D i v i s i o n of T e c h n i c a l I n f o r m a t i o n ) , "Nuclear Reactor and | E a r t h q u a k e s " ,

T1D7Q21*, August 1 9 6 3 .
20.
B i g g s , J.M. " i n t r o d u c t i o n t o S t r u c t u r a l Dynamics", McGraw H i l l 196^.
19^
2 1 . C lough , R.W., "Earthquake A n a l y s i s by Response Spectrum Superposi t i o n s " , l ui l e t i n

of t h e Seismoloîcal^ S o c i e t y of America, J u l y 1962.
- 327 -
22.
Higney, J.T., "Application of Perturbation Techniques to the Navy's Dynamic Design

Analysis Method", The Shock and Vibration Bulletin, Naval Research Laboratory,
December 19^9
23
Ziekiewicz, O . C , "The Finite Element Method in Structural and ^Continuum Mechanics",

McGraw Hill, I967.
2h.
ASME Codes and Standards, Interpretations of the Code for Pressure Piping,
Mechanical Engineering, November 1970.
25.
Hovanessian, S.A., L.A. Pipes, "Digital Computer Methods in Engineering", McGraw

Hill 1969.
26.
Ketter, R.L., S.P. Prawel, Jr., "Modern Methods of Engineering Computation", McGraw
Hill 1969.
27.
Biggs, J.M., Roesset, J.M., "Seismic Analysis of Equipment Mounted on a Massive

Structure", "Seismic Design for Nuclear Power Plants", edited by R.J. Hansen, the
M.I.T. Press. 1970.
28.
Jennings, P.C., G.W. Housner, U.C. Tsai, "Simulated Earthquake Motions", Earthquake
Engineering Research Laboratory, California Institute of Technology, 1968.
29.
Chen, C , J. David, "Dynamic Analyses of Vital Piping Systems Subjected to Seismic

Motion" , GAI Report No. 1729, Gilbert Associates, Inc., Reading, Pa. 1970.
30.
Hurty, W.C., "Dynamic Analysis of Structural Systems Using Component Modes", AIAA
Journal, April 1965.
328
DISCUSSION
K. AKINO, Japan
Do you calculate s t r e s s e s of piping systems due to earthquake loading in your
code through either forces or moments ?

Is there compatibility of your computer code with piping code, or USAS B31. 1 and B31. 7
which include flexibility factors and s t r e s s indices ?
Ch. CHEN, U. S. A.
In B31. 1 code the s t r e s s e s are calculated by moments. The piping program complies with all the requirements specified in the code.
K 4/7
SEISMIC RESPONSE SPECTRA FOR EQUIPMENT DESIGN

IN NUCLEAR POWER PLANTS
J.M. BIGGS,
Massachusetts Institute of Technology, Cambridge, Massachusetts,
U.S.A.
ABSTRACT
Floor response spectra for the seismic analysis of equipment are generated by a very
simple, generalized method based on the ground response spectrum and the results of a
response spectrum analysis of the supporting structure. In this method the effects of the
structure's modes are computed separately and then combined by an empirical procedure. As
compared to the alternative time-history approach to the construction of floor response
spectra, the proposed method is not only much more simple in application, but is believed
to be more reliable than the former when only a few time-histories of ground motion are used.
The validity of the method is.demons trated by comparison with results derived from actual
recorded ground motions.
INTRODUCTION
Presented in this paper is a method for generating floor response spectra to be used in
the analysis and design of equipment or piping mounted in a massive structure. It particularly relates to the problem as encountered in the design of nuclear power plants. The problem is crucial, because in such facilities continued operation of the equipment after an
earthquake is essential to safety. The analysis must be done with care, because in certain
circumstances the maximum acceleration to which the equipment is subjected may be many times
the peak ground acceleration and several times the maximum response acceleration of the supporting structure.
A floor response spectrum, in the present context, is a plot of maximum response acceleration versus period which provides the maximum response of any single-degree system, representing an item of equipment, mounted at the point in the structure for which the response
spectrum has been constructed. It may also be applied to multi-degree equipment by utilizing
conventional methods of modal analysis. It is not directly applicable to items such as piping which are supported at more than one point in the structure. However, this case may also
be handled if one superimposes the effects of the individual support motions.
The method proposed here provides a simple, yet reliable, procedure for generating floor
response spectra. It follows, and is based upon the results of, a response spectrum analysis
of the structure. The method is general in that it is intended to provide an envelope of the
floor response spectra which would be produced by all probable time histories of ground
- 330 motion. It is limited to the case of uncoupled systems, i.e., cases in which the mass of the
equipment is relatively small and does not affect the overall response of the structure. It
would not, for example, apply to the reactor vessel in a reactor building because that item
has appreciable mass and should be included as part of the dynamic model for analysis of the
structure. However, the vast majority of equipment and piping has relatively small mass and
may be considered uncoupled. Because of the large number of pieces of equipment in a power
plant, it is neither practical nor desirable to include them in the model of the complete
building.
A method similar to that presented here was introduced by the author in 1968. [1] The
procedures and numerical functions recommended here represent an updating and improvement of
the original method based upon additional studies of equipment-structure interaction in response to recorded earthquake ground motion.
To illustrate the nature of the problem and application of the proposed method, consider
a typical BWR reactor building. The dynamic model to be used for analysis is shown in Fig. 1.
This is a lumped-parameter model with nodes located on the exterior concrete building, the
concrete containment or drywell, the sacrificial shield and reactor vessel, and the concrete
pedestal supporting the vessel. The exterior building is connected to the interior structures only through the foundation mat, but the drywell, shield and vessel are interconnected
by stabilizer springs. To account for soil-structure interaction, the base mat in the model
is supported on translational and rocking springs. The parameters of the model, i.e., the
mass and stiffness matrices, may be determined by any one of several conventional procedures.
The first step in the analysis is to solve the eigenvalue problem, i.e., determine the frequencies and shapes of the normal modes.
Having established the normal modes, the next step in the building analysis is to determine the maximum response due to the seismic input in terms of displacements, accelerations and inertia forces at the nodes of the model. This is normally done by classical
modal analysis, but the numerical results can be obtained in one of two ways. (1) A ground
motion response spectrum is utilized to produce the maximum modal responses, and the total
response is obtained by combining the modal maxima by root-mean-square or other statistical
device. (2) A numerical solution of the modal equations of motion 1s obtained for a postulated time-history of ground motion and the time-history of total responses 1s obtained
by direct superimposing of the modal responses. The first method is now most commonly used
for analysis of the building because a ground response spectrum can be predicted with some
confidence for a given site, but it is impossible to predict a complete, detailed timehistory of ground motion. If time-history inputs are used, either recorded motions from
actual earthquakes or simulated motions, it is necessary in design to employ several such
inputs to ensure a conservative combination of modes in the multi-degree system. Because of
the uncertainties in predicting the detailed ground motion, and the voluminous computation
required in time-history analysis, the use of ground response spectra is more realistic for
design purposes.
A disadvantage of the response spectrum approach is that it yields only the maximum
modal responses at a point in the structure and not the time-history of the motion which is
required for the direct generation of the equipment response spectrum for that point in the
structure. The purpose of the method described below is to circumvent this difficulty by
permitting the construction of equipment response spectra without a time-history analysis of
- 331 either the building or the equipment.

BASIC CONCEPTS
The maximum acceleration response of the equipment may be considered to be an amplification of either (1) the ground response spectrum or (2) the peak acceleration of the structure
at the point where the equipment is attached. These two approaches are complimentary since
the first 1s more accurate for long equipment natural periods and the second is more accurate
for short periods. Therefore, the proposed method uses both approaches, each in the range
where it is the more accurate. The amplifications factors have been determined empirically.
As will be shown below, the factors are essentially a function of only the ratio of equipment to structure periods (T /T ) and the amount of damping in each. This fact makes possible
a very simple computational procedure.
The maximum equipment response is determined for each mode of the structure. The total
equipment response is then taken to be the root-mean-square of the responses due to the structural modes.
Before the method is described in detail, it will be helpful to identify two limiting
cases of equipment response: (1) Very flexible equipment relative to the structure
(T /T 1 ) , and (2) very rigid equipment (T /T 1 ) . In the first case, the internal distortion of the relatively rigid structure has no effect on the equipment. On the other hand,
those frequencies of ground motion to which the equipment is most responsive are passed
through the structure without modification. Therefore, the equipment responds as though it
were supported directly on the ground.
In the case of rigid equipment, the structure filters out those frequency components of
the ground motion to which the equipment would be sensitive. Therefore, the equipment merely
"rides along" with the structure and the maximum acceleration of the equipment is equal to
that of the structure at the point of attachment.
Considering either of the two limiting cases as a starting point, as the period ratio
approaches unity the equipment response is magnified. In the case of flexible equipment the
structure serves to amplify the ground motion. In the case of rigid equipment, the equipment
Itself amplifies the structure's motion since, in effect, it is being subjected to a harmonic
support motion. In either case, the amplification factor is very similar to the classical
amplification of a damped one-degree system subjected to a harmonic forcing function. This
suggests that the amplification factor is a function of only the period ratio and the damping.
In fact, this is very nearly the case, as will be shown below.
AMPLIFICATION CURVES
Figures 2 and 3 show amplification curves which were developed empirically and which are
used in the proposed method. Fig. 2 provides the ratio of peak equipment acceleration (A )
to the acceleration which would occur if the equipment were supported on the ground (A ).
Fig. 3 provides the ratio of peak equipment acceleration to the maximum acceleration of the
structure at the point of equipment support, (A ). Amplification curves for other combinations of structural and equipment damping have the same shape but different ordinates.
The amplification curves are based on analyses of the model shown in Fig. 4. This system was subjected to four actual earthquake records, and numerical analyses were made to pro-
- 332 vide the two amplification factors for a range of period ratio.
Although the model is a two-
degree system, it should be noted that this does not imply that the structure has only one
degree of freedom.
Instead, the lower mass and its supporting spring represents any one of
the uncoupled normal modes of the structure.

These four particular earthquakes were chosen because they are typical of strong motion
records, and also because they have different frequency content, i.e., the maximum responses
occur in different frequency ranges.
Since the results are not completely independent of the actual value of the periods,
analyses were made for various values of . The points plotted in Figs. 2 and 3 each repre
sent the maximum amplification factor computed for values of T r ranging from 0.05 to 2.50
sees.
However, this does not result in excessive conservatism as may be seen by inspection
of Fig. 5, which shows the variation of peak amplification factor (at /
= 1) with .
Each point plotted is the maximum of the responses due to the four earthquake records.
The
value actually used for the peak amplification (10.4) is only slightly unconservative at cer
tain values of .
It is somewhat conservative for
> 1.0, but such periods do not usually

s
occur in nuclear power plant structures.

and it is concluded that ignoring
'
Plots for other period ratios show similar results
the effect of
does not produce significant error.
The amplification curves adopted (shown in Figs. 2 and 3) are generally upper bounds for
the four earthquakes.
It may be observed in Fig. 2 that there is little scatter in the
points computed for the four earthquakes.
However, for small period ratios ( /T
< 0.7) the
curve becomes irregular, as would be expected because in this range the equipment response
is determined by the motion of the structure rather than the ground motion.
In Fig. 3 there
is little scatter in the lower range of period ratios but considerable scatter at the peak
and for higher ratios, as would be expected since the ground motion is more significant in
this range.
in Fig
Note that the curve in Fig. 2 approaches unity for large period ratios and that
3 approaches unity for small ratios.
Based upon these results, it was decided to use the amplification of the structure's
motion (Fig. 3) below /
.11 > 0.9.
= 0.5 and amplification of the ground motion (Fig. 2) for
Between these two points an interpolation of the two approaches was adopted.
Since for each amplification factor in its applicable range the results are nearly the same
for all four earthquakes, it is believed that these amplification curves are generally reason
able regardless of the detailed time-history of the ground motion.
THE PROCEDURE
A floor response spectrum is generated by computing the maximum equipment acceleration
for a series of values of equipment period within the range of interest.
The structure has
been previously analyzed on the basis of a given ground response spectrum, to provide the
following data:
sn
sn
sn
= natural period of mode

= participation factor for node
= eigenvector for mode at point of equipment support
= maximum acceleration in a mode of structure at point of equipment support
For each value of equipment period (T ) the following procedure provides a point on the floor
333
response spectrum:
1. D etermine A , the equipment response acceleration as if it were supported on the
ground. This is obtained by reading the ground response spectrum for the equipment
period and damping.
2.
For each significant mode of the structure:

(A)
IfTe/Tsn<0.9
Compute A
=A
A
(j ) , where the ratio in parenthesis is obtained from Fig. 2.
sn
(B) If T e /T s n > 0.5

Compute A

= s" sn A
ne
A
(~)> where the ratio in parenthesis is given by
eg
Fig. 3, and
i
T
< 1.25,
C
ne
=1
1.25 < T e /T s < 2.25,

c n e = i+ (^ ^ [ ( r O ^ ) 2 ) 1 ' 2 ! ]
sn
^ > 2.25,
S
C ne = ((
)' 2 ) 1 / 2
sn*sn'
where the summation includes all structural modes being considered.
(C) If 0.5 < ~< 0.9

, / 0.5
Compute A = A + (A A ) n .
y
en en v en en' 0.4
If e'/T s is not within this range,
A êquals either en
A oren
A .
3
en
I
II
3. Repeat Step 2 for each significant mode of the structure.

4. Obtain the total equipment response by combining the effects of the structural modes
as follows,
Ae = U A 2 n ] 1 / 2
The interpolation procedure in Step 2(C) is necessary because the two methods in (A) and
(B) cannot give identical results. This is true because of the approximation in the amplifi
cation curves and also because a smoothed ground response spectrum used for design does not
give consistent values of Aeg and Asn.
- 334 In step 2 () the multiplier . - appears because this is a measure of the effect of
mode on the equipment. For example, if r
= 0, mode does not participate in the seismic
response. If
= 0, mode produces no motion at the equipment support. In either case,
no equipment acceleration is associated with mode n. C is an empirical correction factor
which ensures the correct result when is very large. If 1s much larger than any of the
structural periods, A e. / Aeg
= 1 for all modes and Ae must equal
Aeg . When the modes are com^
bined in Step 4, the construction of C ensures this result. The range 1.25 - 2.25 was se
lected to provide spectra consistent with computed responses to actual earthquake records.
The combination of modal effects by root-mean-square in Step 4 is consistent with the
method most commonly used for analysis of structures based on response spectra. Any other
method of modal combination could also have been used, but it should be consistent with that
used for the structure. Thus, when all /T values are small, A will be equal to the pre
dicted maximum acceleration of the supporting structure, as it should be.
The computations required by this procedure are extremely simple and can even be execu
ted by hand. When a computer is used the calculations are almost trivial.
VERIFICATION OF METHOD
The proposed method is intended to provide a floor response spectrum which is an enve
lope of all spectra which would be produced by reasonable time histories of ground motion.
Whereas any single time-history would produce a spectrum which 1s unconservatlve in some
ranges of equipment period, the simplified method produces in one computation a spectrum
which is reasonable over the complete period range. In order to verify this statement, com
parisons have been made with results obtained by time-history analyses for various earth
quake records. Four such comparisons are shown in Figs. 6 through 9.
In these figures the dashed lines are the result of time-history analysis such as 1s
commonly employed for the construction of floor response spectra. This involves first a
numerical modal analysis of the supporting structure to procure a time-history of the motion
of the structure at the point of equipment support. This time-history 1s then used as input
to a one-degree system in a numerical analysis which provides one point on the floor response
spectrum.
The solid curves in Figs. 6-9 have been computed by the simplified method based upon the
actual, unsmoothed ground response spectrum for that particular earthquake record. In other
words, A
and A
have been taken from the unsmoothed spectrum. In this way,
the effect of
J
eg
sn
smoothing the ground response spectrum, as is normally done for design, has been eliminated
from the comparison.
Figs. 6 and 7 show comparisons for point 46 in the reactor building model shown in Fig.l.
The two ground motion records used are 1940 El Centro NS and 1966 Parkfield -5 NW, both nor
malized to a peak ground acceleration of 0.2g. In both cases the floor response spectrum has
two peaks corresponding to the first and second modes of the structure which have periods of
0.80 and 0.29 sees. It will be noted that the small peaks and valleys of the two curves coin
cide, thus indicating that the simplified method properly reflects the character of the
ground motion. In Fig. 6 it may be observed that the El Centro input is unconservative
throughout, especially at the first mode peak. In Fig. 7, the response to the Parkfield in
put is very slightly more conservative on both sides of the second mode peak, but otherwise
- 335 very similar to that predicted by the proposed method. It should be re-emphas1zed that agreement between the two curves being compared in each plot is not expected since the proposed
general method 1s Intended to be an envelope of all possible seismic inputs.
The Parkfield input is Included in Fig. 7 because that record was not one of those used
1n developing the amplification curves (Figs. 2 and 3 ) . This serves to prove that the proposed amplification curves are Indeed general and not dependent on the detailed nature of the
seismic motion.
The comparisons in F1gs. 8 and 9 are derived from another BWR reactor building which is
similar but not identical to that shown in Fig. 1. The Taft and El Centro earthquakes have
been normalized to 0.08g. In this case the first four modes of the structure contribute significantly to the floor response spectrum. The periods of these modes are 0.28, 0.19, 0.17
and 0.14 sees. The response in this case is therefore quite complicated, but even so, the
general method produces a very reasonable result which is at all points more conservative
than the t1me-h1story results.
As a result of these and many other such comparisons which have been made, it may be
concluded that the proposed method produces conservative, yet reasonable, results throughout
the range of equipment period.
APPLICATION
When a smoothed ground response spectrum is used for design, the resulting floor response spectrum 1s also smooth. Such a spectrum is shown in Fig. 10.
The peaks of floor spectra, which occur at points of resonance with the structural modes,
tend to be quite narrow. Since the natural periods of the structure cannot be known precisely, it is prudent to design the equipment for a range of structural periods. This is particularly true of structures on soft foundations represented by soil springs, the constants of
which cannot be determined accurately. To account for this uncertainty, the final floor response spectrum should be taken as the envelope of all spectra computed for a probable range
of structural periods.
Shown in Fig. 11 is an actual set of floor response spectra for equipment design at a
point in the reactor building of Fig. 1. Each of the four curves is for a particular value
of equipment damping. The plots are completely computer-produced. To allow for uncertainty
in structural periods, each curve is a composite of those computed for assumed upper and
lower limits of the periods. As an approximation, the envelope was obtained by constructing
straight lines between the peaks associated with the upper and lower limits of each period.
The computer cost for producing Fig. 11 was approximately five dollars. Most of this was for
the automatic plotting since the computational cost by the simplified method is almost negligible.
SUMMARY
The simplified or generalized procedure presented herein is a convenient, yet reliable,
method for the generation of floor response spectra to be used for the seismic design of
equipment and piping in nuclear power plants. It eliminates the need for the large number
of costly time-history analyses which would otherwise be required. More importantly, it is
believed to be more reliable than a time-history approach based on a limited number of
- 336 ground motion records. This is true whether the motions are actual earthquake records or
artificial motions mathematically derived from a ground response spectrum. In either case
one cannot be sure that the selected records are conservative for the particular multi-degree
structure supporting the equipment. On the other hand the proposed method is intended to be
an envelope of all probable seismic inputs.
It is hoped that the method presented provides a more realistic approach to the critical
problem of seismic design of equipment in nuclear power plants.
Reference
[1] BIGGS, J.M., ROESSET, J.M., "Seismic Analysis of Equipment Mounted on a Massive
Structure," Seismic Design for Nuclear Power Plants, edited by R.J. Hansen, MIT Press,
Cambridge, Massachusetts (1970).
337 -
FIGURE
I . DYNAMIC
MODEL OF REACTOR BUILDING
FIGURE 2 : AMPLIFICATION
MOTION
Structural
Equipment
Aeo6-
Damping * .04
Damping .005
Toft
El Centra
Helena
Golden Gate
OF
GROUND
- 338
FIGURE
3 : AMPLIFICATION OF STRUCTURE'S! o
MOTION
t
Structural
Equipment
Damping ' . 0 4
Damping . 0 0 5
Toff
El
Helena
Golden
Centro
/'
Gate
/i
ik
^^"^
-^
0.2
0.4
0.6
0.
./ .
10
17
14
STRUCTURE
{Mode nl
77777777777
GROUND
FIGURE
4:
MODEL
MOTION
FOR
TIME-HISTORY
ANALYSIS
1.6
ie
33 9
"1
.Jk/alue
Used
Maximum
of 4 Earthquakes
Toft
El Centro
Helena
Golden Gate
FIGURE
5: PEAK
AMPLIFICATION (*I.O) VS STRUCTURE

Is
PERIOD
340
FIGURE 7 : FLOOR RESPONSE SPECTRUM POINT 46 RESPONSE

TO PA RKFIELD (.20g )
<i
or
Generalized
Method
Time History Analysis
Structural
Equipment
Damping 0 4
Damping.005
in
O
. 2
(n
UJ
<
2
FIGURE 8 : FLOOR RESPONSE SPECTRUM POINT 3 RESPONSE

CENTRO (.08 g )
TO
04
<
a:
LU
Generalized
Method
TimeHistory Analysis
?2
Structural
Camping . 0 4
Equipment
Damping
.005
EL
341
FIGURE
9:
FLOOR RESPONSE
TAFT (.0 8 g )
SPECTRUM
POINT 3
RESPONSE
Generalized
Method
Time-History
< 3-
Structural
Equipment
Z
O
TO
Analysis
Damping = . 0 4
Damping = . 0 0 5
<
2
FIGURE 10: FLOOR
RESPONSE
Based on smoothed
spectrum ( 0 . 2 g )
Structural
Equipment
o
a.
0.8
T.
1.0
sees
SPECTRUM POINT 31
Damping
Damping
ground
=.04
=.005
response,
342
: :: :
::;:
M :
::::: TTTI1
;::;.
|8i
fili f
| I
lit
.B Ef C."
F
5
1n
.
Sf
L .
:
.... ,! .. . :
::::
RH
: - ;:
: . 'i'I :
' t , . rr
: .:
ti
1
b
: ' :
!
.1
!
..::
ii.. 2
:.,|:.
i
"
'
TF lii
F 31 I f
1| | u.
'
P Q i t 4T
rrT
~ ..!
a 11tsl .fa; J S
:L a r" R K JQ
fl
.J
...j:
: lia .:
. :;.Ma
FIGURE 11
A
y.
':|:
v . |
Ef IR Th Q l
J Ri
11 ARI IF Ml
lh| |3. D Ei
F
"
t,D
.......
C|I 1
':
_. V L ;
<I
11
'ir
...
, !
1
'
.... "
? il!
|
.....
::/
//
1
|1
a
.
: !|i;
"":
"
pi :
i":
M;::
':' L
s,,
r
$ ij;
^k
r'
1
i
; : :
v
u.
.... C
j '
!l:;: 1
Q'!
! ' i
,|ia
1
r .01
1/ 1:
-!**
'. |j
1/
/I :
.... ""!:" ...
y
1
Jy f
"1
: a_i
s;
s,
Ij S,
iiii .::
:::3 ?;<:
!
| | i. y];: ...: | s
, : ii: iiiii m l : . :;ii!
\-\::ii.
iiii ui;
;'
.L...
....j....
'T n
| M
"Si
..
... : ..... r .
\ S.^
.:"N |S S N
KjN f
s 41~.
i"
<
li
....
a.
i I Af
1 . D.
ai! ::.: iii .:::

:
u
R
;;::
iii-
1 . it

li
?..n
"
- 343
DISCUSSION
In the comparison of the time history analyses with the spectrum analyses, is
the same mathematical model used of the power plant ? If the model is the same wouldn't
you expect to obtain the agreement shown ?
.
A
The same dynamic model is used in both c a s e s . However, the resulting floor
response spectrum would be different for two r e a s o n s :
1. the amplification factor might not be c o r r e c t for that particular earthquake and s t r u c t u r a l
period,and
2. the use of r o o t - m e a n - s q u a r e for combination of the modal effects might be in e r r o r . Nevertheless, the results indicate that these e r r o r s a r e not serious.
B. NOWOTNY, Germany
Did I c o r r e c t l y understand your method, if I have the following imagination ?:
The floor response in the time domain consists of something like decaying vibrations. I think
you assume the vibration to be exactly a decaying vibration with the maximum acceleration
of the building of this floor and with the damping factor of the building.
A
The shape of the amplification curves is very similar to that for the response to
a damped harmonic input. However, the actual ordinates shown in the paper have been determined empirically from actual earthquake r e c o r d s .
SALES OFFICES
The Office for Officiai Publications sells all documents published by the Commission of
the European Communities at the addresses and at the price given below. When
ordering, specify clearly the exact reference and the title of the document.
GREAT BRITAIN A N D THE C O M M O N W E A L T H

H.M. Stationery
P.O. Box 569
London S E 1
UNITED STATES OF AMERICA

European Community Information
2100 M Street. N.W.
Suite 707
Washington, D.C. 20 037
Service
BELGIUM
Agencies :
00187 Roma Via del Tritone 6 1 / A e 6 1 / B
00187 Roma Via XX Settembre (Palazzo Ministero
delle finanze)
20121 Milano Galleria Vittorio Emanuele 3
80121 Napoli Via Chiaia 5
50129 Firenze Via Cavour 46/R
16121 Genova Via XII Ottobre 172
40125 Bologna Strada Maggiore 2 3 / A
NETHERLANDS
Moniteur belge Belgisch

Staatsblad
Rue de Louvain 40-42 Leuvenseweg 4 0 - 4 2
1000 Bruxelles 1000 Brussel Tel. 12 00 26
CCP 50-80 Postgiro 50-80
Agency :
Librairie europenne Europese Boekhandel
Rue de la Loi 244 Wetstraat 244
1040 Bruxelles 1040 Brussel
GRAND DUCHY OF LUXEMBOURG
FRANCE
Service de vente en France des publications
des Communauts
europennes
26, rue Desaix
75 Paris-15 Tel. (1) 306.5100
CCP Paris 23 96
(FR)
Verlag Bundesanzeiger
5 Kln 1 Postfach 108 006
Tel. (0221) 21 03 48
Telex : Anzeiger Bonn 08 882 595
Postscheckkonto 834 00 Kln
Staatsdrukkerij- en
uitgeversbedrijf
Christoffel Plantinstraat
's-Gravenhage Tel. (070) 81 45 11
Giro 425 300
IRELAND
Stationery Office
Beggar's Bush
Dublin 4
SWITZERLAND
Office for officiai

publications
of the European
Communities
Case postale 1003 Luxembourg 1
and 29, rue Aldringen, Library
Tel. 4 79 41 CCP 191-90
Compte courant bancaire : BIL 8 - 1 0 9 / 6 0 0 3 / 2 0 0
GERMANY
ITALY
Libreria dello Stato
Piazza G. Verdi 10
00198 Roma Tel. (6) 85 09
CCP 1/2640
Office
Librairie Payot
6 rue Grenus
1211 Genve
CCP 12-236 Genve
SWEDEN
Librairie CE Fritte
2, Fredsgatan
Stockholm 16
Post Giro 193, Bank Giro 73/4015
SPAIN
Libreria
Mundi-Prensa
Castello 37
Madrid 1
OTHER
COUNTRIES
Sales Office for official

publications
of the European
Communities
Case postale 1003 Luxembourg 1
Tel. 4 79 41 CCP 191-90
Compte courant bancaire : BIL 8 - 1 0 9 / 6 0 0 3 / 2 0 0
Price (11 Volumes) B.Fr.
5.000
REACTOR TECHNOLOGY
I
S
Solutions for Special Problems
DESIGN CONDIVONS and OPERATIONAL
MECHANICAL/THERMAL
B O U N D A R Y & SOURCE
CONDITIONS
REACTOR CORE: N U C L E A R COMPONENTS
stationary, transient
(Partici. M a t r i x ) , Pelle. Cladding, Cap; Fuel,

Moderator. Reflector, and ControlElements
f u e l Element A ssemblies
Spacer!, Hangen, Shroud*:
Core Support and Grid Structure
P R I M A R Y C O O L A N T CIRCUIT STRUCTURES
Kiping, Junctions. Bellows;

Primary Heat Exchangers;
Special Pumps. Circulators, etc.
(THERMOI
ELASTICITY
STRUCTURAL
MECHANICS
SLABS end PLA TES

REACTOR VESSELS
Calandria Vessels:
Steel Pressure Vessels.
Prpstressed Concrete Pressure Vessels
*
%
*&
R A D I A T I O N SHIELDS
1 REA CTOR C O N T A I N M E N T
S
fi
1 Mechanical Safeguarding Barnen;

1 Sted Shells.
Prestressed Concrete Shells
Metals
I.THERMO)
PLASTICITY
GRIDS and FRA MES ,
Reactor Thermal Shields;

Reactor Biological Shields:
S h a d e d Fuel Element Casks
NUCLEAR M A T E R I A L S
fc
THERMO
AND FLUID
DYNAMICS
cyclic, dynamic
REACTOR CORE: S T R U C T U R A L COMPONENTS
LIMITATIONS
3d intension? I
CONTINUA
S
PRACTICAL EXPERIENCE
SAFFTV
ANO RELIABILITY
ANALYSIS
(THERMOl
" VISCOELA STICITY
S j *
*>"
t
I
I
ENGINEERING
STRUCTURAL MA TERIA LS
FRACTURE
'
Metals
Ceramics
028)

Cdnj04820enc 001

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cdnj04820enc 001

Uploaded by

Copyright:

Available Formats

Proceedings

First International Conference on

5 REACTOR PLANT STRUCTURES

Part K Seismic Response Analysis of Nuclear

Commission of the European Communities, Brussels

Conference Organisation by:

First International Conference on

5 REACTO R PLANT STRUCTURES

Part K Seismic Response Analysis of Nuclear

Commission of the European Communities, Brussels

Survey Lectures : Earthquake

Evaluation of the Requirements of Nuclear Systems for

Philosophy and Practice of the Aseismic Design of

Co., Nuclear Power

Problmes de sismes : techniques utilises pour les

Aseismic Design of Nuclear Power Plant Structures

The Earthquake Response Analysis for a BWR Nuclear

Strong Motion Earthquakes and their Effects on Nuclear

Aseismic Design of Nuclear Reactor Building - Stress

Seismic Loading and Interaction Effects

Statistical Treatment of Seismic Loading on Reactor

Analysis of Soil-Structure Interaction Effects under

Soil-Foundation Interaction of Reactor Structures Subject

Dynamic Calculations Using a Framework Analogy to

Aseismic Design of Nuclear Power Plant Piping

Development of Aseismic Design of Pipings Vessels and

Seismic Design Coefficients of Equipment in Nuclear

Aseismic Design of Asymmetric Structures and

Dynamic Analysis of Vital Piping Systems Subjected to

Seismic Response Spectra for Equipment Design in

Topical Grouping of the Proceedings

SURVEY OF THE CONFERENCE: REACTOR TECHNOLOGY

Pressurized Water Reactor

Fuel and Cladding

Mathematical-Physical Models for Nuclear Fuels Materials

Structural Analysis of Fuel Elements

Shock and Vibration Analysis of Reactor Components

Thermal Shock, Pressure Pulse, and Impact Response

Analysis of Reactor Core Support Structures

REACTOR PRESSURE VESSELS

Steel Pressure Vessels

Mathematical-Physical Characterization of Concrete I

Mathematical-Physical Characterization of Concrete II

Methods for Structural Analysis of Prestressed Concrete

REACTOR PLANT STRUCTURES A N D CONTAINMENT

Analysis of Shell Structures; Containment

Seismic Response Analysis of Nuclear Power Plant

Analysis of Thin-Shell Structures I

Survey Lectures: Earthquake Response Analysis and

Thermal and Mechanical Analysis

Survey Lectures: Solid Mechanics Theory

EARTHQUAKE RESPONSE ANALYSIS OF REACTOR STRUCTURES

The general nature of the principles upon which earthquake resistant

' Published in Nuclear Engineering and Design 20 (1972).

N.M. NEWMARK, U.S.A.

ly by instrumentation. Displacement motion might be l a r g e r than that we know at the moment

EVALUATION OF THE REQUIREMENTS OF NUCLEAR SYSTEMS

Published in Nuclear Engineering and Design 20 (1972).

1. This is so unlikely that it should be ignored for evaluation purposes.

PHILOSOPHY AND PRACTICE OF THE ASEISMIC DESIGN

This report is summarizing guidelines for aseismic design of nuclear power

Published in Nuclear Engineering and Design 20 (.72).

and which time histories do you use ?