You are on page 1of 8

Nonlinear Analysis of a 16-Storey R/C Building

Designed According to EC2 & EC8




G.G. Penelis
Penelis Consulting Engineers S.A.

V.K. Papanikolaou
Civil Engineering Dept., Aristotle University of Thessaloniki, Greece





ABSTRACT:
A sixteen storey building with four basements already constructed in Bucharest has been designed according to
the provisions of EC2 and EC8, using elastic spectral modal analysis. Considering that the building is torsionally
sensitive in the nonlinear range, it was further checked and verified using nonlinear dynamic and static
procedures, using a space frame model. Specifically, time history analysis for seven different excitations, as well
as respective inelastic static analysis taking into account torsional effects were performed. The results are
examined regarding structural (global) and member (local) response and various issues concerning the adequacy
of the original elastic design and the applicability of advanced analysis methods are discussed.

Keywords: Reinforced concrete, Nonlinear static analysis, Nonlinear dynamic analysis, Torsion.


1. INTRODUCTION

The scope of the present paper is the application of advanced nonlinear analysis methods on a complex
reinforced concrete multi-storey building (considered high-rise for southern European standards) that
has been designed using the most modern available Codes in Europe, Eurocodes 2 and 8 (CEN, 2004a,
2004b) and has been already constructed in Bucharest.

The building is a sixteen storey RC dual system (frame wall) with four basements, which has been
designed for high ductility (DCH), PGA = 0.24g, q = 4.0, = 2.75,
2
= 0.4, S = 1.0, T
B
= 0.16 sec,
T
C
= 1.60 sec, T
D
= 2.00 sec (Soil C), according to the current Romanian Seismic Code (P100/2006),
which is a direct translation of EC8 (2004b) with a modified spectrum shape (National Annex to EC8).
The architectural plan and elevation of the building is shown in Figure 1 and it is apparent that the
positioning of the stiff core in the upper central part of the plan creates a torsional sensitivity which
will be observed in the second and third modes of vibration (shown in the subsequent section).

According to the requirements of EC8, the building is not characterized as torsionally sensitive, since
the torsional radius of each diaphragm is larger than the respective radius of gyration. Therefore, a
maximum behaviour factor of q = 5.2 is provisioned for high ductility class (DCH), which has been
reduced by the designers, to q = 4.0. However, the torsional sensitivity of the building arises in the
nonlinear region, where it is partially torsionally unrestrained (T.UR), considering the scenario that the
two facade rectangular walls (with openings) will yield during the excitation, resulting a torsionally
unbalanced system (Pauley, 1997). The assessment of the EC8 provisions in such cases (i.e. when only
elastic analysis is performed), was one of the key issues of the present study.

The elastic design has been performed using the modal spectral analysis feature of ETABS finite
element software (Computers and Structures Inc., 2008), while the design and capacity checks for
beams, columns and walls have been performed with the eTools EC dimensioning software (Penelis
Software Ltd., 2009). The static and dynamic nonlinear analyses have been performed using the
advanced Zeus NL finite element package (Elnashai et al., 2009), which features frame distributed
plasticity elements employing a fibre approach. Although more detailed information on the modelling
procedure is provided in a following section, it should be noted that the computational requirements
for solving the current building were surpassing - to the best of the authors knowledge - previous
applications of the same and similar software.




Figure 1. Typical plan (left) and elevation (right) of the 16-storey building

From the nonlinear numerical approaches applied, several issues have been investigated regarding the
adequacy of EC8 provisions for torsionally sensitive buildings, such as the available overstrength, the
required ductility (global and local) and the expected deformed shape (interstorey drifts at centre of
mass and opposite sides). On the other hand, several issues on the accurate nonlinear modelling of
actual buildings when using advanced scientific software have been handled. Such issues were the
application of diaphragm constraints, modelling of complex R/C cores (either using point hinge or
distributed plasticity models) and adequate mass distribution in plan. The above demonstrate the need
for efficient commercial tools and modelling simplifications (such as point hinge model approaches)
when the main scope is the checking of actual buildings in office practice, in contradiction to
analysing prototype buildings with advanced finite element software for research purposes.


2. FINITE ELEMENT MODELLING FOR NONLINEAR ANALYSIS

In this section, the finite element modelling procedure of the considered high-rise building is described
in detail. The building was modelled excluding the four basement levels and full fixation was applied
at the ground level. It consists of sixteen stories (including a mezzanine) and an elevator machine
room on the top story (Fig. 2). The gross dimensions of the plan are 32.2 18.8 m and the total height
is 58.25 m. The structure was modelled using the Zeus-NL finite element software (Elnashai et al.,
2009), which is capable of representing the spread of inelasticity within the member cross-section and
along the member length, utilizing the fibre analysis approach. It can also predict the behaviour of
three-dimensional frames under static or dynamic loading, taking into account both material and
geometric nonlinear response.

For concrete, a uniaxial constant confinement constitutive model was employed, with a peak
compressive strength f
c
= 32.0 MPa, tensile strength f
t
= 3.0 MPa, crushing strain
c
= 0.0022 and a
constant confinement factor K = 1.2 for the confined concrete section regions. For steel, a bilinear
elastoplastic constitutive model with kinematic hardening was applied, defined by the Youngs
modulus E
s
= 200 GPa, yielding strength f
y
= 500 MPa and a hardening modulus of 5 . All above
material properties were set in accordance with the original design directives, i.e. grades C32/40 for
concrete and B500C for steel.

Reinforced concrete sections (geometry and detailed reinforcement bar topology) were defined
according to the original formwork drawings and were assigned to cubic elastoplastic frame elements,
whose tangent stiffness matrix is integrated using second order Gaussian quadrature (two Gauss
points). Consequently, the size of these elements was critical in order to properly capture the inelastic
response in dissipative zones of the structure. For this reason, all physical structural elements (beams,
columns and walls) were meshed into four frame finite elements using a 15-35-35-15 % proportion
across the element length, which leads to a finer discretization near beam-column connections (where
resulting forces and deformations are expected to be larger).



Wireframe view

Solid view

Figure 2. Finite element model of the 16-storey building

Structural walls were modelled using vertical frame elements along the wall mass centre and their
horizontal kinematic constraint at storey levels with neighbouring beams was modelled explicitly,
using rigid elements (elastic material with E
Rigid
= 10E
s
and 1.0 1.0 m cross section). Rigid
diaphragm action was considered at all storey levels. However, numerical treatment using master-
slave joint constraints was not available in the employed software and hence an explicit representation
was necessary, using end-pinned crossed diagonal rigid links on each quadrilateral slab region.

The total gravity load for each storey was calculated from the G+
2
Q seismic load combination of the
initial design and was applied to the three beam inner-nodes and wall end-nodes, according to their
tributary area. For dynamic analysis, the total mass of each storey m = (G+
2
Q)/g was distributed to
the end-nodes of all vertical elements (columns and walls), according to their tributary area. The
horizontal loading pattern for inelastic static analysis was applied on the node nearest to the centre of
mass (C.M., Fig. 2) of each storey level and, for dynamic analysis, biaxial excitation in both directions
x and y (in the form of time-acceleration history) was applied on all base nodes.

Mode Zeus-NL ETABS
1 1.03 0.98
2 0.81 0.73
3 0.53 0.56
4 0.25 0.22
5 0.21 0.21
6 0.16 0.16
7 0.14 0.12
8 0.13 0.09
9 0.12 0.08
10 0.11 0.08

Mode 1 (1.03 sec)

Mode 2 (0.81 sec)

Mode 3 (0.53 sec)

Figure 3. Eigenperiods and mode shapes (Zeus-NL) of the finite element model

In order to validate the aforementioned modelling procedure, an eigenvalue analysis was performed
and compared qualitatively and quantitatively to the respective analytical results from the original
design, that was performed using the ETABS finite element software. Very good correlation was
observed regarding both eigenperiods and mode shapes. The first mode shape was purely translational
towards the weaker axis of the structure, the second one was mixed (translational/rotational) and the
third was purely rotational (Fig. 3). Considering that Zeus-NL and ETABS are considerably different
computational platforms in both finite element formulation principles and modelling assumptions (e.g.
in ETABS, walls and slabs were modelled with shell elements, rigid diaphragm was formulated using
kinematic constrains and gravity loads are applied directly on slab shells), it is deemed that the present
modelling attempt was successful and credible for the demanding nonlinear static and dynamic
analyses, presented in the subsequent sections.


3. NONLINEAR DYNAMIC ANALYSIS

3.1. General

For the comparison between nonlinear dynamic and static procedures, the initial idea was to derive the
Incremental Dynamic Analysis (IDA) curve (Papanikolaou and Elnashai, 2005). Nevertheless, the
onerous computational and storage cost considering the large scale of the present finite element model
and the large number of required PGA-scaled dynamic analyses redirected the research towards
applying three individual bidirectional dynamic excitations, in order to determine the maximum top
displacement, the deformed shape, the required ductility and the overall dynamic behaviour of the
structure.

3.2. Selected excitations

The excitations selected were three artificial ten second strong motion records (with a standard EC8
compatible frequency content for a = 0.24g and Soil C) and the 1977 Vrancea event (NS and EW
components). The 1977 Vrancea earthquake was introduced unscaled, since it presents an actual event
that seriously affected the city of Bucharest. This was considered of great interest in order to examine
the structural behaviour of an EC8 designed building, especially since the excitation has a significant
spectral amplification around the one second period range, which corresponds to the principal mode of
the structure.

3.3. Analysis procedure

Seven nonlinear time history dynamic analyses were performed, six using the EC8 artificial records
and one using the Vrancea event as follows :

a) EC8 #1 to #3 record : 100 % in the x direction and 30 % in the y direction (x + 0.30y)
b) EC8 #1 to #3 record : 30 % in the x direction and 100 % in the y direction (0.30x + y)
c) Vrancea NS record in the x direction and EW record in the y direction (x:NS + y:EW)

The analysis results that will be presented in the respective section (combined with results from
inelastic static analysis) are selected using (a) the most unfavourable dynamic response (in terms of
top displacement) for the three EC8 excitations and (b) the Vrancea event, as follows :

a) Base shear versus top storey displacement (P- capacity curve).
b) Maximum displacement profiles of building sides.
c) Moment chord rotation curves for a corner base column (C13, see Fig. 2).


4. INELASTIC STATIC ANALYSIS

4.1. General

The procedure adopted in EC8 (Annex B) for the static nonlinear analysis of buildings is the N2
method (Fajfar and Fischinger, 1988), which in the present form does not account for any torsional
amplifications in the nonlinear range, although later developments have been made to extent this
method to take into account the torsional response (Fajfar et al 2005). In the present study and for the
inelastic static analyses, the approach suggested by Penelis and Kappos (2002, 2005) is employed for
calculating the spectral loads as well as the equivalent SDOF oscillator, accounting for both the
translational and torsional components of the building response.
4.2. Analysis procedure

Three inelastic static (pushover) analyses were performed, two using EC8-compatible static loads (for
a = 0.24g and Soil C spectrum) and one using the Vrancea event. The above analyses were performed
in both positive and negative directions in order to capture the expected asymmetric structural
response. Specifically :

a) EC8 spectrum : 100 % in the x direction and 30 % in the y direction (x + 0.30y)
b) EC8 spectrum : 30 % in the x direction and 100 % in the y direction (0.30x + y)
c) Vrancea NS in the x direction and Vrancea EW in the y direction (x:NS + y:EW)

From the previous section it is apparent that different loads (lateral and torsional) were derived for
each static excitation, depending on the spectral shape. The capacity curves extracted from the analysis
were converted to Acceleration-Displacement Response Spectrum (ADRS) and bilinearised
employing the equal areas principle. The target displacement of the SDOF oscillator was in turn
determined iteratively using constant ductility capacity spectra and finally the target displacement of
the MDOF building was calculated.


5. ANALYSIS RESULTS

5.1. Global response

The global response results that were selected for presentation from each excitation and each type of
analysis are the following:

a) Static and dynamic capacity curves (P-) (Figs. 4 and 5) which demonstrate that there is a close
correlation between static and dynamic nonlinear behaviour and that the building shows asymmetric
response due to structural eccentricity. The maximum top displacement from nonlinear dynamic
analysis for the EC8 compatible excitations is approximately 20 cm for the x direction and 30 cm for
the y direction (Tab. 1), which renders the estimation of 36 cm and 68 cm calculated from the elastic
design of the building (EC8), respectively, a safe approach. Moreover, the base shear capacity of the
building is approximately 50 MN for x direction and 40 MN for y direction, while the base shear
design force is approximately 24 MN.

-60
-40
-20
0
20
40
60
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Dynamic
Static
Target displacement
Px (MN)
x (m)
EC8#2 x + 0.3y Direction x
1

%

d
r
i
f
t
-60
-40
-20
0
20
40
60
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Dynamic
Static
Target displacement
Py (MN)
y (m)
`
EC8#1 0.3x + y Direction y
1

%

d
r
i
f
t

Figure 4. Global P- comparison between static and dynamic analysis for the
EC8#2 x + 0.3y (left) and the EC8#1 0.3x + y (right) excitations

b) The storey displacements in the x and y directions for both sides of the building (x : top and bottom
of plan, y : left and right of plan, see Fig. 2), which demonstrate not only the translational response,
but also the rotational (Fig. 6). In the same figure, the derivation of the target displacement for
inelastic static analysis (using the ADRS method) is also shown. The differences between the dynamic
and static nonlinear approach regarding the displacement profiles are presented in Table 1. It is
deemed encouraging that the average difference regarding the response at the centre of mass is about
11 %, while the average difference for the sides response (or rotational response) is about 14 %.

5.2. Local response

The local response results are shown for the corner column C13 (top-right corner of plan, see Fig. 2) in
order to investigate the biaxial nonlinear stress state of the section under a varying axial load. Figure 7
shows the moment - chord rotation curves (static and dynamic) as well as the maximum dynamic
demand in comparison to the static demand calculated at the target displacement step of the inelastic
static analysis (for the excitation EC8#2 x + 0.30y). These results show an acceptable agreement
between dynamic and static approach, considering the different variation of the column axial load
between static and dynamic nonlinear analysis. Furthermore, the nonlinear demands (column
rotations) are well within the design requirements of EC8.

-60
-40
-20
0
20
40
60
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Dynamic
Static
Target displacement
Px (MN)
x (m)
Vrancea x:NS + y:EW
1

%

d
r
i
f
t
Direction x


Figure 5. Global P- comparison between static and dynamic analysis for the Vrancea x:NS + y:EW excitation

0.00
0.05
0.10
0.15
0.20
0.00 0.10 0.20 0.30 0.40 0.50 0.60
S
a
(g)
S
d
(m)
= 3.0
Sd,t = 0.28
dt = 0.295 m
Vrancea x:NS + y:EW Direction x
c1 = 0.947
c2 = 0.946
m* = 25361
0
10
20
30
40
50
0.00 0.10 0.20 0.30 0.40
Dynamic
Static @ dt = 0.295 m
h (m)
x
(m)
Top Bottom
58.25
Vrancea x:NS + y:EW
Direction x


Figure 6. Derivation of target displacement and displacement profiles for the Vrancea x:NS + y:EW excitation


6. CONCLUSIONS AND DISCUSSION

From the analysis results presented in the previous section, the following comments are derived
regarding the adequacy of the EC8 design of the building, as verified by the use of advanced nonlinear
static and dynamic approaches :

a) The building has the required overstrength and ductility to justify the behaviour factor q = 4.0,
employed in the original design.
b) The building global response from elastic analysis according to EC8 (storey displacements and
forces) is conservative by approximately 100 % compared to the nonlinear response, which is
expected since the Code design includes a global safety factor.
c) The local inelastic demand on the corner columns, as checked for column C13, shows that they
have been adequately designed.

-12000
-8000
-4000
0
4000
8000
12000
-0.008 -0.004 0 0.004 0.008
Dynamic
Static
Target displacement
ry (rad)
My (kNm)
Column C13 - Direction x EC8#2 x + 0.3y
-12000
-8000
-4000
0
4000
8000
12000
-0.008 -0.004 0 0.004 0.008
Dynamic
Static
Target displacement
rx (rad)
Mx (kNm)
Column C13 - Direction y EC8#2 x + 0.3y


Figure 7. Local M-r comparison between static and dynamic analysis for the EC8#2 x + 0.3y excitation

Table 1. Comparison of top displacements at C.M. and opposite sides between static and dynamic analysis

EC8#1 x + 0.3y (
x
) 0.181 0.200 10.6%
EC8#2 x + 0.3y (
x
) 0.208 0.185 11.2%
EC8#3 x + 0.3y (
x
)
0.205 0.200 2.5%
EC8#1 0.3 x + y (
y
) 0.308 0.290 6.0%
EC8#2 0.3 x + y (
y
) 0.229 0.290 26.7%
EC8#3 0.3 x + y (
y
) 0.251 0.270 7.6%
Vrancea x:NS + y:EW (
x
)
0.259 0.295 13.8%
Average 11.2%
EC8#1 x + 0.3y (
x
) - Top 0.141 0.151 7.3%
EC8#1 x + 0.3y (
x
) Bottom 0.222 0.250 12.8%
EC8#2 x + 0.3y (
x
) - Top 0.177 0.138 22.2%
EC8#2 x + 0.3y (
x
) Bottom
0.240 0.232 3.4%
EC8#3 x + 0.3y (
x
) - Top 0.175 0.151 13.7%
EC8#3 x + 0.3y (
x
) Bottom 0.238 0.250 5.1%
EC8#1 0.3 x + y (
y
) - Left 0.290 0.227 21.8%
EC8#1 0.3 x + y (
y
) - Right 0.328 0.352 7.3%
EC8#2 0.3 x + y (
y
) - Left 0.219 0.226 3.1%
EC8#2 0.3 x + y (
y
) - Right 0.244 0.352 44.6%
EC8#3 0.3 x + y (
y
) - Left 0.238 0.187 21.4%
EC8#3 0.3 x + y (
y
) - Right 0.264 0.299 13.0%
Vrancea x:NS + y:EW (
x
) - Top 0.222 0.231 4.2%
Vrancea x:NS + y:EW (
x
) - Bottom 0.300 0.362 20.7%
Average 14.3%
Excitation Difference
Response at centre of mass (CM)
Excitation Difference
Response at opposite sides
Dynamic
(maximum)
Static
(at target disp.)
Dynamic
(maximum)
Static
(at target disp.)


From the above it would be a safe conclusion that this high rise R/C torsionally sensitive building,
designed according to Eurocodes, is indeed capable to withstand both the design (EC8) and the
Vrancea earthquake, which is the historically and seismologically critical for Bucharest, event with a
global safety factor of approximately 2.0, rendering the EC8 design provisions, for this type of
buildings, a safe design approach.

From the comparison between dynamic and static nonlinear approaches, it is evident that for either the
global or the local response, the more efficient, less time consuming and computationally less
demanding inelastic static approach applied herein, produces results that are reasonably close to their
dynamic counterparts (and therefore usable) for this particular torsionally sensitive building.
Finally, from the whole experience of attempting to accurately model an actual complex building for
nonlinear analysis, the following issues have been raised :

a) The extent of modelling detail of each section is critical, i.e. reinforcement amount and location,
confined and unconfined regions, wall sections, T-beam sections etc.
b) The modelling of the cores is an important issue, which heavily depends on the nonlinear
approach selected. In a distributed plasticity fibre model, that takes into account the axial
nonlinearity of each element (N- curve), the cores may be modelled as a set of individual walls
(one wall per core flange). On the contrary, in a point hinge model, where the moment cannot be
nonlinearly decomposed to a couple of tension-compression axial forces (since a N- law is not
provided), the core must be modelled as one element, with a moment-rotation (curvature) diagram
defined for the complex core cross section.
c) The modelling of the rigid connections (actual versus numerical rigidity).
d) The modelling of the rigid diaphragm (e.g. using X-bracing).
e) The material models adopted for concrete and steel (especially the hardening of reinforcement
bars, compared to the expected hardening of the reinforced concrete as a composite material).
f) The vertical initial loads and mass distribution.
g) The application and distribution of lateral (pushover) loads and torsional moments, in order to
avoid numerical instabilities.
h) The efficient post processing of the analysis results which - especially for the dynamic analysis -
are produced in enormous amounts.

All these issues demonstrate that currently, the only practical alternative to elastic design of buildings
is the simplified inelastic static analysis using commercial analysis software employing point hinge
models and design recommendations such as FEMA 356 [BSSC, 2000], since more advanced
approaches may conceal serious and unforeseen pitfalls. Nevertheless, it is speculated that the
continually increasing computational power will favour nonlinear dynamic analysis to reach higher
levels of applicability in the future.


REFERENCES

BSSC (2000) Prestandard and Commentary for the Seismic Rehabilitation of Buildings, FEMA-356, Federal
Emergency Management Agency, Washington, D.C.
CEN (2004a) Eurocode 2 - Design of Concrete Structures, European Committee for Standardization.
CEN (2004b) Eurocode 8 - Design of Structures for Earthquake Resistance, European Committee for
Standardization.
Computers and Structures Inc. (2008) ETABS reference manual, Computers and Structures, Berkeley,
California.
Elnashai A.S, Papanikolaou V.K and Lee D.H. (2009) Zeus-NL - A program for inelastic dynamic analysis of
structures, Mid-America Earthquake Center, University of Illinois at Urbana-Champaign.
Fajfar P., Fischinger M. (1988) N2 A method for non-linear seismic analysis of regular buildings,
Proceedings of the Ninth World Conference in Earthquake Engineering, Tokyo-Kyoto, Japan, Vol. 5, 111-
116.
Fajfar,P., Marusic, D. and Perus, I. (2005) The extension of the N2 method to asymmetric buildings,
Proceedings of the 4
th
European Workshop on the Seismic Behaviour of Irregular and Complex Structures,
CD ROM, Thessaloniki.
Papanikolaou V.K. and Elnashai A.S. (2005) Evaluation of conventional and adaptive pushover analysis I :
Methodology, Journal of Earthquake Engineering, 9(6), 923-941.
Paulay, T. (1997) Seismic torsional effects on ductile structural wall systems, Journal of Earthquake
Engineering, Vol. 1, No. 4, pp. 721-745.
Penelis G.G. and Kappos A.J. (2002) 3D pushover analysis: the issue of torsion, Proceedings of The Twelfth
European Conference on Earthquake Engineering, electronic resource , London, UK.
Penelis G.G. and Kappos A.J. (2005) Inelastic torsion effects in 3D pushover analysis of buildings,
Proceedings of the 4th European Workshop on the Seismic Behaviour of Irregular and Complex Structures,
electronic resource, Thessaloniki, Greece.
Penelis Software Ltd. (2009) eToolsEC user manual, Penelis Software Ltd., Thessaloniki, Greece.

You might also like