ML20151C902

Comments on 871021 Final Rept NSAC-113, Epri/Westinghouse Owners Group Analysis of DHR Risk at Point Beach, Per NRC Request.Resources Did Not Permit Exam of Plant,Review of Training Procedures & Emergency Guidelines

ML20151C902
Person / Time
Site:	Point Beach
Issue date:	01/11/1988
From:	Davis P AFFILIATION NOT ASSIGNED
To:	Boehnert P Advisory Committee on Reactor Safeguards
References
ACRS-CT-1911, NUDOCS 8804130256
Download: ML20151C902 (14)

Text

.

. P. R. Davis bb 1935 Sabin Dr.

Idaho Falls, ID Jan 11,1988 Mr Paul Boehnert Senior Staf f Engineer Advisorv Committee on Reactor Saf eguards SUBJECT- Review of 'EPRl/WOG Analysis of Decay Heat Demoval Risk at Dolnt Beach" (NSAC-113), Final Report, Oct. 21,1987.

Dear Paul,

Pursuant to voor request,I have reviewed the subject document and my ,

commer.ts and observat tons are transmitted herewith for information and use by you and the Sub-committee on Decay Heat Removal it should be noted at the outset that the review necessarily suf fers from some rather severe limitatfons in particular, resources did not permit an examination of the plant, a rev:ew of training procedures and emergency guidelines, or verification of much of the data and Information which is referred to but not elaborated upon in the EPRl/WOG study. This limitation was found to be particularly significant in the area of recovery actions which played an important role In the EpRl/WOG study Verification of the quantitative Impact of recovery actions generally requires detailed knowledge of the plant layout as well as training and operating procedures Thus, only an identification of some potentially questionable aspects of recovery could be included as part of the review.

In addition to the subject document, the review made considerable use of two related documents (References 1 & 2) for backup information.

In PRA reviews, I generally find it useful at the outset to Ident',fy and quantitatively rank the accident sequences which are dominant  ;

a: c0ntributors to CMP and risk. This provides valuable perspective in -

3S evaluating which assumptions, data, and other issues are important in '

g deriving the results. This approach is particularly useful when two PRAS  ;

5 are being compared, which is the case in this review (I note in this regard j k that the title of the subject document is somewhat of a misnomer in that j 1

EE- the EPRl/WOG is not, as stated on page 2-1, an Independent evaluation of DHR risk but rather a re-analysis of the NRC etfort, as provided in Ref.1,

(( and is subject to the same limitations and omissions in scopel i

884  !

m n.o l The dominant core melt sequences will be considered first. Table 1 gives l

.. .. the dominant sequences contributing to core melt probability for the two studies, with the sequences listed in decreasing probability from the NRC study. The table provides the ass @ssed probability of each sequence and the percent contribution of each to the overall CMP for the two studies.

The last column in the table provides the ratio of the NRC assessed '

probability to the EPRl/WOG probability for each sequence. This table was derived from inf ormation presented in the EPRl/WOG report.

Table 1 reveals that there are significant differences between the two studies. Not only is the CMP for the EPRl/WOG substantially lower (a f actor of 31, with a total CMP of 1.0E-5), but, except for seismic initiators, the ranking of the dominant sequences is considerably dif ferent. (For some unexplained reason, the Reference 3 review indicates, on pages 5 of the cover letter and 5-1 of the report, that the NRC and EPRl/WOG CMP resu ts for Point Beach are only a factor of 10 different)

For the EPRl/WOG Study, three of the NRC dominant sequences (Sequences 1.5, and 6) were either found to be negligible or, in the case .

of sequence '6, would not occur at all. The most significant difference between the two studies in terms of the secuence that contributes most to the change in core melt freauency is the internal flood sequence ("l). The EPRl/WOG assessment that the probability of this sequence is <E-8 accounts for almost 1/3 of the ? actor of 31 difference in the CMP between the two studies.

The two studies agree, however, in one general aspect. Both show significant contributions to CMP from "external' events. For the NRC study. 55f; of the CMP contribution comes from external events (sequences 1.2, and 5). For the EPRl/WOG study,74r. of the CMP is from external events, all from seismic initiated accidents. The next most significant sequence in the EPRl/WOG study (*7) would need to haw a probability increase of almost a factor of 10 to be an eaulvalent contributor.

It should be noted that a small part of the dif ference In the two results can be attributed to a design change to be made at the plant which was considered by EPRl/WOG but not considered in the NRC study. This design change is the installation of a back-up DC power supply in the form of seismically quallfled batteries. Based on results in the studies, it appears that consideration of these batterles in the NRC study would result in a reduction in the CMP from 3.lE-4 to 2.8E-4, producing a modest reduction in the dif farence in CMP between the two studies from 31 to 28. Two sequences would appear to be af fected by this change; the seismic secuence ("I in Table 1) and the loss of of f-site power sequence (*7). In the former case, it appears the NRC CMP from seismic initiators would be reduced from 6.lE-5 to 3.0E-5, which would reduce the difference between the two s;udies for seismic CMP from a f actor of 8 to a factor of 4.

TABLE 1- COMPARISON OF CMP DOMINANT SEQUENCES BETWEEN THE NRC AND EPRl/WOG STUDIES NRC EPRl/WOG NRC .

Seauence (designator) CMP  % Cont. CMP % Cont. EPRl/WOG ..,

l. Internal Flood 7.7E-5 25 <E-8 negl. >7700

2. Seismic 6. l E-5 20 7.4E-6 74 8

3. Small Loca, ECC recirc. 4 7E-5 16 5.8E-7 5.8 81 Iallure (S 7MH i 'H2')

4 Long term station black- 3.6E-5 12 5.4E-7 5.4 70 out (LTSB)

5. Internal Fire 3.2E-5 10 6.3E-8 negl. 500

6. Transtent LOCA w/fallure 2.5E-5 8 N/A* 0

• of ECC recire (T,0H,'H2')

7. LOOP with loss of feed- 6.7E-6 2 7.7E-7 8 9 water and F&B (T MLE) i

8. All other sequences 2 SE-5 8 6 SE-7 6.5 38 TOTALS 3. l E-4 100 1.0E-5 100 31 l

'Acmrding to the EPRl/WOG stud /, this secuen would not occur because tha transtents mnsidered in this sequen (reactor / turbine trip with feedwater and offsite power evallable) would not reach the rellef valve setpoints.

I i

. $1nce the major thrust of these studies is an evaluation of plant improvements on the basis of value/ Impact, an important aspect of the results is the relative ofI-site impact of the accident sequences. In this regard, the most significant accidents are those which result in early and gross failure of the containment function. According to Sect. 9 of the EPRl/WOG study, the exDetted person-rem /yr. for the EPRl/WOG study would be about a f actor of 30 less than NRC !f the NRC source term methodology were used in the EPRl/WOG study. This dif ference is essentially the same as the CMP dif ference between the two studies.

However, use of the EPRl/WOG base case source terms (an application of the IDCOR methodology) would result in a further reduction of a factor of about 8 between the studies based on Information given in Sect. 9. Thus, the base case results from the EPRl/WOG study on the basis of expected person-rem /yr. would be about a factor of 240 less than the NRC study.

Since both of the studies conclude that none of the suggested improvements are justified on a value/ impact basis, this review will concentrate on evaluating the reasons for the large discrepancy in CMP.

The remainder of this review consists of: a) an examination of the dif ferences for the sequences in Table 1, and b) miscellaneous comments and observations, primarily on the EPRl/WOG study but including some consideration of the NRC study.

A. Examination of Dtfferences in Dominant Accident Sequences

1. Internal Flood (secuence 'l)- This sequence, according to Sect. 3 (Pg. l 3-26) and ADpendix E of the NRC study (Ref.1) Involves flooding of the service water pump room as a result of a rupture in the fire main which  ;

apparently traverses the room. This is assumed to cause failure of all four SWS pumps which subsequently causes f ailure of the PCS, CCW and loss of all safety injection pumps, RHR pumps and the two motor driven AFS pumps. The probability of this sequence is assessed at 7.66E-S/yr, the most dominant core melt secuence (2S"i of the total CMP). This probability !s based on the product of the following f actors:

Flooding probability = 2.2E-2/yr.

SWS pump f ailure probability = 0.1/ flood event (based on judgement) ,

AFW steam turbine driven pump failure probability = 3.4E-2/ demand l l

In the EPRl/WOG analysis, the probability of this sequence Is assessed to be <10E-8 (see Table 1), a negilgible contributor. This large reduction is based on two f actors: a) the EPRl/WOG analysis computes the probability of flooding at 3.73E-S/yr. (Pg. 5-8), and b) according to the EPRl/WOG o ..

.- , study (Pg. 8-10) the HPl system does not depend on CCW, thus the option to provide feed and bleed cooling would still be available even if AFW were lost.

The NRC assessment of flood probability is based on the frequency of auxillary building floods from data (not service water pump room flood data as stated on Pg. 8-10 of the EPRl/WOG study) as given in Table El of .

Appendix E. This data is referenced to be from a 1984 report by Kazarian .

and Fleming. I am not f amilar with this report and can make no judgement at this time regarding its applicability to the scenarlo postulated.

However, the flood probability of 2.2E-2/yr seems high; it would lead to an expectation of some two such flooding incidents /yr for the U.S. nuclear plant population.

The EPRl/WOG flooding frequency is based on 3 correlation by Thomas (Pg.

5-8), and assumes that a rupture of a specific "T" In the fire main piping is reoutred to cause flooding. I am not f amilar with either the configuration of the IIre main piping in the SWS room or the Thomas correlation, and therefore cannot judge the applicability to the scenarlo being evaluated.

However, the explanation of the correlation is not complete in the EPRl/WOG study, and I cannot reproduce the result given by applying the formula. The probability does seem culte low, implying that no such flooding would be expected in the entire U.S. reactor population during their 30 year lifetime with considerable margin.

With respect to the dependence between CCW and HPI, I do not have suf ficlent plant information to make an Independent evaluation. The EPRl/WOG study merely states (pg. 8-10) that the dependence does not exist, while the NRC study assumes that it does. However, if feed and bleed is a vlable means of core cooling for this sequence, even a fac.r of 0.1 in F&B success probability would reduce the NRC sequence probabihty to a minor contributor (2.5%).

In summary, without evaluation of additional Information, I am unable to judge which analysis is more appropriate for this scenarlo. The NRC assessment seems excessively conservative while the EPRl/WOG result appears overly optimistic.

2. Seismic- As Indicated in Table 1, the NRC estimates a CMP of 6.lE-S/yr for seismic initiated events, while the EPRl/WOG estimate is a factor of 8 lower. Page 6-13 to 6-14 of the EPRl/WOG study lists six factors which are alleged to contribute to the dif ference. As Indicated previously, half i of this factor appears to be based on the Installation of new seismic resistant batteries at Point Beach which were not considered in the NRC study. The important remaining dif ferences appear to be 1) the result of a reduced selsmic hazard curve used by EPRl/WOG (Pg. 6-14) and 2) less fragility assumed for the RWST. In the case of the reduced seismic hazard l

curve, the EPRl/WOG study assumed that the NRC curve should be reduced

.' by a factor of 2 for less than 3XSSE and a f actor of 5 for greater than 3XSSE. These reductions are said to be based on discussicas with EPRI (Pg.

6-13).

The f actor of four dif ference in the two results (neglecting the effect of the new batteries) would not generally be considered very significant given the large uncertainties normally attributed to seismic CMP estimates. However, the factor becomes more significant in this comparison because of the overwhelming dominance (74%) of seismic CMP in the EPRl/WOG study. For example, if the EPRl/WOG seismic CMP were raised by a f actor of eight to be consistent with the NRC result, the total CMP for the two studies would be dif ferent by less than a factor of five rather than 31.

I have insufficient Information to judge the fragility of the RWST. With respect to the seismic hazard curve, I performed a comparison of the NRC result with a hazard curve prepared for the Zion site, the closest site to Point Beach for which I have Independent selsmic hazard information (the sites are about 125 miles apart). The Zion hazard curve is contained in the Zion PRA (Ref. 3).

The NRC study does not contain an actual hazard curve, but merely lists return frequencies for ranges of accelerations which are multiples of the SSE (Pg 3-11 of Ref.1). On page C-6, the SSE is listed as 0.12g, thus, the return frequencies can be related to a range of accelerations. If these relationships are plotted on tae Zion seismic hazard curve (Fig. 7.2-1, Pg.

-' 2-18 of Ref. 3) a comparison can be made. The comparison is somewhat complicated by the f act that the Zion curve is actually a family of 9 curves each of which is assigned a probability of being applicable to the Zion site. However, the NRC data f alls above (assuming that the freauencies apply to the midpoint of the acceleration range) the upper bound of all of the curves for accelerations less than about .3Sg., and is above all but two of the curves (with a combined probability of 0.216) for accelerations above .3Sg. On this basis, the NRC data seems quite conservative, and the EPRl/WOG revision would be more consistent with the Zion curve, even 'though the proposed revision produces a change which is inconsistent with the trend of the Zion curve (ie, the NRC curve is further outside the Zlon bounds at lower accelerations where the proposed EPRl/WOG revision is less significant). .

In an attempt to add further perspective on this issue, I have compiled 1 seismic core melt frequencies from a number of PRAs, all for plants east of the Mississippl. The result of the compilation is presented in Table 2.

The NRC and EPRl/WOG results both fall within the range of results in the  !

table. The EPRl/WOG results are quite close to the Zion PRA results, the  :

closest site to Point Beach (note however, that the SSMRP result for Zion l

TABLE 2- COMPARISON 0F SEISMIC CORE MELT PROBABILITES FROM SELECTED PRAS 4

PLANT TOTAL CMP SEISMIC UNCERTAINTIES ACCELERATION Zion 7E-5 5.6E-6 2E-8/3E-5(1) Becomesimportant at.6g Indian Pt.-2 SE-4 1.4E-4 7E-6/SE-4 ( I) -

Indian Pt.-3 2E-4 3 IE-6 4E- 12/2E-5( 1) Major failures start at .6g Millstone-3 IE-4 9.4E - 5 -

Becomes important at .3g

.%Dro$ 2E-4 2.8E o -

Max contribution at.7g Limerid. SE-5 5.7E- 6 IE-9/3E-5(2) -

Oconee 3 3E-4 6.3E 5 - -

SSMRP( 3) -

2E-4 - -

Point 66ach* (NRC) 3 IE-4 6 IE-5 -

Major contribution .36-6g(Pg 3-11)

Point Beach * ( EPRt/WOG)l.0E-5 7.4E- 6 - -

( 1) These values ore described in the PRAs as the 90% confidence intervals (2) These values are the 5% and 95% bounds (3) This stue/ s i the Stesmic Safety Margin Research Program for the Zion plant sponsored by the NRC, see "Applicatfon of ti)e SSMRP Metho@lcQ/ to the Setsmic Risk at the Zion Nuclear Power Plant', Lawrence Ltvermore Natfonal Labs, M. P. Bohn, et al, Me/,1983.

• these studies exclude consideration of large break LOCAs and ATWS .

~-

is considerably higher, even in excess of the NRC result for Point Beach).

Not much can be made of this comparison because of the very wide range in the results. At best, it can be stated that neither the NRC nor EPRl/WOG

, result is outside the (rather considerable) range of selsmic Cf1Ps for other '

nuclear plants east of the Mississippi.

3. Small LOCA, ECC rectrculation failure- As shown in Table 1, the EPRl/WOG probability estimate for this sequence is a factor of 81 lower than the NRC evaluation. The f actors contributing to the difference, according to page 8-4 of the EPRl/WOG study are: 1) frequency of small break LOCA,2) CCW success criteria, and 3) operator actions. These will be considered separately.

The frequency of small break LOCAs in the NRC study was estimated at 2E-2/yr., while the EPRl/WOG study uses 3E-3. This dif ference contributes a f actor of 7 to the total dif ference (a f actor of 81)In this sequence. The NRC frequency is dominated (Pg. 5-3 of EPRl/WOG) by pump seal LOCA events and is based on an Internal NRC memorandum (to D.

Eisenhut from T. Murley) which is based on an LER search. However, the EPRl/WOG study argues (Pg. 5-3) that the majority of these events were f ailures which resulted in very small leak rates which would not trigger containment spray operation and thus would not require ECC recirculation (note that the sequence definition impiles small break LOCAS which reautre ECC recirculation). The EPRl/WOG frequency is also stated to be consistent with the Oconee PRA which is based on industry data, and is consistent with other PRAs. In my opinion, the EPRl/WOG value is more reasonable. The NRC value would imply two small LOCAs requiring ECC recirculation per yr in a 100 reactor popullation. To my knowledge, no such events have yet occurred (also stated by the EPRl/WOG report) In 400 reactor years of experience. The EPRl/WOG estimate is some three times ,

higher than that used in WASH-1400 which seems to adequately account for the possibility of large leak pump seal LOCAs. It is also my understanding that Westinghouse has implemented procedures for recovery of losses of pump seal injection and cooling which are the likely causes of pump seal f ailure. (I note that the NRC study assumes, Pg. A-1, as does the EPRl/WOG assessm:-nt, that pump seal failures as a consequence of station blackout are not considered. I questioned this assumption at a previous ACRS sub-committee meeting and was told that the assumption is based on the presumption that the pump seal LOCA issue under these conditions ,

will be resolved under another unresolved safety issue, and the resolution will eliminate this event from consideration. It seems plausible that if the issue is resolved by some design or operational change, this change will also affect the probability of spontaneous pump seal LOCAs).

With regard to operator actions, the NRC study assumes a human failure probability for f ailure to initiate recirculation of 3E-3 (table 4-4 of EPRl/WOG), while EPRl/WOG uses lE-4, which accounts for a factor of 3.3

.l ., in the probability dif ference for the sequence being considered. However,

. Dage 2-11 of the NRC study Indicates that a human failure probability of IE-3 was used for this s!quence, which would account for a factor of 10 dif ference. Based on the NRC study results, it appears that IE-3 was actually used. Without examining the actual proceedures, their availability, and the timing for operator actions, I have no basis to judge which assessment is more realistic. Given the long times involved for the scenarlo under consideration, and the increased attention which has been placed on recirculation procedures since WASH-1400, I suspect that the EPRl/WOG assessment is more realistic, it also seems consistent with other PRA results as Indicated in Table 4.4 of the EPRl/WOG study.

The NRC study assumes that high pressure recirculation is dependent on the RHR pumps which are in turn dependent on CCW (Pg. 2-15). The EPRl/WOG study apparently assumes that no such dependency exists, but I could find no discussion of this dependency. The only reference to high pressure Injection dependency on CCW which I could find appears on Pg.

8-10, and this apparently is related only to the injection mode of high pressure ECC and not the recirculation mode (see item 1 above). Without detalled plant Information on dependencies, I have no basis to make a judgement on this issue. The discussion on Pg. A-10 of the NRC study Indicates a dependency, but acknowledges that it does not apply for some sequences The small break LOCA event tree in Appendix B (Pg. B-10) does not include CCW as a heading for the sequence being considered here.

4 Long term station blackout (LTSB)- A factor of 70 difference exists between the NRC and EPRl/WOG assessment for this sequence. I was unable to determine the quantitative significance of all of the factors which appear to contribute to the reduction. The f actors which appear to be relevant include.

d Dif ference in LOOP initiating frequency (factor of 1.3 per Table 5-1) l Dif ference in Diesel Generator CC f ailure (factor of 3.3, Table 5-1) l Installation of new station batterles (unknown f actor)

Manual operation of turbine driven AFW pump.(unknown f actor)

The dif ference in LOOP initiating frequency is not significant and will not be considered further.

The diesel generator common cause f ailure for the EPRl/WOG study is estimated at 5.0E-4 on page 5-4 and 1.5E-3 for NRC. The basis for the EPRl/WOG value Is stated (Pg. 5-2) to be that the common cause values were based on "methods consistent with current industry practice', and were derived using the NRC result as a starting point,"compared to other PRAs and further updates were based on plant design and experience as well as opportunity for recovery". Further justification is given on 5-11.

Part of the basis is related to the relatively low common cause probability

- - provided in the Millstone 3 PRA. However, this value appears to be related to the low single diesel generator failure probability used in the Millstone PRA based on plant specific tests of the diesels which would not be appilcable to Point Beach. These low values were criticized in an NRC review of the Mllistone 3 study (5). Based on a survey of diesel generator common cause failure data, the EPRl/WOG value seems low. From Reference 4, the probability of failure on demand for 2 of 2 diesel generators (the Point Beach configuration according to Appendix A, Pg.

A-21 of the NRC study) varies from 2.3E-3 to 7.8E-3/ demand from seven sources based on U S. data (this assumes a 4 week test Interval for the data source dependent on test interval). It is not clear from reference 4 f f or how recovery was accounted for, nor is it stated what time Interval was assumed as the mission time for the f ailures. However, under plant blackout conditions it is my opinion that recovery operations will be quite dif ficult due to several factors, including potential lighting problems, inability to operate repair equipment which depends on AC power, and problems with limited and non-renewable air supplies for repeated start attempts it is my view that the NRC result is more reasonable; it is already below the lower bound of the reference 4 survey.

With respect to the additional batterles, they appear to be of benefit only to matntain steam generator liquid level measurements (Pg. 4-11 and A-9) for use when manually controlling the steam turbine driven AFW pump, which is discussed below.

As stated on Pg 8-2, the largest f actor appears to be the EPRl/WOG assumption that loc 51 manual control of the steam turbine driven AFW pump could be accomplished af ter loss of both AC and normal DC, with communication between the local control personel and the plant operators (Pg. 3-6 and 4-11). I am skeptical that such operation would likely be successful. First of all, according to page 3-6, under these plant blackout conditions, a diesel fire pump needs to be started to provide cooling for the turbine driven pump. This requirement is not acknowledged in the procedure described on page 4-11. It is not stated how long the turbine '

pump can operate without normal cooling (supplied by service water), but It is estimated (Pg. 3-8) that it would take 10-15 minutes for the operator to reach the AFW equipment. It is not clear if the pump would have Biready been automatically started and be operating without cooling or have tripped from overheat. Further, the time to reach the equipment ,

should consider that the plant is in a blackout cond'.tlon, le- no elevators, Ilmited lighting, disabled security systems if dependent on AC power, etc.

Locai operation of the steam admission valve to the turbine would appear very dif ficult unless provisions have been made to protect the operator from the excessive heat, and the noise from pump operation, diesel fire pump, and steam flow could make communication with the operating room, necessary to control steam generator level, very dif ficult. Time is limited to accomplish all of the required operations and stabilize flow (see e

I

,, related comment " 8 following). Without more detailed Information, it 19

, not possible to determine if all of these potential problems have been adequately considered, but they do not appear to have been systematically evaluated in the EPRl/WOG report. Consequently, I am inclined to accept ,

the NRC assessment as being more realistic until a more definitive analysis of the operation can be evaluated.

5. Internal fire- The PRl/WOG fire assessment produces an estimated CMP from fires which is a factor of 500 below the NRC assessment.

Without addittonal plant Information, including layout itformation for critical components, fire protection systems and philosophy, control of combustibles, fire barriers, etc. I cannot make a definitive judgement on the fire assessment for either study. On the basis of Information in the EpRl/WOG assessment (Pg. 6-4 et seq) It appears that 1) neither study provides a comprehensive fire risk assessment utill210g state of the art analysis, and 2) the NRC assumptions with respect to fire risk appear overly conservative.

6. Transient LOCA with failure of ECC recirculation- This sequence, according to EpRl/WQG, will not occur because, as noted previously, this transient would not reach the relief valve set points. I have no basis to judge if the EpRl/WOG evaluation is valid or not.

7. Loss of off-site power with loss of feedwater and feed and bleed- The EpRl/WOG assessment estimates that the probability of this sequence is a f actor of 9 less than NRC. According to page 1-13 of EPRl/WOG, the key reason for the change is the addition of new batterles at the Point Beach site. This assessment appears valid.

B. Miscellaneous Comments and Observations

1. The NSAC Perspective section of the report refers (Pg. V) to a core melt probability target in the NRC safety goal. There is no reference to a core melt probability in the published version of the safety goal. A more ,

I meaningful comparison would be public health risks which are the exclusive quantitative provisions of the goal.

2. Pg.1-1: It is stated here that Point Beach was one of 6 plants selected in the A-45 erfort "to represent a broad range of reactor and DHR system designs, so as to form the basis for a consistent set of generic or group-generic new licensing requirements for DHR at U.S. reactors." This l description of the selection process seems to ignore what appears to be the single most important selection criteria:"Point Beach was identified in the Inltlal qualitative screening as having suf ficient potential vulnerabilities (to decay heat removal) to warrant additional study"(Pg.

1-3 of Ref.1). j

1 l

. '3. Pg.1 ' 2: It is statcl here that "The EPRl/WOG analysis of Point Beach resulted in significant raductions in core melt frequency (of the NRC results) due to special emergencies. In the case of seismic analysis, the major changes occurred as a result of considering a variety of recovery actiors for earthquakes less than three times the SSE." This explanation .

seems inadequate and inconsistent with tne NRC result which Indicates (Pg. 3-11) that the major contribution to core melt from seismic events k comes from accelerations betweeri three and ' +1mes the SSE.

4. Pg.1-15; It is indicated here that recovery actions including use of water from the spent fuel pool have t:een added to the NRC model. If spent fuel is stored in the spent fuel pool which has to be immersed to prevent overheating, this may not be a reasonable alternative to consider.

5. Pg 3-4 and 3-5: The basis is not provided for atsumptions and statements given nere. In particular, the assumptions regarding the time ,

that PORV block valves are closed and the statement that relief valve set points would not be reached for the transients considered.

6. Sect. 3: Numerous assumptions which are stated to be conservative are listed in tr.is section (cf. Pg. 3-11,3-12,3-14,3-15). However, no information is provided to enable an tvaluation of the effect of these conservatisms. Such information would be useful in providing additional perspective on the results.

7. Pg 5-1: It is Indicatad here that Point Beach has operated 16 years, thus providhg valuable plant specific data. However, neither the EPRl/ WOC or NRC studies appear to e 'ount for plant aging effects. An ongoing NRC program to examine this issue has found that plant aging ef fects can increase the CMP for older plants.

8.Pg A-9: The times used here to cvaluate human error probabilities for operating the turbine driven AFW pump upon 'oss of DC appear optimistic.

It is stated that 10-15 m nutes would be required to manually start tne pump, including local operation of the DC steam admission valves. This is ,

stated to leave 15-20 minutes available for recovery before steam l generator dryout. However, on page 3-8 it is stated that it 4kes the personnel 10-15 minutes just to reach the pumps af ter belr,,,. lspatched by the p! ant operating staf f. ihus, it would appear that only 5-10 minutes ..

Is availabie for recovery. Furthermore, since this scenarlo appears only to be important in conjunction with loss of AC (otherwise the motor driven AFW pumps would be available), the diesel driven fire pumo must also be started to provide cooling to the turbine driven pump.

9. Pg. 9-1: It is not clear why the EPRl/WOG expected dose for case 1 (Interpretation of NRC calculations) is so much lower (a factor of 15) than  !

l l

the NRC Case Study results. Further. It is also not clear why the IDCOR source term results are higher than the interpretation of the NRC calculations when the IDCOR source term essentially elimina'.?s all early

. containment failures.

10. Pg C-2. It is not clear why the entries under containment failure mode probability are identical for all accident types, nor is it clear why the failure mode probabilities do not sum to 1.0.

I1. Pg.C-10: It is not clear why the La releases for the DH (etc.) failure mode have not been reduced by a f actor of 2, nor why the I-Br has not been released by a f actor of 2 for the MT cases.

12. Pg. C-11: The Xe-Kr release fractions are greater than unity (1.7) for Accident Type 4

13. Appendix C, general: It is not clear what evacuation assumptions were made in the EpRl/WOG conse":nce analysis. However, it should be noted ,

that the overwhelming domt.n.. ice of earthquake initiated accidents to CMP may mean that evacuation is degraded or impossible for most risk significant sequences due to of f-site damage (communications, road ways, etc.) from the event. Whlle this factor may not be significant for value-Impact analysis which consider on-site costs (since they will dominate as stated on Pg. 4-1 of Ref. 2)it may be important for analyses which do not consider such costs. (Inclusion of on-site costs has been a controversial issu?, for example, see Pg. 5-2 of Ref. 2) ,

14 As an observation, it is worth emphasizing the discussion in Sect. 7 .

which brings out some important points relevant to adding additional l Safety systems which are consistent with my own concerns. In summary, it is noted that consideration of such systems must be done carefully, t

with emphasis given to the details of the sequences vhich they are being Installed to prevent, and recognition of interf aces and dependencies with other plant systems. Such systems can impose a detrimental burden to the '

operators, can introduce additional accident sequences, and can exacerbate some sequences prevlously found to be Insignificant conttrbutors. (Ref 2 ,

l also provides some cogent observations in this regard.)

1 hope this review is of use to you and the Sub-committee Sincerely, l

P. R. Davls CC: Dave Ward, ACRS i

I . _. __ , ,

(_ __

.- .s ,

REFERENCES I. "Shutdown Decay Heat Removal Analysis of a Westinghouse 2-Loop ,

Pressurized Water Reactor ~, NUREG/CR-44S8, W. R. Cramond, et al, Sandla National Labs, Mar.1987.

2. Letter, Gerald Neils Chairman NUMARC Working Group on DHR to D.

Ericson, Jr., Sandla National Laboratories, June 22,1987.

3. Zion Probabilistic Safety Study, Commonwealth Edison Co.,1981.

4. "Common Cause Failure Data. Experience from Diesel Generator Studies". S. Hirschberg and U. Pulkkinen Nuclear Safety Magazine, Vol. 26, No. 3. May-June 1985.

S. 'A Review of the Millstone 3 Probabilistic Safety Study', NUREG/CR

-4142. A. A. Garcia, et al, Lawrence Livermore National Laboratory, Oct.

1985.

f I

4 9

9 s , l $ e