Sie sind auf Seite 1von 11

-

8.

A Replaeement for the 30%, 40%, and SRSS Rules


for Multicomponent Seismic Analysis
Charles Menun and Armen Der Kiureghian, M.EERI

A response spectrum rule for combining the contributions from three orthogo
nal components of ground motion to the maximum value of a response quantity is
presented. This rule, denoted CQC3, is compared to the 30% and 40% rules and
the squarerootofsumofsquares (SRSS) rule currently specified in many design
codes. It is shown that these current rules are special cases of the CQC3 rule,
when certain conditions regarding the nature of the ground motion or the struc
tural response are satisfied. Because these conditions are not always satisfied, it is
argued that the CQC3 rule should be adopted as a general rule for the multicom
ponent combination problem. The CQC3 rule additionally offers a simple formula
for determining the most critica) orientation of the ground motion components for
each response quantity of interest. The CQC3 rule is computationally simple and
easy to implement in standard dynamic analysis codes.
INTRODUCTION
In the design and analysis of structures subjected to seismic loading, multicomponent
ground motions should be considered. In general, the translational ground motion can be re
solved into three components, two in the horizontal plane and one in the vertical direction.
Rotational ground motions are usually neglected. Penzien and Watabe (1975) observed that
during the strong motion phase of an earthquake the ground motion components are uncor
related along a welldefined orthogonal system of axes whose orientation remains reasonably
stable over time. This system of axes, denoted the principal axes of the ground motion, are
oriented such that the major principal axis is horizontal and directed toward the epicenter, the
intermediate axis is in the transverse horizontal direction, and the minor principal axis is ver
tical. The components along any other system of axes obtained by rotation of the principal
axes are naturally correlated.
The rnagnitude of a response quantity due to each component of the ground motion is
commonly evaluated by means of the response spectrum method. In this method, individual
modal responses are obtained using the specified response spectrum for the component and
are combined to compute the response. The rules for combining the modal responses due to a
single component of ground motion (e.g., the CQC rule) and their accuracy are well
documented (Der Kiureghian, 1981; Wilson et al., 1981 ). In the case of multicomponent ex
(CM, ADK) Dept. ofCivil & Environmental Eng., University ofCalifomia at Berkeley, 94720
153
Earthquake Spectra, Volume 14, No. 1, February 1998

154

C. Menun and A Der Kiureghian

citation, the contributions from the individual ground motion components must somehow be
combined to compute the total response. The procedures specified in the current design codes
for this purpose do not have sound bases, as they do not properly account for the correlation
between the ground motion components mentioned above. As a result, there has been an in
terest to either establish the extent of the validity of these procedures or provide improved
combination rules.
The objective of this paper is to examine the shortcomings of the current combination
rules for response to multicomponent excitation and to propose a replacement that has a
sound theoretical basis. The proposed rule was originally derived by Smeby and Der Kiureghian (1985) and is an extension of the CQC rule commonly used to combine the modal
responses resulting from a single component of ground motion. For this reason, the rule is
named "CQC3." This rule properly accounts for the correlation between the components of
seismic input in accordance with the PenzienWatabe idealization. The rule can be easily im
plemented in any standard linear dynamic analysis program, as the required terms are rou
tinely calculated by such programs. It is shown that the rules in current use are in fact special
cases of CQC3, when certain assumptions are made with regard to the nature of the ground
motion and the structural response. These assumptions however are not always valid. Hence,
it is argued that CQC3 should be adopted as a general rule for the multicomponent excitation
problem.
In practice, the principal axes of the design earthquake may not be known. In that case, it
is prudent to design for the worst case, i.e., for that orientation of the principal axes that
yields the maximum response. With the current rules, such an approach requires a tedious
numerical procedure involving a series of dynamic analyses with varying orientation of the
principal axes. The CQC3 rule provides a simple formula for determining the critica) angle of
the ground motion and the corresponding maximum response.
To date, examples ofthe application ofthe CQC3 rule (Smeby and Der Kiureghian, 1985;
Lopez and Torres, 1996) have focussed on buildingtype structures with rectangular geome
try. To the authors' knowledge, more complex threedimensional structures have not been
studied. As a numerical example in this paper, a curved bridge is considered. This example
serves to illustrate the advantages of the CQC3 rule and to evaluate the errors incurred by the
combination rules currently specified by design codes.
CURRENT MUL TICOMPONENT COMBINATION RULES
Most design codes that address the multicomponent ground motion problem specify that
the contributions to a response quantity from the orthogonal components of seismic input be
combined either by the squarerootofsumofsquares (SRSS) rule or by a percentage rule.
The 1997 edition of the Uniform Building Code (UBC 97) requires the use of either the
SRSS rule or the "30%" rule, but only for structures having certain types of irregularities
(ICBO, 1997). The current California Department of Transportation (Caltrans) bridge design
specifications require the 30% rule for all structures (Caltrans, 1990). The Caltrans code does
not specify the SRSS rule as an altemative. The ASCE Standard for the Seismic Analysis of
SafetyRelated Nuclear Structures (ASCE 486) specifies the SRSS rule or the "40%" rule
(ASCE, 1986). Finally, the Applied Technology Council (ATC32) has recently recom
mended the 40% rule or the SRSS rule for the design and analysis of bridge structures (Nutt,

A Replacement Rule for Multicomponent

Seismic Analysis

155

1996). These rules disregard the issue of correlation between the components of ground mo
tion.
lf R1, R2 and R3 denote the contributions to a response quantity respectively from the
two horizontal and one vertical component of ground motion, then the estimate of the total
response, R, according to the SRSS rule is

R = (R2 +R] +RJ)Yi

(1)

The basic assumption of this rule is that the response quantities R1, R2 and R3 are statisti
cally independent. Considering the PenzienWatabe model described earlier, this rule would
be accurate if the assumed horizontal ground motion components were along the principal
axes of the earthquake. However, in practice, the directions of the components normally are
chosen to coincide with the assumed axes ofthe structure. In sorne cases, such as the Caltrans
specifications for bridges (Caltrans, 1995), the directions of the components are prescribed
based on the geometry of the structure. These assumed directions may not coincide with the
principal axes of the earthquake. Consequently, the input components of the ground motion
in general would be correlated. The SRSS rule does not properly account for this correlation.
Of course it is possible to orient the structure axes along the principal directions of the earth
quake, where the components are statistically independent. This approach, in effect, leads to
the CQC3 rule described later in this paper.
The percentage rule is believed to have originated from work done by Newmark (1975)
and by Rosenblueth and Contreras (1977). The rule crudely approximates R as the sum of
100% ofthe response resulting from the input in one direction and sorne percentage, a, ofthe
responses resulting from the inputs in the other two directions. The combination that yields
the most critica! estimate of the total response is used for design. Hence, three cases must be
considered:

R = R1 + ctR2 + ctR3

(2a)

R = ctR1 +R2 +ctR3

(2b)

= ctR1 +ctR2

+R3

(2c)

Rosenblueth and Contreras (1977) suggested a= 30% in order to minimize the errors intro
duced by this linear approximation. This is the value adopted by UBC 97 (ICBO, 1997) and
by the Caltrans bridge design speciftcations (Caltrans, 1990); however, these codes only
specify (2a) and (2b) with the term ctR3 omitted from the combination. While both codes re
quire consideration of the vertical component when its effect is signiftcant, neither explicitly
instructs the designer how to combine its contribution with those of the horizontal compo
nents. Newmark (1975) recommended using a= 40%, arguing that the resulting estimate in
general will be conservative compared to the SRSS result. This recommendation is adopted
by the ASCE 486 Standard (ASCE, 1986). Presumably, this is also the basis for the recent
recommendation by ATC32 (Nutt, 1996), although Newmark's paper is not cited there.
Neither of these percentage rules properly account for the correlation between the ground
motion components.

156

C. Menun and A. Der Kiureghian

z1, z2 =structure axes

z; =principal

z],

axes of ground motion

Structure

zi

+;;~:;~,

,'

_, ..

Figure 1. Definition of 0 in horizontal plane.

PROPOSED MULTICOMPONENT

COMBINA TION RULE

It is clear from the preceding discussion that the multicomponent combination problem
involves two sets of axes: one set defining the principal directions ofthe ground motion, and
one set defining the structure model. W e assume that the third axis of both coordinate sys
tems is vertical. Hence, the relative orientation of the two coordinate systems is defined in
terms of an angle 9 between the two sets of horizontal axes. We consider 9 to be positive
when it defines the position of the earthquake principal axes relative to the structure axes in
the counterclockwise direction, as shown in Figure 1.
The specification of the input excitation is assumed to be in terms of design response
spectra associated with the principal directions of the earthquake. The spectral ordinates for
mode i of the structure are denoted as S1; for the major principal axis (from the earthquake
source towards the site), S2; for the intermediate principal axis (the transverse horizontal di
rection), and S3; for the minor principal axis (the vertical direction). In general, the spectral
shapes for the three directions can be different. However, in practice the design response
spectra in the two horizontal directions are usually proportional such that the ratio of their
ordinates is a constant for ali modal frequencies, i.e.,
(3)
where O ~ y ~ 1 is a constant. While rules for the more general case of different horizontal
spectral shapes can be easily developed (Smeby and Der Kiureghian, 1985; Lopez and Tor
res, 1996), their use in design practice is not common. In the remainder of this paper we as
sume that the above proportionality rule holds.

A Replacement Rule for Multicomponent Seismic Analysis

157

Smeby and Der Kiureghian (1985) originally derived the following combination rule for
multicomponent seismic input:
3nn
[

+ 2 (1y2

)nn

{,

R = ~tr~p.akiakjSkiSkj

)trn ~n

- ~ -y2 tr~p1(aaj -a2;a2j)S1;S1j

],Vi

sin 0
(4)

P1;2jS1;S1j sin 0 cose

In the above expression, n denotes the number of modes of the structure, Py denotes the
correlation coefficient between the responses in modes i and j , aki denotes the effective
participation factor associated with the input direction k and mode i , and y and 0 are as de
fined above. We note that the definition of 0 adopted in this paper is opposite in sign from
that assumed in (Smeby and Der Kiureghian, 1985). The coefficient Py is identical to that
used in the CQC rule (Der Kiureghian, 1981; Wilson et al., 1981); it is a function ofthe mo
dal frequencies and damping ratios. The factor aki is a function of the structure properties
and the specific response quantity of interest (Smeby and Der Kiureghian, 1985). As is usual
in practice, these factors are defined with respect to inputs along the structure axes, which are
different from the axes ofthe seismic input here. Thus, aki represents the response in mode i
when the spectral ordinate in the structure direction k is unity. Note that, with this definition,
the first subscript on Ski refers to the k th principal axis of the ground motion, whereas the
first subscript on aki refers to the kth axis of the structure. The above combination rule is
based on elementary principies ofrandom vibration theory and correctly accounts for the cor
relations between the modal responses of the structure as well as the correlations between the
components of ground motion (Smeby and Der Kiureghian, 1985).
We define
Rt = LLPifk;akjSk;Skj
i

(5a)

Rk1 = LLP!ikiljskislj

(Sb)

Based on the preceding definitions, Rk represents the contribution to the response of the
structure from the kth ground motion component, if it were acting along the kth structure
axis. Equation (Sa) is the conventional CQC rule as used in practice. Rkl represents a cross
term between the contributions to the response from the kth and l th ground motion compo
nents, if they were acting along the kth and l th structure axes, respectively. lt arises from the
correlation between the ground motions along the structure axes. It should be clear that

Rt1 = Rk

and Rkk = Rf . Note that Rkl involves the same terms that enter into the definitions
ofthe ordinary responses Rk and R1. Using these definitions in (4), we obtain

C. Menun and A. Der Kiureghian

158

R ~ [ {Rf + Rf + Rj )-(l-y2 { Rf-

y12 R? )sin 2 8 +2( l-/

)R12 sin 8cos8

(6)

Equation (6) shall be referred to as the CQC3 rule. The rule contains three distinct terms. The
first term is identical to the SRSS rule and represents the response if the principal directions
ofthe ground motion were to coincide with the structure axes, i.e., if 0 =O. The second and
third terms, which involve the angle 0, account for the fact that the principal directions of the
ground motion do not coincide with the structure axes. The contribution from the correlation
between the components of ground motion along the structure axes is also implicit in these
terms.
lt can be seen from (5) and (6) that the CQC3 rule is not computationally demanding: it
requires calculation of aki, Ski and Py, which are readily available if the conventional CQC
rule is implemented in the dynamic analysis software, and sin 0 and cos0 for a given angle 0
between the two coordinate systems. Most current software for multicomponent seismic
analysis compute the terms R1, R2 and R3. To implement the CQC3 rule, it is necessary to
also compute the cross term R12 and evaluate (6) for any given 0.
In actual practice, 0 may not be known. In that case, it is prudent to design for that value
of 0 that gives the largest response. The CQC3 rule provides a simple means of calculating
the critica! value of 0 for each response quantity of interest. Differentiating (6) with respect
to 0 and setting the resulting expression equal to zero, the critica! orientation of the principal
axes of the earthquake with respect to the structure axes is found to be

ecr = tan1

=2
R1

3_R12
y -1

--R2

(7)

y2

Equation (7) has two roots between O and 7t radians corresponding to those values of 0 that
give the maximum and minimum values ofthe response. Interestingly, one can show that

ecr

is actually independent ofthe ratio y between the two horizontal components. To see this, !et
~
R2 = L;LJP!i2;a21S1;SIJ and R12 = L;LJP!i1;a21S1;SIJ. Clearly, these define values

of R2 and R12 if the spectral ratio y were to be 1. Noting that R2 = y R2 and R12 = y R12, we
obtain from (6) and (7)

R = [(Rf +y2 R.J +Ri )(ty2 )(Rf

-R.J )sin 2 0 + 2(1y2 )R.12 sin 0cose]Yz

(8)

(9)

A Replacement Rule for Multicomponent Seismic Analysis

Hence,

159

acr is obtained from (9) independent of y by using the spectral ordinates S1j along

both horizontal structure axes. Note that no significant computations beyond that required for
the conventional CQC rule are required to determine acr frorn (7) or (9).

COMPARISON OF CQC3 WITH EXISTING RULES


Ifthe principal components ofthe ground motion coincide with the structure axes with di
rection 1 being the major axis, i.e., if'B =O, equation (6) reduces to

R = (Rf + Ri + Rj }Vi

(10)

This is the SRSS rule. This equation also arises when the intensities ofthe horizontal compo
nents of ground motion are equal, i.e., when y= l. Thus, the conventional SRSS rule is accu
rate only when either ofthese two conditions is satisfied. When the orientation ofthe princi
pal axes of the earthquake is unknown, the SRSS rule can be used provided the effective par
ticipation factors are defined relative to these axes. Naturally, since these axes remain un
known, the effective participation factors will be functions of a. One can easily verify that
this approach leads to the same results as in (6) and (7). In the current practice the SRSS rule
is not used in this manner and, hence, it could lead to unconservative results when the above
conditions are not satisfied.
The 30% and 40% rules can be seen as further simplifications of the SRSS rule. Suppose
for the response of interest the contribution from the loading along structure axis 1 is larger
than the contribution from an equal loading along structure axis 2, i.e., R2 s R1. Then, setting

R2 = y R2 = y bR1, where O s b s 1, and neglecting the contribution R3 from the vertical


component, one obtains from (10)

R=R1(1+y2b2}Yi

(11)

Clough and Penzien (1993, page 658) examine the difference between (11) and the 30% rule

= R1+0.30R2 = R1(1+0.3b)

(12)

For y= 0.85, they show that the 30% rule is a conservative linear approximation to (11) with
a maximum difference of approximately 5%. This favorable comparison, however, only
holds under the assumption that the horizontal ground motions directed along the structure
axes are uncorrelated with the weaker component having 85% of the intensity of the stronger
component. In general, this assumption may not be satisfied and the error associated with the
30% rule can be substantially higher. The 40% rule would naturally be even more conserva
tive. It is noted that these two rules are the predominant rules adopted in many current design
codes.
Expressions for the maximum response, R, and the critica) orientation, acr, different
from those given in (6) and (7) were derived by Wilson et al. (1995). However, their expres
sions are incorrect since, as pointed out in (Lopez and Torres, 1996), they do not properly
account for the cross term R12. Wilson et al. (1995) also recommend the use ofthe SRSS rule
with y= 1 to produce "structural design which has equal resistance to seismic motions from

160

C. Menun and A. Der Kiureghian

..

40 ft

..

ELEVATION
\

Abutment Sprin:5 ft

t/

~~nge~

\45

Abutment Spring

Figure 2. Example bridge analyzed.

ali directions." However, such a design can be overly conservative since ground motions of
equal intensity in the two horizontal directions are generally not observed.
NUMERICAL EXAMPLE
To compare the results obtained from the CQC3 rule to those obtained from the 30%, the
40% and the SRSS rules, the curved bridge shown in Figure 2 is considered. The "compres
sion" model ofthis bridge, as defined in (Caltrans, 1995), is analyzed. Only horizontal com
ponents of ground motion are considered. The components are assumed to have identical
spectral shapes defined by the elastic acceleration response spectrum given in the Caltrans
bridge design specifications for a site with less than 1 O ft alluvium and a peak rock accelera
tion of 0.3g (Caltrans, 1990). Fifteen modes ofvibration are included in the dynamic analy
ses, accounting for approximately 90% of the total effective mass in each direction. Ali
modes of vibration have an assumed damping ratio of 5%.
The maximum bending moments about the x and y axes of column 4 as computed by the
CQC3 and SRSS rules are shown in Figures 3a and 3b for several values ofy along with the
results from the 30% and 40% rules. The CQC3 results are presented for ali possible orienta
tions of the principal axes of the ground motion relative to the assumed structure axes
shown in Figure 2. The results presented for the SRSS and percentage rules, however, corre
spond to what would normally be done in practice, namely they assume the principal axes of
the ground motion coincide with the structure axes. It should be apparent that the results ob

A Replacement Rule for Multicomponent

Seismic Analysis

161

~
?

1600

= 1.00

(CQC3 & SRSS)

= 0.85

= 0.67

1400
1200

40% Rule
30% Rule

(a)

r = 0.50

1000

tt

CQC3

SRSS

e ( radians)
40% Rule

= 1.00 (CQC3 & SRSS)


__/.+

-y

1200

30% Rule

1100
(b)

1000
900

e ( radians)

Figure 3. Maximum bending moments in column #4.

162

C. Menun and A. Der Kiureghian

tained from the SRSS and percentage rules depend upon the assumed orientation of the
structure axes. If these axes are oriented differently from that shown in Figure 2, different
values for the bending moments will be computed by the SRSS and percentage rules.
It can be seen in Figure 3 that, except for the case when y= 1, the SRSS rule underesti
mates the maximum bending moments in column 4 relative to the CQC3 rule. The difference
arises from the fact that the SRSS rule ignores the correlation between the horizontal compo
nents of the ground motion acting along the assumed structure axes. As the value of y de
creases, the effect of the correlation between the components of ground motion becomes
more pronounced. Recall that when y= 1, the CQC3 rule reduces to the SRSS rule. This in
fact can be seen in Figure 3. However, horizontal ground motions of equal intensity implied
by y = 1 are typically not observed. Consequently, the CQC3 and SRSS results for y = 1 are
conservative estimates ofthe maximum bending moment.
The SRSS rule can be used to generate the curves produced by the CQC3 rule in Figure 3.
Each point on a curve, however, corresponds to a different orientation of the structure axes
(coinciding with the angle of the principal axes of the earthquake) and hence requires a
modification of the computer model of the bridge and a new set of dynamic analyses. The
advantage of the CQC3 rule is that it does not require this repeated analysis. Furthermore, it
provides a simple formula for determining the critica! orientation, ecr' ofthe earthquake and
the corresponding maximum value for each response quantity of interest.

Like the SRSS rule, the response quantities computed by the 30% and 40% rules are a
function of the assumed axes of the bridge. The critica! orientation of the structure axes pre
dicted by these rules can only be found by tria! and error, requiring many dynamic analyses.
As for the SRSS rule, this procedure can be tedious and expensive. Note in Figures 3a and 3b
that the lower bounds predicted by the 40% rule are higher than the results obtained by the
CQC3 and SRSS rules for y= 1, which as mentioned earlier are conservative estimates of the
two responses. Except for the rare occasion when the two horizontal components of ground
motion give rise to nearly equal contributions to the response quantity of interest, the 40%
rule will always be more conservative than the SRSS and CQC3 rules with y = l. However,
because of the crude nature of the approximation involved, it is not possible to predict how
the results of the 30% rule will compare to the CQC3 results in general. It can be seen in Fig
ure 3a that the maximum bending moment about the x axis ofthe column computed using the
30% rule is comparable to, but slightly higher than, the result obtained by the CQC3 rule for
y = l. However, the 30% rule estmate for the maximum bending moment about the y axis
can be seen in Figure 3b to lie between the CQC3 results for y= 0.85 and y= l.
CONCLUSIONS
The CQC3 rule is a much more general and rational procedure than either the percentage
rules or the SRSS rule currently specified in many design codes for the multicomponent
ground motion problem. By incorporating the Penzien and Watabe model of ground motion
into its formulation, the CQC3 rule can explicitly account for the correlation and relative in
tensities of the components of the ground motion. The 30% and 40% rules do not rationally
address either of these aspects ofthe seismic input. An implicit assumption ofthe SRSS rule
is that the components of ground motion are uncorrelated. If the structure axes are properly
chosen, the SRSS rule will yield the same result as the CQC3 rule for the maximum value of

A Replacernent Rule for Multicornponent Seisrnic Analysis

163

a given response quantity. However, for the SRSS rule, the critica! orientation of the structure
axes can only be found by tria! and error, requiring many dynamic analyses. Furthermore, the
critica! orientation of the axes will be different for different response quantities. In contrast,
the CQC3 formulation provides a simple and computationally efficient means of finding the
critica! orientation of the structure axes relative to the principal axes of the ground motion
and the associated maximum value of a given response quantity. In light of these observa
tions, it is recommended that the CQC3 rule be adopted as a general method for combining
the contributions from orthogonal components of ground motion to the maximum value of a
response quantity.
Support for this work was provided in part by the California Department of Transporta
tion under Contract No. R TA59 Al 31, whose support is gratefully acknowledged. The views
and opinions presented are those of the authors and do not necessarily reflect the official
views or policies ofthe sponsor.

REFERENCES CITED
ASCE Working Group on Seismic Analysis of Safety Related Nuclear Structures, 1986, Seismtc
Analysts o/ Scfety Related Nuclear Structures and Commentary on Standard for Setsmic Analysis
o/Safety Related Nuclear Structures, American Society ofCivil Engineers, New York, 91 p.
California Department ofTransportation (Caltrans), June 1990, Bridge Design Spectftcattons.
California Department of Transportation (Caltrans), March 1995, STRUDL modeling guidelines,
Memo to Designers 20-4 Attachment A.
Clough, R. W. and Penzien, J., 1993, Dynamics o/ Structures, McGrawHill, New York, 738 p.
Der Kiureghian, A., 1981, A response spectrum method for random vibration analysis ofMDF sys
tems, Earthquake Engineering and Structural Dynamics, 9, p. 419435.
lntemational Conference ofBuilding Officials (ICBO), 1997, 1997 Uniform Building Code, Volume
2, Structural Engineering Design Provistons, Whittier, CA, 492 p.
Lopez, O. A. and Torres, R., 1996, Discussion of 'A clarification of orthogonal effects in a three
dimensional seismic analysis', Earthquake Spectra, 12, p. 357361.
Newmark, N. M., 1975, Seismic design criteria for structures and facilities, TransAlaska pipeline
system, Proceedings o/ the US. National Conference on Earthquake Engineering, Earthquake
Engineering Institute, p. 94103.
Nutt, R. V., 1996, ATC32: Provisional recommendations for the seismic design of bridges, The
Fourth Caltrans Seismic Workshop, California Department ofTransportation, Sacramento, CA.
Penzien, J. and Watabe, M., 1975, Characteristics of 3dimensional earthquake ground motion,
Earthquake Engineertng and Structural Dynamics, 3, p. 365374.
Rosenblueth, E. and Contreras, H., 1977, Approximate design for multicomponent earthquakes, Journal o/ Engineering Mechanics Dtvision ASCE, 103, p. 895911.
Smeby, W. and Der Kiureghian, A., 1985, Modal combination rules for multicomponent earthquake
excitation, Earthquake Engineering and Structural Dynamics, 13, p. 112.
Wilson, E. L., Der Kiureghian, A., and Bayo, E. P., 1981, A replacement for the SRSS method in
seismic analysis, Earthquake Engineertng and Structural Dynamics, 9, p. 187194.
Wilson, E. L., Suharwardy, l., and Habibullah, A., 1995, A clarification ofthe orthogonal effects in a
threedimensional seismic analysis, Earthquake Spectra, 11, p. 659666.

Das könnte Ihnen auch gefallen