ML20011A446

Applicant Evaluation-Luco Rept.

ML20011A446
Person / Time
Site:	Summer
Issue date:	10/31/1981
From:	SOUTH CAROLINA ELECTRIC & GAS CO.
To:
Shared Package
ML20011A444	List: ML20011A443 ML20011A446
References
NUDOCS 8110130362
Download: ML20011A446 (19)

Text

_. ._ .- .__ . _ _ . __

4 4

APPLICANT EVALUATION LUC 0 REPORT 4

l l

VIRGIL C. SUMMER NUCLEAR STATION DOCKET No. 50/395 SOUTH CAROLINA ELECTRIC & GAS COMPANY OCTOBER, 19El i

i-gOQh[oh p A

APPLICANT EVALUATION OF LUC 0 REPORT ON VIRGIL C. SUMMER NUCLEAR STATION SEISMI'ITY STUDIES Prof. J. Enrique Luco has reviewed the " Supplemental Seismologic Inves-tigation",' including Appendix XI, portions of the FSAR (361.13, 361.17.4, 361.21) and portions of the Safety Evaluation Report. His response is contained la a report entitled, " Comments on Estimates of Scrong Ground Motion for the V. C. Summer Nuclear Station, Unit I", dated September 23, 1981. The issues raised by Luco result from misinterpretations of the studies which have been performed by the Applicant or from the use of incorrect parameter values in his analyses. This deserves a direct response; the form of the response follows the issues raised by Luco, in order.

ON THE HANKS - MCGUIRE METHOD TO ESTIMATE PEAK ACCELERATION Luco 1.s correct ia pointing out that in the usual characterization of earthquakes via the Brune model (which is done through observations of spectral amplitudes in the frequency domain), stress drops vary greatly and corner frequency and spectral decay at high frequencies are the subject of current discussions. However, the Applicant is not using the Brune model in this usual, frequency-domain application but in the method propostu 'oy Hanks and McGuire (1981). This method (which uses observa-j ti .ans of ground accelerations in the time domain) provides remarkably stable 1stimates of stress drops for past earthquakes (in fact, this is one of the major points of Hanks and McGuire, 1981). That this is the case is recognized later in Luco's report when he states, "The stress drop paramet.er appearing in the estimate of peak acceleration obtained by Hanks and McGuire has no relation with the stress drop determined by standard seismologic methods. In particular, Hanks and McGuire found that the peak accelerations for events in California could be approx-imated by a constant stress drop of 100 bars, independent of the stress i

drops calculated for these events by standard seismological methods."

Hanks and McGuire also point out that, regardless of the accuracy or 1

E

-- -ww-, - . -..---,y ~-..,.-,-v,--.~,,-r.----e+.. -,.-c.,-.-- ..--.-e _ _ - - - - . - - -

insecuracy of this methodolcgy in characterizing corner frequency and high frequency spectral decay, the model works in predicting both root-mean-square (r=s) and peak acceleration. Thus Luco's concerns about uncertainties in str'ess drop, corner frequency, and spectral decay are applicable to frequercy domain methods, not to the Applicants' time domain method. Applicants' method does not lead to estimates of peak acceleration which are highly uncertain, as Luco implies, but rather leads to peak acceleration estimates with confidence as high as available by using other state-of-the-art methods.

ESTIMATES OF STRESS Dr.0P Luco implies that stress drops estimated (in the frequency domain) by standard seismological methods and presented by the Applicant are irrelevant. This is not the case. The Applicant has presented such data in Appendix VII of the Supplemental Seismologic Investigat. ion to give as complete a picture as possible about the data which have been gathered at the site. In the context in which Luco views these data (that of the experience of Hanks and McGuire with California data), the standard stress drop data presented in Appendix VII are entirely consistent. In California, standard stress drop values range from 6 to 140 bars, and the values appropriate for ras acceleration estimates is 100 bars; at Monticello, standard stress drop values range from 1 to 5 bars and the value appropriate for rms acceleration is 25 bars.

The derivation of rms acceleration a by Luco is slightly differ-ent from that of the Applicant because Luco explicitly ine.ludes the term (1 + (f,/f)2)-1 in the integral, whereas the Applicant does not. The Applicant's derivation, which was und to calculate values of a and peak acceleration presented to the ACRS Seismic Subcommittee on February 26, 1981, is given in the attached Appendix. To be sure, the term (1+(f /f) )~

9 appears in equations in the Applicants' FSAR section 361.17.4, but the question of whether more accuracy is gained by discarding the term and integrating from f=f , or including the term and integrating from f=o, is moot:

the available spectra from Monticello can be fit either way with equal accuracy. The more important point is that it makes little difference

to conclusions gained by comparison of estimates to data at high digi-tization rates (e.g. 500 points per second, which are presumably most accurate) where fu=40 or 50 hz is appropriate. This is shown in the attach-ed Appendix (a typical effect on stress drop estimates for these records is 15%). At lower digitization rates there is more effect which accounts in part for the results Luco obtains in comparison to data presented by the Applicant in FSAR Table 361.17.4-1. In any event, conclusions from FSAR Table 36'. 17.4-1 are obsolete: relocation of the August 27, 1978 earthquake now indicates a source-to-site distance of 0.67 km rather than 0.8 km, and digitization at 500 points per-second have brcome available. A more enlightened conclusion is obtained by comparison of predictions with records digitized at 500 points per second. Table I shows the appro-priate parameters, observations, and predictions made by the Applicant that were presented at the ACRS Subcommittee meeting February 26, 1981.

It also shows estimates made by Luco's equation (2) which indicate that a stress drop of 26 bars explains the observations (using Luco's preferred value of f =50 u hz) . Thus the analysis sad equations developed by Luco fully support the stress drop value of 25 bars used by the Applicant for the most recent accurate data available on the August 27, 1978 earth-quake. The value of 100 bars obtained by Luco in his report is based on an erroneous source-to-site distance (calculated from preliminary depth estimate) of 1.6 km; all investigators familiar with the data includ-ing USGS (Fletcher, personal communication,1981) now agree that a source-to-site distance of about 0.67 km (as used in Table 1) is accurate.

Luco's calculation of stress drop for earthquakes at Hsinfengkiang Reservoir (his Table 2) is incorrect on two counts, thus rendering his con-clusions invalid. First, Luco uses an upper frequency of 20 hz, whereas the Chinese strong motion instruments provide linear response up to 35 hz (the Applicant states this in its Appendix XI). Thus f u=35 hz or I greater would be more appropriate. Second, Luco used surface-wave magnitude M, in place of local magnitude g . For the Chinese data j (M

L = 3 to 5 ) , M, is less than g by about one unit. As a result, smaller source sizes and larger stress dror: are obtained than is correct l 1

1

.. ~4 for thaso dete. In cny ccao, tha important ras.nlt from tha Chin 2co data is that stress drops determined from peak acceleration do not increase with magnitude; the Applicant made this point in Appendix XI, and Luco apparently agrees with this conclusion. One further point is that the seismological stress drop calculated for the M,=6.1 main shock. of Hsinfengkiang was 7.5 bars (Sheng et al., 1973), a value which is not inconsistent (given the above discussion) with the ras acceleration stress drops reported by the Applicant in Appendix XI.

Luco conclades that a stress drop of 150 bars is appropriate. Not onl:, is this view unsupported by data, it is contradicted by data at Monticello, at Hsinfengkiang, and in California. For the firs t two locations, stress drops less than 25 bars are indicated; California data .

are irrelevant to the issue of very shallow induced seismicity and, in any case, indicate a stress drop for ras accelerations of 100 bars. Luco has presented no data which indicate that a stress drop of 150 bars is appropriate to use with the Hanks and McGuire method.

ESTIMATES OF PEAK GROUND ACCELERATION The peak acceleration values shown by Luco in his Table 3 are invalid for the Virgil C. Summer Nuclear Station because they are based on a 100 bar stress drop. It is not surprising that Luco's Table 3 values agrees with the equations of Joyner et al. (1981) at R=7.3 km (zero epicentral distance)* because these equations are bas d on Californic data and Hanks and McGuire have shown that 100 bars e appropriate for Cali-fornia earthquakes.

Data shown by Luco in his Table 4 and his Figure 1 are misleading.

He states, that "This sample may be biased towards the largest peak accelerations," (emphasis added), but in fact the sample g biased. For the Oroville data which Luco finds of particular interest, the average of the larger peak accelerations on each record for 4.0 < g < 5 is 382 _

cm/sec , whereas, the mean of the Oroville af tershock peak accelera-tiona on bedrock sites for the same magnitude range is 164 cm/sec (Seekins and Hanks, 1978). Thus th= data presented by Luco are very much It appears that Luco has misinterpreted the meaning of the parameter R=7.3 km used by Joyner et al (1981); this le not a depth estimate, so that R is not hypocentral distance as Luco states. Joyner et al (1981) simply use constant R as a parameter to fit their data.

l- . _ - - - - . - - . ~ . - - - - -- - -- -- - - - - - -~ ~ '~"

biased toward higher accelerations and should not be used to determine peak acceleratior levels. Further, the Oroville af tershocks are charac-terized by an ras-acceleration determined stress drop of 100 bars which the Applicant has shown is inappropriate for Monticello Reservoir earth-quakes.

VERTICAL PEAK ACCELERATIONS l

The Applicant agrees with Luco's observation that vertical peak accel- '

erations are generally less than horizontal peak accelerations during earth-quakes of magnitude less than 6. Data supporting this have been presented by the Applicant in Section 361.17.4 of the FSAR.

ROCK VERSUS SOIL SITES The SMA instrument is located on the abutment between Monticello Dams B and C. An examination of the topography of this region indicates that the instrument site is located a1.most at the top of a hillock that is partly man-made and partly natural. The SMA recording, in all likeli-hood, represents amplification of the motion of the hillock relative to motion that would be observed in the free-field at either a soil or rock site. Nonetheless, the Applicant has conservatively assumed that no such amplification has occurred in its use of the SMA recording of the A2 gust 27, 1978, earthquake to evaluate seismic source parameters. The Applicant maintains, however, that the accelerograph records are not' strictly repre-sentative of free-field motion, a distinction that Luco fails to draw.

In the free-field, and for short epicentral distances, peak ground accelerations are comparable for rock and soil sites (Campbell, 1981; Joyner and Boore,'1981).

RESPONSE SPECTRUM AT FOUNDATION LEVEL Luco states that it is not appropriate to compare the 5% and 7% SSE i spectra with the 2% M =4.5 spectrum to study the effects on equipment at the louer levels of the plant. This statement (which points to the lack of effect of structural damping for fcunlation equipment) would be true if a fixed base model were used in the analysis. Since foundation compliance was taken into account in the soil structure interaction analysis and the base mat response was amplified 10% relative to the input motion, it is appropriate to compare the 5% and 7% SSE spectra with the 2% M =4.5 spectrum for the effects on equipmant at the lower levels of the plant.

. The conclusions reached by Luco regarding the level of conservatism of response spectra are incorrect. The velocity amplification factor used by Luco (a value of 1.9) is in fact a mean-plus-one standard-deviation (mean + o) amplification factor, not a mean factor. This is apparent from comparisons with Regulatory Guide 1.60 spectral amplification factors and with the (mean + a) amplification factor deve-loped by the Appifcant for the velocity range (see Table 3). Thus Luco's pseudo-velocity spectral amplitude for 5% damping of 0.29 ft./sec. is a (mean + o) amplitude, not a mean amplitude.

Further documentation of the Applicant's methodology is provided as follows. The response spectra developed to represent vibratory ground motion from reservoir-induced earthquakes at Monticello Reservoir were derived following requirements indicated in Reg. Guide 1. 60, "Desiga Response Spectra for Seismic Design of Nuclear Power Plants." Specifically, Reg. Guide 1.60 staces that the standard design response spectrum procedure ". . .does not apply to sites which (1) are relativtly close to the epicenter of an expected earthquake or (2) which have physical charactcr-istics that could significantly affect the spectral combination of input motion. The Design P.esponse Spectra for such sites should be developed on a case-by-case basis." The Virgil C. Summer Nuclear Station would be close to the epicenter of any reservoir-induced seismicity of concern; hence site-specifie response spectra were developed to represent grcund motion for these events.

This procedure consisted of using response spectrum shapes for earthquake ground motions recorded at magnitudes, distances, and site.

conditions representative of reservoir-induced earthquakes at the Virgil C. Summer facility. These respense spectrum shapes, for magnitutes in the range of interest, were than compared to other available data to ensure their applicability.

The shapes for th*.;e spectra were ta. ken from the publication of Johnson and Traubenik (1978). These spactral shspes reprasent ground motions based on records obtained on rock sites for earthquakes with magnitudes (Q) between 4.7 and 6.5, with source-to-site distances of less than 20 kilometers. The derived spectra for 5 percent de.mping for g = 4.0, 4.5, and 5.3 events scaled to 0.15 g peak acceleration a:w

7-labeled "RIS" in Figure 1. These are mean + c spectra, based on the amplification factors reported by Johnson and Traubenik (1973). Use of the mean + o spectrum is consistent with the procedure defined as accep-table for standard design response spectra in Reg. Guide 1.60.

Also shown in Figure 1 are the Reg. Guide 1.60 spectrum for 5 percent damping, and the Virgil C. Summer Nuclear Station SSE spec-trum for 5 percent damping, both scaled to 0.15 g acceleration (the SSE ar.celeration at the facility). It la apparent that the derived RIS spectra generally match both the Virgil C. Summer spectrum and the RG 1.60 spectrum at the highest frequencies, but deviate at intermediate and low frequencies, the extent depending on both the earthquake magni-tude and the frequency of interest. The reason for this deviation is that broad-banded design spectra typically represent ground motions for earthquakes of magnitude around 6-1/2 (they are derived from recorded ground r9tions during seismic events with an average magnitude of 6-1/2).

The RIS spectra, on the other hand, logically reflect the lack of intermediate and low frequency energy which will be generated during magnitude 4.0 to 5.3 earthquakes with sm 11 source to site distances.

Two steps are required to generate site-specific spectra of the type shown in Figure 1, and in comparing these spectra to other results available it is convenient to break the comparison into these two steps.

The first step is the estimation of a peak velocity and a peak displace-ment which are consistent with the peak acceleration of the earthquake of interest. In the present application, the peak velocity-to-acceleration ratio is most critical because it determines the upper corner frequency of the spectrum. The peak displacement is not important in the present application because the Virgil C. Summer SSE spectrum greatly exceeds the RIS spectra in the displacement-controlled region (at lower frequencies).

To evaluate the peak velocities derived by Johnson and Traubenik (1978), we compare them to results derived from other studies. Table 2 v'-e'- -e-W v'w-'-- -"----*e g' 8P "" - * - ' " " ' irr "----r =w- '

@--"-* - - w'--' *'9"'m' *-=*u"* ' ' -' '

1. 7 ri