You are on page 1of 10

EngineeringStructures,Vol.

ELSEVIER

PIh S0141-0296(96)00142-3

19, No. 8, pp. 655-664, 1997


1997 Elsevier Science Ltd
All rights reserved. Printed in Great Britain
0141-0296/97 $17.00 + 0.00

A comparison of the present


SEAOC/UBC torsional provisions
with the old ones
M. De Stefano
Dipartimento di Analisi e Progettazione Strutturale, Universitd di Napoli Federico 11,
P. le Tecchio, 80125, Napoli, Italy

A. Rutenberg
Department of Civil Engineering, Technion-lsrael Institute of Technology, Haifa 32000,
Israel
(Received April 1996; revised version accepted October 1996)

The seismic response of single storey asymmetric structures modelled as two-degree-of-freedom elastic-perfectly plastic oscillators
and designed using the design eccentricity formula of the present
Uniform Building Code (UBC) torsional provisions, which first
appeared in the 1988 edition, is compared with designs based on
the earlier UBC 1979. Changes in the total strength requirement
and its distribution among the resisting elements are examined,
and the effects of these changes on the ductility demand of threeand five-element systems are evaluated. It is shown that for two
levels of the strength reduction factor R (= 4;1) the ductility
demands predicted for systems designed according to the UBC 79
provisions are in most cases similar to those for UBC 88 systems,
with the exception of torsionally flexible systems - mainly mass
eccentric and, to a much lesser extent, stiffness eccentric systems.
In view of the relative complexity of the UBC 88 provisions, a modification of the simpler UBC 79 ones is proposed which results in
the UBC 79 torsionally flexible systems behaving in a similar and
often better manner than their UBC 88 counterparts. 1997
Elsevier Science Ltd.

Keywords: earthquake resistant design, seismic codes, asymmetric


structures, nonlinear response

1.

visions have been given since 1988 in the UBC 2 and in


the NEHRP Recommended Provisions 3. In the following
sections, these provisions will be referred to as UBC 88.
Accounting for the effects of asymmetry in these provisions is a two-step procedure. In the first step uncertainties in the location of the horizontal forces are
accounted for by displacing their line of action from the
mass centre in each direction by an accidental eccentricity
assumed to be 5% of the building dimension b perpendicular to the direction of excitation, or

Introduction

The seismic provisions of the Structural Engineers Association of California (SEAOC) are usually adopted by the
Uniform Building Code (UBC). In addition to their adoption in many states in the USA, these regulations are taken
in many countries worldwide as a model for national codes.
Therefore, changes in the UBC are important to structural
engineers involved in seismic design and, of course, to code
writing bodies.
The torsional provisions under review first appeared in
the 1988 edition of the SEAOC Recommended lateral force
requirements and tentative commentary I. These were
mainly motivated by the need to reduce the vulnerability
of torsionally flexible asymmetric systems as manifested in
the 1985 Michoacan (Mexico) earthquake. Similar pro-

ed

e + O.05b

( 1)

where ed = design eccentricity, e -- eccentricity of mass


centre CM from the centre of rigidity Cn (Figure 1).
The second step consists of checking for torsional irregu-

655

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg

656

xible side

stiff side

e-~

e~

ai

4
(
Figure 1 System parameters and design eccentricities

larity. For this purpose, an accidental eccentricity amplifier


A~ is introduced, which is obtained from the formula
1 . 0 <--A ~ = (. : - ; . ......
2 ~3.0
~.
\l.ZO,,,,~/

(2)

where 6m,~ = maximum lateral displacement of the floor


at the level considered, 8,,,,~ = average of the horizontal
displacements at extreme positions of the floor at the same
level, i.e. 6,,,~ = (6, ..... + 6,,,~)/2, and 6,,,/,, = minimum lateral
floor displacement. Note that 8 ...... and 6,,~ are computed
based on the value of ej obtained from equation ( 1 ). When
A~ --< 1.0 (i.e. when 6,.... <-- 1.56,,,,), there is no torsional
irregularity, and equation (1) applies. When A~ > 1.0 torsional irregularity exists and
ea = e 0.05Axb

(3)

For resisting elements located at the flexible side of the


floor slab (Figure 1) the '+' sign always governs, whereas
for the stiff (or rigid) side elements
e~ = e - 0.05A~b G 0

or

e~ = e + 0.05A~b
(3a, b)

whichever gives the larger absolute force on the element


under consideration. Equation (3a) also states that forces
on the stiff side element should not be smaller than those
acting on a similar but torsionally balanced building, i.e.
when e = 0. The latter restriction applies also to stiff side
elements computed with equation (1) (Ax = 1.0). Since
many buildings are not sufficiently stiff torsionally, it is
often necessary to apply the above two-step procedure.
It is of interest to note that the formulation given above
to e5 (Equation (3a)) is not the only possible interpretation
of the actual wording of the current editions of the three
codes (SEAOC, UBC, NEHRP). In fact they all suggest
that equation (3a) could read
ed

e - 0.05A~b

(4)

without the upper bound < 0 . The common US interpretation, as given in equation (3a), is probably based on the
'General design requirements' of the UBC (Chapter 23,
1988 edn) in which the statement 'forces shall not be
decreased due to torsional effects' - deleted from the 1988
and subsequent editions of the SEAOC and the UBC torsional provisions - still persists. Results based on the alternative interpretation, as given in equation (4), can be found
in Reference 4.
The design eccentricity in earlier SEAOC codes 5 and
UBCs
6 was
based only on the first step, namely A~ = 1.0
always and no force reduction for the stiff side elements.
The rationale for increasing the strength of asymmetric
structures is given by the NEHRP Commentary 3 as based
on indications that the 5% accidental eccentricity may not
be adequate to protect the structure against torsional instability. The examples given there are the top storeys of nominally symmetric buildings, structures where calculation of
relative stiffness is problematic, e.g. masonry walls or vertical elements made of different materials and structures laterally supported by a central core. However, the present
provisions are also meant to be applied to asymmetric systems that are not so handicapped.
Several relevant studies have recently been published:
Chandler et al 7 have compared the response of systems
designed according to several seismic codes, including the
two interpretations of UBC 88, but not with the earlier edition. W o n g and Tso ~ have presented a comparison of UBC
88 with the older one for bidirectionally excited systems
having a natural period of 0.5 s. They concluded that the
provisions of UBC 88 (and later editions) are superior to
those of its 1979 counterpart, since they reduce the ductility
demand on the rigid-edge element of some torsionally flexible systems.
The purpose of the present paper is to compare the nonlinear seismic response of single storey asymmetric structures designed by the present provisions with those
designed by the old ones, and to show that the new code
provides only minor improvement over the old one, and
that it suffers from similar limitations: inadequate, albeit
often improved, control of ductility demand ( D D ) in torsionally flexible, particularly highly eccentric, systems. It
is shown that the difference in behaviour is only partly due
to the higher total strength allocated by the new code.
Based on the observation of system behaviour and, particularly, of element ductilities, a simple modification to the
UBC 79 torsional specification is proposed in order to obviate the need to compute the displacements with ea as per
equation (1) and to evaluate response with e,~ as per equation (3), as required by the UBC 88 procedure. With this
modification the UBC 79 torsional provisions guarantee the
same, or an even better, level of protection as the more
complex UBC 88 provisions.
2.

Structural

systems

Two classes of eccentric systems (Figure 2) were analysed:


stiffness eccentric models (Figure 2a) and mass eccentric
ones (Figure 2b). Systems were symmetric in mass, stiffness and strength along the x-axis. The stiffness eccentric
models are of the CM type, i.e. the central element is
located at the mass centre CM. The mass eccentric models
have a symmetric stiffness disposition with respect to the
centre of rigidity CR and eccentricity is introduced by shifting the mass centre by a distance e from C~. Therefore,

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg

(a)

Co)

-,4

ICR

3.1

STIFFNESSECCENTRICCM MODEL
Figure2

lY

t .... ~1-.... J~

Rt ..... I- ~3 5

[. . . . . . .

657

,Cu

I-

;. . . . . .

-.4

MASSECCENTRIC SME MODEL

Three- and five-element CM and SME models

this model is sometimes called the 'shifting' mass model


(SME).
The S M E model has been popular with investigators
since its stiff-edge element is located at a larger distance
from CR than in its C M model counterpart, and hence is
affected more by rotations. However, it tends to become
less realistic with increasing e, as elements 1 and 3 can no
longer be located at the edges of a rectangular slab with
uniformly distributed mass. To overcome this difficulty,
nonuniform storey mass distributions have been proposed,
e.g. a very large linear mass located along the flexible edge
of the slab s . However, these are not particularly realistic
in regular buildings. Therefore, although results from SME
models with relatively large e are investigated in the present
study to show behavioural trends, it should be realized that
they are less likely to be representative of realistic structures.
Mainly three- and also a limited number of five-element
models in each class were studied. Three-element models
represent systems in which the x-direction resistance is concent'rated along the axis of symmetry, whereas five-element
models represent systems having at least part of their xdirection resistance located at the periphery (Figure 1).
Evidently, these x-direction elements also contribute to the
torsional rigidity and resistance o f the system.
The force Fi,. acting at yield on the ith element along the
y-direction (Figure 1) is obtained from the well known linear static formula

[kiy

+ edaikiy]

F~.,.= Fv. LXKy- "ZKo J

(5)

where Fro = design base shear or yield strength of the associated symmetric system (i.e. with ed = 0), k~y= lateral stiffness of the ith element along the y-direction, "~Ky = total
y-direction stiffness, "ZKo = total rotational stiffness with
respect to CR and ag = perpendicular distance of the ith
element from Ck. It is convenient to normalize the dimensions with respect to the mass radius of gyration about CM,
p, So that the results are not dependent on the dimensions
of the floor slab. Letting a~ = a/p, ed = ed/p and 1) 2 =
]LKo(pZ~,Kv), where 12 = torsional-to-lateral frequency ratio,
equation (5) takes the form

F,.= F,.,, EK, ' 1 + 1221

(6)

When normal elements (No. 4 and 5 in Figure 2) are


present, the forces acting on them can be obtained from
the expression

kjx e~a;

Fix = +Fro XKy D,2

(7)

As noted, the strength of these elements is usually governed


by the x-direction base shear, which is another independent
parameter. To simplify matters it was assumed that ]~Fj, =
F~o, i,e. that the basic strengths in the two directions are
equal,
To make the comparison between five-element and threeelement models meaningful it is necessary to keep their
elastic properties equal. Note that, since part of the torsional
rigidity ~Ko of the five-element models is due to elements
4 and 5, the contribution of elements 1 - 3 to ZKo must be
adjusted accordingly; although the contribution to "~Ko of
x-direction elements 4 and 5 is an independent parameter,
it was arbitrarily set at 30%.

3.

Eccentricity overstrength ratio

The fact that e~ > e~ leads to "ZFiy > F:, i.e. the total
strength of structures designed on this basis is larger than
the total base shear. The ratio "ZFJF,.o is the eccentricity
overstrength ratio (OSR). This overstrength is code-dependent since e j varies from code to code and, as expected, it
usually increases with increasing e* (e* = e/p) and with
falling 12. The behaviour is described in Figure 3 which
gives OSR versus e* for three- and five-element CM and
SME models. Note that for five-element C M models only
curves for ~ = 1.0 and 1.25 are included, since, under the
design assumptions in this paper, such models cannot be
constructed when f~ = 0.8.
The overstrength at e* = 0 is of some interest. It is seen
that, due to accidental eccentricity, the strength of symmetric structures is larger than the code base shear. As can
be seen from Figure 3, such OSR is of the order of 10%
for three-element models and somewhat lower for fiveelement ones.
The different effects of the two codes on the total
strength of the models becomes immediately apparent. As
expected, the UBC 88 OSR is always larger than that of
UBC 79. CM models are designed with higher overstrength
than SME models, and systems with no orthogonal
elements possess larger strength than systems with orthogonal elements. It can be seen that the ratios ROV =
OSR(88)/OSR(79), showing the increase in strength
obtained by the UBC 88 torsional provisions with respect
to the UBC 79 ones, are relatively small, reaching 15% for
three-element SME models with large eccentricity and
= 0.8.
Figures 4 and 5 show the ROVs for each of the three
elements of three-element C M and SME models, respect-

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg

658
OSR

CM MODEL - 3 ELEMENTS

1.8

OSR[

SMEMODEL-3~

1ffi0.80 [

U B C 79
- - - - - U B C 88

LeO
1.25

1.8 l"

1.6

1.6

1.4

1.4[

1.0(
OSR

0.'2

o.~,

0.8

o:6

CM M O D I ~ - 5 ELEMENTS

1.8

UBC 79
- -- -- UBC

1.6

0.2

- -- -1.6

0.4

SMEMODEL-5FA.EblENTS

1.8I

88

0
OSR I[

O=LO0]
1.25 1~

LeO
1.25

~... - ~=~..--~

12L

1.21

[1=0"801

UBC 79
" - - - - U B C 88

0.6

0.8

Q=0.80 [

1.00 .[
1.25

UBC79
UBC88

1.4

1.2
C*

0/2

1.(

0;4

0~6

0.8

1.00

0.2

0.4

0.6

0.8

Figure 3 System overstrength factors for CM and SME models

ROV

f2= 0.80

1.3

ROV
'

1.00

]L"

f2= 0.80

OV

1.00

1.25

f2=

=t
.
.

1.2

1.1

1.0~ 3 CM- Element


0.9

3 CM- Element 2

[ [ 3 CM- Element 3
I /

0.'2

0.'4

0.'6

0.80

0.'2

0.4

0.'6

e*ll

0.80

0;2

0.~4

0;6

e*

0.8

Figure 4 Element overstrength factors for three-element CM models

1.8

ROV

t~= 0.80

ROV

f~= 0.80 1
1.00
1.25

1.00

1.00 l
1.25

1.6

ROV

Q = 0.8O |

1.25
3 SME -

Element 2

0.~

0.'4 o.6

3 SME - Element 3

1.4

1.2

0.'2 0.4

e*

0~6 o.so

Figure 5 Element overstrength factors for three-element SME models

e*

o.so

o~2 o.'4

o.6

e*

0.8

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg


ively. It can be seen that element 3 is always protected
more by the UBC 88 provisions than by the UBC 79 ones,
whereas the critical rigid element 1 is designed with larger
strength by UBC 88 mainly when II = 0.8. As II increases,
element 1 is provided with larger strength only for small
eccentricity; furthermore, in those cases, when fl = 1.0 and
1.25 the increase in strength provided by UBC 88 does not
reach 10%. The central element 2 of SME models does not
receive any increment in strength from the newer provisions.
It is well known that higher total strength is likely to
result in lower ductility demand and lower expected damage at least for low to medium period systems. Primafacie,
therefore, the new code is an improvement, but higher total
strength usually involves higher structural costs. This, however, is not a major problem in the present case in view of
the rather small differences in strength between the two
design procedures. Also, good inelastic performance
depends to a large extent on strength distribution among
the resisting elements, rather than on total strength alone.
To better understand behaviour, it may be useful to separate
these two effects. In earlier studies this has been effected
by normalizing the total strength to a common value, usually to Fro9 '~. To achieve this in the present study, the
strengths of the UBC 79 designs have in some cases also
been normalized to those of their UBC 88 counterparts, i.e.
element strengths have been factored by the ROVs given
in Figures 4 and 5. In this way the effect of relative costs
has been neutralized. However, the analysis essentially
focuses on the response of actual (i.e. non-normalized)
code designs.

4.

Dynamic parametric study

The study considered a relatively wide range of parameters:


three values for e* = 0.25, 0.50, 0.75, three values for fl
= 0.8, 1.0, 1.25 and uncoupled lateral vibration periods
ranging from /~,, = 0.10 s to T~ = 1.50 s at 0.10 s intervals.
Note that e* = 0.75 (i.e. e = 0.25 b, being the assumed
mass radius of gyration p equal to 0.333 b) is a very large
eccentricity. It was considered herein mainly to study
response trends, and also extreme cases, such as eccentric
cores. Five percent damping and elastic-perfectly plastic
force-displacement relation were assumed. As is known, a
simple bilinear relation does not realistically model
reinforced concrete behaviour. However, there is some evidence t2 that the ductility demand, as based on more faithful
degrading models, is similar. Two values of the force
reduction factor R were considered: R = 4, which is a typical value for yielding systems, and R = 1, representing the
design strength which is sufficient to keep the corresponding symmetric system elastic, but results in some inelastic
excursions in asymmetric systems. The latter can be
regarded as the serviceability limit state.
The four models: CM and SME models with three and
five elements each were unidirectionally excited by the time
histories of three representative earthquakes: Imperial Valley 1940 (El Centro record - comp. S00E) with ratio a / v
(a = peak ground acceleration in units of g, v = peak ground
velocity in m/s) very close to one and Northridge 1994
(Newhall record - comp. 34N/118W) with a / v = 0.62, representing earthquakes with energy in the medium to low
period range, and Mexico Michoacan 1985 (SCT record comp. N90W) with a / v = 0.28, representing an earthquake
having energy in the long period range. Note that, in order

659

to save computational time, only 50 s of the latter record


were considered in the analysis (from the 20th to the 70th
second). It will be observed that different spectral shapes
are usually associated with different ground motion intensities. However, for the purpose of comparison between
codes, the use of scaled records for the two levels of R
appears appropriate. Finally, as already noted, UBC 79
designs, strength normalized to the levels of their UBC 88
counterparts, were considered in addition to the actual
designs, (i.e. with different overstrengths).

5.

Results

Inelastic time history analyses were carried out in order to


compare the nonlinear seismic behaviour of three- and fiveelements single storey buildings designed by the static procedures of the two codes. The peak displacement ductility
demand ( P D D ) among the three elements oriented along
the seismic excitation direction was chosen as the main criterion. The P D D is not an ideal damage indicator, but it
is very easy to compute and, for the type of comparison
undertaken in the present study, appears to be quite
adequate. To isolate the effects of asymmetry the resulting
PDDs were normalized with respect to (i.e. divided by)
the DDsy m (ductility demands) of the reference symmetric
systems, i.e. single-degree-of-freedom (SDOF) elastic-plastic oscillators having a natural period equal to the
uncoupled lateral period of the asymmetric system and a
total strength of Fy o. The value of F,.o was defined by dividing the strength of the reference symmetric systems when
designed just to remain elastic through a reduction factor
R=4.
Some results are also given for R = 1, i.e. the strength
level of the corresponding symmetric system which is just
sufficient to keep it elastic. This strength level may be
regarded as the serviceability limit state. However, R = 1
does not mean that P D D = 1.0 also for asymmetric systems,
and the purpose herein is to compare the plastic excursions
of the two code-designs. Bounds on lateral displacement
and interstorey drift are usually imposed by codes. Therefore, some comparisons of peak displacements are also
presented.

6.

Three-element systems (no normal elements)

The results of Figures 6 - 8 compare the normalized peak


ductility demand R D D = P D D / D D w , of UBC 88, UBC 79
1.4 -RDD
1.2
1.0 ~ * = 0 . 2 5
0.8
0.6
"
1.4

3 Element C M m o d e l - f ~ = 1 . 0 0

EL CENTRO

1.2 ie*-Q,50

1.0
0.8
0.6

1.4
1.2
1.0

0.8
0.6
0.4
0.2 0

"'-

~
U B C 79
. . . . . . . . . . . . U B C 79/88
- - - - " 7 U B C 88

~ g -..Q,.~5 . . . . . . . . . . . . . . . . . . . . . . . .

"--

e
c)
0.'2

0;4

0.'6

0~8

1~0

1.'2

1.'4

1.6

Figure 6 Effect o f n o r m a l i z e d eccentricity e* on RDD f o r threee l e m e n t CM models, El Centro record (R = 4)

660

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg


3 Element CM model - f~= 1.00

1.4 -RDD
1.2
1.0
0.8

1.4

1.2

e * = 0.25

o6O
'

NEWHALL
1.2
1.0
0.8
0.6

EL CENTRO /

0.60"8
.

1.0
0.8
0.6
0.4

" ~

............

- ....

88

T (sec)
0.2

0.4

0.6

0.8

1.0

1.2

Figure 7 Effect o f n o r m a l i z e d eccentricity e* on


e l e m e n t CM m o d e l s , N e w h a l l record (R = 4)

1.4 -RDD

1.4

RDD f o r

1.6
three-

3 Element CM model - fl= 1.00 / ' ~

0.8
0.6

~.~

"
UBC 79
............ UBC 79/88
- .....
UBC 88
.
-~'~"~

0.40

0.'2

0.'4

0.'6

0.'8

1.'0

0.5

........ J

~.-

""-----

0.6 .
|

'o f

..........A

..........................
: ............................

Effect of normalized

~2.0
eccentricity

2.'5
e* on

UBC 88

RDD f o r

three-

e l e m e n t CM m o d e l s , M e x i c o record (R = 4)

"

"

NEWHALL
-,

.............A

......................

o.,[ / "
0d

. . . . . . . . . . . . . .

0.'2

0.'4

r(sec)

0 '6

0 '8

1.'0

1 '2

1 '4

Figure 10 Effect o f n o r m a l i z e d eccentricity e* on


t h r e e - e l e m e n t SME m o d e l s , N e w h a l l record (R = 4)

1.2
1.0 .
0.8
0.6

l:o ......................... ~

for

,~~

............ ~

0.8 C "-._':~z~z~z~z~z~z~ ~
~
0"6 ~ .... ~ ~ . . ~ . . . .
16 ~

11:4 e*=0.75
0.6

RDD

MEXICO

4ii~ -i f ~ = 0 . 5 01

1.0

16

3 Element SME model - f~= 1.00 h

1.6 -RDD

and UBC 79 strength normalized to that of UBC 88 (or


UBC 79/88) systems for a range of uncoupled natural periods 7",.. When using the Mexico record as input ground
motion, whose dominant period T~ is about 2.0 s, the range
of T,. has been extended up to 3.0 s in order to cover values
of Z~./T~ significantly larger than one even in this case. The
comparisons are given for three-element CM models with
torsional-to-lateral frequency ratio ~ = 1.0, having small,
intermediate and large eccentricity, namely: e* = e/p =
0.25, 0.50, 0.75 with R = 4. It can be seen that for the three
selected records the differences between UBC 88 and UBC
79/88 designs are very small, and in many cases the UBC
79/88 R D D s are lower than those of UBC 88. The nonnormalized UBC 79 results are usually only marginally
larger than the other two. Similar results are predicted in
Figures 9 - 1 1 for the three-element SME models. Here, in
many cases the actual UBC 79 and the UBC 79/88 designs
lead to lower D D s than those of UBC 88. The exceptions
are at the lower period range (T~. < 0.5 s) at large, and for
El Centre also intermediate, eccentricities. Note, however,
that SME models with large e* are not particularly realistic.
It can also be seen that neither of the two codes is able to
control effectively the increase in ductility demands due to
asymmetry for long period SME systems under the
Northridge Newhall record. The response of the two
designs to the Newhall record in the long period range is
disturbing: inordinate amplifications of peak ductility
demands with respect to their symmetric counterparts even
for systems with large torsional rigidity (f~ --> 1.0) and

/'

UBC79 1
............ UBC79/88/

...............~

~.5 ~ e*= 0.75

. . . . . . . . . . . . . . . . . . . . .

T (sec)
3.'0
3.5

]
"

/88
......

0.4 ~ -**"~'z::'z"a'~
"~U
0.2;
015
1;0
115
Figure 8

1.5
1.0
0.5

three-

/-~

"': .....................

RDD f o r

3 Element SME model - f~= 1.00

MEXICO
7......................................................... .c':...................

T (sec)
1.'4
1.6

1.'2

Figure 9 Effect o f n o r m a l i z e d eccentricity e* on


e l e m e n t SME m o d e l s , El C e n t r e record (R = 4)

2.0.2"5
0 " 2:-e*=
51.5

0.8

~t

1.0 !

1.0
0.8

f'x
t~'~_~

3.0 :RDD

1.6

1.4
1.2 "--e*= 0.75

e* = 0.75

0.20

UBC 79/88

2~tmc

A I~
* ..................
0.50
~ , - - ,.. ~,. . . . . . . . . . . . . . . . ~ . ~,. .-. ......
, ........................... ~,..............

1.0 e
- -

~,~

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.4
1.2

"-" -"

3 Element SME model - f~= 1.00

"RDD

-~-

~v

,-'',~

............... V ................

- - "
UBC 79 t
............ UBC 79/881
......
UBC88
El

",,

""

...................................................................

""

Figure 11

(sec)
0.5

Effect

1.0

of

1.5

normalized

2.0
eccentricity

2.5

3.0
e*

on

3.5

RDD

for

t h r e e - e l e m e n t SME m o d e l s , M e x i c o record (R = 4)

small to moderate eccentricities (e* = 0.25 and 0.50) are


predicted.
The effect of torsional-to-lateral frequency ratio II on the
R D D of three-element systems excited by the three considered records for an intermediate eccentricity e* = 0.50
(corresponding to e = 0.167 b) has also been evaluated.
Figures 12 and 13 show R D D s obtained for CM and SME
models under the E1 Centre record. As expected, R D D
tends to fall with increasing [~, with values which tend to
oscillate about unity. It can be seen that only for torsionally
flexible systems, i.e. [~ = 0.8, UBC 88 designs consistently
lead to better inelastic performance than UBC 79 ones,
especially for SME models.

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg


2.0
1.5

-_RDD

t~= 0 . 8 0 ~ ~

Element C M
,s

model -

1.4
1.2
1.0
0.8
0.6
0.4

e*= 0 . 5 0

1.0
0.5
EL CENTRO

1.00

~=

1.5
1.0
0.5

UBC 79

0
f~= 1.25

1.5

.......

U B C 88

1.0
0.5

.T(sec)

06

0.'2

0.'4

0.'6

0~8

l.'0

172

1.4

2.0 .
1.5

RD~.50
-

1.0

~-~......

1.0

.........

............

1.2
1.0
0.8
0.6
0.4

e . ~ : . = . . 0 . 7 5 ............................................................................

~
0

e
0.2

0.4

0.6

:-7

~ - ~ . ~ . ~

f~= 1.25

........

- -

UBC 79

.......

U B C 88

1.0

1.2

1.4

...........................

0.5

,T(sec)
0.'2
0.2

0.'4

0.'6

0.'8

170

1.'2

1.4

5 E l e m e n t S/VIE m o d e l - f~= 1.00


...........................

EL CENTRO

1.o
0.8
0.6
0.4 .
0.8
0.6
0.4

1.6

Figure 13 Effect of torsional-to-lateral f r e q u e n c y f l ratio on


RDD f o r t h r e e - e l e m e n t SME models, El Centre record (R = 4)

Five-element systems (with normal elements)

The study of five-element models under undirectional excitation is controversial. It has been suggested t3 that such
models should realistically be studied under biaxial excitation, when intense x-direction (Figures 1 and 2) ground
motion leads to frequent yielding of elements 4 and 5,
which, as a result, offer little resistance to y-direction
motion induced torques. However, these elements contribute to the resisting torque when the x-direction excitation
is not severe. They also contribute to the torque when the
x-direction seismic forces are resisted by a stiff and strong
central x-direction element (not shown in Figure 2). Therefore, in this section some results for systems with orthogonal elements under unidirectional excitation are presented.
Figures 14 and 15 show the variation in RDD with the
normalized eccentricity e* for five-element CM and SME
models under the E1 Centre record. The results fully confirm those obtained for the corresponding three-element
systems. In particular, when ~ = 1.0 or is larger (not
shown), UBC 79 and the newer UBC 88 lead to quite similar inelastic performance. It can also be seen that in many
cases, the UBC 79/88 curves present the lowest values, thus
demonstrating implicitly that the UBC 88 strength distribution is not superior to that of UBC 79 which has a clearly
simpler formulation and requires less computational effort.
In order to shed further light on the relative efficiency
of the strength allocation among the three y-direction
elements, element behaviour is examined in the following
section.

1.6

Figure 14 Five-element CM model: effect of normalized e c c e n tricity e* on RDD (R = 4)

1.0
.......

.......

)
0.8

1.2 . e * = 0 . 5 0
.

UBC 79
UBC 79/88

::.-::._.~7..8.8

...................................................................

1.5

7.

EL CENTRO

0.4

EL CENTRO

06

..............................

0.8
0,6

~ = 1.00

0.5

lo

m o d e l - t~= 1.00

1.2 : e * = 0.25
1.0 .......~

0.5
1.5

Element C M

1.4 - R D D

1.2
1.0
0.8
0.6
0.4

1.6

Figure 12 Effect of torsional-to-lateral f r e q u e n c y .Q ratio on


RDD f o r t h r e e - e l e m e n t CM models, El Centre record (R = 4)

RDD

661

~4,

) . ~ . . ~ ........

~,s
LrBC 79/88,

:. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

e
0.2

0.4

....

c
0.6

0.8

1.0

Figure 15 Five-element SME model:


eccentricity e* on RDD (R = 4)

8.

1.2

effect

s"
1.4

of

)
1,6

normalized

Element ductility demands

Curves of the element ductility demand EnD to DD~>m


ratios (REND) are given in Figures 16 and 17 for threeelement C M and S M E models with e* = 0.75 excited by
the E1 Centre record. Results for the strength normalized
UBC 79 designs have been omitted in order to highlight
the different behaviour of each of the three elements due
to use of the actual code specifications. It is evident that
the band of the results, showing the nonuniform distribution
of plastic demands on the resisting elements, becomes
increasingly narrow as the system torsional stiffness
increases.

RDD

1.4
3 Element C M m o d e l - e*= 0.75 ,,
1.2
f
l
=
~
1.0 0.8
0.6
0.4
0.2
0
Elem. 1
UBC 79 . . . . . . . . . . Elem.2
1.2 :ztl= 1.00
.
.
.
.
.
Elem. 3
1.0
0.8
0.6
0.4
0.2
0
Elem. 1
UBC 88 . . . . . . . . . . Elem. 2
1.2 i ~ = 1.25

.
.
.
.
Elem.
3
1.0
0.8
~ ~a,..
....
"
=.-" ~
~=- "~ =':.~.==--~-~x~
0.6
0.4
0.2
00
0..2
0..4
0.'6
0:8
1.'0
1~
1..4

T(sec)

1.6

Figure 16 Element ductility d e m a n d s ratios (REDD) o f threee l e m e n t CM systems (R = 4)

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg

662
2.5
2.0
1.5
1.0
0.5
0
2.0
1.5
1.0
0.5
0
2.0
1.5
1.0
0.5

R D D

3 Element SME model -

2.5

e*= 0.75

PDD

3 E l e m e n t C M m o d e l - e*= 0 . 5 0

2.0

Elem. 1
. . . . . . . . . . E.lem. 2
.....
Elem. 3

UBC 79
f~= 1.00

1.5

1.0

Elem. 1
. . . . . . . . . . Elem. 2
.....
Elem. 3

UBC 88
f i = 1.25

0.5
UBC79

00

0.2

0.4

0.6

0.8

1.0

Figure 17 E l e m e n t d u c t i l i t y d e m a n d s
e l e m e n t SME s y s t e m s (R = 4)

1.2

1.4

1.6

r a t i o s (REDD) o f t h r e e -

For torsionally flexible systems ( ~ = 0.8), which behave


in the worst manner, neither of the two codes is able to
optimize the strength allocation among the three resisting
elements. In particular, the stiff side element 1 is allotted
insufficient strength in CM models; the situation, however,
appears to be less critical for SME models designed according to UBC 88.
As the torsional stiffness increases, i.e. f~ --> 1.0, the
maximum REDDs shift to element 2 and to element 3.
Therefore, the usual assumption that element 2 is not a critical one is questionable for torsionally stiff systems,
especially if they are mass eccentric. Note, however, that
when elements 2 and 3 have the maximum values of ductility demands, the corresponding REDDs are in most cases
lower than one, meaning that, due to the code overstrength,
they still behave better than their symmetric counterparts.
On the other hand, when element 1 experiences the largest
yielding, the corresponding REDDs are well above unity,
thus indicating a really problematic situation.
The above observations confirm, from a qualitative point
of view, that adopting the UBC 88 provisions does not
result in a major improvement in the inelastic behaviour of
torsionally flexible systems, whereas they may even lead
to an increase in the inelastic demands of systems having
larger torsional stiffness.
9.

Peak displacements

Comparisons of the peak displacements RPKD, nondimensionalized to those of the corresponding symmetric systems, for the R = 4 level are given in Figure 18 for three-

UBC88

f~= 1.25

0.'2

'
0.4

fl= 0.80
. . . . . . . . . . fl= 1.00
.....

fl= 1.25 T(sec)

'
0.6

'
0.8

'
1.0

I .'2

1.4

1.6

Figure 19 E f f e c t o f t o r s i o n a l - t o - l a t e r a l f r e q u e n c y r a t i o ,O, o n
PDD f o r t h r e e - e l e m e n t CM m o d e l s a t s e r v i c e a b i l i t y l i m i t state
(R = 1)

element CM models excited by the El Centre record. As


expected, the different planwise strength distributions provided by the two codes have practically no effect on this
response parameter, except in the case of very short period systems.
10.

Quasi-elastic response: R = 1

Figures 19 and 20 compare the peak ductility demands


(PDDs) for three-element CM and SME models designed
using the two editions of UBC, with the strength level
defined through R = 1. The influence of the uncoupled torsional-to-lateral frequency ratio [ l is evaluated for systems
having an intermediate normalized eccentricity (e* = 0.50)
under the E1 Centre record. Due to asymmetry, the strength
level corresponding to R = 1 is not sufficient, in many
cases, to maintain both CM and SME systems in the elastic
range. This appears to be a main shortcoming of the two
codes, since both UBC 88 and UBC 79 provide significant
overstrength (more than 50% for systems with 12 = 0.8
designed to UBC 88).
When considering CM models, it appears that the two
versions of the UBC lead to very similar PDD values at
the serviceability limit state. In particular, the same performance (in fact, the curves for the two codes coincide)
is predicted for torsionally flexible systems, 12 = 0.8, for
which, when R = 4, the UBC 88 designs were shown to be
superior to the UBC 79 counterparts. This is due to the fact
that, when e* = 0.50, both codes allocate the same strength
3.0

PDD
3 Element C M m o d e l - ~'l= 1.00

00

3.0

RPKD

~ = 0.80
. . . . . . . . . . ~ = 1.00
.....

c)

3 Element S M E m o d e l -

e*= 0 . 5 0

2.5

2.0

,, (," \ I~

2.0

~,.s

xl

.-

.&

1.5

1.5
1.0

0.5 - UBC 79

e*= 0.25
. . . . . . . . . . e*= 0.50
.....

UBC 88

e*= 0.75

.....

0.2

014

0;6

018

110

1.4

1.6

Figure 18 Peak d i s p l a c e m e n t s ratios (RPKD) f o r UBC 79 and


UBC 88 t h r e e - e l e m e n t CM s y s t e m s (R = 4)

--a=
UBC 79

e*= 0.75
,T (sec)
112

......

0.5

e*= 0.25
. . . . . . . . . . e*= 0.50

......

0.2

0.80

. . . . . . . . . . t~= 1.00
.....

0.4

a = 1.~25
0.6

n--- 0.80
UBC g8
i

0.8

. . . . . . . . . . fl= 1.00
. .i . . .

O= 1.25 ,T (sec)

1.2

1.4

1.6

F i g u r e 2 0 E f f e c t o f t o r s i o n a l - t o - l a t e r a l f r e q u e n c y r a t i o ~Q o n
PDD f o r t h r e e - e l e m e n t S M E m o d e l s a t s e r v i c e a b i l i t y limit state
(R = 1)

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg


to the stiff edge element 1, which experiences the largest
ductility demands.
When considering SME models, both codes lead to similar PDD values for systems having 1) --> 1.0. For torsionally
flexible systems, the UBC 88 specifications appear to be
more adequate than the UBC 79 ones; this is essentially
due to the larger strength with which UBC 88 provides
element 1. Values of the peak ductility demands for SME
models are larger than those for CM models, in many cases
exceeding 2 for torsionally flexible systems designed
according to UBC 79.
For systems having large eccentricity (e* = 0.75) UBC
88 designs offer some advantage relative to UBC 79 in the
low period range, even when 1~ --> 1.0. However, due to
their larger overstrength, the PDDs are in most cases lower
than or very close to one, and the difference between the
two codes does not exceed 10%. Also, when UBC 79 systems are provided with the same total strength as their UBC
88 counterparts, the differences between the two codes
practically disappear (not shown).
11. A proposal for improving UBC 79 torsional
provisions
From the results presented in the preceding sections it is
evident that the seismic provisions of UBC 1988 or later
editions do not constitute a significant improvement over
the older ones, i.e. UBC 79 and earlier provisions, which
are much easier to implement. This is because, as noted, the
UBC 88 procedure requires that an additional displacement
calculation be made in order to evaluate the amplifier Ax.
The provisions of both codes are in need of improvement.
In this section a proposal for improving the UBC 79
specifications in a simple manner is made. It has been found
that UBC 79 specifications are inadequate for torsionally
flexible systems, i.e. when [l < 1.0. As can be seen from
Figures 4 and 5, the stiff edge element (element 1 ) is in
need of additional strength. One simple way of achieving
this is by dividing the strength of element 1, (as per the
UBC 79), through 1~, when 1) < 1.0 (only).
The increase in the total overstrength due to the above
modification is small. For example, for three-element CM
models with ~ = 0.8 and e* = 0.50, the UBC 79 overstrength
factor increases from 1.37 to 1.46, still being lower than that
of UBC 88 (= 1.52). Figures 21a and 21b compare RDDs
and PDDs resulting from the proposed modified UBC 79
provisions with those of the actual UBC 79 and UBC 88
2.0

RDD

R= 4

(a)

663

provisions for R = 4 and for R = 1. It can be seen that the


modification of UBC 79 appreciably improves the response:
for R = 4 the corresponding RDD curves predict in most
cases the lowest values; for R = t these systems have the
lowest PDDs over the whole period range considered.
For three-element SME models, the proposed modification of UBC 79 still results in a significant improvement
in behaviour. Namely, for 12 = 0.8 and e* = 0.50, the UBC
79 OSR increases from 1.13 to 1.25, whereas the UBC 88
OSR is equal to 1.18. As a consequence, the response of
systems designed with the modified UBC 79 is characterized by RDDs and PDDs (Figures 22a and 22b,
respectively) which are similar to and in most cases lower
than those of UBC 88 designs. Note, however, that in this
case the actual UBC 79 designs have poorer performance
than the UBC 88 ones.
The improvement in response of torsionally flexible systems is obtained over the full range of realistic normalized
eccentricities. As an example, for CM models with fl = 0.8
and e* = 0.25, designed with R = 4, it was found that the
modified UBC 79 designs present in many cases the lowest
RDDs (not shown). For SME models it was also found that
the proposed modification makes the UBC 79 designs as
good as the UBC 88 ones.

12.

Conclusions

In this paper, stiffness eccentric and mass eccentric systems


designed according to the current UBC torsional provisions
(UBC 88) and to the previous ones (UBC 79) have been
studied in order to assess their relative performances.
The results presented in the preceding sections show, for
realistic designs at strength levels corresponding to R = 4,
that the ductility demands predicted for UBC 79 designs
are in most cases similar to those for UBC 88 designs, with
the exception of torsionally flexible systems. For the latter
systems, the present torsional specifications lead to a significant reduction in the peak ductility demands compared
with the UBC 79 provisions, mainly for SME models which
become less realistic with increasing eccentricity. However,
when the UBC 79 strength is increased to the UBC 88 level,
the difference between the two codes diminishes. Note also
that neither code can adequately control the PDD of torsionally flexible systems with large eccentricities. Furthermore, the response of the two designs to the Northridge
Newhall record is disturbing, since it is characterized by
inordinate amplifications of the peak ductility demand rela2.0
1.5

1.0

0.5

~'

1.0

. . . . . . . .

0
20 . P D D

"

0.5

" ~ " "

3 Element C M model
R= 1 ,~

"

(b),

1
)11

0.5

1~6

............................

0.5 ['~"
i

00

UBC 79
.

0.'2

Modified UBC 79
.

0.'4

0.'6

0~8

UBC 88
.

T(

1.'0

1~2

1.'4

see

2.0

Figure21 Effect o f p r o p o s e d m o d i f i c a t i o n to UBC 79 on peak


d u c t i l i t y d e m a n d s o f CM m o d e l s w i t h D. = 0.8 and e* = 0.50: (a)
R = 4, (RDD); (b) R = 1, (PDD)

1.0

DD~D

--

3 Element S M E model

R=

1 / / / ~ (b)

............................................................................................................................................................

UBC 79
0.'2

Modified UBC 79

............

0.'4

0.'6

0~8

UBC 88
1~0

,T (sec)
1~2

1.4

1.6

Figure22 Effect of p r o p o s e d m o d i f i c a t i o n to UBC 79 o n p e a k


ductility d e m a n d s of S M E m o d e l s with fl = 0.8 a n d e* = 0.50:
(a) R = 4, (RDD); (b) R = 1, (PDD)

664

Seismic torsional provisions in UBCs: M. De Stefano and A. Rutenberg

tive to their symmetric counterparts. The particular features


of the input leading to this behaviour should be explored.
Systems having a strength level corresponding to R =
1 were analysed in order to evaluate the response at the
serviceability limit state. It has been found that they may
sustain relatively large inelastic excursions, even when they
possess significant overstrength compared with their symmetric counterparts. UBC 79 and UBC 88 designs lead to
practically equal PDDs for systems having ~Q -> 1.0. For
torsionally flexible systems, it has been found that the UBC
88 leads to better performance than UBC 79 for mass
eccentric (SME) models only.
In view of the above, it is evident that the seismic provisions of UBC 88 or later editions thereof do not constitute
a significant improvement over the older ones, e.g. UBC
79, which are much easier to implement. Only torsionally
flexible systems benefit from the newer UBC edition, and
this, of course, is very important. However, it has been
shown that a significant improvement in performance of
UBC 79 torsionally flexible designs can be achieved by
dividing the strength of the stiff edge element 1 by {L when
~Q is lower than unity. In this manner, torsionally flexible
systems designed with the simpler UBC 79 procedure
behave in a similar and often in a better way than their
UBC 88 counterparts. However, further studies are needed
to evaluate fully the benefits achieved with this minor
modification to UBC 79.
This study was mainly concerned with the relative merits
of the two codes. As is evident from the R = 1 results, as
well as for the R = 4 ones (not shown) the peak ductility
demand may be larger than the target ductility, irrespective
of the code chosen. This problem, however, is beyond the
scope of the present study.
Finally, the adequacy of using unidirectional ground
motions for code calibration may be questioned. However,
it has been shown H that analysis under unidirectional input
of three-element systems provides a good estimate of
response under bidirectional excitation for systems having
resisting elements in two orthogonal directions. Moreover,
a very recent large parametric study ~5 of systems under
bidirectional excitation has confirmed the conclusions about
the adequacy of several seismic codes that were previously
drawn from unidirectional analysis.

Acknowledgments
The authors are thankful to the reviewers for their useful
comments and suggestions. The financial support provided

by the Italian CNR, the Israeli NCRD and the National


Building Research Institute at the Technion, Haifa is gratefully acknowledged.

References
1 Recommended lateral force requirements and commentary, Seismology Committee, Structural Engineers Association of California,
SEAOC, Sacramento, CA, 1988
2 Uniform Building Code, ICBO, International Conference of Building
Officials, Whittier, CA, 1988
3 NEHRP Recommended Provisions Jot the Development of Seismic
Regulations for New Buildings 1994 edn, Federal Emergency Management Agency, Part 1 Provisions FEMA 222A, and Part 2
Commentary FEMA 223A, 1995
4 De Stefano, M. and Rutenberg, A. "The present SEAOC/UBC torsional provisions: are they better than the old ones?', Report, National
Building Research Institute, Technion, Haifa, Israel, February 1995
5 Recommended lateral force requirements and commentary, Seismology Committee, Structural Engineers Association of California,
SEAOC, Sacramento, CA, 1974
6 Uniform Building Code, International Conli~rence of Building
Officials, Whittier, CA, 1976
7 Chandler, A. M., Correnza, J. C. and Hutchinson, G. L. 'Perioddependent effects in seismic torsional response of code sysems', J.
Struct. Engng, ASCE 1994, 120, 3418-3434
8 Wong, C. M. and Tso, W. K. 'Evaluation of seismic torsional provisions in Uniform Building Code', J Struct. Engng, ASCE 1995,
121, 1436 1442
9 Rutenberg, A., Benbenishti, A. and Pekau. O. 'Nonlinear seismic
behavior of code-designed eccentric systems', Proe. lOth World
Cov![. on Earthquake Engineering, Madrid, 1992. vol 10, pp 57515756
10 Rutenberg, A., Eisenberger, M. and Shohet, G. qnelastic seismic
response of code-designed single storey asymmetric structures',
Engng Struet. 1992. 14(2), 91 102
11 De Stel~_no, M. Faella, G. and Ramasco, R. 'Inelastic response and
design criteria of plan-wise asymmetric systems', Earthquake Engng
Struct. Dyn. 1993, 22, 245-259
12 Tso. W. K. and Bozorgnia. Y. "Effective eccentricity for inelastic
seismic response of buildings', Earthquake Engng Struet. Dyn. 1986,
14, 413-427
13 De la Llera, J. C. and Chopra, A. K. "Understanding the inelastic
behaviour of asymmetric-plan buildings', Earthquake Engng Struet.
Dyn. 1995, 24, 549 572
14 Correnza, J. C., Hutchinson, G. L. and Chandler, A. M. 'Effect of
transverse load-resisting elements on inelastic earthquake response of
eccentric-plan buildings', Earthquake Engng Struct. lkyn. 1994, 23,
75-89
15 De Stct~tno, M., Faella, G. and Ramasco, R. 'Inelastic seismic
response of plan-asymmetric systems under directional ground
motions', Proc. European Workshop on Seismic Behaviour ~f'Asymmetric and Setback Structures, Anacapri, Italy, 1996 (to appear)

You might also like