ML19345B964
| ML19345B964 | |
| Person / Time | |
|---|---|
| Site: | San Onofre  |
| Issue date: | 11/05/1980 |
| From: | Bernreuter D LAWRENCE LIVERMORE NATIONAL LABORATORY |
| To: | |
| References | |
| NUDOCS 8012030379 | |
| Download: ML19345B964 (6) | |
Text
,
4
[Q 4
Review of the Reports Dealing with the Simulaticn of Earthquake Ground Moticns for San Onofre Nuclear Generators Staticn Unit I for the Nuclear Regulatory Camissicn.
by
)
D. L. Bernreuter, Leader Engineering Geoscien s Group At the outset, 'I would like to reafirm my overall support for the general approach usel by 'IERA-DELM in attenpting to estimate the strcng ground motion characteristics at the San Onofre site. I believe that the approach is useful and sheds light cn a nunber of important questicns. A number of negative coments follow. Dese arise because 'IERRA-DELTR is attenpting to solve a very difficult problem - a problem for whicts most (if not all) researchers cgree that we cb not yet even understand the basic physics of the earthquake ~
rupture process much less know hcw to solve the problem. TERA-min'A has attempted to overecme our lack of knowledge by use of a simplified fault vrtura process and the introdteticn of randmness. It is interesting to note ecch of ne " improved" models developed by 'IERA-DELTA that are of the major iaprovements was the introducticm of more 'randeness. This seems counter-productive to the basic gcal of a calculational effort. In my view of the usefulness of a calculational effort lies in our ability to say that we c.derstand therbasic Ehysics of the problem, that our model incorporates in a reasonably correct way the basic important parameters and physics. Then by.
bcurx31ng the various parameters we can calculate reasonable worst case results apprcpriate for - in our case - the earthquake in the posulate offshores zone cf deformaticn at the San Onofre site.
I accept that there is censiderable randemness in the earthquake fan 21 ting process. However, the continued introdtx:ticn of more and more randmness in place of understanding the #1ysics of the problem is very troublesme - thus I cannot agree that the model has been properly calibrated.
In fact cne key problem that I have with 'IERA-DELTA's wrk is over the criteria they used to argue that they have adequately calibrated their model.
They argue that approximate equality. of peak acceleraticn and "scxne match" to mocthed response spectra is more than adequate to shcs that their model cniservatively models the possible earthquakes at the San Onofre site.
I, on the other hand, feel that scme match of the more important phases of the time series are necessary before we can argue that the maSel used is reasonable.
If, for example as I strcngly feel is the case, the initial slip velocity behin3 the rupture front is highly variable rather than uniform as modeled by_
'ILtA-DEUIA, then it is inportant to know where the ene gy which gives rise to,
the peak ground moticn originates before cne can use sech a match to judge what values of slip velocity are reasonable. This can only be done by matching the important features of time series.
The next report by Day, (Ref.1.), shcus that the introducticn of ~
variaticns in tectonic stress along the fault has a very significant effect cn 801203.0N7 f
both rupture velocity and slip velocity. This report by Day also shows that (at least fcr one reasonable mnrk1) that the slip velocity is a function of the stress drop and introduces rapid changes in rupture and slip velocity.
These rapid changes in rupture and slip velocities could have significant effect cn the time series of the ground acceleraticn and on the high frequency end of the respcnse spectra.
b In my past reports (References 2 and 3), I also indicated that in my view l
the major deficiency of the 'IERA-DELM's reports is that they have not esFahlished c:nservative bounds for the key parameters of their utdel j
a @ropriate for the postulated SSE for the San Onofre site. 'IERA-DELTA's sttxiles show that the really important parameters of model that control the higher frequency ground moticn ($z) are:
)
1.
Vo = initial slip velocity (dynamic stress drop) 2..
Rupture velocity 3.
Both the micro-incoherence and the maraS-rardinness introduced.
4.
Q values used
~
~
5.
Geologic structure 4
l Not included in the 'IERA-DELM study, but in my view of ccmsiderable inportance, is the ocnlinear behavice of the soils at the various sites used for calibraticn puposes..
I In nry M reports I discussed why I-felt that 'IERA-DELM's calibraticn of -
Vo was inadequate. My views are still much the same as I do not feel that Susplements II and III address my prime concerns, which were with the modeling j
of Vo, Q and the randcznness introduced into the model. We study by ray (Ref 1) serves to underline these concerns and introduces a new concern which deals with law rupture. velocity was modeled. My understanding of the
'IERA-DELM report is that the rupture velocity is always less than the shear wave velocity. 'Ihe time of rtpture was randcaly chosen, but is always slower than 0.9 $ ( p=. shear wave velocity).. The study by day and the' work of DAS and Aki suggest that the rtpture velocity can be larger 'than the shear wave-velocity - in fact in Day's study it generally was larger. In addition, we can expect randcza variaticn in rupture time, but at least Day's study suggests that as the rupture grows the rupture velocity grows. This potential correlaticn does rot seem to be included in the 'IERA-DEL'IA model.
Of potential inportance is the fact that the slip functicn studied is always a menber of the same family of functions and very smooth except at the start an! stopping of the rupture. The slip velocities mlmlated by Day show additicnal character which could have significant influence cn the high frequency content of the' spectrum. The. potential inportance of such variaticns in the slip functicn seens to need study. All of the randemness introduced by 'IERA-DELM may cover tlw ranges of variation in the slip I
l l
functicn that might be postulated including different functicnal form of the slip function, but this is not at all evident rcan the sensitivity studies presented.
In Supplements II and III TERA-DELTA attuff,ted to provide added verificaticn for their choice of Vo = 800 cgse. In Supplement II the Lcng I-Beach earthquake of 1933 and the San Fernando earthquake were modeled, and in Supplement III the recent Imperial Valley earthquake was modeled.
The first problem that I have with the Lcng Beach earthquake modeling is that the resultant time series is shorter than the recorded time series. I carmented cn this point with regard to 1 ERA-DELTA's modeling of the 1940 Imperial Valley record at El Centro. This shortness could ccxne frca the Q's used and cr lack of ccuplexity in the rupture model such as variaticns of Vo, starting and stoj; ping of the rupture etc. In additicn, the modeling seems to give too high a peak -leratien. It is hard, to make jirigements abcut the mcdeling because of the late start of the instruments and lack of cther studies to pin down scxne of the important parameters of t' rupture process.
The San Fernando earthquake has received ocnsiderable study'.
Many of the studies of this event suggest that the rupture process was highly variable with the most energetic.part of the rupture occurring cn the laser part of the. fault. The only way to check to see if in fact the uniform modeling used by TERA-DELTA provides a calibraticn for Vo is to be able to determine if higher values of Vo should have been used cn the lower fault ard icwer values cn the u);per fault under the recording isite.- We~ also need to -
assess what impact this might have cn the TERA-DELTA model. However, as we do not have the can];uted time series to ccurpare with the recorded data it is very difficult to assess how reasonable the mcdel is. In additicn, it is very hard to assess the correctness of the topographic amplificaticn factor used. The 1
problem is cnce again related to a questicn. of needing a crapariscn of where the ene gy is coming frcan for both the model and the recorded data.
The Imperial Valley earthquake of 1979 provides a number of staticms as well as staticns with obsolete time which could be used to calibrate _
7 ERA-DELTA's model. They have made lfttle use of such data and cnce again dse a unifcxm stress drop mcdel with ~ranckxn rupture velocity and other randcza parameters. As I discussed in Ref. 4 the locaticn of the center of energy release could be a number of kilcmeters frczn the El Centro array. Also as I~
pointed out in Ref. 4 there is some evidence that a major barrier could have existed several kilcmeters south of El Centro array. The TERA-DELTA model
.ould put the effective center of every release much closer to the El Centro array. It seems to me that this can, be resolved because we know when in time the energy arrived. The initial wave shapes should alsa provide additicnal insight into the correctness, of any model.
I fird it noteworthy to crntrast the difference in duration between the records recorded cn the El Centro array ard by the Bonds Corner Staticn.
Also, the Mexican is much longer duraticn (like Bands Corner) than that
recorded by the El Centro array. This indicates to me the rupture process was very complex ard ncnuniform. Once again this could only be resolved by amparison of wave shapes and arrival time (which for the first time are available).
One of the imiutant questicns that the 'IERA-DELTA Jeports attempts to address is the focusing of the seismic energy towards the site. Ccnsiderable y_
randcnness is introduced in the model to reduce the focusing effects. Day raises one interesting objecticn to the 'IERA-DELTA model on Page 30 of this report. Also, the nonliner behavice of the soil is of possible importance when ccx: paring calculated linear results to recorded results. Without s:xne assessmelt of the nonlinear effect and ccanparisen of wave shape, type and arrival time I find it impossible to judge hcw real and necessary both the micro and marco randcaness used are. In a3dition, the ccanputed spectra for Imperial Valley seems to lack energy in the pericd range of 1 to 4 seconds.
These larger period waves seem to show upcn the recorded accelerograms at a number of staticns suggesting that there may be.large coherent zones of rupture. 'Ihe potential implicaticn of this is that there might be too much -
)
ranchnness in the 'IERA-DELTA model - or the scale' (1 km) is Mo small for scxne zcnes of the fault. 'Ito bad 'IERA-DELTA didn't compute the M, frcan recorded moticn data to cranpare to the Mn'~s of their simulated earthqua. The M 's of the simulated earthquake appear to be low suggesticns that more L
1 see wave energy is needed.
~
'Ihe geologic strix:ture and Q's are impcetant parameters. For example, Fig. 4-13 of May,1978 shows a censiderable variaticm (factor of 2 or more) in' the ccanputed spectra as a functicn of geologic structure for several different sites. The importance of Q is hard to determine as. to sane extent _the value of Vo was chosen relative to the Q model used. But what isn't clear is how much the Q model might effect the spectrtxn for lcnger and larger fault _ rupture sequences. If irdeed Q is independent of frequency, then changes in Q would most likely be a second order effect relative to same of the other concerns discussed above. On the other hand,.0 is not independent of frequency but increases with frequency, then how Q is modeled could be important. The sensitivity sttriies do not really address the role and importance of
~
variaticns in Q.
In order. to provide added cernparisons TERA-DELTA used the data and a reg essicn analysis to determine how ~the ground moticn varies with distance.
In my view, the regressicn model given in Chapter 2 of Supplement 'III aise appears to use a questicn of metrics, of distance frcm fault trace and yet was an attenuaticn of 1/CRv20)l.75 for horizcntal acceleraticns and 1 CR+10)l.75 for vertical. Such a model might suggest that the main source
/of energy is at same depth. Very near the fault changes in distances of a few kilcmeters are very important. To simply use the closest distance to the fault trace can introduce cchsiderable.confusicn. This is discussed in scxne detail by Shakal (Ref. 5).
e --
In claing, I still feel that ERA-mmA has not yet properly calibrated their model. If the strcmg ground motion recorded at the various sites used for cmparism and calibraticn was che to:
(1) high stress drop earthquakes (2) the regicn of large stress drop was very near the recording site as modeled (3) nonlinear behavior of the soil is of seccnd order of iuportance (4) the scale of rand::mness used by ERA-DELM and the amount of randomness introduced to reduce focusing is appropriate then the results man play the role ERA-DELM wasn't there to play. On the other hand I feel -hat several of the earthquakes used wre not high stress drop earthquakes w that the zone of high stress drop was at sme distance from the recording site. I also suspect that the nonlinear behavior of the -
soil is inportant. - In addition, I think that the manner in which randcnness is introduced and its scale needs more careful calibraticn. Overall I think that to ac ptably calibrate their model, ERA-DELM must look at wave shapes.
and arrival times in order to address the above points.
If my analysis given in Ref. 4 (Table VII, 'which updates Table I of Ref.
2), is correct, then the May 1940 Imperiil Valley event for the event nearest El Centro, Parkfield, t51979 Imperial Valley and Coyote Lake earthquakes are' Icw stress drop earth W.es.
I applied the same analysis to the 1933 earthquake an3, also found it to be a low stress drop event of the same order as the other earthquakes. This cnly leaves San Fernando as a high' stress drop event used to calibrate the model. However, as discussed above there is-considerable questicn about the way ERA-DELTA modeled San Fernando. Ebr -
these reascns I feel that we must consider the results of ERRA-DELM's x
modeling to be mean values. I muld think that we could well expect a factor of 2 uncertainty at all frequencies including peak acceleration. This factor of 2 muld correspcnd to the cne sigma level. 'Ihis is assuming that high dynamic stress drop earthquake is possible in the offshore zone of dezenaticn. If cnly low dynamic stress drop event like Parkfield or Imperial Valley can occur then the Housner spectra at 0.679 seens reasonable in light of ERA--der,ges modeling.
g e
9 q,
e 4
\\
.i.
N 1.
Day, S. M., "Three-Dimensional Finite Difference Sinulaticm of Fault Dynamics." System, Sciona and Softwarte Report SES-R-80-4295. Deced+r,
1979.
2.
Bernreuter, D. L., " Review of Repcrts cm Simulation cf Ground Moticns for
,p the San Onofre Nuclear Generating Station - Unit I."
LINL Letter Report to NRC, December 4, 1979.
3.
Bernreuter, D. L., "Comnents cm the Report Simulaticm of Earthquake Groun3 Moticm for San Onofre Nuclear Generating Station - Unit I."
LINL Letter Report to NRC, Sept.auims 7, 1978.
4.
Bernreutar, D. L., " Scaling and Estimaticm cf Earthquake Groun3 Moticn as a Functica of the Earthquake Source Parameters and Distance." Draft Repcrt to NRC, August 14, 1980.
Shakal, A. " Analysis of Peak Ground at Near-Source Distance." Draft S.
Report to NRC, October, 1980.
~
223GJ
- =w=****
e e
O
=.
\\O e e e
- H e
e OO w '
- i
,