Engineering Structures 69 (2014) 111

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

An equivalent accidental eccentricity to account for the effects of

torsional ground motion on structures
Dhiman Basu a,b,, Michael C. Constantinou b, Andrew S. Whittaker b
a
b

Department of Civil Engineering, Indian Institute of Technology, Gandhinagar, India

Department of Civil, Structural and Environmental Engineering, University at Buffalo, State University of New York, Buffalo, New York, United States

a r t i c l e

i n f o

Article history:
Received 3 April 2013
Revised 26 February 2014
Accepted 28 February 2014

Keywords:
Torsion
Eccentricity
Earthquake ground motion
Seismic design
Seismic isolation

a b s t r a c t
The seismic design of buildings and other structures should include provisions for inherent and accidental torsion effects. Procedures developed decades ago for use with equivalent lateral force (static) analysis
have been often used for response-history analysis with no investigation of whether the procedures
achieve the desired result; namely, robust framing systems with limited susceptibility to excessive torsional displacement. The utility of procedures presented in ASCE 7 for treating accidental eccentricity as
means for accounting for the effects of torsional ground motion is examined by analysis of simple linear
and nonlinear systems. Results indicate that these standards-based procedures do not achieve the desired
trends when torsional ground motion effects are considered, namely, increased component demands
with increasing accidental eccentricity. An alternate approach for using accidental eccentricity concepts
in accounting for torsional ground effects is then proposed and veried in representative examples for
simple linear and nonlinear systems. In each case, component demand increases monotonically as the
accidental eccentricity increases.
2014 Elsevier Ltd. All rights reserved.

1. Introduction

1.1. ASCE 7 standard

Calculation of the seismic response of buildings and other structures requires consideration of torsion. Standards of design practice recognize the importance of torsional contributions to
horizontal displacement response and simplied procedures have
been proposed to estimate these contributions. Two types of torsion are considered: natural (or inherent) and accidental. Natural
torsion is the product of non-coincident centers of mass (CM)
and rigidity (CR) at one or more oor levels in a structure.
Accidental torsion is used to indirectly account for: (a) plan
distributions of reactive mass that differ from those assumed in
design, (b) variations in the mechanical properties of structural
components in the seismic force-resisting system, (c) non-uniform
yielding of components in the seismic force-resisting system, and
(d) torsional ground motion.
Seismic analysis and design of buildings require consideration
of natural and accidental torsion. Rules are presented in ASCE
Standard 7, Minimum Design Loads for Buildings and Other Structures
[1] for use with Equivalent Lateral Force (ELF) or static analysis,
and dynamic analysis, as summarized below.

Section 12.8.4 of ASCE 7 presents rules for addressing torsion if

the ELF procedure is used to analyze a building. Section 16.1
presents rules for use with dynamic analysis. Specically,
Section 16.1.5, Horizontal Shear Distribution, states The distribution of horizontal shear shall be in accordance with Section 12.8.4
except that amplication of torsion in accordance with
Section 12.8.4.3 is not required where the accidental torsion effects
are included in the dynamic analysis model. That is, ASCE 7 allows
the analyst to include accidental torsion in the models for dynamic
analysis but does not provide guidance as to how to do so. It has
been common practice to include these effects: (a) by ignoring
them in the dynamic analysis and then considering those in accordance with Section 12.8.4.3 of ASCE 7, or (b) by explicit consideration of the effects through the use of accidental eccentricity in a
manner similar to that used in the ELF procedure but in dynamic
analysis.
A number of studies on accidental torsion have been reported in
the archival literature. De-La-Llera and Chopra [810] calculated a
value of the accidental eccentricity for use with the ELF procedure
by studying the dynamic response of single and multistory buildings subjected to torsional ground motion. The ground motions
were calculated from records of horizontal acceleration at the
foundation level of instrumented buildings by dividing the

Corresponding author at: Department of Civil Engineering, Indian Institute of

Technology, Gandhinagar, India. Tel.: +91 9925433861.
E-mail address: dbasu@iitgn.ac.in (D. Basu).
http://dx.doi.org/10.1016/j.engstruct.2014.02.038
0141-0296/ 2014 Elsevier Ltd. All rights reserved.

D. Basu et al. / Engineering Structures 69 (2014) 111

difference in accelerations recorded at adjacent accelerographs by

the distance between them. These studies validated the ASCE-7
procedure (see above) for including accidental eccentricity in the
ELF procedure. The studies also concluded that results obtained
from the study of single-story buildings provide approximately
correct results for multistory buildings. A limitation of these
studies is that the torsional ground motions used for analysis are
specic to the instrumented buildings and do not represent freeeld rotational motions that could be used for the analysis of any
structure. De-La-Llera and Chopra [11] also analyzed one-story
systems using (a) the ELF, and (b) response- history analysis procedures with only horizontal ground excitation. Five percent accidental eccentricity was added to the actual eccentricity for analysis. A
comparison of results showed signicant differences. De-La-Llera
and Chopra [12] studied the variation of the torsional amplication
(see above) as a function of the ratio of the uncoupled torsional to
translational frequencies in one-story systems and proposed an
enveloping procedure to compute member forces for design using
the ELF procedure. De-La-Llera and Chopra [13] also developed a
multi-step procedure to determine the increase in edge displacement due to accidental torsion by utilizing data recorded at the
foundation and oor levels of an instrumented building. The procedure was then extended and evaluated from data recorded in 12
buildings [17]. Chandler [5] studied the effect of accidental torsion
on inelastic response of buidings. Torsional component of ground
motion can also be contributed from the spatially varying horizontal components and its effect on structural response was investigated by Hao [15]. Experimental investigations have also been
reported studying the effect of torsional ground motion (De-LaColina et al. [7] and Wolff et al. [22]). Recently, Sheikhabadi [19]
investigated the adequeacy of code specied accidental eecntricity.
The ELF denition of accidental eccentricity has been used for
response-history analysis to account for multiple effects, including
uncertainty in mass distribution and torsional ground effects but
its technical basis has not been demonstrated. The effect of torsional ground motion on the torsional response and its relation
to accidental eccentricity is studied in this paper. The effects of
uncertainty in mass distribution are not studied in the paper as
they are appropriately accounted for by the use of accidental
eccentricity. It is shown in this paper that the shifting of the center
of mass results in torsional response that does not necessarily
increase as the accidental eccentricity increases, therefore it does
not properly account for the effects of torsional ground motion.
Consider the simple three degree-of-freedom system shown in
Fig. 1. The plan dimensions of the single story structure are a  b.
The oor plate is supported by six columns that have lateral stiffness K1, K2 and K3 as shown in the gure. The CR is located a distance e from the CM. The calculation of CR is simple for this
structure. The offset of the CR from the CM produces the natural
or inherent torsional moment, Mt. The total torsional moment for
this structure, including dynamic amplication, could be calcu-

Fig. 1. Analytical model.

lated as the product of the translational inertial force and a design

eccentricity, ed, as follows:

ed ae  bb

where a is a dynamic amplication factor, b is a decimal fraction

(set equal to 0.05 in ASCE 7), and b is the building plan dimension
perpendicular to the applied translational force. The product bb is
the accidental eccentricity. In ASCE 7, the dynamic amplication
factor is applied to the accidental torsion only, and a = Ax, where
Ax is the torsional amplication factor and has values in the range
between 1 and 3. The approach adopted in ASCE 7 is straightforward and does not require explicit calculation of the CR at each oor
level, since the natural torsion is directly taken into account in the
analysis of a mathematical model by applying the code-specied
lateral load prole through the CM at each oor level. The calculation of the CR at each oor level is not straightforward for a multistory building and is dependent on the lateral force prole used for
the ELF procedure as discussed in Hejal and Chopra [16] and Basu
and Jain [3]. Although procedures have been developed to account
for the dynamic amplication of natural torsion when using an
ELF procedure (e.g., [20,14,3] they have not been adopted in ASCE 7.
The ASCE 7 rules for imposing accidental torsion in multi-story
structures do not provide guidance on whether a shift in the CM of
each oor is to be ordered or random, and if the same shift is to be
used at each oor level. If the dominant contributor to accidental
torsion is torsional ground motion, an ordered shift in the CM of
each oor plate is reasonable. If this is not the case, random shifts
are reasonable, with the net effect of accidental torsion likely being
small in the lower stories of a medium-to-high-rise structure.
Shifting the CM at each oor level to consider the effects of accidental torsion alters the modal properties of a structure and its
modal damping ratios if Rayleigh damping is used with a standard
software [6] to describe the inherent damping in the structure.
Note that actual torsional ground motion does not produce these
changes. The impact of shifting the CM on the modal properties
and structural response has not been discussed in the literature
and is studied below.
Single-story three degree-of-freedom systems are subjected to
translational and torsional components of seismic excitation to
study the conventional treatment of accidental torsion. The study
shows the limitations of the conventional approach when used
with response-history analysis. An alternative denition of accidental eccentricity is proposed and veried by a series of analyses
of single story elastic systems and nonlinear seismic isolation
systems.
2. Mathematical model for dynamic analysis
The one-story singly symmetric system shown in Fig. 1, composed of a rigid deck of mass m supported on six massless lateral-load-resisting elements, is used for analysis. The CM of the
deck is located at its geometric center and its radius of gyration
about a vertical axis passing through the CM is r. Each lateralload-resisting element has identical translational stiffness in the
two orthogonal directions but no torsional stiffness. The system
is symmetric about the x axis but has an eccentricity e about the
y axis. This system could represent a seismic isolation system supporting a rigid superstructure or a single story singly symmetric
building. The system is subjected to both translational seismic
excitation along the y axis and torsional ground excitation.
The parameters used to characterize the model are: (1) Ky =
total lateral stiffness in the y direction (equal to that in the x
direction), (2) xy = (Ky/m)0.5 = uncoupled lateral frequency, (3)
KhR = torsional stiffness about the CR, (4) xh = (KhR/mr2)0.5 = uncoupled torsional frequency, and (5) X = (KhR/r2Ky)0.5 = ratio of
uncoupled torsional frequency to translational frequency. For a

D. Basu et al. / Engineering Structures 69 (2014) 111

given aspect ratio and location of the elements with respect to the
CM, the lateral stiffness of each of the elements may be expressed
as

"

#

2
1  e  1 sy sa
1  e 2
;

X2 R2a 2
2sx b
4 sx
sx b
" (
#

2 )
1
sy sa
2  e 2
K2 Ky
 2X2 R2a  2
1
2
sx b
sx

K1; K3 Ky 

p
0:5
In Eq. (2), sx = b*/b, sy = a*/a, sa = a/b, Ra 1 s2a  =2 3sx . Assuming a unit mass, Ky in Eq. (2) may be replaced by x2y . Given the
dimensions of the deck and location of the elements, this elastic
system is uniquely described by three normalized parameters:
xy, X and e/b.
Specic to this six-element model, the arbitrary selection of the
three normalized parameters does not lead to a physical or real
system (which requires K1, K2 and K3 to be positive). Note that
xy can be chosen regardless of X and e/b for a physical system.
Further, if X is specied, e/b cannot be arbitrary. For example,
when X = 1.0, the range of e/b for a physical system is 0 6 e/
b 6 0.35. Similarly, 0 6 e/b 6 0.30 and 0 6 e/b 6 0.10 are for
X = 1.25 and 1.50, respectively.
Note that X P 1.0 implies a system with uncoupled translational frequency less than the uncoupled torsional frequency. Such
a system is denoted herein as a torsionally stiff system. Conversely,
a torsionally exible system is characterized by X < 1.0. Physically,
a torsionally stiff system has stiffer members located towards the
periphery of the structure, whereas in a torsionally exible system
these members are located towards its center. In practice, torsionally exible systems are uncommon and may be considered unrealistic, particularly for seismically isolated structures. From the
mathematical model considered here, it may be noted that 0 6 e/
b 6 0.03 for X = 0.8.
The range of e/b discussed above is specic to the mathematical
model considered here. A different range can be obtained if the
number and location of the elements are different. Nevertheless,
the mathematical model considered here covers a wide range of
torsionally stiff systems and is sufcient for the purposes of this
study.
3. Conventional calculation of accidental eccentricity
It is common practice to shift the CM at each oor level by a distance equal to the accidental eccentricity to amplify the maximum
translational response when performing response-history analysis.
This approach is studied herein and its effect on the displacement
demand is examined. For convenience, the two sides with respect
to the CR of the model are denoted as Side A and Side B as shown in
Fig. 1. In a torsionally stiff system (X P 1.0), elements located on
Side A are expected to sustain more displacement than those on
Side B. In a torsionally exible system (X < 1.0), elements located
on Side B are more critical than those on Side A. Since torsionally
exible systems are not very common in practice, the procedures
followed in this paper are rst formulated for torsionally stiff systems. Torsionally exible systems are then also analyzed with
appropriate modications.
The CM is rst shifted away from the CR (increasing the actual
eccentricity) and denoted here as Shift 1. The CM is then shifted to
each side in turn (increasing and then decreasing the actual eccentricity) and denoted as Shift 2.

cannot be directly recorded at this time and need to be extracted

by analysis of translational time series (e.g., [4]). Herein, a procedure based on earthquake acceleration time series (M6.1 event of
January 16, 1986, source-to-site distance of 20 km) recorded at
the Large Scale Seismic Testing (LSST) array in Lotung, Taiwan, is
utilized to extract the rotational ground motions. The Lotung-LSST
(LLSST) site is a part of the SMART-1 array. The fteen free-surface
accelerometers at the LLSST are positioned along three arms at
approximately 120-degree intervals (Fig. 2). Each arm extends for
about 50 m and the spacing between the surface stations varies
from 3 m to 90 m. Each arm contains ve stations that are designated as 15 starting from the center of the array. That is the three
stations numbered 1 in the array represent tier 1; the three stations numbered 2 in the array represent tier 2; and so on.
Furthermore, each station is identied herein as FAi_j, where i is
the arm (13) and j is the station (15). Further details can be obtained from [18].
The Surface Distribution Method (SDM) [4] is used to compute
the torsional ground motion. The following adjustments were
made to the procedure to develop an upper bound torsional spectrum: (i) The recorded EW (y) and NS (x) components are considered as the horizontal acceleration eld instead of the SH wave
component, (ii) The shear wave velocity at the surface layer
(=140 m/s per [21]) is used instead of the apparent SH wave velocity computed by Basu et al. [4] (=249 m/s), and (iii) Only stations
up to tier 4 are included in the analysis. Since the SDM yields
one torsional ground motion for each surface station considered,
the torsional acceleration history with the highest peak torsional
acceleration is used as the torsional ground motion input in this
study. The translational acceleration history is that recorded at
the interior station FA1_1, which is not the station where the peak
torsional acceleration is computed. The translational acceleration
histories along directions x and y, the torsional acceleration and
their respective 5-percent damped response spectra are shown in
Fig. 3. As the system is symmetric about the x axis, the translational acceleration in the NS direction is not input to the model.
3.2. Procedure for calculating the accidental eccentricity in
conventional approach
The procedure described below represents a systematic approach to quantify the accidental eccentricity for use in dynamic
response history analysis of structural systems excited with only
translational seismic excitation. The accidental eccentricity is

3.1. Ground motion considered

This study of accidental torsion includes an explicit consideration of torsional ground motion. Time series of torsional motions

Fig. 2. Station layout in Lotung array [18].

D. Basu et al. / Engineering Structures 69 (2014) 111

0.1
0
-0.1
-0.2

10

20

30

40

0.6

0.2

Acceleration (g)

Spectral acceleration (g)

Acceleration (g)

0.2

0.4

0.2

Acceleration (rad/sec 2)

Spectral acceleration (g)

0.6

0.4

0.2

Period (sec)

(d) EW (y) direction spectrum

0.1
0
-0.1
-0.2

10

20

30

40

Period (sec)

Time (sec)

(b) NS (x) direction spectrum

(c) EW (y) direction history

Time (sec)

(a) NS (x) direction history

0.2
0.1
0
-0.1
-0.2

10

20

30

40

Spectral acceleration (rad/sec2)

0.4
0.3
0.2
0.1
0

Time (sec)

(e) Torsional history

Period (sec)

(f) Torsional spectrum

Fig. 3. Input ground acceleration data.

established so that the peak corner displacement (along the y

direction) response of the system with only translational excitation
is equal to or greater than the peak corner displacement (along the
y direction) response of the actual system (without accidental
eccentricity) excited by translational and torsional ground excitation. The steps of this procedure are:
1. Select values for the normalized parameters xy, X and e/b that
uniquely dene the elastic system. Note that these parameters
describe the uncoupled translational frequency, the ratio of
uncoupled torsional to translational frequency and the actual
normalized eccentricity, respectively.
2. Simultaneously apply the translational and torsional acceleration histories and calculate the absolute maximum displacement at the farthest element on Side A, U.
3. Repeat Step 2 but apply only the translational acceleration history; let the absolute maximum displacement for the same element be U fl ; compute the torsional amplication factor as
R1 U fl =U fl :
4. Repeat Steps 2 and 3 but reverse the direction of the torsional
acceleration history and compute the torsional amplication
factor R2; select the target torsional amplication factor as
R = max (R1, R2). Note that the target torsional amplication is
the actual torsional amplication.
5. (a) Shift 1: Shift the CM away from the CR by an offset ea (the
accidental eccentricity) and analyze the system by applying
only the translational acceleration history; let the absolute
maximum displacement at the furthest element on Side A be
U fl .
(b) Shift 2: Repeat step 5a but shift the CM in the opposite
direction and compare the two values of U fl ; record the greatest
value.
6. Dene the torsional amplication factor associated with offset
ea as R U fl =U fl : This represents the computed torsional amplication factor. Repeat Step 5 for a range of values of ea and generate the associated torsional amplication factor. The required
accidental eccentricity for the system considered is given by the
offset ea for which R* P R.

3.3. Results and discussions

The procedure outlined above was applied to a variety of elastic
systems selected by varying the three normalized parameters
xy, X and e/b. For torsionally exible systems (X < 1.0), the procedure presented above is appropriately modied so the element response on Side B is chosen for the calculation of the torsional
amplication. In each system, the aspect ratio of the deck and
the location of the elements with respect to the CM of the deck
were selected to be sx = 1, sy = 1, and sa = 0.5. Damping in this three
degree-of-freedom (DOF) system was described by Rayleigh damping with 5% damping ratio in the rst and third modes. Analysis of
each system was carried out and the calculated target (or actual)
and computed torsional amplication factors were compared to
calculate the required accidental eccentricity ea. Results for two
cases (X > 1.0 and X < 1.0) are presented below; other results are
presented in Basu et al. [4].
For the presented case, the uncoupled translational period is
1.0 s and the ratio of the uncoupled torsional to translational frequency (X) is 1.25. The target (actual) torsional amplication calculated in the procedure with Shift 1 is presented in Fig. 4 for
actual eccentricities in the range of 00.3. It may be seen that
the target (or actual) torsional amplication is nearly constant at
just less than 1.1. Fig. 4 also presents the computed torsional
amplication (for only translational excitation) at various values
of accidental eccentricity in the range of 0.01 6 ea/b 6 0.05 (panel
a) and 0.46 6 ea/b 6 0.49 (panel b). It may be seen in Fig. 4 that
the computed torsional amplication uctuates randomly and it
is apparent that it is not possible to select a single value of accidental eccentricity ea/b to match the target and the computed torsional amplication factors.
It is not possible to select a value for the accidental eccentricity
and perform response-history analysis with only translational
excitation to correctly capture the effects of translational and torsional ground excitation. The increase in displacement response is
not monotonic with increasing accidental eccentricity as shown in
Fig. 4c. (A shift in the CM of greater than 0.5b is meaningless for
nearly all framing systems).

D. Basu et al. / Engineering Structures 69 (2014) 111

1.3

Target

1.2

e a /b=0.01
1.1

e a /b=0.02
e a /b=0.03

e a /b=0.04
e a /b=0.05

0.9
0.8

Torsional amplification

Torsional amplification

1.3

1.2

Target

1.1

e a /b=0.46
e a /b=0.47

e a /b=0.48

0.9

e a /b=0.49

0.8
0.7

0
0.1
0.2
0.3
Normalized eccentricity (e/b)

0.1

0.2

0.3

Normalized eccentricity (e/b)

(a) 0.01 ea b 0.05

(b) 0.46 ea b 0.49

Torsional amplification

1.4
e/b=0.002
e/b=0.05
e/b=0.10
e/b=0.15
e/b=0.20
e/b=0.25
e/b=0.30

1.2
1
0.8
0.6

0.1

0.2

0.3

0.4

0.5

Accidental eccentricity ( ea /b)

(c) 0 e b 0.30
Fig. 4. Variation of torsional amplication, Shift 1, conventional approach.

1.6

1.2

Target
ea/b=0.01

1.1

ea/b=0.02
ea/b=0.03
ea/b=0.04

ea/b=0.05
0.9

0.1

0.2

Torsional amplification

Torsional amplification

1.3

1.4

Target
ea/b=0.12
1

ea/b=0.13
ea/b=0.14

0.8
0.6

0.3

ea/b=0.11

1.2

ea/b=0.15
0

0.1

0.2

0.3

Normalized eccentricity (e/b)

Normalized eccentricity (e/b)

(a) 0.01 ea b 0.05

(b) 0.11 ea b 0.15

Torsional amplification

1.6
e/b=0.002
e/b=0.05
e/b=0.10
e/b=0.15
e/b=0.20
e/b=0.25
e/b=0.30

1.4
1.2
1
0.8
0.6

0.1

0.2

0.3

0.4

0.5

Accidental eccentricity (ea/b)

(c) 0 e b 0.30

Fig. 5. Variation of torsional amplication, Shift 2, conventional approach.

Fig. 5 presents the results of Fig. 4 but for Shift 2, which better
represents the mandatory language of Section 12.8.4.2 of ASCE 710. It is clear from the results of Figs. 4(c) and 5(c) that there is

no trend of increasing torsional amplication with increasing accidental eccentricity in response-history analysis using only translational excitation.

D. Basu et al. / Engineering Structures 69 (2014) 111

Figs. 6 and 7 present a similar study on a set of torsionally exible systems characterized by X = 0.8 and with the uncoupled
translational period being unity as in the previous studies. Note
the maximum range of e/b is now 0 6 e/b 6 0.03 for the system

to have practical signicance. The torsional amplication is now

based on the response of the element located at Side B. Observations are similar to those presented above for the torsionally stiff
systems.
1.1

Target

1.3

ea/b=0.01
ea/b=0.02

1.2

ea/b=0.03
ea/b=0.04

1.1

ea/b=0.05

0.01

0.02

Torsional amplification

Torsional amplification

1.4

ea/b=0.46

0.9

ea/b=0.47
0.8

ea/b=0.48

0.7

ea/b=0.49

0.6
0.5

0.03

Target

0.01

0.02

0.03

Normalized eccentricity (e/b)

Normalized eccentricity (e/b)

(a) 0.01 ea b 0.05

(b) 0.46 ea b 0.49

Torsional amplification

1.6

1.4

e/b=0.002
e/b=0.01
e/b=0.02
e/b=0.03

1.2
1
0.8
0.6
0.4

0.1

0.2

0.3

0.4

0.5

Accidental eccentricity (ea /b)

(c) 0 e b 0.03

Fig. 6. Variation of torsional amplication in Side-B element, shift-1, conventional approach.

1.12

1.3

Target
ea/b=0.01

1.2

ea/b=0.02
ea/b=0.03
ea/b=0.04

1.1

ea/b=0.05
1

0.01

0.02

Torsional amplification

Torsional amplification

1.4

1.08

ea/b=0.46
1

ea/b=0.47

0.96

ea/b=0.48
ea/b=0.49

0.92
0.88

0.03

Target

1.04

0.01

0.02

0.03

Normalized eccentricity (e/b)

Normalized eccentricity (e/b)

(a) 0.01 ea b 0.05

(b) 0.46 ea b 0.49

Torsional amplification

1.6
e/b=0.002
e/b=0.01
e/b=0.02
e/b=0.03

1.4
1.2
1
0.8

0.1

0.2

0.3

0.4

0.5

Accidental eccentricity (ea /b)

(c) 0 e b 0.03
Fig. 7. Variation of torsional amplication in Side-B element, shift-2, conventional approach.

D. Basu et al. / Engineering Structures 69 (2014) 111

The relationship between torsional amplication and accidental

eccentricity shown in these gures (panel c of Figs. 47) is independent of the torsional acceleration history used for analysis
and is rather the result of changes in the dynamic characteristics
of the system when the CM is shifted. Shifting the CM in response-history analysis does not deliver the expected increase in
displacement response.
4. Alternate denition of accidental eccentricity
Recognizing that the current representation of accidental eccentricity does not achieve the desired goal with response-history
analysis, an alternate denition is proposed here that can be schematically described through Fig. 8 in terms of application of the
inertial force. A torsionally stiff system is considered in Fig. 8 for
the purpose of illustration. However, the concept applies to torsionally exible systems as well, but with appropriate
modications.
Fig. 8a shows the inertial force and moment acting through the
CM of the considered system when subjected to the translational
gy along the y direction. The inertial force is comacceleration u
y and mu
gy in the Y direction and moment
prised of components mu
h , where r is the radius of gyration of the slab. To account for
mr 2 u
the effect of accidental torsion in the elements located on Side A
gy is shifted away from the CR by a distance ea
only, the force mu
(see Fig. 8b). This is equivalent to applying a torsional moment
gy (see Fig. 8c). The inertial force and moment shown
equal to mea u
in Fig. 8c can be considered as resulting from a set of equivalent
ground motions acting on the original system as shown in
Fig. 8d. The equivalent ground motions consist of the original
translational motion and a torsional motion calculated by multiplying the translational motion by an arm ea/r2, where ea is the
accidental eccentricity to be determined.

(a) Without accidental torsion

(c) Equivalence of (b)

4.1. Procedure for calculating the accidental eccentricity in proposed

approach
The steps for calculating the accidental eccentricity in the proposed approach are identical to those described in the conventional approach except that step 5 of shifting of the CM is
replaced by the application of a torsional acceleration history as
dened above. Note that the procedure uses two types of torsional
acceleration histories: (i) an actual record (described above under
the heading of Ground Motion Considered) to compute the target
(or actual) torsional amplication, and (ii) an articial record derived by multiplying the translational acceleration history by a factor that is a function of the newly dened accidental eccentricity
(ea/r2). The intensity of the articial torsional acceleration history
is increased by incrementing the accidental eccentricity until the
target torsional amplication is obtained.
Note that for torsionally exible systems, the direction of the
articial torsional acceleration is opposite to that for the one for
torsionally stiff systems.
4.2. Scaling of accidental eccentricity
The articial torsional acceleration history described in Fig. 8d
gh ea =r 2 u
gy , where ea is the accidental eccentricity and u
gy
is u
the translational acceleration history. Alternatively, the torsional
acceleration history may be expressed as

gh ea =r2 b =r2 u

gy
u

where b is the greater of the two plan dimensions, a and b, and

ea ea =b =r2

4
*

Note that quantity (b /r) is a multiplier larger than unity (for a

square plan, it is equal to 6).

(b) Shifting one part of inertial force to account

for accidental torsion in elements located on
Side A

(d) Equivalent ground motion inputs

Fig. 8. Schematic representation of the alternative denition of accidental eccentricity.

D. Basu et al. / Engineering Structures 69 (2014) 111

4.3. Results and discussions

Target
ea/b=0.01
ea/b=0.02

ea/b=0.03
ea/b=0.04
ea/b=0.05
ea/b=0.06

ea/b=0.07
ea/b=0.08
ea/b=0.09
ea/b=0.10

1
0

0.01

0.02

0.03

Normalized eccentricity (e/b)

(a) 0.01 f 0.10

4

e/b=0.002
e/b=0.01

Torsional amplification

The set of systems analyzed previously by shifting the CM is

reanalyzed using the proposed denition of accidental eccentricity
(ea in remainder of this paper). Fig. 9 presents results for the set of
torsionally stiff systems considered earlier. Fig. 9 presents the torsional amplication calculated by analysis with translational and
the articial torsional ground motions. This is compared to the target torsional amplication, which is calculated by analysis with
translational and the actual torsional ground motions. Fig. 10 presents similar results to those of Fig. 9 but for the set of torsionally
exible systems considered earlier. Note that Fig. 9(b) and
Fig. 10(b) show increasing torsional amplication with increasing
accidental eccentricity. In other words, if the structure is able
achieve the required torsional amplication with 2% accidental
eccentricity, it will meet the same with any higher value of accidental eccentricity. Such a property was not evident with conventional procedure.
Therefore, the proposed procedure can be used to develop design recommendations for accidental eccentricity for use with response-history analysis to account for the effects of torsional
ground motion. The recommendation should be based on a comparison of calculated torsional amplication and the target torsional amplication obtained using actual torsional ground
motion input. Table 1 presents values of accidental eccentricity
(ea =b) as a percentage of the plan dimension normal to the direction of excitation, for use in response-history analysis. Note these
values are for elastic systems and based on analysis using only
one record of torsional ground motion.

Torsional amplification

e/b=0.02
e/b=0.03

1
0

0.02

0.04

0.06

0.08

0.1

Accidental eccentricity (f)

(b) 0 < e b 0.30

Fig. 10. Variation of torsional amplication in torsionally exible system, proposed
approach, f ea =b.

Torsional amplification

5
Table 1
Accidental eccentricity ea =b for an elastic system.

Target
f=0.01
f=0.02
f=0.03
f=0.04
f=0.05
f=0.06
f=0.07
f=0.08
f=0.09
f=0.10

4
3
2
1

0.8
1
1.25
1.5

0.1

0.2

0.3

Normalized eccentricity (e/b)

(a) 0.01 f 0.10

Torsional amplification

e/b=0.002
e/b=0.05
e/b=0.10
e/b=0.15
e/b=0.20
e/b=0.25
e/b=0.30

4
3
2
1
0
0

0.02

0.04

0.06

0.08

1.0

1.5

3.0

4.0

0.01
0.01
0.01
0.01

0.01
0.01
0.03
0.03

0.01
0.02
0.03
0.02

0.03
0.03
0.03
0.01

0.03
0.03
0.02
0.02

5. Accidental eccentricity in nonlinear isolation systems

0
0

Tn (s)
0.5

0.1

Accidental eccentricity (f)

(b) 0 < e b 0.30

Fig. 9. Variation of torsional amplication in torsionally stiff system, proposed
approach, f 
ea =b.

The system shown in Fig. 1 is assumed now to represent a rigid

structure that is seismically isolated. The isolation system consists
of six isolators and the system is symmetric about the x axis but
has an eccentricity about the y axis. The mass of the rigid superstructure is lumped at the CM (and geometric center) of the deck.
The isolators have the bilinear hysteretic forcedisplacement relationship shown in Fig. 11, where Q is the strength (force at zero
displacement), Qy is the yield strength, Y is the yield displacement
and Kd is the post-elastic stiffness. Analysis of this nonlinear isolation system can be performed using well-established procedures as
outlined in Basu et al. [4].
A broad range of behavior of the isolation system is considered.
These systems are characterized as described above in case of elastic systems with the exception that the post-elastic stiffness, Kd, is
used instead of the elastic stiffness, Ky. The uncoupled translational
period based on the post-elastic stiffness is denoted as

T d 2p=xd

Three isolation systems are considered with periods of Td = 3 s,

Td = 4 sand Td = 5 s. The ratios of the uncoupled torsional frequency

D. Basu et al. / Engineering Structures 69 (2014) 111

Torsional amplification

3.5
Target
f=0.005
f=0.010
f=0.015
f=0.020
f=0.025
f=0.030
f=0.035
f=0.040
f=0.045
f=0.050

3
2.5
2
1.5
1

0.1

0.2

0.3

Normalized eccentricity (e/b)

Fig. 11. Forcedisplacement relationship for a typical isolator.

Fig. 12. Variation of torsional amplication in Side A element of nonlinear isolation

system, Td = 4 s, X = 1.25, Q/W = 0.05, Y = 5 mm, f 
ea =b.

to translational frequency, based on the post-elastic stiffness, X, are

considered as 1.0, 1.25 and 1.5 to cover a wide range of torsionally
stiff isolation systems. The normalized natural eccentricity (e/b) is
increased from zero, in increments of 0.05, until the maximum value possible for which the system has physical meaning. Further,
X = 0.8 is also considered to study the torsional exible isolation
systems and in such a case e/b is increased from zero, in increments
of 0.01, until the maximum value possible for which the system has
physical meaning. The yield displacement of each isolator was set
equal to 1, 5 and 10 mm. The ratio of characteristic strength to supported weight (Q/W) for each isolator was set equal to 0.04, 0.05,
0.06 and 0.07. The combination of parameters used to describe
the isolation system represents a range of Friction Pendulum and
lead-rubber bearing isolation systems, but using yield displacements appropriate for each type of system: = 1 mm for Friction
Pendulum bearings and Y = 10 mm for lead-rubber bearings.

5.1. Results and discussions

The accidental eccentricity computed for all systems considered
in this study are reported in Tables 24. Results from one of these
cases, Td = 4 s, Y = 5 mm, X = 1.25 and Q/W = 0.05 are presented in
Fig. 12. This is an example of torsionally stiff isolation systems,
where it is demonstrated that the accidental eccentricity is nearly
independent of the actual eccentricity. A similar observation was
made for a torsionally exible isolation system with Td = 4 s,
Y = 5 mm, X = 0.8 and Q/W = 0.05, for which results are presented
in Fig. 13. The results presented here and additional results in Basu
et al. [4] suggest that the required normalized eccentricity per Eq.
(3) is in the range of 0.0050.01 with the yield displacement

Table 2
Accidental eccentricity ea =b in nonlinear isolation system, Y = 1 mm.

Td = 3 s
j = Q/W (%)

Td = 4 s
j = Q/W (%)

Td = 5 s
j = Q/W (%)

0.8
1
1.25
1.5

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

0.005
0.005
0.005
0.005

Table 3
Accidental eccentricity ea =b in nonlinear isolation system, Y = 5 mm.

Td = 3 s
j = Q/W (%)

Td = 4 s
j = Q/W (%)

Td = 5 s
j = Q/W (%)

0.8
1
1.25
1.5

0.005
0.01
0.01
0.01

0.005
0.005
0.005
0.005

0.01
0.01
0.01
0.01

0.01
0.01
0.01
0.01

0.005
0.01
0.01
0.01

0.005
0.005
0.005
0.005

0.01
0.01
0.01
0.01

0.01
0.01
0.01
0.01

0.005
0.01
0.01
0.01

0.005
0.005
0.005
0.005

0.01
0.01
0.01
0.01

0.01
0.01
0.01
0.01

Table 4
Accidental eccentricity ea =b in nonlinear isolation system, Y = 10 mm.

Td = 3 s
j = Q/W (%)

Td = 4 s
j = Q/W (%)

Td = 5 s
j = Q/W (%)

0.8
1
1.25
1.5

0.01
0.01
0.01
0.01

0.01
0.01
0.01
0.01

0.005
0.01
0.01
0.01

0.005
0.005
0.005
0.005

0.01
0.01
0.01
0.01

0.01
0.01
0.01
0.01

0.005
0.01
0.01
0.01

0.005
0.005
0.005
0.005

0.01
0.01
0.01
0.01

0.01
0.01
0.01
0.01

0.005
0.01
0.01
0.01

0.005
0.005
0.005
0.005

10

D. Basu et al. / Engineering Structures 69 (2014) 111

7. Conclusions

Torsional amplification

3.5
Target

f=0.005
f=0.010
f=0.015

2.5

f=0.020
f=0.025
f=0.030

f=0.035
f=0.040
f=0.045

1.5

f=0.050

1
0

0.01

0.02

0.03

Normalized eccentricity (e/b)

Fig. 13. Variation of torsional amplication in side B element of nonlinear isolation
system, Td = 4 s, X = 0.8, Q/W = 0.05, Y = 5 mm, f 
ea =b.

having the largest inuence on its value. Additional studies with

other torsional motions are needed to better dene the values of
normalized eccentricity.

6. Discussion on the proposed denition of accidental

eccentricity
This paper presents an alternative denition of accidental
eccentricity to account for the effects of torsional ground motion.
This new denition of accidental eccentricity applies a torsional
ground motion, which is the product of a translational ground motion and a factor that is a function of the proposed accidental
eccentricity. The required accidental eccentricity is that which
achieves the actual torsional amplication as determined by use
of the actual torsional ground motion. The displacement of a corner
of the building is used to dene the torsional amplication, as has
been used in many prior studies (e.g., [811]. For multistory systems, the number of response quantities that could be used to dene the torsional amplication increases rapidly. Each quantity
would produce a different accidental eccentricity in a multistory
building.
While the calculation of the required accidental eccentricity in a
multistory system should consider all important response quantities and utilize that which produces the greatest demand, it has
been common to dene the accidental eccentricity using single
story models and then verify the validity of the results by analysis
of selected multistory structures. For example, De-La-Llera and
Chopra [9] showed the torsional amplication computed for a
one-story system was identical to that in any story of a special
class of multistory buildings [16,2] if the contributions from the
higher modes are negligible. The design recommendations for accidental eccentricity for this special class of structures are then identical to those for one-story systems.
The denition of the accidental eccentricity proposed herein
also applies to sources of accidental torsion other than the torsional component of ground motion. However, the procedure to
quantify the value of the accidental eccentricity will differ by
source. For example, the source could be uncertainty in the inplane stiffness of lateral load resisting elements. Then the procedure of De-La-Llera and Chopra [9] could be used together with
the denition of accidental eccentricity proposed here.
Further studies are needed if the denition of accidental eccentricity presented above is accepted. Such studies would require the
use of a large number of torsional ground motions and structural
systems to provide the body of knowledge necessary for the development of design recommendations and code language.

When only considering the effects of torsional ground motion

and without consideration of uncertainty in mass distribution,
the conventional approach of accounting for the effects of accidental torsion by shifting the CM does not produce the desired effect in
response-history analysis because the shift changes the dynamic
characteristics of the structure so that it is possible to have reduced
displacement demands with increasing eccentricity. An alternative
denition of accidental eccentricity is proposed wherein accidental
torsion is accounted for by simultaneously applying torsional and
translational acceleration histories. The torsional acceleration history is computed as the product of the translational history and a
scale factor, which is a function of the proposed accidental
eccentricity.
The proposed procedure has been studied for a broad range of
single story elastic systems and nonlinear isolation systems. Torsional amplication is predicted correctly, namely, increasing torsional response with increasing eccentricity. The study
demonstrates that for nonlinear isolation systems, the required accidental eccentricity increases as the yield displacement increases.
Values of the required accidental eccentricity to account for torsional ground motion effects are presented (tabulated in Tables 1
4) but are specic to the ground motions considered in this study
and should not be considered as the design recommendations. This
paper proposes only a methodology of accounting for the accidental eccentricity due to the torsional component of the ground motion. Quantication of such accidental eccentricity requires a
rigorous application of the proposed methodology based on a large
number of ground motions recorded at various geologic/site
conditions.
Acknowledgements
The nancial support for the studies described herein was provided by MCEER (www.mceer.buffalo.edu) under Thrust Area 3,
Innovative Technologies, through a grant from the State of New
York. The Institute of Earth Science, Academia, Sinica, Taiwan provided the strong motion data. The nancial support, technical review and provision of data are gratefully acknowledged. Any
opinions, ndings, conclusions or recommendations expressed in
this paper are the authors and do not necessarily reect those of
either MCEER or the State of New York.
References
[1] ASCE. Minimum design loads for buildings and other structures. Standard
ASCE/SEI 7-10, American Society of Civil Engineers, Reston, VA; 2010.
[2] Basu D, Gopalakrishnan N. Analysis for preliminary design of a class of
torsionally coupled buildings with horizontal setbacks. Eng Struct
2008;30:127291.
[3] Basu D, Jain SK. Alternative method to locate centre of rigidity in asymmetric
buildings. Earthquake Eng Struct Dynam 2007;36(7):96573.
[4] Basu D, et al. Characterizing the rotational components of earthquake ground
motion. MCEER-12-0005, Multidisciplinary Center for Earthquake Engineering
Research, SUNY, Buffalo, NY; 2012.
[5] Chandler AM. Inuence of accidental eccentricity on inelastic seismic torsional
effects in buildings. Eng Struct 1995;17(3):16778.
[6] CSI. SAP 2000: static and dynamic nite element analysis of structures.
Berkeley, California, Computers and Structures, Inc; 2009.
[7] De-La-Colina J, Gonzalez JV, Perez CAG. Experiments to study the effects of
foundation rotation on the seismic building torsional response of a reinforced
concrete space frame. Eng Struct 2013;56(11541163).
[8] De-La-Llera JC, Chopra AK. Accidental in buildings due to stiffness uncertainty.
Earthquake Eng Struct Dynam 1994;23:11736.
[9] De-La-Llera JC, Chopra AK. Accidental torsion in buildings due to base
rotational excitation. Earthquake Eng Struct Dynam 1994;23:100321.
[10] De-La-Llera JC, Chopra AK. Evaluation of code accidental-torsion provisions
from building records. J Struct Eng, ASCE 1994;120(2):597615.
[11] De-La-Llera JC, Chopra AK. Using accidental eccentricity in code specied static
and dynamic analyses of buildings. Earthquake Eng Struct Dynam
1994;23:94767.

D. Basu et al. / Engineering Structures 69 (2014) 111

[12] De-La-Llera JC, Chopra AK. Estimation of accidental torsion effects for seismic
design of buildings. J Struct Eng, ASCE 1995;121(1):10214.
[13] De-La-Llera JC, Chopra AK. Evaluation of seismic code provisions using strongmotion building records from the 1994 northridge esarthquake. UCB/EERC-97/
16, Earthquake Engineering Reserach Center, University of California,
Berkeley; 1997.
[14] Goel RK, Chopra AK. Seismic code analysis of buildings without locating
centers of rigidity. J Struct Eng, ASCE 1993;119(10):303955.
[15] Hao H. Torsional response of building structures to spatial random ground
excitations. Eng Struct 1997;19(2):10512.
[16] Hejal R, Chopra AK. Earthquake response of torsionally-coupled buildings.
UCB/EERC-87/20, Earthquake Engineering Research Center, University of
California, Berkely; 1987.

11

[17] Lin WH et al. Accidental torsion in buildings: analysis versus earthquake

motions. J Struct Eng, ASCE 2001;127(5):47581.
[18] LLSST. <http://www.earth.sinica.edu.tw/~smdmc/llsst/llsst.htm>.
[19] Sheikhabadi MRF. Simplied relations for the application of rotational
components to seismic design codes. Eng Struct 2014;59:14152.
[20] Tso WK. Static eccentricity concept for torsional moment estimation. J Struct
Eng, ASCE 1990;116(5):1199212.
[21] Wen KL, Yeh YT. Seismic velocity strucure beneath the SMART-1 array. Bull
Inst Earth Sci, Acad Sin 1984;4:5172.
[22] Wolff ED, Ipek C, Constantinou MC, Morillas L. Torsional response of
seismically isolated structures revisited. Eng Struct 2014;59:4628.