ML20236A785

Safety Evaluation Providing Risk Perspective of Util May 1983 Proposal to Change Tech Specs Re Allowed Outage Times from 3 to 7 Days for Nine Safety Sys,Including Charging, Safety Injection & RHR Pumps & Emergency Diesel Generators

ML20236A785
Person / Time
Site:	Byron
Issue date:	01/15/1986
From:	NRC
To:
Shared Package
ML20236A787	List: ML20236A785 ML20244D633
References
NUDOCS 8710230128
Download: ML20236A785 (26)

Text

% +,

\

ENCLOSURE Safety Evaluation Report Related to Proposed Change in Maintenance i  !

Allowed Outage Times For the Byron' Station. -)

i

]

1

1. Introduction '

1 In May 1983 Commonwealth Edison Company (CECO) proposed that the Allowed 1 Outage Times (A0Ts) set forth in the Technical Specifications for tho Byron. plant be changed from.3 days to 7 days.for the following nine safety systems:' charging (CHG) pumps,' safety injection (SI) pumps, residual heat removal (RHR) pumps, essential service water (ESW) pumps, j

. auxiliary feedwater (AFW) pumps, component cooling water.(CCW) pumps, fan coolers, containment spray (CS) pumps, and emergency diesel )

,;: ~ generators (DGs).U)

L The A0T defines the maximum period of time during reactor operation that

, a failed safety system component is allowed to remain inoperable pending

-! completion of repairs before the plant is required to be shut down. The <

l

-l A0T specified in the technical specifications is primarily based on engineering judgment, system-specific licensing review considerations, -

and on related component repair times.*

1 o

j. .

• The staff considers that probabilistic risk assessment methodology can provide a quantitative basis for the settir.g of technical specifications. To this end, an NRC research program, " Procedure for Evaluating Technical Specifications" (PETS), has been undertaken to establish appropriate licensing guidelines for the specification of A0Ts and surveillance intervals.(2)

.I

' 7' 8710230128 860115 PDR ADOCK 05000454 P PDR v

.~......w..#.... . ,,.w,,_, _ -. ..

g . _ -

n. '

l I

l

[  ;

For the Byron A0T proposal, CECO suggested that a probabilistic risk l analysis could be the basis for showing that the proposed modification l

in A0Ts would not significantly increase the risk to the public. Sub-  !

sequently, on May 2, 1984, CECO submitted a formal request for the j i

proposed change in A0Ts together with a report (WCAP-10526) providing l a quantitative' analysis of the A0T risk impact.(*) This' report was prepared by Ceco's consultant, Westinghouse Electric Corporation.

The Byron A0T analysis was largely based on PRA models and data developed i for the Zion Probabilistic Safety Study (ZPSS),(4) the rationale for this being the general similarity of the two CECO plants in regard to NSSS'and 3

containment design, and operation. The Applicant's submittal provides i

estimates of the increase in core melt frequency (CMF) and offsite health j consequences at the Byron site due to an increase in the A0T from 3 days to 7 days for the proposed group of safety systems. The WCAP-10526 report l concludes that the associated increase in_ risk to the public is insignifi-  ;

I cant compared with the baseline risk. Further, it is suggested in

]. '

WCAP-10526 that the proposed extension in A0Ts could provide positive j safety benefits in effecting not only a reduction in the number of shut- I down challenges to the plant safety equipment but also a possible overall l

1 increase in safety syst'em availability, in that the. increased A0T for the 1

1 occasional long duration maintenance job would allow for a more thorough l and effective repair and preventive maintenance and, thereby, a lowering j in the expected frequency of such maintenance events.

i l

I i

1

,1.,- 4,

+ -

y; 3-

'The staff initiated its review of the Byron /,0T _ risk analysis in March i.

I 1985, with the technical assistance of the Brookhaven National Laboratory I i

]l (BNL). It is to be noted that the BNL analysis was performed in the con- 1

. text of .the existing operational status of the Byron station (i.e. , opera-j tion of Unit 1-by itself) and was based on the plant system information  !

(e.g., success criteria) provided in the Applicant's submittal. A draft I

report of the BNL analytical results was reviewed by the staff in October

! 1985. As subsequently revised, the BNL calculational results indicated a high core melt frequency for Byron 1,^due primarily to the two pump ESWS in use, which, in the absence of the potential capability for inter-

1. unit sharing of essential service water, exhibits a relatively high fre-quency of failure, with attendant risk to the~ plant. )

These results were j discussed with the Applicant on December 6, 1985, and resulted in a ,

commitment by Ceco for establishing an acceptable Unit 2 ESW backup to l Unit 1 prior to resumption of Byron 1 operation. ) The final report of i

! the BNL review of the A0T proposal is scheduled for issuance in December

)

i 1985 as a NUREG report (NUREG/CR-4404).

This SER provides the staff's risk perspective on the Byron A0T proposal.

! It is intended as supporting information to the Division of PWR Licensing-A for decision making on the Ceco proposed A0T change.

1

[ 2. Review Approach i

The approach used in reviewing the A0T proposal included examination of the following aspects of the WCAP-10526 analysis:

I ew. *...-o,eha....,., ..~.,c.wm, ,.%, 8

n '. > , j i

I (a) Appropriateness of the Static Fault Tree Method: The Applicant's risk study made use of various PRA models and data developed for the Zion Probabilistic Safety Study (ZpSS) on the basis of the  ;

overall similarity of the Zion and Byron plants. The actual plant system differences between the two plants were taken into t

account by a redoing of the system fault tree analyses. The methodology used entails a static (time-independent) treatment.

An initial effort of the BNL review involved examining the need for using a time-dependent reliability technique (e.g., Markov, Frantic),

as against a static fault tree approach in regard to the specific j question of allowed outage times. The BNL findings (in agreement with results arrived independently through the PETS research program) I indicated the feasibility of the static fault tree method for i

i evaluating A0T changes, the static method tending to give some- j what higher (more conservative) values of core melt frequency than i

j those obtained using a time-dependent method such as Markov.

! f i

(b) Risk Index Consideration is required in selecting an appropriate l 1

risk index at which to assess the impact of a change in A0T, i.e. ,

l whether to judge the impact at the level of component / system un- -l availabilities, core melt frequencies, or offsite health consequences.

The conclusion in the Byron submittal that the impact of the proposed l

A0T change is statistically insignificant was drawn from the perspec-tive of offsite health consequences, for which quantitative assessments must take into account the complexities associated with obtaining 1' .

l I i 4

==______. - -

t .; . 'n c ~ .

-5r r

plant damage: frequency estimates,'the analysis of'co'ntainment-response and radionuclides'releasea,'and the analysis of site: con--  ;

4: ..

l a, . .

sequenc'es. Accordingly, the assessment of health consequences is .{

only indirectly related to a change in a'particular plant system .!

I

-parameter such as' allowed outage' time for component repairs'. In- '

. addition, assessments of offsite consequences are invested with -

.a

'1 large uncertainties. A more direct, sensitive measure of the A0T.'

q' impact can be obtained at the levels of component or system un -

availability, or, better, at the level of accident sequence'fre-l

)j : quency, since at the sequence level possible component / system i 1

interactions.that might be missed at the systems level would be

! .taken int 9 account at the sequence level. O The results of the n i g Brookhaven analysis permit evaluation of the'A0T impact-using the~

i 3 i risk index of core melt frequency,'which the staff believes to be i

1 an appropriate basis for management decision on proposals regarding l

.j' changes in maintenance A0Ts.

-) 1 4 i It is noted that the A0T impact was evaluated in terms of the j q

] differences in the values of the core melt frequency obtained for j the 3-day'and 7-day A0Ts. Taking the arithmetical difference rather than the fractional change in these values ensures taking into l account only those sequence contributors that are associated with 3

the A0T-related minimal cutsets. l

'! l e

I I

_,.u_,.__.__....,.._.__._._.__

. _ . ~ -

.! 4 ..

l 1 -m -

(c)-Single-UnitOperation: The Brookh'aven revi u was based on the current operational ' status of the Byron Stat; ion, i.e. , operation of Unit 1 only,' without capability for inter-unit sharing of the ESWS

and-the CCWS. The existing capability for an electrical crosstie 4

between Busses 141 and 241,.which provides for additional reliability 1 a

f 1 , of Train A of the AFWS, was not modeled in the Applicant's analysis u,

a nor in the Brookhaven analysis. An indirect estimate of the effect of the~ bus crosstie on core melt frequency can'be derived from the ,

BNL cutset. data and found to be small. On the other hand,.the non-4 availability of inter-unit _ sharing of essential service water was 1 found to have an important effect on the estimated Unit 1 core

j. melt frequency. This arises because in the. Byron design (in con- ]

.J

+

trast with Zion) it is the two pump ESWS that provides for direct .)

~

cooling of most of the A0T components of interest (diesel generators, i

high pressure injection pumps, motor-driven AFW pumps coolers, con-tainment spray pumps and fans, etc.). For these components, loss

, of ESW cooling can result in rapid failure or unavailability of the components within a period of several minutes, so that the loss of a

1- ESW with assumed non-recovery enters into a number of dominant

! l 1 sequences,' including, in particular, the loss of seal cooling of the .!

l. 3 reactor coolant pumps with a consequent seal LOCA. Consideration  !

q.

9 of the loss of ESW as an accident initiator was considered in the i l l' BNL analysis, but not in WCAP-10526.

L s

.5 t

ln ,

T  !

i ,

l

'i- .c -uimv .-=m., ,,*m.,

, e e a . .mm =ue - === m ei- .

l-  !

. (d) Applicability of ZPSS Models and Data: This review area was con-

-! s idered in some detail in the light of the dependence of the l .l Applicant's analytical results on the ZPSS. In the context of the i j A0 Tissue,thelistofaccidentinitiatorswasjudgedtobereason-t ably complete except for twa important initiating events: loss of-i

t essential service water, discussed above, and loss of a DC bus.

The latter event was found to be of significance in the NRC review of the ZPSS (a CMF contribution of about 7x10 8/yr at Zion) in connection with failure of bleed and feed operation because of the dependence of PORV operation on DC power. Its significance in the case of Byron was not clarified in the BNL analysis because of the unavailability of the requisi'te plant information on the DC bus loading scheme. We note further that the significance of this  !

t

event at Zion is predicated upon the assumption that both PORVs i

I are required for feed and bleed; this is likely to be a j

j conservative assumption. Thus, the staff believes that the loss of ^

j a DC bus at Zion is likely not an important initiator.

l l The event trees used in WCAP-10526 are considered by BNL to be 1 improved versions of those used in the ZPSS. Fault trees for the I Byron A0T systems of interest were developed in detail in WCAP-10526 and are considered generally appropriate to the problem under con-

! sideration. As discussed below in regard to the analysis of the ESWS, i

the success criterion given in WCAP-10526 for the ESW cooling tower

.; fans was subsequently indicated by the Applicant (in a meeting with i

. - . . . - - . - . . . - . - - . - . . ~

c. ,

! ,. i.  ;

l l.

l l

l NRC on December 3, 1985) to be overly conservative and an amended j criterion was submitted. )

The final BNL calculations incorporate this amended criterion. The general technical specifications re- -

1

_ quire, by their definition of operability of a component, that all support system trains required for operability of that component i (e.g., service water or DC power) be operable. This was not modeled in WCAP-10526, tending to yield conservative estimates of unavail-abilities. In the BNL analysis, this operability requirement was taken into account. '

I i

(e) Maintenance Frequency and Duration l.

Estimates of the frequency of maintenance events and their mean

! duration is required fo'r estimates of system unavailabilities. In

, the Applicant's analysis, data on maintenance frequency for the A0T

systems of interest were obtained from the ZPSS data base, which j was developed on the basis of a Bayesian approach involving the

! incorporation of Zion plant data with a prior distribution of generic data, to yield a posterior distribution of maintenance frequency data

" specialized" for the Zion plant. With two exceptions, the Zion i

frequency data, obtained for the general Zion A0T condition of 7 days, 'I was used for the corresponding Byron components for both the 7-day l and 3-day A0T calculations. The two exceptions were: (a) the diesel- )

I l

! driven AFW pump, where the data on maintenance frequency was obtained l from the Torrey Pines AFWS reliability study,( ) and (b) the Byron ESW pumps, where the maintenance frequency data applicable to the I Zion CCW pumps, and not the Zion ESW pumps, were used in WCAP-10526, 1

o,

_g. i on the basis (unexplained) of the difference in ESW pump types be-tween Zion and Byron. These same data on component maintenance frequencies were also used in the Brookhaven analysis. The estimated i

frequency of maintenance for the ESW pumps at Zion was less than the estimated frequency of maintenance on the CCW pumps; hence, if the ,

Zion ESW pump frequency of maintenance had been used, in the Byron 3 study, for the ESW pump maintenance frequency, the maintenance un-availability of an ESW pump would have been less.

i 3

With regard to duration of maintenance events, the Applicant's analysis did not make use of the published ZPSS plant data on repair a

l times for the different systems. Instead, a lognormal prior dis- i tribution of me'an repair duration times was assumed, yielding an estimated Bayesian mean of these means_of 19 hours2.199074e-4 days <br />0.00528 hours <br />3.141534e-5 weeks <br />7.2295e-6 months <br /> for the 3-day A0T condition, this value being equally applied to all of the A0T j systems of interest. For the 7-day A0T condition, the corresponding I

l estimate was 34 hours3.935185e-4 days <br />0.00944 hours <br />5.621693e-5 weeks <br />1.2937e-5 months <br />.

-I: i 1

4 The validity of using a single mean maintenance duration time p

estimate for all the A0T systems of interest, and of basing this l

estimate only on an assumed prior distribution of mean times without

} reference to available Zion plant data is not clear and is open to ll question. In the light of these and other uncertainties bearing on the question of system unavailability due to maintenance, a 1

j sensitivity approach was adopted for the Brookhaven evaluation 1

of WCAP-10526.

i

,-(

j u _ _ _ _ . _ _ . _ _ ._. _

(f) BNL Sensitivity Approach The Brookhaven sensitivity approach involved obtaining estimates of the core melt' frequency for two bounding values of mean maintenance duration time (T,): a lower value of 19 hours2.199074e-4 days <br />0.00528 hours <br />3.141534e-5 weeks <br />7.2295e-6 months <br /> to correspond to the i 3-day A0T condition and an upper bound value of 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> (7 days).

al The calculational method is shown to yield core melt frequency results that are approximately linearly dependent on T,over the parametric range of from 19 to 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br />, so that by interpolation 1

j. the core melt frequency may be obtained for values of T,less than

168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br />.

g (g) System-Specific A0T Risk Impact: The Applicant's results obtain for

the case where the A0T change is applied to all nine proposed systems at the same time. The Brookhaven analysis covers this case, as well j as those in which the A0T change is applied only to the individual i

system, the others remaining unchanged. This yields information on the relative A0T risk impact posed by each of the various systems.

l )

! (h) Seal LOCA Probability: An additional consideration regarding the i

i staff's evaluation relates to the question of the loss of cooling i  !

!.I of the reactor coolant pump (RCP) seals and the probability of a l 1 3

consequent seal LOCA. This as yet unresolved safety issue is of l f importance in the Byron analysis in that for the WCAP-10526 assumed value of 0.5 for the probability of an induced seal LOCA given loss Y ?_ N ? ? N_____-----_--__-_-__------___-_----------------------------------------------------

s. ,

i I

of seal cooling, about 88% of the cutsets are found to involve seal LOCA. In view of the indicated sensitivity of the CMF estimate to i l

the assumed seal LOCA probability value and the uncertainties sur-l rounding this value for Westinghouse RCPs, BNL performed a parametric study of the CMF dependence on seal LOCA probability. The results I l

are discussed in Section 3.2. '

3. Results 3.1 Core Melt Frequency: For the current 3-day LCO for Byron 1, the estimated CMF obtained for the BNL reference calculation is:

CMF = 7.0x10 4/yr. for T, = 19 hrs.

The important assumptions underlying this reference calculation are restated for purposes of clarity. 1 No inter-unit sharing of ESW; '

The assumed probability for a seal LOCA given loss of seal cooling = 0.5 (as assumed in WCAP-10526);

The mean maintenance duration time for the 3-day A0T condition was taken to be 19 hours2.199074e-4 days <br />0.00528 hours <br />3.141534e-5 weeks <br />7.2295e-6 months <br /> (Ref. WCAP-10526); i The initiating event frequency for loss-of-ESW in Byron 1 is estimated by BNL to be equal to 9.5x10 4/yr for T, =

19 hrs, and equal to 7.2x10 3/yp for 7m = 168 hrs.

The changes in CMF obtained in increasing T, from 19 hourt to the full 7-day A0T period of 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> are shown in Table 1. Plots of the CMF as an approximately linear function of T, over the range from I

l l

l 1 E _ __ _

- - ~ -_ _ _ _ _ - . . . . ._

j

j '.' . .

I t

12 -

19 to 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> are shown in Fig. 1. The CMF resu'.i.s show clearly l

that the main contributor to CMF is due to the ESWS. The change in the CMF due to a change in T, for the diesel generators from 19 to 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> is appreciable. Tha same is true, to a lesser degree, for the AFW system. The other systems of interest indicate (within I the 1010/yr cutset truncation level assumed by BNL) no significaitt A0T impact on core melt frequency.

In the list of the dominant sequences provided in the BNL report (NUREG/CR-4404), the top event involves the loss-of-ESW initiator at a frequency of 9.5x10 4/yr. , combined with a seal LOCA probability of 0.5. This yields for this single sequence a CMF contribution of about 5x10 4/yr., which is approximately 70% of the overall CHF of about 7x10 4/yr. for the two pump ESW Byron configuration.

t n

• m- --w .,,y- , , . . w g.-.g-i

-- ) , -- .-.,

I.

TABLE 1 Dependence of CMF on Maintenance Duration Time (T,)

A CMF Case CMF(yr 1) (CMF)168 (CMF)19 ATm Base Case: Nine Systems @ 19 hrs. 7.0(-4)* -- --

Nine Sytems @ 168 hrs.** 4.8(-3) 4.1(-3) 2.8(-6)/hr.

-! Only ESWS @ 168 hrs. 4.5(-3) 3.8(-3) 2.5(-5)/hr.

Only Diesel Generators @ 168 hrs. 8.4(-4) 1.4(-4) 9.7(-7)/hr.

Only AFWS @ 168 hrs. 7.5(-4) 5.6(-5) 3.8(-7)/hr.

Only CF + CS 9 168 hrs. 7.0(-4) <1(-6) <1(-8)/hr.

Only CHG + SI @ 168 hrs. 7.0(-4) <1(-6)' <1(-8)/hr.

Only RHR 0 168 hrs. 7.0(-4) <1(-6) <1(-8)/hr.

Only CCWS @ 168 hrs. 7.0(-4) <1(-6) <1(-8)/hr.

j! .

• Exponential notation is indicated in abbreviated form as 7.0x10 4 = 7.0(-4).

.: )

4 i i ** The initiating event frenuency for loss of the two-pump ESWS in Byron 1 is  !

-l- increased from 9.5(-4) yr 1 for T, = 19 hrs, to 7.2(-3) yr 1 for T,= 168 hrs.

)

't i

I e

d

.,. f~ts. .f.

B >n osi .1 Cone Msir faemurocy vs. ~11,

, Woue Systens BNL MoDet Assunprions

• ESWS Swae-Umr Orev nw yo. {~I~WO- VUhP GSWS)

• loss' O P - ESW AcciceprIN/r/kTDA

• $EAL }.CCA 09OBAb/L /7y = 0,S~ }

l t

n- l l

dMF 2 -

CASE StofE

(15hyr. Mae Syrrens .2.s5s_-)/ Hz.,

ESws LS~(-r.)/Hz.

.D G s 9. 7 (-7)/ HR..

l ,, AFWS 3.2(-7)/HR i orHERs : < t (-8)/ n z.

• df+CS CMG+SE

. *RHR

,,

• CC toS DGs -

A FWS o 7,,,g g .

g.. 6 7 nesmy- Estimreb P' .19- Mr. m Az 3 Dny Aor (Ke( WCAP-tosa)

$ IN

} in So 8'o (Eo th') ' /$O MEAN MA/NTSNANCE .DORAT/0N 7 / M E , Tm (Nouts)

i 1

3.2 Seal LOCA Probability : C'mmon o mode failure of the reactor coolant pump l (RCP) seals and a consequent seal LOCA, arising as the result of the loss of seal cooling, as might occur under station blackout condition or loss of the ESWS, is a continuing, unresolved concern of the NRC staff. Un-certainties surrounding this issue include the questions of recovery '

times, operator recovery actions, and the intrinsic capability of the seals to survive.for a sufficient period of time under high temperature conditions. These uncertainties have been typically treated by a prob-ability factor which can vary between zero and unity. In the Applicant's submittal, this probability was assumed equal to 0.5. The BNL reference calculation was based on the same value.

In the BNL reference case, the cutsets involving the occurrence of seal LOCAs from both non-LOSP and LOSP initiators were found to contribute about 88% of the overall CMF of 7.0x10 4/yr for the T, = 19 hrs. case, with about the same fraction for the .T,= 168 hr. case.

j To obtain a measure of the sensitivity of the estimated Byron-1 CMF to l the assumed value for the seal LOCA probability, BNL reran the reference calculation with the seal LOCA probability set equal to zero ("Model 2,"

NUREG/CR-4404). In Fig. 2, the plotted results of CMF vs. T, show not only a substantial reduction in the initial (3-day A0T) CMF as the assumed probability is reduced, but also a significant reduction in the slopes of the curves over the rcnge of T, considered. We note that the l

BNL analysis does not include the effects of battery depletion on station blackout. The effects become more important as the seal LOCA probability is decreased.

_ . ,v .

L + b 1 0 4 l

l

. , 6 e e .

, . , :i e. .a  ;; .

///NE SyffSHS 1

, , , PAC 4N .f l

l

' 5 y . .

1 CME NEW PREQurucy vs. C,

. i_.t i o L: .

-r;3e . - .  : 4 . .,

t# ReBABatry LX y. 2/FEGKGPr fMOFS OF SEML' doc 4 m

l

, .g. e s ,

9 e ..- h e .- 9

. J . .-o 2- ... .- 6 ..

i

}

! 1 I

l . .

.f , . .

_ gyg _

._ ,_ _ .0 f

- 4 .

y...y y. . . .__. . ..

_ _ t___. . .jp . _.__. ,

_L__..__.4__....L...._._..._-. .

i

! i i

i i  :

___. l __,i I . . .,i . _ 4. . _

i 4... . . .. .

! I g I

.__..I_ _

_ _..__ g' -._. f_ I 7.: . _. .i __.l .. 7_4__i

[  ! .

.4_.._.

i i  ; I

._ L. L.. . . .

. J, _ . .l . .._ !. __;.._

r-

, . . _ L,. _.,s,,!

1

, _ . ..&. 9.. 6 W. - .-..-. _.e *- ,'.={ . . y- + . & t I '

l

. . . .. -- { . _4 - ... . .. J t . L. __

, 1

...--.,'..-._._...l.-_

i i

! i e ) i

, ._f . - .

.. .r ' . . h. -. .- -

! I i 1 ;_j i

i  !

L!

i l  !

l J __; _ J._ . _ . . . - ,

r '

l i

' i' l  ! l ' i  !

! .;. j .

l

'_. . 1I e . ., . _

i I .

i  !

j i [  ; + i l '

'-+--*--+

l

[.  ; t' + -- - - ' - - ' -

- . I i ii . 4_* _ y_

l

[ l q jl.. _ ,_  ; j ._. . .,._ t ._ . . _.l . . . _" _ .

~ ~ ~ ~ ~ ~ ~

%_ . . _._/ , _j ._.m ,__. _'-- T r c..q . .j._ [ }. . . . J. . t i pi_.i q .p ..L.2

. . N/w Sysymf

,1

6. . _ ..

-[ ,- . ,

, . 6.. , .

t y n .j ._l. ., . ..,-. , k.  ! I , . .. .

. . . -. .,7 ..co f"..-

.r s

..A7espo.eawg, . , . _ _ . . . . . ., . . . -

f*p

.. 7- t f5 -3'-h. y A CT 1  :- i ,

..A.../IN,..E. S. ysrs/gf

* ,,,,,,\

,  ! i. l \ i i

- r _ 9._

, .t.

l L '

l

1. ~/

n' _

fr _ .._

l' l , . ' .-~""~~  !

1. =m'"

l l

s. - _ _ __ _ _ ._

0- , r , , , , -

g ,

7

,- . 0 y y 2o A d. /Ao_ _

._ _. /.p_ _ _ _ .._d./0_..,

L Yf N Y l

m .,

s In the December 6, 1986 meeting with the Applicant in regard to the high values of CMF obtained for Byron 1 under the existing 3-day LCO, Westinghouse reviewed the results of recent seal tests performed in France and suggested that the results gave indication of seal sur-vivability under loss of seal coolant conditions for time periods of several hours. Additional detailed discussion of these results and related considerations are scheduled in future meetings with the staff; however, at the time of preparation of this Safety Evaluation Report, the competent NRC staff regarding this unresolved issue does not believe acceptably conclusive evidence has been obtained to date to permit resolution of the issue in regard to Westinghouse RCP seals.

For this reason, we believe justification for a change in the assumption of the seal LOCA probability value of 0.5 would not be warranted in the context of our evaluation of the Byron A0T proposal.

An additional consideration that argues against a reduction in the 1

assumed probability of a seal LOCA is the following: Even if we accept the Westinghouse analysis indicating that under loss of all seal cooling, the RCP leakage flow rate (without 0-ring seal failure) would level out at about 20 gpm per pump, core uncovery can be estimated to occur in a

{

j time period of about 16 to 20 hours2.314815e-4 days <br />0.00556 hours <br />3.306878e-5 weeks <br />7.61e-6 months <br />, assuming the operator takes action

)

to reduce system pressure and temperature, but fails to recover injection capability. For the station blackout situation, the probability of re-covery of AC and of coolant injection in a 20-hour period is sufficiently high so that this is not a concern. For the loss-of-ESW initiator in a t

1 J

i l L _'__ ___

u .

- 16'-

.s Byron'two ESW pump configuration, this is a concern,-since the estimated mean time to repair an ESW train is about 19 hours2.199074e-4 days <br />0.00528 hours <br />3.141534e-5 weeks <br />7.2295e-6 months <br />, if the allowed outage time is 3 days. (We note that this estimated-19 hr. value given in the-WCAP-10526 was derived using Bayesian, methods; however, the Zion plant-specific ~ data on mean time to repair of service. water pumps is approxi-mately equal to the value of 19 hours2.199074e-4 days <br />0.00528 hours <br />3.141534e-5 weeks <br />7.2295e-6 months <br /> used in the BNL analysis.) More-over, for a leakage flow rate of 20 gpm per pump, the pressurizer. level  !

would decrease to the point where high pressure injection could be expected to be initiated at a time relatively early to restoration of ESW, with the result that the high pressure injection pumps could fail from loss of cooling unless the operator takes special action to secure the

] pumps. (The high pressure injection pumps can run for only a few minutes without lube oil cooling before failing.) Hence, even without a gross leakage from the reactor coolant pump seals, there is a substantial chance I

of core melt, given a loss of ESW initiator. Of course, if the RCP seal j leakage exceeds 20 gpm per pump (because of some degradation of seal per-I i- formance), the operator has less time to take action, and the time to

~f

, recover ESW is less. With availability of a Unit 2 ESW pump as additional 1 l

j backup to the Unit 1 system, the loss of ESW initiator is of much less Concern.

i l

l 4. Discussion l.:!

j 4.1 Essential Service Water System 1 1

i' 4.1.1' Initiating Event Frequency For loss of CSW: The BNL estimate of l

.y l

the annual frequency for loss of the ESWS in Byron 1 was based on a fault l

W G s WN CA CH S S NC T

I N OX N E E

M T E

N N MA A O OE 5 O

NP CH O O M X O RO C

+ S 2

b J _

^ 8 8 3

3

• 4 4

' I 1 MAS

}g

1. F A

$ S X

I U

f 2

l M O E X X T

T A S B S B S 2 2 2 2 Y O O O O O O S x X x S S fS R A 2 8 I 8 2 A P

- E T

P 1

P 1

O P

1 0

1 O

- A 0 X x X W x S [ S S S 2 i 2 i

E

- C I

O A

I A

1 8

1 O

8 1

- V O O O O o M O R X x X X

_ 1 E

S S

I W2 JM S S I

S 2

_ 3 L

= 3 A A 8

- e, I

T N

0 3

e,X 8

3 1

H D 2

8 3

1 X

3 u E 6 S 6 6 S G

I F

S S

E mS 1 X M

/ mS$X '

1 x

1 M p a C a G W

O n 3

6 1

h a 3

6 1

M L

F a X A a X 8 D

Sh 2 6

2 S$ 2 6

8 E x X1

- I a S h S N 8 A F I F

I Un A

3 6

1 N

i S

Mu 3

6 1

S A

B L

P M

Fu X Sh A

e u X S' '

M h

I S

a A un 3 6

E 3

6 M/

A MMS 8 1 N

Fu x A X M

\-

Sh

\ >

d I

N wi O o: _ R UR O 0x . T E x 5 E D S

i 2 RA g- E

- H N N u H n m

_ U . (

1 T

E i

2 g u

rmt a Y R 6

j'f N'A U" /" Wa ,

,s '> i

,y lJ ' .L L i~

i tree analysis of the two pump system (see. Fig. 3), using-the Applicant's l modification of the success criterion for the'ESW cooling tower fans: 1 E from availability of one pump and two fans of the cooling. tower associated

.with the operating train to availability of one pump and any one of.the >

four Unit-1 fans from both cooling towers. In addition, the BNL estimate for the failure rate of an operating ESW pump was based on a generic L value for Westinghouse pumps of 9x10 8/hr (*) The results indicated : i frequency.of failure of running pump = 9.4x10 2fyp,  !

'E * '

probability of unavailability of standby. pump = 1x10 2/ demand, '

l so that for assumed independent failures of the two pumps, the overall e

j' failure frequency for the two pump EWS is seen to be about 9.4x10 4/yr.

l With cross-tie availability of a third ESW pump from Unit 2, the frequency .

j s

~

of the loss of ESW initiator is obtained by multiplying the frequency 1 .

j (with only two pumps available) by the probability of non-availability n .

M of a pump from Unit 2. This probability was estimated as .04, as follows.

7[. The probability of human error of failing to effect the transfer in 1/2 j hour was estimated at .02. The hardware failure probability assumed is

]ij about twice the hardware failure probability of the Unit 1 standby train,

j. to take into account the fact that certain valves must change state to

• l effect the transfer, and to take into account increased maintenance un-i availability of the Unit 2 ESW pump. On this basis, the overall result

.]

for the 3 pump configuration is an annual frequency for loss of ESW of about 4x10 5/yr.., or about a factor of 25 lower than the corresponding frequency for the two pump configuration.

I~

l 4

e

a ..

1 J

1 l

(It-is to be noted that the above estimates for loss of ESW frequency may ,

be somewhat non-conservative in that the probability of common cause failure between the running and standby pumps has not been includad. A rough estimate, assuming a 1% probability the standby train fails, '

t given failure of the operating train, yields a factor of two i

increase in the loss of ESW frequency for the two pump system.)  !

j t

f 4.1.2 E' stimate of Reduction in CMF With Addition of Third ESW Pump: The BNL review was originally directed to the Byron 1 two-ESW pump configura- l tion, so that an evaluation of the effect of adding another standby Unit-2 ESW pump (as proposed by Ceco on Deceniber 6,1985) was not included as i i

,1 part of the BNL contractual effort. However, an approximate estimate of '

i the reduction in CMF may be obtained from examination of the ESW-related i l

cutsets and single event importances obtained as part of the BNL analysis.  !

The major part of the ESW-related contributions to CMF is the. top sequence

, consisting of the loss-of-ESW initiating event frequency combined with a

.i

50% seal LOCA probability. Other ESW-related contributions to CMF are  ;

a 1 i' associated with the initiation of other transients (e.g, turbine trip, I reactor trip) combined with the estimated unavailability of the ESWS and I

a 0.5 seal LOCA probability. Taking into account the factor of 25 reduc- -

tion in the estimated frequency of loss of ESW obtained with the addition i

of the third ESW pump, the Byron CMF is estimated to be reduced from l  ! 7x10 4/yr. to a value a little less than 2x10 4/yr. , of which roughly one-

! i tenth is ESW related, i

I I

r

• 4.1.3 , Estimate of A0T Impact for Case of Three Pumps: For the ESW two-pump case, the increase in CMF due to an increase in T, from 19 hrs. to 168 hrs. for the group of nine systems was about 4x10 3/ year, with a related slope of about 3x10-5/yr for each additional hour in T, (Table 1).

The main contribution to the CMF is that associated with the loss of ESW initiating event sequence, and taking account the above-mentioned factor of 25 reduction in initiating event frequency obtained with the additic) of the third pump, it is estimated that the increase in CMF for the three pump case as T,is increased from 19 hrs. to 168 hrs. is about 1.4x10 4/yr. , corresponding to an increase in CMF of about 1x10 8/yr.  !

for each additional hour of maintenance.

As an example, if we assume a mean time-to-repair for an ESW pump for the 7-day A0T condition to be 50 hours5.787037e-4 days <br />0.0139 hours <br />8.267196e-5 weeks <br />1.9025e-5 months <br />, the increase in CMF above the baseline

, value of nearly 2x10 4/yr. for the three pump case is seen to be AT,x slope = (50-19) x 1x10 8 or about 3x10 5/yr. An increase in CHF of this i

magnitude for maintenance events on the ESWS is not considered acceptable.

J

, It may be further noted that while use of a parameter such as mean main-tenance duration time is useful analytically, a particular maintenance -

job requiring a full 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> would not be ruled out given a 7-day A0T technical specification. Therefore, if the A0T is considered to be means for constraining the risk, events extending the full A0T must be taken into consideration. This consideration is underscored in the case s

where the slope of the CMF vs. T, curve is appreciable and the system in question represents a vital support system such as the essential

. service water system.

L,-  :

)

i l

5. Conclusions 5.1 The increase in CMF associated with changing T, from 19 to 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> is small for the following systems: charging pumps, safety . ._

i injection pumps, RHR pumps, CCW pumps, containment spray pumps, and the fan coolers. These results support a recommendation to permit an increase in the Bryon A0TS for 3 days to 7 days for these systems.

i 5.2 For the two pump ESW Byron configuration, the BNL results indicate '

the rate of increase in CMF of 2.5x10 s/yr. for each additional hour in mean maintenance duration time to be unacceptably high. Even for the three pump configuration, the estimated slope of 1x10 6/yr. per hr. is considered to be excessively high for a vital support system of this type. Accordingly, the results obtained do not support a recommendation for an increase in the A0T for the Byron-1 ESWS, even j with a third pump from Unit 2 available as backup.

I l 5.3 The results in Table 1 show the slope ACHF/AT, for the diesel 1

generators in Byron 1 to be about 1x10 8/yr-hr. For reasons similar to j l

those indicated above for the ESWS, we believe this rate is excessively ]

high and does not provide a supportable basis for increasing the A0T in an important b'asic support system such as the diesel generators.

1 l

l 1

l l

~

~

~'

__ _ _ _1_

j V Ja; .

5.4 The slope of 6CMF/AT, for the AFWS is approximately 4x10 7/yr per hr.,

which while less than that for the ESWS or diesels is still appreciably high. From the ZPSS, the Zion plant specific data for mean duration cf  ;

maintenance events on_the motor driven AFW pumps is found to be about 50 hours5.787037e-4 days <br />0.0139 hours <br />8.267196e-5 weeks <br />1.9025e-5 months <br />. ifthisvalueisassumedapplicabletotheByronAFWpumps,

~

the increase in CMF would be AT, x slope = (50-19)'x 4x10 7/yr-hr. or

,- about.1x10 s/yr. We note, too, that the Byron AFWS is a two pump system, with one of the pumps driven by a diesel engine - an AFW pump type for which relatively little generic information on maintenance  ?

and failure rates is available. The attendant uncertainties regarding 4

this pump type, plus the A0T risk impact results obtained for this two pump system do not support approval of the the Applicant's proposal to increase the A0T for this system.

I j S.5 The BNL reference calculational results indicated that as a single event, the seal LOCA contributed about 88% of the overall CMF of 7x104/yr. for the current 3-day A0T condition. As indicated in Fig.

-i 2, a reduction in the assumed value of 0.5 for the seal LOCA probability j would appear to reduce the initial (3-day A0T) value of CMF and the slope 1

of ACMF/aT,. However, a reduction of the seal LOCA probability for the

]

j case of the loss-of-ESW initiator is probably not warranted, even if

.]_ the issue of the large seal LOCA (e.g. , 300 gpm per pump) is resolved.

j As discussed in Section 3.2, an expected seal leakage rate of about

20 gpm per pump (under loss of seal cooling conditions without a seal l 1

{

l _

failure) can lead to core uncovery before the ESWS can be repaired. In the case.of station blackout-initiated seal LOCAs, resolution of the i.

large seal LOCA issue would lead to a reduction in the importance of p

the diesel generators, and the slope of the CMF vs. T, curve would then decrease considerably. Ilowever, station battery depletion time would then become an important consideration; battery depletion effects were '

not considered in the staff analysis.

I 5.6 We believe it-is important to note that the risk perspective '

obtained in our evaluation is strongly dependent on the assumptions of the frequency and duration of maintenance events given in the Applicant's submittal. A 7-day A0T conditions would permit the utility to take up to a full 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br /> to complete repair of a component, and .

although the statistical mean duration time might be low compared with i

168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br />, the possibility of requiring the full A0T to complete the

• )

maintenance event does exist, and from the plant-specific data given in )

l

! the Zion PSS long maintenance duration events are not rare. The Byron j i

[ technical specifications do not, in general, provide a control on the frequency of maintenance events, and therefore on the cumulative outage j times for the various systems, The staff believes that if the A0Ts for f I

certain systems were to be increased, reasonable prudency would appear  !

to warrant self-monitoring actions by the utility to obtain trending information on the frequency and duration of maintenance events, especially for the risk important ESW and DG support systems, and for l the two pump AFW system.

l l

3

~

1 T_L_-_l_ -

~ ~ '^~

1

s'  %

f References  !

1. Letter from L. O. Del George, Commonwealth Edison Company, to Harold Denton, May 2, 1984.

, -2. Procedures for Evaluating Technical Specifications (PETS) Program, f

3. " Byron Generating Station Limiting Conditions For Operation Relaxation-Program," Westinghouse Electric Corporation, April 1984.

4. " Zion Probabilistic Safety Study," Commonwealth Edison Company of Chicago, Fall 1981.

l

5. Letter from K. Ainger, Commonwealth Edison Company, to Harold Denton, December 23, 1985.  !

6. W. E. Vesely and J. L. Boccio, "The Use of Risk Analysis for Determining Technical Specifications: Issues and Review Considerations," Draft Report, December 20, 1984.

7. " Byron AFW Reliability Study," Prepared by Torrey Pines Technology l Company, 1982.

l

?

l t

i I

i fW -

e-*- v e n m.

L______.._ __