ML19260B213
| ML19260B213 | |
| Person / Time | |
|---|---|
| Site: | Dresden  |
| Issue date: | 10/26/1979 |
| From: | FUGRO, INC. |
| To: | |
| Shared Package | |
| ML17174A225 | List: |
| References | |
| NUDOCS 7912070420 | |
| Download: ML19260B213 (7) | |
Text
.
CRITIQUE OF LLL/ TERA REPORTS ON SEISMIC HAZARD ANALYSIS Prepared for:
NUCLEAR SERVICES CORPORATION 1700 Dell Avenue Campbell, California 95008 i
Prepared by:
FUGRO, INC.
Consulting Engineers and Geologists 3777 Long Beach Boulevard Long Beach, California 90807 October 26, 1979 1514 156 ren oro
.//20
INTRODUCTION At the request of Commonwealth Edison and Nuclear Services Corporation, Fugro has reviewed the following three draft reports prepared by Lawrence Livermore Laboratory (LLL) and Tera Corporation on Seismic Hazard Analysis:
o
" Seismic Hazard Analysis:
A Methodology for the Eastern United States," Tera Corporation, August 23, 1979.
o
" Seismic Hazard Analysis:
Solicitatio.' of Expert Opinion,"
Tera Corporation, August 23, 1979, o
" Seismic Hazard Analysis:
Site Specific Response Sectra Results," by D.
L.
Bernreuter (LLL), C.
P.
Mortgat (Tera) and L.
H. Wight (Tera), August 23, 1979.
These reports were submitttd to the Nuclear Regulatory Commission (NRC) as part of their Safety Evaluation Program (SEP) of older nuclear power plants.
This document presents a critique of these reports.
Because of the large amount of material contained in these reports and the limited amount of time available for review, the comments below are mostly general in nature.
CRITIQUE General.
The three LLL/ Tera reports basically present:
(1) a 1
probabilistic model for estimating uniform risk spectra at a site, (2) the solicitation of expert opinions an-2 the interpre-and incorporation of these opinions in the probabilistic model, 1 Uniform risk spectra are response spectra whose ordinates have the same probability of being exceeded during some time period.
1514 157
2 and (3) the results of the application of this model to nine nuclear power plant sites included in the NRC's SEP.
The results for each site consist of a set of uniform risk spectra, each spectrum derived using one expert's opinions, and an overall spectrum representing the synthesis of all expert's opinions.
Comparisons are also made between the synthesized spectrum and spectra obtained from more deterministic approaches.
The reports make no recommendations as to which spectra are to be used in the SEP or the corresponding performance criteria of structural components and local soils.
However, some discussion is given on the merits and limitations of the use of each type of spectrum.
An important criticism of uniform risk spectra that was mentioned is that the spectra essentially represent the contribution of all possible earthquakes in the site region..
On the other hand the facility has to resist the shaking from only one earthquake at a time.
Therefore, the uniform risk spectra may be overly conservative in this regard, and other types of spectra may be more appropriate for the SEP.
The critiques of each of the three reports are given below.
Coi. ion.
Ten experts in the fields of Solicitation of Expert i
seismology and tectonics responded to specific questions in these areas in order to provide the necessary inputs to the probabilistic model.
Ideally, the experts should be very knowledgeable in all the problem areas and ha,ve plenty of time to seriously study the questions and all of the relevant data 1514 158
.C...,
3 and respond in an unbiased and objective manner.
However, for various reasons, it appears that these ideals were not achieved.
The experts generally were ceismologists with some expertise in tectonics.
To cur knowledge none of the experts has done extensive research on the attenuation of strong ground motion or the response of Jocal soils during earthquake shaking, two important considerations in any site-specific analysis.
The limited expertise in the latter area was evidenced by the response to Question 4-20 (p.11-106), which asked what was the maximum acceleration that various types of soil deposits could sustain.
Only three experts responded and their answers indicated a limited knowledge in this area.
Some of the experts emphacized their low confidence in their answers.
Even within their fields of expertise, the experts are usually most knowl-i edgeable about the seismicity and tectonics in the region in which they are located.
In this regard it appeared that the experts concentrated their efforts on the data they were most familiar with and spent little time with other data.
Sometimes only two or three experts rcsponded to a particular question or group of questions.
More feedback between the experts and LLL/ Tera would have been desireable during the course of the study.
The experts did not have a chance to judge the reasonableness of the LLL/ Tera interpretations of their responses to the questions, nor did the experts review the methodology or the results of the Tera /LLL probabilistic analyses for each site.
The experts may well have 1514 159
4 modified their responses to some questions if they had understood the way in which those responses were going to be used.
The method of self-ranking of the experts' opinions and the inclusion of the information by LLL/ Tera to weight all of the experts' opiriions is questionable.
The resulting synthesized spectrum from the application of the weighting technique is biased toward experts who have a high degree of confidence in their opinions.
A high self-confidence in one's answer does not necessarily mean that that answer is likely to be correct.
Conversely, a lower self-confidence ranking does not necessarily reflect a lesser understanding of the problem.
Some type of cross-ranking between the experts might have established whose opinions were most respected.
Methodology for the Eastern United States.
The major concern 4
with the methodology was the treatment of the uncertainties associated with input parameters of the probabilistic model, and the impact of these uncertainties of the results.
A detailed analysis of the uncertainties in order to fully evaluate their impact was not given.
However, LLL/ Tera did question whether or not the rather large uncertainties in the attenuation relations based on statistical analysis of empirical data and the uncertain-ties associated with the expert's opinion were accurate represen-tations of reality.
Limited sensitivity studies performed on the attenuation uncertainty showed in one particular example that an 80 percent increase in the standard deviation had a similar increase in the spectral level for a given return period.
Although 1514 160
5 the uncertainty in the final probabilistic distribution for the response spectral ordinates was not estimated, it is important to know if the model is combining the uncertainties of the input parameters in such a way so as to produce a grossly overestimated uncertainty in the final distribution.
If the model is doing this, then at long return periods the corresponding uniform risk spectra will be overestimated.
Site Specific Response Spec'ra Results.
The reasonableness of the results could be better evaluated if a more complete docu-mentation were given on the effects the various input parameters had on the results.
For example, the relative contributions of the various earthquake magnitudes, in addition to the seismic sources, would have been.useful.
More analysis of the effects of the uncertainties of the input parameters, through formal statistical methods and sensitivity studies, would also be useful.
A review of the results for the Dresden nuclear power plant was made based on our knowledge of the seismicity and tectonics of that site region.
The LLL/ Tera results showed that the New Madrid seismic zone contributed anywhere from 35 to 47 percent to the 1000-year peak ground acceleration while the Central Stable Region seismic source contributed 45 to 61 percent.
The relative contribution from the New Madrid zone seems unreasonably high simply because the zone is 300 to 500 km from the site (depending on the expert's choice of the appropriate northern boundary) at its closest approach.
Expected peak ground a'ccelerations at these distances from an earthquake of epicentral intensity equal 1514 161
6 to XII (MM-) should be less than 0.01g according to the formula given on Table 5-5 (p. 5-21) of the LLL/ Tera report, " Site Specific Response Spectra Results. "
The report offers no concrete explanations or presents any analysis to explain why the New Madrid seismic zone should contribute so heavily to the results and, until this is investigated, the reslts for Dresden should be interpreted with reservations.
i P
1514 162
-