BNL-NUREG66410

EVALUATION

OF MODAL

COMBINATION

METHODS

FOR

SEISMIC RESPONSE SPECTRUM ANALYSIS

>aper ID K4-A4-US Morante, Richard* Brookhaven National Laboratory Building 130,32 Lewis Road Uptonj New Yorlcj 11973-5000 USA Wang, Yung Brookhaven National Laboratory Building 130,32 Lewis Road

Upton,New York 11973-5000 USA

Chokshi, Nilesh United States Nuclear Regulatory Commission Mail Stop T-10 L1 Washingto% DC 20555 USA Kenneally, Roger United States Nuclear Regulatory Commission Mail Stop T-10 L1 Washington DC 20555 LEA Norris, Wallace United States Nuclear Regulatory Commission Mail Stop T-10 L1 Wa&ington, DC 20555 USA

* Corresponding Author - Phone (5 16) 344-5860, Fax (5 16) 344-3957, e-mail: morante@bnLgov. 4

DISCLAIMER
This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, make any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or manufacturer, or service by trade name, trademark, otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by-the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

DISCLAIMER
*

Portions of this document maybe illegible

in electronic image products. Images are

produced from the best available original

document.

ABSTRACT Regulatory Guide 1.92 Combining Modal Responses and Spatial Components in Seismic Response Analysis was last revised in 1976. The objective of this project was to reevaluate the current regdatory guidance for combining modal responses in response spectrum analysis, evaluate recent technical developments, and recommend revisions to the regulatory guidance. This paper describes the qualitative evaluation of modal response combination methods. INTRODUCTION The United States Nuclear Regulatory Commission MC) issues Regulatory Guides (RG) which describe methods acceptable to the NRC staff for satisfying regulations. RG 1.92, Combining Modal Responses and Spatial Components in Seismic Response Analysis: (Reference 1) was last revised in 1976, prior to a number of significant technical developments for combining modal responses. The 1989 revision to Standard Review Plan (SW?) Section 3.7.2, Seismic System AnalysisJ (Reference 2) recognized a number of recent technical developments by reference, and stated that t$eir application to nuclear power pkmt seismic analysis is subject to review on a case-by-case basis. Also incorporated into SRP Section 3.7.2 as Appendix A was a procedure to address high frequency mode effects, developed by Kennedy (Reference 3). , The initial phase of this program focused on review of the technical literature and selection of candidate modal response combination methods for more detailed numerical evaluation. Acceptable methods in RG 1.92 were also included to provide a comparison to more recent technical developments. References 3 and 4 provided an exceIIent starting point. The methods selected for evaluation were those which have been subjected to the greatest level of prior review and assessment. The methods evaluated for combining out-of-phase modal response components were Square Root of the Sum of the Squares (SRSS), NRC Grouping, NRC Ten Percen; NRC-DSC, Rosenblueths DSC (Reference 5), and Der Kiureghians CQC (Reference 6). The methods evaluated for separating in-phase and out-ofphase modal response components were Lindley and Yow (Reference 7), Hadjian (Reference 8), and Gupta (Reference 4). The method evaluated for including high frequency mode effects was Kennedy (Reference 3). DESCRIPTION OF MODAL RESPONSE COMBINATION METHODS

The major application of seismic responsespectrum analysis is for systems and components attached to building structures. A building-filtered in-structure response
spectrum depicting spectral acceleration vs. frequency is the typicaI form of seismic input for
Note: This work was performedunder the auspicesof the U.S. Nuclear RegulatoryCommission.

..

-.

such analyses. This type of spectrum &mlly exhibits a sharp peak at the fimdamental frequency of the buildingkoil dynamic system. An idealized in-structie response spectrum is shown below, the spectral regions and key frequencies are indicated. low-frequency
(Predominantlyout-ofphase response) a

mid-frequency -+ from out-of~~Transition

i

~ --+
1 I I :

high-frequency
(In-phase, pseudostatic response)

A.
!

phase to in-phase response)

I I

frequency@) frequency at which the peak spectral acceleration is reached typically the -f& = fimdarnental frequency of the buildingkoil system faA . frequency at which the spectral acceleration returns to the zero period acceleration frequency above which the SDOF modal respo&es we in-phase with the time fm = varying acceleration input used to generate the spectrum The high frequency region of the spectrum (>fipJ is characterized by no amplification of the peak acceleration o~~theinput time history. A SDOF oscillator having a frequency > fmAis accelerated in-phase and with the same acceleration magnitude as the applied acceleration, at each instant in time. A system or component with fimdamental frequency >~mAis correctly analyzed as a static problem subject to a loadiig equal to mass times ZPA. This concept can be extended to the high frequency (>~ZPJ modal responses of multi-modal systems or components. The mass not participating in the amplified modal responses (i.e., missing mass) multiplied by the ZPA is applied in a static analysis, to ob&n the response contribution from all modes with frequencies > fzp~ In the Iow-frequency region of the spectrum (<~p) the modal responses of SDOF oscillators are not in-phase with the applied acceleration time history, and generally are not in-phase with each other. These are designated out-of-phase modal responses. Since a response specti provides only peak acceleration vs. frequency, with no phasing itiorrnation, the out-of-phase peak modd responses for a multi-modal structural system require a rule or methodology for combination. Based on the assumption that the peak modal responses are randomly phased, the square root of the sum of the squares (SRSS) method was developed and adopted. Modifications to SRSS were subsequently developed, in order to account for potential phase correlation when modal frequencies are numerically close (i.e., closely spaced modes). In the mid-frequency region ~sp tofZpJ, it has been postulated that the peak SDOF oscillator modal responses consist of two distinct and separable elements. The first element is the out-of-phase response component and the second element is the in-phase response
JiP f~ fm.4

.,.-

....

..

..

..

..

.-

component. It is fbrther postulated tit there is a continuous transition from out-of-phase response to in-phase response. If~p ~m* can be defined, then the mid-frequency region can Sub-wjons:fip <f<fi ZUIdfW dgkpy be further divided intoIWO past ractice in the nuclear power industry has been to assume that individual modal p responses in the mid-frequency region& <~<&J me out-of-ph~e> ad combination methods applicable to the low-frequency region are applicable to the mid-frequency region. Terms used in the following sections are: Spectral Acceleration for mode i Sai = Response of mode i &= = In-phase response ratio for mode i In-phase response component for mode i %i = Rpi = Out-of-phase response component for mode i Total in-phase response component from all modes Rr= Total out-of-phase response component horn all modes Rp = Total combined response from all modes Rt = Modal response correlation coefficient between modes j and k. Cj~ = Combination of Out-of-Phase Modal Res~onses Components In generalized form, all of the out-of-phase modal response combination methods can be represented by a single equation: Yz Rp = ~ [ jsl ~ Cj~ Rpj Rp~ ksl 1
(Eqn. 1)

The coefficients Cj~can be uniquely defined for each method. Square Root of the Sum of the Squares (SRS$) At the foundation of all methods for combining uncorrelated modal responses is SRSS. All of the methods for combination of the out-of-phase response components are equivalent to SRSS if there are no closely spaced modes. In the case of SRSS, (Eqn. 2) Cj~ = 1.0 for j =k Cj~ = 0.0 for j # k SRSS Combination reduces to:
(Eqn. 3)

NRCGroupingMethod The NRC Grouping Method (Reference 1) is the most commonly applied method of
accounting for closely spaced modes in the nuclear power industry. The system modal

responses are grouped and summed absolutely before pefiorming SRSS combination of the groups. The modal responses are grouped such that the lowest and highest frequency modes in a group are within 10%and no mode is in more than one group. Rp = $

The major criticism of the NRC Grouping Method is the use of absolute summation within each group. If modal responses are assumed to be correlated because they have closely spaced frequencies, then summation should be algebraic within each group. In terms of the

[1

!4

GI+2

(Eqn. 4)

...

..

..

coefficients, Cj ~,the NRC Grouping l~ethod is defined as: c. = 1.0 forj =k c;: = 0.0 forj # k, not in same &oup (Eqn. 5) . Cj~ = 1.0 for j* k, in the same group, Rpj * Rp~ >0 Cj~ = 1.0 forj # k, in the same group, Rpj * Rp~ <0 NRC Ten Percent Method The NRC Ten Percent Method (Reference 1) is a generally more conservative variation of the NRC Grouping Method. Closely spaced modes are defined as modes with frequencies within 10/0of each other and absolute summation of the closely spaced modal responses is specified. The difference is that modal responses are not grouped.

In terms of the coefficients, Cj~, the NRC Ten PercentMethod is defined as: c. = 1.0 forj=k

c;: = 0.0 for j* k, and~ and~~ separated by > 10%

Cj~ = 1.0 for j # k, and~ andfi separated by < 10XO; * Rp~ >0 Rpj (Eqn. 6) Cj~ = -1.0 for j #k, fi andfi separated by< 10%; Rpj * Rp~ <0 The NRC Ten percent Method will always produce results ~ NRC Grouping Method. NRC Double Sum Combination (NRC-DSC) The NRC-DSC Method (Reference 1) is an adaptation of Rosenblueths method. The coefficients Cj~are defined by Equaticm 7. A conservative modificatio~ consistent with the NRC Grouping and Ten Percent methods, is that the product Cj~Rpj Rp~ is always taken as positive. In Rosenblueths method, the product maybe either positive or negative, depending on the signs of Rpj and Rpk Consequently, NRC-DSC will always produce results ~ Rosenblueths method. Rosenblueths Double Sum Combination (DSC) Rosenblueth (Reference 5) provided the first significant mathematical approach to evaluation of modal correlation for seismic response spectrum analysis. It is b-reed on the application of random vibration theorjj, utilizing a finite duration of white noise to represent seismic loading. A formula for calculation of the coefficients Cj~~ a fhnction of the modal
circular

frequencies (())j,Q modal damping ratios.(~j,~J, and the time duration of strong

earthquake motion (tD)-was derived. lJsing the form of the equation from Reference 1,

c. Jk

m
1

(Eqn. ?)

Numerical values of Cj~were tabulated for the DSC Method as a fimction of frequency, frequency ratio, and strong motion duration time for constant modal damping of lVO,2XO,5?40 and 10O/O.The most significant result is that Cjkis highly dependent on the damping ratio. For 2Y0, 5V0 and 10% damping, Cj~ ~ 0.2,0.5 and 0.8 respectively, at a frequency ratio of 0.9 (modal frequencies within 10%). The definition of closely-spaced modes used in the NRC Grouping and Ten Percent Methods is not damping-dependent.

Der Kiureghian h Complete Quadratic Combination (CQC)

Der Kiureghian (Reference 6) presents a methodology similar to Rosenblueths Double Sum Combination for evaluation of modal correlation for seismic response spectrum analysis.

..

It is also based on application of random vibration theory, but utiliies an iniinite duration of white noise to represent seismic loading. A formula for calcdation of the coefficients CJ~as a fimction of modal circular frequencies and modal damping ratios was derived:

While the form of Equation 8 differs significantly fiomEquation 7, the WO equations produce equivalent results if t~ is assumed very large in Eq~tion 7. : ,
Separation of Modal Responses into Out-of-Phase Components and In-Phase Com~onents

Three methods have received considerable prior review and evaluation Lindley-Yow (Reference 7), Hadjian (Reference 8), and Gupta (Reference 4). The mathematical statement of each method is not restricted to the mid-frequency range &sp<~<&J of the response spectrum. However, as discussed previously, it is in the mid-frequency range that the separation of individual peak modal responses into out-of-phase and in-phase modal response components is applicable. The similarities and differences, as well as the limitations, of the three methods are described in the following sections. Lhzdley-Yow Method Mathematically, the Lindley-Yow method (Reference 7) is defined by the following equations: (Eqn. 9) Osaisl.O ai = ZPA/Sai Rri =Ri *ai (Eqn. 10) Rpi = Ri * ~~
(Eqn. 11) (Eqn. 12)

RP =
[

~
j=lk=l

Cjk Pj

Pk

Rt = ~-

V2

(Eqn. 13)
(Eqn. 14)

The following characteristics of the Lindley-Yow method are observed: ai - I-O ~fi ~~=A (s% = ZPA). Therefore,~P ~m& me in-phase component of modal response for every mode has an associated acceleration equal to the ZPA. The out-of-phase component of an individual peak modal response has an associated modified spectral acceleration given by
q q q

~ai = ( Sa~ - ZpA2 )x c

q

(Eqn. 15)

& = (Rp~+ I@~~; which infers that the in-phase and out-of-phase response components of an individual peak modal response are uncorrelated. ai attains its minimum value at~ =~sp, but increases for~ <~sp until it attains a value of 1.0 when S%= ZPA in the low I12equency region of the spectrum. Values becomes imaginary. of ai >1.0 have no meaning because (1 - a~)% #m obvious limitation of the Lindley-Yow method is in the low frequency rmge V<J$P) of the response spectrum. Therefore, the Lindley-Yow method is applicable to structural

.-

-.

systems which do not have significantmodal responses with~ < fsp. Circumventing this limitation in the Lindley-Yow method is straightiomvard apply it only to those modes withj >fsp set ais Ofor~ <f~p. and While in theory the Lindley-Yow method includes the in-phase contribution from modes above~ZPA,ts practical application is for modal responses below~ZpA,coupled with the i missing mass method for modal contributions above~zPA. Ha~ian Method The Hadjian Method (Reference 8) is similar in formulation to the Lindley-Yow method, with two notable differences: Equation 11 is replaced by @qn. 16) R]~=&*(l _ ai)
q q

Equation 14 is replaced by l?t=]Rpl+]Rrl (Eqn. 17)

The modified spectral acceleration is given by

~ai = Sai - ZPA

(Eqn. 18)

The Hadjian method has the same limitation as the Lindley-Yow method in the low frequency range, because the definition of ai is identical. However, the Hadjian Method possesses internal contradictions with respect to the assumed phase relationships between inphase and out-of-phase response components. Combining Equations 10 and 16 yieIds &= Rpi+Rri (Eqn. 19) This implies that the in-phase and out-of-phase response components for each mode are in-phase with each other. However, all kis are in-phase and summed algebraically, per Eqn.

12, to obtain Rr. Therefore,it would fcdlowthat all Rpis are also in-phase and should be
summed algebraically to obtain Rp. This contradicts Equation 13, in which the Rpis are assumed to be predominantly out-of-phase. Kennedy (Reference 3) previously identified this contradiction. On this basis, the HadjiCanmethod is not recommended and was not included in subsequent numerical studies.

Gupta Method The Gupta Method (Reference 4) is identical in form to the Lindley-Yow method. The one very significant difference is the definition of ~i. Equations 10 through 14 remain the same. In the Gupta method, ai is an explicit fimction of frequency. The original basis for definition of ai is semi-empirical, derived from numerical studies using actual ground motion records. A best fit equation, which defines ai as a continuous fbnction of frequency, was developed horn the results of the numerical studies. Two spectrum-dependent frequencies ~, fJ are first defined as follows: S a~u J= 27r Svmm
(Eqn. 20) and velocity, respectively.

where S~.T and SV~~y the maximum spectral acceleration are

4 (J UI%J3
Guptas

(Eqn. 21)

deftition

of ai is given by: ai=Ofor~<$ !n(~/fi] i = Qn (&/j ) for~ <$<$ (Eqn. 22)

ai = 1.() for~ >~z

For a sharpIy peaket in-structie response spectrum, X& because SV- = Max (S@~i) = S% / co~P; nd the Gupta method hM the following characteristics: a Forfi ~&, ai = O;treated as out-of-phase. Forfi ~~ S~ZPA, i = 1.0; this infers thatjp =fi in the Gupta method. a
q q

Only modal responses with~P <J <A are separated into out-of-phase and in-ph~e response components. The potential limitations of the Gupta method iie in the semi-empirical basis for definition of ai as a fiction of~. In Reference 4, Gupta indicates that ~ can be numerically evaluated if the time history used to generate the response spectrum is known. It is implied that numerical evaluation of ai is more accurate than the semi-empirical definition of ai given by Equation 22. The overaI1 structure of the Gupta method is superior to the Lindley-Yow method because there is no limitation for modal responses withj <f&. In addition, any value of~p ~&p*can be accommodated by setting~ YP, in lieu of Equation 21.
q

Contribution of I&h Freauencv Modes Missing Mzss Method The Missing Mass Method is a convenien~ computationally efficient and accurate method to (1) account for the contribution of all modes with frequencies above the frequency &pJ at which the response spectrum returns to the Zero Period Acceleration (ZPA) and (2) account for the contribution to support reactions of mass which is apportioned to system support points. It constitutes the to,tal effect of all system mass which does not participate in (i.e., missing from) the modes with fi-equencies belowj&W The system response to the missing mass is calculated by petiorming a static analysis for applied loads equal to the missing mass multiplied by the spectrum ZPA. This method is mathematical~ rigorous and is considered the onIy acceptable method to account for high frequency modal contributions (fz&pJ and mas apportioned to system support points. Kemedy (Reference 3) documented this method and recommended that it be included in Regulatory Guidance. The 1989 revision to the SRP Section 3.7.2, Seismic System Analysisfl (Reference 2) incorporated Kennedys recommendation as Appendix A. The mathematical details are presented in both References 2 and 3. Comnlete Solution for Resuonse Suectrum Analysis Two methods are defined for constructing the complete response spectrum analysis solution. The coefficients Cj~are defined by one of the out-of-phase combination methods. Method 1 Method 1 represents the common method applied to response spectrum analysis since the 1980s. Amplified modal responses (f<fipJ are combined by SRSS with a correction for closely spaced modes. The contribution of unamplified modal responses (f>&J is calculated by the missing mass method. These two components are then combined by SRSS to produce the total solution. Mathematically, this is represented by

[E$cjkRj4
n = no. of modes below fzPA

(Eqn. 23)

Rt = ~Rp2 +Rr2

- .

~.-

I
:

Method 2 Method 2 introduces the concept of in-phase and out-of-phase modal response components for the amplified modes (~<~m~. Mathematically, the complete solution is represented by
Rpi = Ri *(1 -u;)% Rri =Ri *ai RP = 5 S CjkRpjRpk [ =1 k=l

n= no. of modes below&PA

..

(Eqn: 24)

Method 2 is equally applicable tc)both the Lindley-Yow and the Gupta methods. Only the definition of (ti changes. ~
SUMMARY

OF REWLTS

The qualitative evaluation of modal response combination methods provided the foundation for subsequent numerical studies, which quantitatively evaluated the combination
methods by comparison to time history analysis results. Together, this provided the basis for technical conclusions and recommendations for revision of regulatory guidance.

REFERENCES
1.

USNRC Regulatory Guide 1.92, Combining Modal Responses and Spatial

Components in Seismic Response Analysis, Revision 1, February 1976. USNRC Standard Review Plan, NUREG-0800, Section 3.7.2, Seismic System Analysis, Revision 2, August 1989. Kennedy, R.P., Position Paper on Response Combinations, Report No. SMA 1221 1.02-R2-0, March 1984. Gup@ A.K., Response Spectrum Method in Seismic Analysis and Design of Structures, CRC Press, Inc., 199:3.

2. 3. 4.

5.

Rosenblueth,E., and Elorduy, J., Responses of Linear Systems to Certain Transient Disturbances,Proceedings of the Fourth World Conference on Earthquake
Engineering, Santiago, Chile, 1969. Der Kiureghian, A. A Response Spectrum Method for Random Vibrations, UniversiU of California at Berkeley, June 1980. Lindley, D.W., and Yew, T.R., Modal Response %unrnation for Seismic Qualification, Proceedings of the Second ASCE Conference on Civil Engineering and Nuclear Power, Vol. VI, Paper 8-2, Knoxville, TN, September 1980. Hadjian, A.H., Seismic Response of Structures by the Response Spectrum Method, Nuclear Engineering and Design, Vol.66, no. 2, pp. 179-201, August 1981.

6. 7.

8.

Disclaimer otice:Thispaperwaspreparedin partby an employeeof theUnitedStatesNuclearRegulatory N Commission.lt presents information thatdoesnotcurrentlyrepresentan agreed-upon staffposition.NRChas neitherapproved ordisapproved itstechnicalcontent.