EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2005; 34:1449–1465

Published online 1 June 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.486

Inelastic torsion of multistorey buildings under

earthquake excitations

K. G. Stathopoulos‡; § and S. A. Anagnostopoulos∗; †; ¶

Department of Civil Engineering; University of Patras; 26500 Patras; Greece

SUMMARY
The inelastic earthquake response of eccentric, multistorey, frame-type, reinforced concrete buildings is
investigated using three- and
ve-storey models, subjected to a set of 10, two-component, semi-arti
cial
motions, generated to match the design spectrum. Buildings designed according to the EC8 as well as
the UBC-97 code were included in the investigation. It is found that contrary to what the simpli
ed one-
storey, typical, shear-beam models predict, the so-called ‘exible’ side frames exhibit higher ductility
demands than the ‘sti’ side frames. The substantial dierences in such demands between the two
sides suggest a need for reassessment of the pertinent code provisions. This investigation constitutes
one of the
rst attempts to study the problem of inelastic torsion by means of realistic, multistorey
inelastic building models. Additional studies with similar or even more re
ned idealizations will certainly
be required to arrive at de
nite results and recommendations for possible code revisions. Copyright
? 2005 John Wiley & Sons, Ltd.

KEY WORDS: asymmetry; eccentricity; torsion; multistorey buildings; earthquake inelastic response;
plastic hinge model

INTRODUCTION

Earthquake torsion in buildings can be induced from a variety of causes such as stiness and
or mass eccentricities, eccentric arrangements of ‘non-structural’ yet load-bearing elements,
asymmetric yielding and input variations at the support points. Because some of these factors
cannot be known in advance and hence cannot be explicitly accounted for in design, torsion
can make the response of the building more severe and thus increase damage or even con-
tribute to collapse. Modern codes (e.g. UBC [1], EC8 [2], Greek Code [3]) deal with torsion

∗ Correspondence to: S. A. Anagnostopoulos, Department of Civil Engineering, University of Patras, 26500 Patras,
Greece.
† E-mail: saa@upatras.gr
‡ E-mail: kstathop@upatras.gr
§ Civil Engineer, PhD.
¶ Professor.

Received 9 March 2004

Revised 22 December 2004
Copyright ? 2005 John Wiley & Sons, Ltd. Accepted 25 January 2005
1450 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

by placing restrictions on the design of buildings with irregular layouts and also through the
introduction of an accidental eccentricity that must be considered in design.
Most of the published work on inelastic earthquake response of non-symmetric buildings
is based on simpli
ed, highly idealized, one-storey models, with simple shear-beam elements
designed for lateral load resistance (e.g. References [4–7]). The majority of such models
consists of three elements parallel to the y-axis, have single eccentricity (ex
= 0, ey = 0)
and have been subjected to simple component earthquake motions (e.g. Reference [8]). In
recent years, similar systems but with elements along both the x and y axes and with
double eccentricity (ex
= 0, ey
= 0) have also been investigated for two-component seismic
input (e.g. References [9, 10]). It is worth to mention here that the inelastic shear-beam
model has been used since 1972 for earthquakes response analyses of asymmetric multistorey
systems [11, 12].
A key issue that many published papers address is whether the torsional provisions of
various codes are adequate (e.g. References [13–16]). The answers to this question, however,
vary considerably, depending upon the type of models and assumptions used by the authors in
the respective studies. As a result, conicting conclusions have often been obtained, identi
ed
and discussed (e.g. References [6, 17, 18]). A more important issue, however, that seems
to have escaped the attention of many researchers, at least until recently, has to do with the
question of how appropriate is the simpli
ed, one-storey, shear-beam model to approximate the
inelastic earthquake response of real multistorey buildings, through which the pertinent code
provisions should be evaluated. Some recent comparisons have shown signi
cant qualitative
dierences in the predicted responses between simpli
ed and detailed models, thus raising
questions about code assessments based on the oversimpli
ed one-storey inelastic models
[19–22]. These dierences and their causes are as follows: (a) The stiness and strength
of the resisting elements of the simpli
ed shear-beam model are speci
ed and calculated
independent of each other and only for seismic loads. In real buildings, member stiness,
strength and yield deformation are related to each other directly in a way that a change in
one parameter entails changes in the other two. This problem has been addressed recently,
using again simpli
ed one-storey systems with two types of lateral load resisting elements, and
interesting conclusions have been drawn [23]. (b) In real buildings, members are designed
for lateral and vertical loads and hence their stiness and strength, in absolute or relative
terms, are dierent from the corresponding values of the resisting elements of the simpli
ed
shear-beam models. Thus, the percentage changes of these quantities caused by applying Code
provisions for torsion in real buildings are much smaller than the respective changes in the
idealized one-storey, shear-beam models and the same should be expected for the pertinent
eects on the corresponding responses. (c) Yielding of an end-element of the simpli
ed model
implies the practical elimination of the stiness in that position. A corresponding case in a
real building would be the formation of a mechanism at the same side of the building, i.e.
the simultaneous yield of all beam and column ends in a given oor of the corresponding
frame, which modern codes prevent through capacity design provisions. In real buildings, the
post-elastic stiness of any given frame is a signi
cant fraction of its elastic stiness, as it
is controlled by the substantial number of members, typically columns, that are elastic at
any given instant. Thus, there are great dierences in the post-elastic eccentricities between
real buildings and the shear-beam models. (d) Higher mode eects have also an inuence,
although similar dierences were observed when the comparisons were made with one-storey
buildings as well [20, 21].

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1451

In order to eliminate uncertainties associated with extrapolations to multistorey buildings,

the comparisons reported in References [10, 21] were restricted to one-storey, three-degree-of-
freedom models, whose only dierence was in the type of the lateral load resisting elements.
Given, however, that results from the highly idealized one-storey, shear-beam models have
routinely been used to assess code torsional provisions in general (thus deemed applicable
to multistorey buildings as well), and in view of the aforementioned shortcomings of the
simpli
ed models even for one-storey buildings, a need exists for an assessment of inelas-
tic earthquake induced torsion in multistorey, frame type buildings, based on more realistic
structural models (e.g. those using plastic hinges). To the best of the authors’ knowledge,
there are only few studies based on such models. The pertinent publications are limited to the
investigation of an existing building [24], to the study of highly idealized framed buildings
with one axis of symmetry and under one-component motion [25, 26] or to some idealized
eccentric frame buildings, designed as symmetric [27]. Not only are these studies very limited,
compared to those based on the simpli
ed one-storey shear-beam model, they also address
other aspects of the problem than the work herein.
The present paper presents results from an investigation of this problem, based on three-
dimensional, multistorey building models with biaxial eccentricity, excited by a set of two-
component earthquake motions. The post-elastic behaviour is idealized using plastic hinges at
the ends of the yielding members. Results for one-storey eccentric buildings using the plastic
hinge model can be found in Reference [21].

SYSTEMS AND MOTIONS USED

The buildings selected for this study (Figure 1) are three- and
ve-storey reinforced concrete
space frames, formed by three or four plane frames along the x and y directions. The two end
frames parallel to the y direction are labelled Fr1 and Fr3, those parallel to the x direction
are Fr4 and Fr6, while Fr2 and Fr5 indicate interior frames parallel to the y and x directions,
respectively. Note that for the y direction of the three-storey building and for both directions
of the
ve-storey building, Fr2 and Fr5 designate pairs of interior frames. Both buildings have
a typical storey height of 3:0 m and a ground storey height of 4.0 m. Using appropriate dis-
tributions of the oor loads, e.g. through non-symmetric live load distribution, non-symmetric
brick partition wall placing and non-symmetric balconies (common causes of mass eccentric-
ity in typical Greek buildings, not shown in the given layout), non-symmetric joint masses
were assigned at each oor and thus biaxial mass eccentricities were introduced in all oors
(CM on the diagonal, see Figure 1). The models used for both design and analyses were
3-D, lumped mass models with masses lumped at the joints. For both buildings, the cases
of em = 0 (symmetric), em = 0:10d and 0:20d (em = mass eccentricity, d = length of diagonal)
were examined, while for the three-storey building the value em = 0:30d was also included.
Both buildings were designed as spatial frames for gravity and earthquake loads according
to early versions of EC2 (reinforced concrete), and EC8 (earthquake resistant design), which
requires an accidental eccentricity eacc = 0:05L to be considered in each direction (L = length
of the building side perpendicular to the earthquake direction). The dimensioning of the frame
members took into account the uneven distribution of member forces due to the mass eccen-
tricities, and hence stiness eccentricities were also generated, as it happens in actual practice.
We should also note here that dierent variants of the buildings were produced by designing

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1452 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

1 2 3 4 1 2 3 4
FR4 FR4
y
y
CM
CM
5 6 7 8
5 ey 6 7 8 ey
Ly=12m FR5 FR5
GC x Ly=15m GC x
CR
CR
ex 9 10 11 12
ex
9 10 11 12 14 16
FR6 13 15 FR6
FR1 FR2 FR3 FR1 FR2 FR3
Lx=18m Lx=21m

Figure 1. Three- and
ve-storey reinforced concrete buildings.

them with both the dynamic (multi-modal, response spectrum) method and by the equivalent
static method as per EC8-94, and also by the dynamic method as per UBC-97 [19]. Response
results for all these designs were similar, so only those from the dynamic design per EC-8
will be presented here. Results pertaining to the designs according to UBC-97, in which an
ampli
ed accidental eccentricity is used, can be found in Reference [22].
Since in multistorey buildings, the so-called centre of rigidity (CR) cannot be really de
ned,
except under very restrictive conditions, an approximate CR was computed herein for reference
purposes, on a oor-by-oor basis as follows:

m
n
i=1 Kf−iy xi i=1 Kf−ix yi
esx =
m ; e sy =
n (1)
1 K f−iy 1 Kf−ix
 −1
24E 2 1 1
Kf−i = 2
+
+
(2)
h Kc Kba Kbb

where esx , esy are the x and y co-ordinates of the approximate centre of rigidity CR, Kf−i
designates the approximate storey stiness of frame i; x and y the directions of the frame
axis, m and n the number of frames along the y and x axes, respectively, E the modulus
of elasticity, Kc = EIc =h; Kb = EIb =‘, Ic , Ib the section moment of inertia of columns and
beams, respectively, h the storey height and ‘ the beam length. The indices a and b in the
summations for the frame storey stiness designate the upper and lower beams of the frame

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1453

Table I. Estimated physical eccentricities.

Storey 1 Storey 2 Storey 3 Storey 4 Storey 5

Number of
storeys
m
x
y
x
y
x
y
x
y
x
y

3 0.10 0.07 0.09 0.09 0.10 0.09 0.10

0.20 0.13 0.20 0.17 0.20 0.16 0.20
0.30 0.24 0.28 0.27 0.29 0.25 0.28
5 0.10 0.07 0.07 0.08 0.09 0.09 0.09 0.08 0.09 0.10 0.10
0.20 0.15 0.17 0.18 0.19 0.18 0.19 0.16 0.18 0.17 0.19

Table II. Fundamental periods of buildings.

Code stiness Secant stiness to yield

Number of
storeys
m Ty (s) Tx (s) T (s) Ty (s) Tx (s) T (s)

3 0.00 0.69 0.61 0.61 1.10 1.01 0.98

0.10 0.70 0.62 0.53 1.05 1.01 0.88
0.20 0.67 0.58 0.45 0.94 1.00 0.78
0.30 0.70 0.54 0.36 0.90 1.00 0.71
5 0.00 0.82 0.76 0.74 1.32 1.31 1.23
0.10 0.83 0.75 0.64 1.28 1.30 1.10
0.20 0.84 0.69 0.54 0.90 1.00 0.65

in the considered storey. Table I lists the total approximate eccentricities of each building
(see Figure 1) corresponding to mass eccentricities of 0:10; 0:20 and 0.30 (
m = 0:30 was not
considered for the
ve-storey building).
The two sets of buildings, modelled as 3-D structures, were subjected to a group of semi-
arti
cial, two-component motions and were analysed using a modi
ed version of the ANSR
program [28]. Beams and columns were idealized as non-linear members by means of the
well-known plastic hinge model, for which the modi
ed moment—rotation relationship of
Takeda [29], with a strain hardening ratio of 0.05, was used. Bending–axial interaction in
columns is treated approximately, through separate interaction diagrams in the two planes of
bending. However, this approximation is not very important because the column axial loads
are small fractions of the limit loads and also because most of the yielding takes place in the
beams due to capacity design provisions.
Two levels of member stiness were considered: one corresponding to operational conditions
as speci
ed by the code for the design of the building and the second, a secant stiness to
the yield point based on antisymmetric bending. The latter is computed as EI = My ‘=6y ,
where My is the yield moment, ‘ the member length and y the yield rotation according
to the semi-experimental equation by Park and Ang [30]. The secant stiness at yield is
quite lower than the code stiness and leads to substantially softer models. This is seen from
the three lower periods Tx , Ty , T of the two buildings, listed in Table II for the various
eccentricities.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1454 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

Figure 2. Typical force-deformation curve of Frame 1 from push-over analysis

of the
ve-storey building for
= 0:20.

For an easier assessment of code design eects on strength distribution among the lat-
eral load carrying elements, as well as for a subsequent comparison with simpli
ed, one-
storey, shear-beam models, pushover analyses were also carried out to determine the force–
deformation curves for each of the plane frames forming the symmetric and the eccentric
variants of the two buildings. These curves were obtained for a typical static, triangular,
code-prescribed earthquake loading and were subsequently approximated with bilinear dia-
grams as indicated in Figure 2 for one of the frames. Table III lists the yield force, stiness
and yield displacement of each plane frame corresponding to the various eccentricities. It is
recalled that the mass distribution was selected so as the centre of mass lay on the diagonal
(Figure 1), which implies equal mass eccentricities along the two main axes. We observe
that as the mass centre moves away from the geometrical centre, the strength and stiness of
the elements towards which it moves generally increase, since the corresponding gravity loads
also increase. However, these increases do not follow a totally consistent pattern because these
two properties are also aected from the seismic loading (which depends on the three periods
Tx , Ty and T , and the design spectrum) as well as by drift limitations and other code imposed
constraints. Moreover, the approximation introduced by linearizing the push-over curves may
also have a small eect.
For this study, a group of
ve two-component, semi-arti
cial motions were used. These were
generated from a group of
ve, two-component, real earthquake records, to match closely the
EC8 design spectrum (spectrum with descending branch ∝ 1=T 2=3 ). The method, based on trial
and error and Fourier transform techniques [31], gave excellent results as Figure 3 indicates.
A good match is necessary so that the results are not masked by any eects of over- or
under-loading the buildings. Each motion pair was applied twice by mutually changing the
components along the x and y system axes. Thus, each design case was analysed for 10
sets of two-component motions and mean values of peak response indices were computed. In
this manner, the eects of individual motions are smoothed and the conclusions become less
dependent on speci
c motion characteristics.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1455

Table III. Strength, stiness and yield displacements from push-over analyses of the plane frames
forming the three- and
ve-storey buildings.

m Fr1 Fr2a Fr2b Fr3 Fr4 Fr5 Fr6

Three-storey building
0.00 Fy (kN) 408 376 376 408 571 529 571
K (kN/m) 3025 2765 2765 3025 4497 4006 4497
y (m) 0.135 0.136 0.136 0.135 0.127 0.132 0.127

0.10 Fy (kN) 445 378 342 455 657 549 632

K (kN/m) 3502 2762 2535 3636 5312 3952 4978
y (m) 0.127 0.137 0.135 0.125 0.124 0.139 0.127
0.20 Fy (kN) 395 296 347 539 709 509 667
K (kN/m) 3095 2404 2977 4784 5766 4161 5512
y (m) 0.129 0.125 0.117 0.113 0.123 0.122 0.121
0.30 Fy (kN) 338 303 389 700 870 499 621
K (kN/m) 2817 2747 3272 5968 6926 4198 5671
y (m) 0.12 0.11 0.119 0.125 0.126 0.119 0.110

Five-storey building

m Fr1 Fr2a Fr2b Fr3 Fr4 Fr5a Fr5b Fr6

0.00 Fy (kN) 601 603 603 601 682 629 629 682
K (kN/m) 3193 3059 3059 3193 3513 3164 3164 3513
y (m) 0.188 0.197 0.197 0.188 0.194 0.199 0.199 0.194
0.10 Fy (kN) 689 582 610 671 725 667 617 731
K (kN/m) 3552 3023 3338 3519 3837 3462 3280 3748
y (m) 0.194 0.193 0.183 0.191 0.189 0.193 0.188 0.195
0.20 Fy (kN) 553 586 622 903 823 786 710 662
K (kN/m) 3051 3455 3726 5093 4435 4657 4041 3354
y (m) 0.181 0.170 0.167 0.177 0.186 0.169 0.176 0.197

RESULTS

Three sets of results will be given: top storey displacements, ductility factors and damage
indices. They will be presented only for the secant to yield member stiness, since the con-
clusions from the stier models (those with member stiness according to the code) are for
practical purposes almost the same. Computation of ductility factors as well as damage in-
dices was decided because the
rst have been used widely with all types of inelastic structures,
while the second have been developed as better measures of inelastic deformation for con-
crete members. Listed displacements are mean values of the peak displacements from the 10
earthquake sets, while ductility factors and damage indices are maximum values for all beams
and columns (separately) in any storey of a plane frame, selected from the set of values at
both ends of all respective members in a storey, where each value at a given location is the
mean of the maxima for the 10 sets of earthquake motions.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1456 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

0.8
EC8 - Soil B - Ag,max=0.24g
0.7
Mean Sa
0.6

0.5
Sa (g)

0.4

0.3

0.2

0.1

0
0.00 0.50 1.00 1.50 2.00 2.50 3.00
Period (sec)

Figure 3. EC8 Design spectrum and mean spectrum of the ten semi-arti
cial motions.

The ductility factor used herein is the typical index, based on the maximum plastic hinge
rotation at any of the two member ends. It is de
ned as
 
p
# = 1 + (3)
y
where p is the maximum plastic hinge rotation at the ends of a member (beam or column)
and y is a normalizing ‘yield’ rotation, typically set equal to y = My ‘=6EI . The damage index
is based on energies computed from the hysteresis loops at the ends of yielding members and
is de
ned as follows [32]:

max Ed + 0:02 d Eh
DI = (4)
Ed; u
where Ed is the deformation energy of the member during its peak response, Eh the total
hysteretic energy consumed during the response and Ed; u the deformation energy when the
member is loaded monotonically to failure.
Figure 4 compares the mean of the peak top storey displacements of the two buildings at
their ‘sti’ and ‘exible’ edges as functions of the natural eccentricity. The two upper graphs
are for the three-storey buildings and the two lower graphs are for the
ve-storey buildings.
Moreover, frames 1; 3 are in the y direction and frames 4; 6 in the x direction (see Figure 1).
We see that the ‘sti’ sides in both directions are always experiencing less displacement than
their ‘exible’ counterparts as a result of natural eccentricities, thus justifying the ‘sti’ vs
‘exible’ terminology. However, the displacements of each frame do not exhibit a consistent,
monotonic change, as functions of the natural eccentricity, although the same trend appears
for corresponding frames of both buildings.
Mean, peak, rotational ductility demands for the members of the three- and
ve-storey
buildings are given in Figures and 6, while damage indices are shown in Figures 7 and 8.
The values given are means of the peak values in all beams or columns of a frame at a given

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1457

3 story - Frames in y direction 3 story - Frames in x direction

Fr1 Fr3 Fr6 Fr4

22.0 22.0
20.0 20.0
18.0 18.0
(cm)

(cm)
16.0 16.0
14.0 14.0
12.0 12.0
10.0 10.0
0.00 0.10 0.20 0.30 0.00 0.10 0.20 0.30
= e/L = e/L

5 story - Frames in y direction 5 story - Frames in x direction

Fr1 Fr3 Fr6 Fr4
30.0 30.0
28.0 28.0
26.0 26.0
(cm)

(cm)
24.0 24.0
22.0 22.0
20.0 20.0
18.0 18.0
16.0 16.0
0.00 0.10 0.20 0.00 0.10 0.20
= e/L = e/L

Figure 4. Peak edge displacements of the three- and
ve-storey buildings.

Solid lines: sti edges–frames 1,6; Dashed lines: exible edges–frames 3,4.
Frames 1,3: y direction. Frames 4,6: x direction.

oor, for the 10 earthquake excitations. The top two rows of graphs in all
gures are for
beams and the lower two are for columns. Each graph has two sets of data, corresponding to
a dierent frame. Solid lines are for frames at the ‘sti’ edge (frames 1 and 6), while dashed
lines are for frames at the ‘exible’ edge (Frames 3 and 4). Moreover, in each group of two
rows (top two for beams and lower two for columns) the
rst row is for frames in the y
direction, while the second row is for frames in the x direction. Labelling of these frames can
be seen in Figure 1. Graphs in the same column correspond to the same mass eccentricity,
0:00; 0:10 and 0.20 for both three- and
ve-storey building.
To comment briey on the response of the symmetric buildings, we see that ductility
demands in the beams are in the range of 2.0–4.5, while the columns remain essentially
elastic except for the top oors, where capacity design rules are not applied, and also for the
base where yielding is typically expected. In the
ve-storey buildings, column yielding at the
top and base is minimal.
The eect of eccentricity is seen by comparing beam and column ductility of the two
end frames with corresponding values in the symmetric cases. As the eccentricity increases,
we observe a substantial increase in beam ductility of the ‘exible’ side frames and a less
pronounced decrease in the beam ductility of the ‘sti’ side frame. This is the case for both
buildings and in both directions x and y. At some levels the ‘exible’ side frame exhibits

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1458 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

[Frames 1-3 - Beams] Floors

ε=0.00 ε=0.10 ε=0.20

[Frames 6-4 - Beams] Floors

ε=0.00 ε=0.10 ε=0.20

[Frames 1-3 - Columns] Floors

ε=0.00 ε=0.10 ε=0.20

[Frames 6-4 - Columns] Floors

ε=0.00 ε=0.10 ε=0.20

Figure 5. Rotational ductility factors for frames of the three-storey buildings.

Solid lines: sti edges–frames 1,6; Dashed lines: exible edges–frames 3,4.
Frames 1,3: y direction. Frames 4,6: x direction.

more than two times the ductility demands of the ‘sti’ side frame. Ductility factors of the
beams in the middle frames, not shown here, have milder increases than the ‘exible’ side
beams, compared to the symmetric case. The eects of eccentricity on column ductility are

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1459

[Frames 1-3 - Beams] Floors:

ε=0.00 ε=0.10 ε=0.20

[Frames 6-4 - Beams] Floors:

ε=0.00 ε=0.10 ε=0.20

[Frames 1-3 - Columns] Floors:

ε=0.00 ε=0.10 ε=0.20

[Frames 6-4 - Columns] Floors:

ε=0.00 ε=0.10 ε=0.20

Figure 6. Rotational ductility factors for frames of the
ve-storey buildings.

Solid lines: sti edges–frames 1,6; Dashed lines: exible edges–frames 3,4.
Frames 1,3: y direction. Frames 4,6: x direction.

not as pronounced as on the beams, mainly because columns are designed to remain elastic,
except for the top storey and base. Nevertheless, we see that in the three-storey building,
column ductilities of the ‘sti’ side have either remained practically unchanged or they have

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1460 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

[Frames 1-3 - Beams] Floors:

ε ε ε
[Frames 6-4 - Beams] Floors:

ε ε ε
[Frames 1-3 - Columns] Floors:

ε ε
ε
[Frames 6-4 - Columns] Floors:

ε ε ε

Figure 7. Damage indices for frames of the three-storey buildings. Solid lines: sti edges–frames 1,6;
Dashed lines: exible edges–frames 3,4. Frames 1,3: y direction. Frames 4,6: x direction.

decreased, with only exception at the base of frame 3 for
= 0:20. In the ‘exible’ side, we
see ductility increases at all locations except at the top of frame 3 in the y direction. In the

ve-storey building, all columns at the ‘sti’ side have remained elastic, while at the ‘exible’
side there has been a substantial ductility increase at the top oor for
= 0:10.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1461

[Frames 1-3 - Beams] Floors:

ε ε ε
[Frames 6-4 - Beams] Floors:

ε ε ε
[Frames 1-3 - Columns] Floors:

ε ε ε
[Frames 6-4 - Columns] Floors:

ε ε ε

Figure 8. Damage indices for frames of the
ve-storey buildings. Solid lines: sti edges–frames 1,6;
Dashed lines: exible edges–frames 3,4. Frames 1,3: y direction. Frames 4,6: x direction.

Similar observations are made if damage indices rather than ductility factors are used
as measures of inelastic response, as can be seen in Figures 7 and 8, where the same
graph arrangement as in Figures and 6 has been used. We see again that damage in-
dices in the eccentric buildings are substantially increased in the frames at the ‘exible’

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1462 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

edges and decreased in the ‘sti’ edges, compared to those of the corresponding symmetric
buildings.

DISCUSSION

The fact that both indices of inelastic response—ductility factors and damage indices—for
both structural directions—x and y—exhibit similar patterns, renders more credibility to the
results and indicates a real pattern of behaviour that constitutes a code shortcoming, namely
the uneven distribution of ductility demands in eccentric buildings. To explain it we can
go to Table III and to Figure 4. Table III indicates that as eccentricity increases, the push-
over estimated stiness and strength generally increase, but this is done in a more or less
proportional manner, so that the equivalent yield displacements remain relatively unchanged.
We see that the change in these yield displacements for any of the edge frames of both
buildings is less than 13%, as the eccentricities increase from 0 to 30% and from 0 to 20%
in the three- and
ve-storey buildings, respectively. On the other hand, Figure 4 indicates that
as the eccentricity increases, the peak displacements at the ‘exible’ edges (frames 3 and 4)
increase noticeably, compared to the displacements at the ‘sti’ edges (frames 1; 6). Higher
global inelastic displacements imply even higher local inelastic eects, leading to the observed
substantial dierences in rotational ductility factors and damage indices at the two edges.
So, if the objective for a well-balanced design would be to have a similar level of inelastic
action throughout the structure, thus avoiding premature local failures, it seems that current
design provisions in EC8 do not meet such an objective when applied to eccentric frame
buildings. We will also note here that since damage indices are energy-based parameters, they
could be considered as more ‘stable’ than peak ductility demands. Given, however, that both
indices are mean values from 10 earthquake motions, this makes them quite reliable for more
general conclusions. Similar results have been obtained and the same conclusions are reached
with the stier models of the buildings based on the higher member stiness speci
ed by the
codes for design purposes [19].
We note here that the simpli
ed, one-storey shear-beam model predicts the opposite be-
haviour: higher ductility demands at the ‘sti’ rather than the ‘exible’ sides (e.g. References
[5, 7, 20, 33–35]). This must be attributed to the shortcomings of the simpli
ed model for
predicting the inelastic response of multistorey, frame-type buildings, as listed in the intro-
duction. These
ndings raise a cautionary ag about the way the simpli
ed models have been
used to assess code provisions for torsion. Of course, they do not imply that every conclusion
based on such models is not valid. Their simplicity will still keep a role for them as useful
tools for identifying the key factors aecting inelastic torsional behaviour, at least for simple
buildings, and for preliminary testing of new recommendations for code revisions. However,
for the reasons reported earlier, any such recommendations ought to be tested with as realistic
models as possible, before they are adopted in codes.

CONCLUSIONS

Based on realistic structural models that were subjected to a group of 10 earthquake,

two-component motions, the results reported herein provide a more reliable test of inelastic

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1463

torsional response of multistorey frame buildings and consequently of the pertinent code pro-
visions for torsion, than the simpli
ed one-storey shear models used in the past. Nevertheless,
their generalization should be made with care and any implications or recommendations about
code provisions for torsion should be con
rmed with further studies, including a wider sam-
ple of non-symmetric buildings. They do, however, suggest that long standing conclusions
or code torsional provisions based on simpli
ed, one-storey models, need to be re-examined.
We must also note here that for the type of design analyses used in this work (spatial,
modal), the pertinent torsional provisions in the subsequent revisions of EC8 and UBC [1, 2]
have not changed and hence the conclusions of this study should be considered applicable
to the latest editions of these two codes. In summary then, the main conclusions are the
following:
(a) Eccentric frame-type buildings designed in accordance to EC8 and subjected to two-
component earthquake excitations do not experience similar levels of inelastic defor-
mation in all their members. Using the response of the associated symmetric building
as the basis for comparison, it was found that frames at the ‘exible’ sides experience
increased inelastic deformations and those at the ‘sti’ sides decreased deformations.
As a result, inelastic response measures, such as ductility factors and damage indices,
at the ‘exible’ side have reached values more than twice those at the ‘sti’ side.
Obviously, such uneven distributions are undesirable as they can lead to premature
member failures. A desirable design for torsion ought to lead to members in which the
design earthquake would generate ductility demands or damage indices not signi
cantly
dierent from those in the corresponding symmetric building.
(b) The
ndings herein should be an alert that the EC8 and UBC provisions for torsion for
eccentric, frame-type buildings, need re-examination. If the same
ndings are con
rmed
by additional studies covering a wider spectrum of cases, it will be necessary to modify
these provisions in a way that will strengthen the frames on the exible sides of
the eccentric building and thus reduce the corresponding ductility demands in strong
earthquakes.
(c) The above result is the opposite of what has been reported in the past, on the basis
of investigations using simpli
ed, one-storey, shear-beam-type systems. For several
important reasons, such systems are not adequate models of multistorey, frame-type
buildings responding in the inelastic range. A re-examination of the basic assumptions
under which the properties of such systems are determined and the way they should
be used for testing code provisions for torsion appears to be in order.
It is hoped that the results in the present paper will alert researchers about a real design
problem and also that they will raise a cautionary ag about extrapolating conclusions from
studies based on the simpli
ed, one-storey shear beam models to frame-type multistorey
buildings.

ACKNOWLEDGEMENTS
The authors would like to thank Prof. M. N. Fardis for making available to them his version of the
ANSR program, Dr T. B. Panagiotakos for his substantial contribution to its modi
cation and Ms C.
M. Alexopoulou for carrying out the push-over analyses.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
1464 K. G. STATHOPOULOS AND S. A. ANAGNOSTOPOULOS

REFERENCES
1. UBC-1997. Uniform Building Code. International Conference of Building O
cials, Whittier, California, U.S.A.,
1997 (Latest Edition: International Building Code-IBC-2000).
2. EC8 —Eurocode 8. Design provisions for earthquake resistance of structures. European Prestandard; CEN,
Doc.CEN/TC250/SC8/N, 1994 (Latest Edition: Eurocode 8: Design of structures for earthquake resistance,
European Standard EN1998-1:2004).
3. EAK2000. Greek Code for Earthquake Resistant Design. Greek Ministry of Environment, City Planning and
Public Works, Greece, 2000.
4. Tso WK, Sadek AW. Inelastic seismic response of simple eccentric structures. Earthquake Engineering and
Structural Dynamics 1985; 13(2):255 – 269.
5. Goel RK, Chopra AK. Inelastic seismic response of one-storey, asymmetric-plan systems. Report No.
UCB/EECR-90/14, Earthquake Engineering Research Center, 1990.
6. Rutenberg A. Nonlinear response of asymmetric building structures and seismic codes: A state of the art.
European Earthquake Engineering 1992; VI(2):3 –19.
7. Stathopoulos KG, Anagnostopoulos SA. Elastic and inelastic torsion in buildings. 11th European Conference
on Earthquake Engineering, 1998.
8. Bozorgnia Y, Tso WK. Inelastic earthquake response of asymmetric structures. Journal of Structural Division
(ASCE) 1986; 112(2):383 – 400.
9. Goel R. Seismic response of asymmetric systems: energy-based approach. Journal of Structural Engineering
(ASCE) 1997; 123(11):1444–1453.
10. Stathopoulos KG, Anagnostopoulos SA. Inelastic earthquake response of buildings subjected to torsion. 12th
World Conference on Earthquake Engineering, 2000 (id.0781).
11. Anagnostopoulos SA. Non-linear dynamic response and ductility requirements of building structures subjected to
earthquakes. Report No. R72-54, Civil Engineering Department, Massachusetts Institute of Technology, U.S.A.,
1972.
12. Anagnostopoulos SA, Roesset JM, Biggs JM. Non-linear dynamic analysis of buildings with torsional eects.
5th World Conference on Earthquake Engineering, Rome, Italy, 1973.
13. Tso WK, Zhu TJ. Design of torsionally unbalanced structural systems based on code provisions I: ductility
demand. Earthquake Engineering and Structural Dynamics 1992; 21:609 – 627.
14. Zhu TJ, Tso WK. Design of torsionally unbalanced structural systems based on code provisions II: strength
distribution. Earthquake Engineering and Structural Dynamics 1992; 21:629 – 644.
15. De La Llera JC, Chopra AK. Evaluation of code accidental-torsion provisions from buildings records. Journal
of the Structural Division (ASCE) 1994; 120(2):597– 616.
16. Chandler AM, Duan XN. Performance of asymmetric code-designed buildings for serviceability and ultimate
limit states. Earthquake Engineering and Structural Dynamics 1997; 26(7):717–735.
17. Chandler AM, Duan XN, Rutenberg A. Seismic torsional response: assumptions, controversies and research
progress. European Earthquake Engineering 1996; 10(1):37–51.
18. Rutenberg A. EAEE Task Group (TG) 8: behavior of irregular and complex structures—state of the art report:
seismic nonlinear response of code-designed asymmetric structures. 11th European Conference on Earthquake
Engineering, 1998.
19. Stathopoulos KG. Investigation of the inelastic response and earthquake resistant design of asymmetric buildings.
Ph.D. Dissertation, University of Patras, Greece, 2001 (in Greek).
20. Stathopoulos KG, Anagnostopoulos SA. Inelastic earthquake induced torsion in buildings: results from realistic
models. Proceedings; 12th European Conference on Earthquake Engineering, London, U.K., 2002; Paper No.
453.
21. Stathopoulos KG, Anagnostopoulos SA. Inelastic earthquake response of single-storey asymmetric buildings:
an assessment of simpli
ed shear-beam models. Earthquake Engineering and Structural Dynamics 2003; 32:
1813 –1831.
22. Stathopoulos KG, Anagnostopoulos SA. Earthquake induced inelastic torsion in asymmetric multistorey
buildings. 13th World Conference on Earthquake Engineering, Vancouver, Canada, 2004; Paper No. 558.
23. Tso WK, Smith RS. Re-evaluation of seismic torsional provisions. Earthquake Engineering and Structural
Dynamics 1999; 28:899 – 917.
24. Boroschek RL, Mahin SA. Investigation of coupled lateral-torsional response in multistorey buildings. 10th
World Conference on Earthquake Engineering, vol. 7, 1992; 3881–3886.
25. De Stefano M, Faella G, Realfonzo R. Seismic response of 3D RC frames: eects of plan irregularity.
In European Seismic Design Practice, Elnashai (eds), Rotterdam, Balkema, 1995; 219 – 226. (ISBN
90.54.10.588.7).
26. Ghersi A, Marino E, Rossi PP. Inelastic response of multi-storey asymmetric buildings. 12th World Conference
on Earthquake Engineering; No. 1716, 2000.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465
 INELASTIC TORSION OF MULTISTOREY BUILDINGS 1465

27. Fajfar P, Marusic D, Perus I. Inuence of ground motion intensity on the inelastic torsional response of
asymmetric buildings. 13th World Conference on Earthquake Engineering, Vancouver, Canada, 2004; Paper
No. 3496.
28. Panagiotakos TB, Stathopoulos KG. ANSRDYN—a program for non-linear static, modal and dynamic analysis
of three-dimensional structures. User’s manual. Civil Engineering Department, University of Patras, Patras,
Greece, 2000.
29. Otani A. Inelastic Analysis of R=C frame structures. Journal of the Structural Division (ASCE) 1974; 100(7):
1433 –1449.
30. Park YJ, Ang AMS. Mechanistic seismic damage model of reinforced concrete. Journal of the Structural
Division (ASCE) 1985; 111(4):722–739.
31. Karabalis DL, Cokkinides GJ, Rizos DC, Mulliken JS, Chen R. An interactive computer code for generation
of arti
cial earthquake records. In Computing in Civil Engineering (ASCE), Khozeimeh K (ed.), New York,
1994; 1122–1155.
32. Fardis MN. Damage measures and failure criteria for reinforced concrete members. 10th European Conference
on Earthquake Engineering 1995; 2:1377–1382.
33. Rutenberg A, Eisenberger M, Shohet G. Reducing seismic ductility demands in asymmetric shear buildings.
8th European Conference on Earthquake Engineering, vol. 3, 1986; 6.7/57– 64.
34. Chandler AM, Duan XN. Evaluation of factors inuencing the inelastic seismic performance of torsionally
asymmetric buildings. Earthquake Engineering and Structural Dynamics 1991; 20:87– 95.
35. Diaz-Molina I. Dynamic torsional behaviour of inelastic systems. M.Sc. Thesis, Civil Engineering Department,
Carnegie Melon University, Pittsburg, 1991.

Copyright ? 2005 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2005; 34:1449–1465