Download - NRC Staff Responses to Public Comments on Draft Regulatory ...

-1-

Resolution of Public Comments on Draft Regulatory Guide DG-1127, “Combining Modal Responses and Spatial Components in Seismic Response Analysis”

(Proposed Revision 2 of Regulatory Guide 1.92)

During the comment period for Draft Regulatory Guide DG-1127, which ended on April 15, 2005, the NRC received comments fromWestinghouse (W), Brookhaven National Laboratory (BNL), DST Computer Services S.A. of Switzerland (DST), and James Annettof New Jersey (JA). The following table summarizes the comments, the staff’s related responses, and the resultant changes (if any)made to DG-1127 to address the comments.

Source Comment Formal Staff Response Change to DG-1127

JA-1 Consider clarifying how the100-40-40 rule may be used inconjunction with determininggeometric resultant responsesusing SRSS.

An example might involve aconcern with sliding offoundations, where friction is aconcern. For this example 100%of the vertical acceleration is usedand a resultant horizontalacceleration determined by SRSSof 40% of the two horizontalaccelerations.

The Square-Root-of-the-Sum-of-the-Squares(SRSS) and 100-40-40 methods for combining theeffects of three directions of seismic excitation areintended to be applied to structural responsequantities, not to the external loading. This isclearly specified in DG-1127. The 100-40-40method applies only when the response spectramethod is used.

The 100-40-40 method, as defined in DG-1127,has been accepted as an alternative to SRSS forestimating the likely maximum absolute value of aninternal response component (e.g., force, moment,deflection, rotation) at a specific location in astructural element, attributable to the combinedeffect of three directions of seismic excitation(horizontal E-W, horizontal N-S, vertical). Thebasis for acceptance is a numerical study, whichcompared the 100-40-40 prediction of maximumresponse to the SRSS prediction of maximumresponse, for the complete range of possible ratiosof responses R1, R2, and R3. The resultsdemonstrate that the 100-40-40 prediction isessentially equal to or higher than the SRSS

None


-2-

prediction for all response ratios. In most structuraldesign evaluations, the maximum response toexternal loads is desired for comparison to codifiedacceptance criteria.

The example cited (sliding of foundations wherefriction forces are important) warrants specialdesign considerations.

First, the potential for liftoff of the foundationcaused by combined vertical and horizontalexcitation needs to be considered. If liftoff isexpected, a nonlinear analysis of the structuralresponse of the foundation/building on thesupporting medium, including liftoff and slidingbetween the foundation and the supportingmedium, is appropriate. The potential for liftoff canbe estimated by linear analysis of thefoundation/building for each of the three directionsof seismic excitation, assuming completefoundation fixity to the supporting medium. AfterSRSS or 100-40-40 combination of the responses,if the prediction of normal forces at thefoundation/supporting medium interface exceed thecompressive contact force attributable todeadweight over a significant area of thefoundation, then liftoff should be considered.

In the absence of liftoff, the total contact forcebetween the foundation and the supporting mediumtimes the coefficient of friction between thesurfaces defines the horizontal load limit beforesliding will occur. The selection of combinations oftotal contact force and total horizontal force that


-3-

may exist simultaneously due to the three-directional seismic excitation is not addressed byeither the SRSS or 100-40-40 methods. Thecritical condition for sliding may exist when none ofthe responses are at their extreme values, becauseof the coupling between the contact force and theallowable horizontal force. A conservativeapproach would be to use the minimum totalcontact force (deadweight minus maximum verticalforce attributable to three directions of seismicexcitation), and the maximum total horizontal forceattributable to three directions of seismic excitation. If sliding is precluded for this condition, there is noneed for further analysis. However, if sliding isindicated by the conservative analysis describedabove, the analyst may need to conduct a lineartime history analysis using three statisticallyindependent seismic excitations [see Paragraph2.2(2) of DG-1127], and to check for sliding atsuitably selected points in time.

DST-1 My comments concern “AppendixA: General Discussion of theResponse Spectrum Method.”

GENERAL COMMENT:I find the formulation ambiguousand I suggest that a morerigorous mathematical approachcan lead to improved engineeringsolutions.

RESPONSE TO GENERAL COMMENT: The material in Appendix A was taken from Section3.2.2.2, “Linear Methods,” of ASCE Standard 4-98,“Seismic Analysis of Safety-Related NuclearStructures and Commentary,” promulgated by theAmerican Society of Civil Engineers (ASCE). The formulation is limited to uniform supportexcitation, typical for analysis of building structuresand equipment, and a very commonly usedassumption in the analysis of multi-supportedsystems (e.g., piping). The intent of Appendix Awas to provide an overview, rather than a rigorousgeneral treatment, of modal superposition time

The staff deletedAppendix A to DG-1127and added a discussion ofuniform support motion(USM) vs. independentsupport motion (ISM) in theBackground portion ofSection B, “Discussion.”That discussion includes astatement that RG 1.92 isonly applicable to USM,and a reference to thecurrent staff position for


-4-

history analysis and .

The commenter clearly highlighted the limitationsand ambiguities contained in Appendix A. The staffwill consider deleting Appendix A from the final RG1.92, and replacing it with references to moreformal treatments of the subject. The staff wouldconsider recommendations for such references.

As an alternative, the comments will be factoredinto a revision of Appendix A. Definitions will beclarified and inherent assumptions (e.g., uniformsupport excitation) will be clearly identified.

ISM analysis.

DST-2 COMMENT #1:My first comment refers toequations (A.1). It is stated that“X=column vector of relativedisplacements (mx1).” The firstquestion is: relative to what?

RESPONSE TO COMMENT #1:If Appendix A is retained, the definition of X will beclarified.

The staff deletedAppendix A to DG-1127.

DST-3 COMMENT #2:The second question is: exactlywhat is the meaning of the term

, defined as the “ground&&ugacceleration”?

RESPONSE TO COMMENT #2:If Appendix A is retained, the definition of &&ugwill be clarified.


DST-4 COMMENT #3:My next comments refer toequations (A.2) and (A.3).

RESPONSE TO COMMENT #3:If Appendix A is retained, the distinction between mand n will be clarified and the matrix dimension of Φwill be corrected. While n=m only if all modes are



-5-

We note that X is defined in (A.1)as an (mx1) vector, where m =number of dynamicdegree-of-freedom. For equation(A.2) to be coherent, Φ and Ymust be (mxm) and (mx1)respectively; that is, they mustinclude all m dynamic degrees offreedom. They cannot bedimensioned as defined by “n =number of modes considered.”

However, both the orthogonalityprinciple, ΦTM Φ =I, and theequation (A.3) are valid if Φ isdefined as the rectangular matrixwhose columns are a subset ofthe mode shape vectors. Then Φwill be (mxn) and Y will be (nx1),where m = number of dynamicdegrees-of-freedom and n =number of modes considered.

retained, only roughly half of the modes areconsidered usable, and the intent of the proceduresoutlined in the RG is to minimize the number ofmodes that need to be calculated. Therefore,

will be the typical case.

DST-5 COMMENT #4:(Because of its length and thenumber of symbols and/orequations used, Comment #4 isnot reproduced here.)

RESPONSE TO COMMENT #4: The Appendix A presentation is limited to uniformsupport excitation. If Appendix A is retained, thiswill be clearly stated.


DST-6 COMMENT #5:(Because of its length and thenumber of symbols and/orequations used, Comment #5 is

RESPONSE TO COMMENT #5: The Appendix A presentation is limited to uniformsupport excitation. If Appendix A is retained, thiswill be clearly stated.



-6-

not reproduced here.)


-7-

DST-7 COMMENT #6:To end my comments on apositive note, based on theformulation in the attacheddocument, I derive an alternateform for the solution.

(Because of the length and thenumber of symbols and/orequations used, Comment #6 isnot reproduced here.)

RESPONSE TO COMMENT #6: It is the staff’s understanding that the commenter isproposing an alternative approach to the Gupta andLindley-Yow methods. If this is the case, the staffwould be interested in an example of the proposedmethod (e.g., piping model BM3 fromNUREG/CR-6645).

None

Note: The commenterprovided additionalinformation to the staff. The information providedwill require in-depth staffevaluation before it can beconsidered for possibleinclusion in a subsequentrevision of RG 1.92.

BNL-1 The importance of the “residualrigid response” is not sufficientlyemphasized. See theNUREG/CR-6645recommendations for bothresponse spectrum analysis(RSA) (Recommendation 6) andmode superposition time historyanalysis (Recommendation 7). The term “significant” used in theDG is very vague and subject tobroad interpretation. It is lessprescriptive than the currentguidance in the SRP. The currentguidance in the SRP can lead to a10% underprediction of supportforces, as discussed inNUREG/CR-6645. The “residualrigid response” should always becalculated and algebraicallycombined with the in-phasecomponents of the amplified

The staff did not formally respond to this BNLcomment; however, the staff did consider thiscomment in developing the final version ofRG 1.92, Revision 2.

The staff revisedRegulatory Position C.1.4in Revision 1 of RG 1.92,to include additionaldiscussion regarding theimportance of includingthe residual rigid responseof missing mass modesfor both response spectrumanalysis and modalsuperposition time historyanalysis.


-8-

modal responses in RSA. Inmode superposition time historyanalysis, the “residual rigidresponse” time history shouldalways be included as anadditional response mode andalgebraically combined with theamplified modal response timehistories, at each instant in time.

BNL-2 Rev. 2 to RG 1.92 needs toinclude a note that RegulatoryPosition 1.4 replaces orsupercedes the procedure in SRP3.7.2, Appendix A, for calculatingthe missing mass contribution tototal response. There is an errorin the SRP equations. This waspointed out to the staff during theNUREG/CR-6645 effort. Appendix I of NUREG/CR-6645contains the correct equations forimplementing the SRP 3.7.2,Appendix A, procedure.


Appendix A to the finalRevision 2 of RG 1.92 isreproduced from Appendix Ito NUREG/CR-6645and contains the correctequations for implementingthe procedure set forth inAppendix A to SRP 3.7.2. However, Revision 2 of RG1.92 does not referenceAppendix A to SRP 3.7.2.

Note: The draft revision(1996) of Appendix A toSRP 3.7.2 corrected thenoted error.

BNL-3 There is no discussion of directintegration time history analysisas an alternative to modesuperposition time historyanalysis, nor any reference to thecomparisons and


None.

Revision 2 of RG 1.92recognizes directintegration time historyanalysis as an alternative to


-9-

recommendations included inNUREG/CR-6645, Appendix E.

mode superposition timehistory analysis in theBackground portion ofSection B, “Discussion.”

Revision 2 of RG 1.92 doesnot reference Appendix Eto NUREG/CR-6645,because the staffconsidered the comparisonof direct integration resultsto modal superpositionresults presented inAppendix E to be outsidethe scope of RG 1.92.

BNL-4 Generally, DG-1127 includes therecommendations of NUREG/CR-6645, Section 5.2, for combiningmodal responses. One exceptionis the NUREG/CR-6645recommendation(Recommendation 3) that, if the“closely spaced modes” methodsof RG 1.92, Rev. 1, were going tobe retained, their applicabilityshould be limited to dampingratios of 2% or less. A secondexception is the NUREG/CR-6645recommendation(Recommendation 5) that the useof Method 1(as defined inNUREG/CR-6645) should bestrongly discouraged or identified


None.

The staff determined thatthe level of conservatismachieved by using Revision 1of RG 1.92 is acceptable. Therefore, it is stillacceptable to use themethods delineated in thatrevision, with the exceptionthat the “residual rigidresponse of the missingmass modes,” as describedin Regulatory PositionsC.1.4 and C.1.5, should beincluded in all analysessubmitted in support oflicensing decisions, after


-10-

as unacceptable. official issuance of Revision 2.

BNL-5 While the NUREG/CR-6645,Appendix F, procedure fordetermining ƒ2 has been includedas Appendix B in DG-1127, thereis no mention that ƒ2 is a functionof the damping ratio. In using thisprocedure, a unique ƒ2 needs tobe calculated for each dampingratio of interest. The numericalexample in NUREG/CR-6645,Appendix F, illustrates this.


The staff revised Appendix Bto DG-1127 to indicatethe dependance of f 2on the damping ratio.

BNL-6 As the author of NUREG/CR-6645, I find the editorial changesthat were made to the text ofNUREG/CR-6645, in developingDG-1127, generally to be a stepbackwards in clarity ofpresentation and readability. Perhaps an independentassessment of both presentationsby typical users of RG 1.92 wouldbe useful.


As appropriate, the staffmodified the text ofDG-1127 to improve itsclarity and readability. Nonetheless, the authorof NUREG/CR-6645understandably believes hisoriginal presentation isbetter. No independentassessment wasconducted.

W-1 Lindley-Yow’s method has alimited range of validity thatprecludes a consistent applicationto all structures and componentsof a given plant or within a givenproject scope.

Addressed by Response E in the NRC’s formalresponse to Westinghouse, dated November 16,2005:

The staff revised DG-1127,consistent with the formalresponse to Westinghouse.

In particular, the staffrevised Regulatory PositionC.1.3.2 and addedRegulatory PositionsC.1.4.2 and C.1.5.2.


-11-

W-2 Gupta’s method requires thedetermination of two attributes foreach response spectrum, namelythe “highest significant frequency”fl and the so-called “rigidfrequency” f2 that “may not beuniquely determined’ (as stated inpage 12 of DG-1127). The authorhimself has proposed andpublished several versions of thebest fit equations and formulas tocalculate approximate values ofthe fl and f2 bounds (see Ref. 6).Itwould be very difficult for anon-expert user to apply themethod in a consistent andreliable manner in severalpractical cases listed hereafter.

The staff revised DG-1127,consistent with Response K.

The staff also revisedRegulatory


-12-


-13-

W-3 Similar reason is already invokedin the DG-1127 itself foreliminating the influence ofanother physical parameter of theseismic input, namely the finiteduration of the strong motionmentioned in Section 1.1.2-(2)(tD in previous version of the draftRevision 2 - DG-1108) because‘the duration value was oftenarbitrary selected.’

The NRC’s formal response to Westinghouse,dated November 16, 2005, noted that the staffhas revised Regulatory Position C.1.1.2 to restorethe original Rosenblueth equation, as presentedin DG-1108.

The staff revisedRegulatory Position C.1.1.2to restore the originalRosenblueth equation,as presented in DG-1108.


-14-

W-4 The use of refined methods forcalculating the response in themedium-to-high frequency rangeseems in contradiction with thesimplification of the responsespectrum check in thepost-earthquake evaluationcriteria of Regulatory Guide 1.166(Ref. 2). The Regulatory positionC 4.1 - Response spectrum check-requires to compare the realspectral accelerations with theOBE response spectrumaccelerations only between 2 and10 Hz, and ignores the spectralaccelerations above 10 Hz thatare considered as non damagingto the nuclear power plantstructures and components. Theimplementation of the proposedmethods will also requiresignificant modifications to theexisting computer codes used forthe seismic analysis of structures,components and piping.

While the programming of newalgorithmic formulas is a one-timeeffort, the definition of newattributes of the response spectrainput will have a repetitive impacton the cost of new projects andwill introduce a degree ofuncertainty in the use of existing

Addressed by the introductory paragraphsin the NRC’s formal response to Westinghouse,dated November 16, 2005:

None.


-15-

spectra input in analysis revisionsor plant upgrade projects

W-5 Addressed by the introductory paragraphs;Responses B, C, D, and K; and closing paragraphin the NRC’s formal response to Westinghouse,dated November 16, 2005:

(Introductory Paragraphs)See W-4 comment for text of the introductoryparagraphs.

(Responses B, C, D)See W-2 comment for text of Responses B, C,and D.

(Response K)

(Closing Paragraph)A licensee or applicant still has the option to utilizetechniques that are accepted in Revision 1of RG1.92, with the single addition of the missing-mass effect. This may be the appropriate approachfor any new analysis using old design bases. As previously noted, Revision 2 is “forward looking.” Some of the concerns Westinghouse has raisedshould not be as significant for new design.

The staff revised DG-1127,consistent with Response K.

The staff also revisedRegulatory


-16-


-17-

W-6 Algebraic combination has beenimplemented in the three pipinganalysis codes used in the pastten years by Westinghouse inseveral projects of seismicrequalification or upgrades ofnuclear plants in Europeancountries, including Switzerland,Spain and Slovenia. It uses analgebraic white-noise modalcombination rule (GAC)developed by P. Mertens (Ref. 7)that has been reviewed andapproved by the Swiss Safety

None.


-18-

Authority (Ref. 8). That method islisted in NUREG/CR-6645 asreference 5, but not evaluated. Ityields coupling factors similar toDer Kiureghian’s CQC methodwith a damping dependentfrequency range for closelyspaced modes. Some resultsobtained with that method arepresented and discussed in apaper on snubber eliminationproject in a Spanish nuclearpower plant (Ref. 9).

W-7

The staff has revised the figures and Regulatory


-19-

For the case of unbroadened, multi-peakedresponse spectra, f1 is the highest frequency atwhich a spectral peak occurs. After broadeningof the spectral peak, f1 is the highest-frequencypoint on the broadened plateau.

W-8

See W-2 comment for text of responses B, C,and D.

None.

W-9 None.


-20-

W-10 None.


-21-

W-11 None

W-12 None.


-22-