ML20245A414
| ML20245A414 | |
| Person / Time | |
|---|---|
| Issue date: | 03/20/1987 |
| From: | Congel F Office of Nuclear Reactor Regulation |
| To: | Minners W Office of Nuclear Reactor Regulation |
| Shared Package | |
| ML20236K304 | List:
|
| References | |
| FOIA-87-380, TASK-082, TASK-82, TASK-OR TAC-M53347, NUDOCS 8703270329 | |
| Download: ML20245A414 (22) | |
Text
_ _ _ _ - _ _ _ _ _ _
- /,,,,,o %,1 UNITED STATES 3'
NUCLEAR REGULATORY COMMISSION 7
3 o
\\
r$
M5mG TON. D. C. 20555 l'
~
/
'Wm 2 0 MAP, ISI/
MEMORANDUM FOR: Warren Minners, Chief 3
Reactor Safety Issues Branch Division of Safety Review and Oversight FROM:
Frank J. Congel, Chief Reliability and Risk Assessment Branch Division of Safety Review and Oversight
SUBJECT:
GI-82*BEYONDDESIGNBASISACh1DENTSIN SPENT FUEL STORAGE p00LS" (TAC NO. M53347)
Reference:
Memorandum from W. Minners to F. Congel, dated February 2,1987, on the above subje,ct.
In accordance with the request of referenced memorandum, we have performed the review of Sections 2, 4, and 5 (with emphasis on subsections 2.2 and 2.4) of a draft report from BNL on the subject issue.
Our detailed comments are provided in the enclosure. These comments are divided into the following four categories: (1) Methodology;(2) Seismic Hazard; (3) Fragility; and (4) Consequence analysis.
In sumary, our major conclusions are as follows:
1.
A system'atic approach is needed to assure appropriate coverage of the subject.
In addition, use of logic diagrams is needed to support some of the reliability estimates; 2.
The assumed high seismic hazard and high fragilities (low capacities) may l
lead to overestimation of seismic induced frequency of structural failure i
of pool; and 3.
More information is needed in the area of the consequence analysis.
Finally, the seismically induced spent fuel pool accidents must be viewed in conjunction with the likelihood of other accidents such as core damage ecciderts.
seismic capacities of the spent fuel pool structures are expected to be comparable or higher for most plants than the components which would govern core dtmage sequences. Therefore, examining spent fuel pool accidents alone may lead to a distorted view of the importance of the seismic event. This is not to say that the seismic initiators should be disregarded for the spent fuel pool accidents, but if the outcome is controlled by major structural catastrophes rather than failures associated with low level earthquakes, the risk from fuel pool failure must be examined in the context Y}'_
of concurrent core melt accidents.
e~~ T f
,d[
l
,f l
I
/
ks I
J Warren Minners -
Please contact N. Chokshi (x28347) of my staff if you need additional infomation or clarification.
D')d,J Frank J. Congel, Chief Reliability and Risk Assessment Branch Division of Safety Review and Oversight
Enclosure:
As stated cc:
T. Speis B. Sheron R. Barrett H. VanderMolen E. Throm A. El-Bassioni L. Reiter S. Acharya N. Chokshi C
l ENCLOSURE Reliability and Risk Assessment Branch Connents on Draft Report "Beyond De' sign-B6 sis Accidents in Spent Fuel Pools" Methodology 1.
Our review of the reference report and the methodology has indicated the following:
The adopted approach is not systematic enough to assure the reader that all of the identified aspects of Generic Issue 82 are covered.
Functional failure of the spent fuel pool involves one or more of 1
the following off normal conditions:
a.
Loss of pool cooling function; b.
Loss of pool inventory; and c.
Loss of configuration / integrity of fuel assemblies.
These conditions can be related to their causative factors, and then consequences should be addressed systematically to evaluate their significance. As an example, loss of configuration or loss of integrity of fuel assemblies can be induced by an energetic event like an earthquake or heavy object being dropped.
In this case.
radioactivity will be released from the pool or possibly configu-rational deformation may result in criticality. Evaluation of the likelihood and significance of such consequences should thus proceed in a systematic fashion tc 9ssure an appropriate coverage of this issue.
The report also lacks rigorous application of established Probabil-istic Risk Assessment (PRA) methodology.
Estimates of unavail-abilities for cooling trains or frequencies of events involving pool heatup were used without any analysis to support these estimates.
Loss of pool inventory can occur slowly or in a rapid fashion.
In case of seismic events or possibly fires, conditions might exist that deprive the operator of control or surveillance capabilities. Such scenarios were not considered in the reference report.
Seismic Hazard 1.
Weighted estimates of seismic hazard proposed for the Millstone site are typical of the high range of hazard for that site as discussed in NUREG 1152 (Millstone 3 Risk Evaluation Report). Utility and EPRI estimates however would be approximately an order of magnitude less at a given acceleration (alsoseecomment5).
2.
The choice of an upper bound cut-off to acceleration has not been justified. The report indicates that this choice is important.
o 2.
The weighting scheme using both test estimates and different percentiles is unwarranted. However, use of a more appropriate weighting scheme (as in NUREG 1152) would appear not to be significant.
l 4
The extrapolation of the Seismic Hazard from Millstone to Ginna appears to be in error. The error however is small compared to other considerations.
5.
The use of a single hazard curve in an absolute manner is unwarranted given the large uncertainties. A better way to use hazard curves is to examine the sensitivity of conclusions to the choice of different hazard curves.
In this case sensitivity tests could include:
a.
a mean hazard curve an order of magnitude less than has been proposed b.
an upper hazard curve a factor of 3 higher than that has been proposed.
(roughly equivalent to the LLNL 85th percentile curve.)
c.
a lower hazard curve two orders of magnitude lower than has been proposed (roughly equivalent to the 15th percentile of the utility sponsored hazard curve.)
Conclusions which are unaffected by these sensitivity tests may be considered robust with respect to uncertainty in hazard. Those which are affected by different hazard assumptions need to be reexamined.
6.
Additional comments and corrections appear in the margins of the text (Attachment I to this enclosure).
It should be rewritten without endorsing any particular hazard curve.
Hazard could vary substantially from site to site particularly if western U.S. sites are being considered.
Fragility 1.
Seismic fragility of pool structures are based on the following two estimates:
I a.
The fragility curve for the Oyster Creek reactor building; and 1
b.
The fragility curve for the Zion auxiliary building shear wall. '
The fragility or the capacity for the Oyster Creek reactor building is expressed in terms of the peak ground acceleration level; however, in the Zion report the fragility of the shear wall is expressed in terms of the local floor acceleration (see attached Fig. 22 from NUREG/CR-3558,
- to this enclosure) at an elevation in the structure. The current study does not account for this difference in these two fragility curves and treats them both as expressed in terms of ground acceleration.
A correction factor which accounts for the response amplification should be aoplied to the Zion shear wall fragility curve to convert it to the ground acceleration level.
Capacity of the shear wall expressed in the ground acceleration level will be less than that expressed in the floor acceleration level.
4 2.
We believe that the seismic capacities used in this study are overly conservathe. The Oyster Creek reactor building represents an early vintage plant. The Zion shear wall is a composite wall consisting of braced steel framing with in-fill reinforced concrete panels. As such, it is not a good representation of a typical shear wall (In fact, the fe41ure of this wall is initiated by the failure of shear studs which are required for composite construction). Spent fuel pools generally have much heavier walls and are located at lower elevations. Attached Table C-10 (Attachment 3 to this enclosure) from NUk/CR-4334 lists shear wall capacities (both shear failure mode and flexural failure mode) for a number of plants.
These capacities are generally much higher than those assumed in the current study.
Further, as noted in NUREG/CR-4334, the shear wall capacity estimates were deemed to be conservative for the b
following reasons:
The limits on inelastic energy absorption capability (ductility a.
limits) were judged to correspond to the onset of significar.t structural damage. This damage level represents a conservative lower bound on the level of inelastic structural deformation which might interfere with the operability of components (i.e., piping and
.L equipment) attached to the structure.
b.
The capacities noted in Table C-10 are for individual shear walls, j
usually corresponding to the lowest capacity wall among a number of i
walls utilized to resist lateral seismic forces in a structure.
Typically, no redistribution of lateral loads to the other walls has l
l
been considered in estimating the lowest capacity in these PRAs.
In some instances, such redistribution analysis was found necessary to arrive at realistic capacities of shear walls (Plant E).
We recommend that an attempt to estimate more realistic fragilities (capacities) for the spent fuel pool walls be made. A number of simple approaches are available to do this (For example, NUREG/CR-4293 -
Reliability Analysis of Shear Wall Structures). A simple survey of a few pool structures may reveal whetiier these failure modes are possible or not.
3.
A number of assumptions are made in Section 2.4 - Pool Structural Failure Due to Heavy Load Drop - regarding the structural failure estimates.
If this is deemed to be an important sequence (as it appears from Table 2.9) a more detailed analysis may be necessary.
It is not clear whether seismic induced crane failures (e.g. dislodging of trolly or structural failure of crane members) have been considered in this study.
In the past this issue has come up at the spent fuel pool expansion hearing of a few plants (e.g.BigRockPoint).
Consequence Analysis Information provided in the Consequence Analysis chapter is inadequate. The following items should be properly add: essed:
{
I It is stated in the chapter that the CRAC2 code was used fer offsite a.
radiological consequence calculations.
However, besides the curies of radionuclides released, the other parameters characterizing the release i
scenarios, namely: the timings, height, and energy of release which are also required for input to the CRAC2 code are not provided. They should be evaluated and listed.
1 b.
Curies of I-131 listed in Table 4.3, Col. 2 is inconsistent with the estimate prescribed in the paragraph for Case 1 in page 4-4.
How were the radionuclides Rh-106, Sb-125, Te-125m, I-129, Ba-137m, c.
Pr-144, Sm-151, Eu-154, and Eu-156 listed in the source tem tables in the chapter treated in the CRAC2 framework? The radiological significance of b
these radionuclides relative to the other radionuclides of the CRAC2 list should be evaluated. Ordinarily, these radionuclides cannot be input to the CRAC2 code as part of the release, although CRAC2 will process some of these radionuclides only as daughter products during the environmental transport of their parents, if the parents were in the source term, The uncertainty range of Te release fraction in Table 4.2 is missing and d.
should be shown, The Ba release fraction shown in Table 4.2 is not consistent with the e.
value stated in Section 4.2.1 in page 4-2.
f.
Table 4.2 does not include results for Cases 3 and 4 of Table 4.5 which should be provided.
Types of consequences that should be shown should be expanded.
Since the g.
magnitudes of some of the releases are quite large, the consequence items such as offsite dose versus distance, early health effect potentials as
~
well as other types of consequences calculated by the CRAC2 code should be provided. Use of one full year worth of the actual site meteorology and site data for population distribution, land use and economic data of the two sites analyzed in the CRAC2 code should provide a fuller spectrum of consequence results which should be displayed.
s k
ATTACR r4G A/r 1
To f_aJc.t.M U $C-24 unproductive to refine the structure and equipment capacity calculations to accuracies which are inconsistent with the hazard uncertainty."6 The specific applicability to spent fuel pools of Reed's assertion is discussed in Section 2.2.1.3.
2.2.1.1 A Review of Seismic Hazard Data f
The primary difficulty in characterizing the seismic hazard at specific hY Isites in the Eastern United States (EUS), i.e., sites to the east of the Rocky
? j'Hountains is that severe earthquakes are rare events in the EUS.
t' A systematic
'f
/
analysis of recorded earthquakes and their relationship to geological features 9
D i
as yielded seismic zonation maps of the EUS.7 However, such information can-not readily be translated into the type of seismic hazard functions needed as
,1 input for PRA, Consequent 1) available historical data are insufficient for obtaining meaningfu site specific estimates of the frequency of sever g,
events.
0g 4
h A different approach to seismic hazard analysis has been developed at Lawrence Livermore National Laboratory (LLNL) by D.L. Bernreuter and his col-leagues under NRC sponsorship. The initial study was a part R s]y tematic Evaluation Program (SEP).e The methodology has been. Sf a P m in a subsequent
- study,
- EUS Seismic Hazard Characterization Project"
( 5440 ). '
- 1 0
( suce)
Since the SEP and SHC results will be used for the seismic hazard esti-44 6 mates, some further discussion of the uci niciter methodology is appropriate.
Three basic steps are involved:
1.
the elicitation of expert opinion to delineate and characterize seis-micall active zo es in the EUSt a -
(,. 4 m e N8Wda btR. f'[ f 5 4f'N0lI fu"Wh'h 'N y
'wolfN!"'
2.
sing the i put data of each expert,4 the co-ion of the seismic hazard u t specific reac ites using several alternative ground motion attenuatio h site corrections, and integrat-ing over each of e delineated seismic zones; and, finally, Sek,an.%,.seli.wf[uobAtm J ~ h ~ ~f A *"
D Is r& y ln n,}
$ af UN br L +k m,/6w JA u.tpeJr vaA4 ae t
& & ded A & 6- ( d s d E U erdphoAS A,u,AG,4,
2-5 3.
t comb ning of the separate expert data accompanied by the gen ra-ion
- ncertainty 1. t s f om the spr xpert om ons nd fr the self-eva tions of ch ex on the e of confi a
ntheirg inion.
The various steps are carried out in a highly disciplined and systematic manner.
Provision is made at various stages for peer review of the methods and input opinion, feedback to the experts and critical evaluation of the re-sults.
N s
In step 1, each expert prepares a "best estimate" map which delineates the seismic zones.
Each zone is characterized by a' set of parameters that give the maximum earthquake intensity to be expected for that Zone (upper mag-nitude cut-off), the expected frequency of earthquakes, and the magnitude re-currence relation.
For each input (zone boundries, seismic parameters), the expert provides a measure of his degree of confidence.
Also each expert is given the option of submitting alternative maps of differing zonations and characterizations (up to as many as 30 maps).
The data froc each expert are
{,
evaluated separately through step 2.
In step 2, the contribution at a given site from each zone is integrated over the zone area and then over all zones.
This requires the use of ground motion models for which a range of alternative models are employed to yield a
/.
set of alternative hazard curves.
A " Ground Motion Panel" of experts have
(, r,, S lh selected several alternative models to be used, each having a weighting factor (see Ref. 9 App. C).
Also each ground motion model incorporates a site spe-cific correction to account for local geology.
~
F./yfM
, ytep 3, the results of the individual experts are combined to obtain a
> "best estimate" hazard curve and the uncertainty bands are computed in ternative ways.
It is obvious that the methodology requires a massive data collection and computer ef fort.
In its present state, the final results are not in a form to be easily applied to a specific PRA by a non-expert in seismology.
Further work is needed to develop a more convenient fonnat for presenting the final
2-6 results.Il In particular, numerical tabulations of the sets of hazard curves (such as those shown in Figs. 2.1 and 2.2) and their derivatives, dH/da j,
f or each reactor site would be helpful.
Also, it appears that the local site geology needs more rigorous consideration in the derivation of the hazard r,.urves (see below).
Members of the Peer Review Panel have suggested several ways in which the tr.ethodology could be refined (see Ref.10, Section 7 and Appendices 0.1-D.a).
Many of these suggestions were implemented in the final feedback process and were included in the final results reported in Ref.10.
n gen ral, the re]
l f viefers agrepd that the Jesults are "c-edible and as goo as pre ent sc nti -
L ic understanding of eastern U.S.
(EUS:
seismicity pro bly all ws" (R }. 1 i
A
. D.1).
(
[
Core nts from NRC licensees and their consulta s indicated objections t
[
applicati n of results to s ific sites, tin that the site specific' correcti n f tors in the gr und otion mod s were too simplif d to ade-quately ak local eologic factor into ccount (Ref.
pp. D.6).
Al so,
the crit 4 ism was that the res s were not adequately tested agains nt historical records.
In order to illustrate the hazard curves, their range of uncertainties and comparison with other studies, a series of figures taken from Ref.10 for the Millstone site is reproduced in Figs. 2.1-2.4 Figure 2.1 is the hazard curve obtained from the "best esti-mate" results for all experts in the SHC study (including the seismic and the ground motions panels).
The curve plots frequency of exceedance per year vs.
peak ground acceleration.
Figure 2.2 illustrates the uncertainties in the hazard curve (15, 50, and 85 percentiles) derived from the spread in expert opinion and the self-confidence factors in the input parameters.
It can be seen that the spreac between the 15 and 85 percentiles is about a factor of 20 at low PGA increasing to about 350 at the high PGA.
Comparison of Figs. 2.1 and 2.2 shows that
2-7 the "best estimate" curve is considerably higher than the 50 percentile, i.e.,
the mean > median.
j Figure 2.3 illustrates the spread in the "best. estimate" hazard curves for all of the experts participating in either the SEPe or the SHC' studies, or both (6 experts participated in both studies).
The spread ranges from about one order of magnitude at lower PGA to about 1.5 orders of magnitude at the higher PGA. The curve marked 'A," which falls considerably below the main grouping, was derived from data input intheSEPstudybyogeofjhe, experts who participated in both gujies.
ThIs revised input for $a"SHE project g
the derived curve by an order of magnitude at the low acce?erations and se h
-- 3 two orders of magnitude at the higher PGA, this raises. the obvious question of whether the experts were somehow influenced by the opinions of their colleagues, or whether the revision resulted from a more careful consid-eration of the various geological factors that were taken into accour.t in pre-paring the input parameters. The question of testing the results for inadver-tent biases of this nature was addressed by the Peer Review Panel members, but their recommendations could not be fully implemented in the final report due h,
to limited time and budget (Ref.10, pg. 7-3).
Figure 2.4 compares the "best estimate" hazard curves for the individual 3
SHC experts with curves generated from zonation maps prepared by the U.S. Geo-logical Survey (USGS)12 and historical data of the past 280 years. As can be seen, the USGS hazard curve (denoted by "X")
lies above the SHG data.
Bernreuter et al. attribute the difference between the SHC and the USGS curves to the variations in the equations used for conversions from intensity to mag-nitude and in the values for the rate of earthquake recurrence (Ref.10, pg.
8-1 et seq.). As would be expected the 280 year, historical hazard curve (de-noted by "H") falls below the SHC dat6 because it does not include postulated stronger earthquakes with return times much greater than the time span of the historical record.
It should be noted that recent research has raised significant questions concerning the frequency of strong earthquakes in the coastal zone of the EUS.13 The speculation has arisen from paleeseismic field studies originally focused on the region of the strong earthquake near Charleston, SC, in 1886,
which produced isny " sand blows "I"i l5 These result from the liquef action and venting to tie surface of sub-surf ace water-saturated sediment.
Several sand blow craters have been found for which radiocarbon dating indicates that moderate to larte earthquakes have recurred in the Charle*, ton region on an average of about every 1800_ years.18 The latest (prior to 1886) occurred about 1100 years ago.16 Sand blows from prehistoric earthquakes have been un-earthed recently in.the region extending from near Savannah, GA as far north as Myrtle Beach, SC.17 The broad extent of sand blows suggests that, Charleston-type earthquakes might be associated with some tec%nic featu which extends for some distance along the east coast and not uniquely centered near Charleston.
Up to the present time, no systematic field search has been made for sand blows outside of the Savannah to Myrtle Beach region.18 Recently Thorson et al. reported the existence of apparent sand blow craters in eastern
(
Connecticut.lS These craters were recently examined by a USGS field team and assessed as not being of the same nature as those observed in South Carolins.le 2.2.1.2 Seismic Fazard Estimates for the Millstone and Ginna Sites b
The "best estimate" and the median,15 and 85 percentile seismic hazard curves developed by the SHC project for the Millstone site are shown in Figs.
2.1 and 2.2.10 These four hazard curves were used to develop the estimates of the seismic failure probabilities of Millstone 1, as described in Section 2.2.1.4 below.
Hazard curves, such as shown in Figures 2.1 and 2.2, arMcted to y
some upper limit cutoff, i.e., PGA's which dould never be exceeded.
We have assumed that the upper limit cutoff for the Millstone site occurs at approxi-mately I g (980.7 cm/sec ), but a different cutoff would give a substantirdy 2
different pool failure frequency.
Seismic hazard curves for the Ginna lite were not generated in the SHC project;20 however, the SEP project included data for Ginna.e Unfortunately, the fortnat of the SEP results, which were directed primarily at obtaining site specific spectra, cannot readily be translated into a "best estimate" hazard In want of a better procedure, we have synthesized a hazard curve for curve.
l
2-9 Ginna from the Millstone curve, using ratios of PGA's for 200, 1000, and 4000
(
year return times, tabulated for the two plants in the SEP study.e The hazard f'
I curve resulting f rom this synthesis is shown in Fig. 2.5.
Because of the go higher kipper magnitgp+o" at the Ginna site, as perceived by experts,
j j
d
.(
stone': M id 8.0 vs. GinnaJ MMI =
we have assumed the upper cut 6 g
[ ( off PGktWThTTiazard curve to be 7though this is recognized to be a somewhat pessimistic assumption, it serves the useful purpose of illustrating the sensitivity of the calculated seismic risk to the upper cutoff of the haz-ard curve.
2.2.1.3 Seismic Fragility of Pool Structures Fragility curves specifically for spent fuel pools have never been devel-oped.21 It is necessary therefore, to rely on fragility assessments for other structures which appear to be of similar construction to spent fuel storage pool s.
It must be recognized that this procedure introduces an additional element of uncertainty in the final risk estimates -- an uncertainty that is dif ficult to quantify.
Another source of uncertainty is the degree to which the stainless steel lining of a pool would enhance the seismic strength'capac-ity (i.e., reduce the fragility).
Conceivably, the reinforced concrete struc-ture of the pool could crack without loss of integrity of the pool lining.
The dilenma of selecting an appropriate fragility for a BWR plant is aggravated by the fact that the pool structure extends typically from the 60 to the 100 foot elevations above grade with the resultant amplification of the seismic bending stresses relative to the lower elevations of the structure.22 For the present analyses, two, somewhat diverse sets of fragility esti-anates, have been used:
- 1) the fragility curve developed by R.P. Kennedy et al.23 for the Oyster Creek reactor building; and
- 2) the fragility of the Zion plant auxiliary building shear walls (north-south ground motion).2*
2-10 In each case, the fragility curve is defined by the following equation:
F(a) = c [(in a/h/eR3,
(2.1) where F(a) is the probability of structural failure given a peak ground accel-eration, PGA = a.
- (.) is the normal distribution function, is the median fragility level (i.e., the acceleration at which there is a 50% probability of failure) and SR is the logarithmic standard deviation expressing the random-nen in the value of A third parameter, 8u, is used to express the un-certainty in the median value and is used to generate upper and lower confi-dence limits.
For example, it can be shown that the substitution for in
+ 8u and
= e-8u Eq.
2.1 of
=
e generate respectively the 84 and 16 percentile curves.
Thus, a set of fragility curves can be generated from three parameters, 8R and Su, The data used for generating the
- Kennedy" and the " Zion" curves are given in Table 2.4 Kennedy notes that the estimated median fragility value of about 0.75 g is considered applicable to plants designed in the U.S. in the mid 1960's.
The Kennedy fragility curve is shown in Fig. 2.6, with the 84 and 16 percen-tile limits.
The corresponding Zion curves appear in Fig. 22, pp. 3-35 of Ref. 24 2.2.1.4 Seismically-induced Failure Probabilities The convolution of the derivative of a seismic hazard curve (e.g., Fig.
2.1) with a fragility curve, yields the annual probability of a seismically-induced failure. This can be expressed by the equation:
hj(a))da, P
F (2.2) 4,3 =
where Pj,j is the failure probability obtained from the convolution of hazard curve i with fragility curve j, dH/da i is the derivative of the hazard curve 1 (i.e., the annual frequency of occurrence of peak ground accel-
- eration, a,
and F(a)j is failure probability at acceleration, a,
for
2-11 fragility curve j. Tne integration is cut of f at the upper limit expected for the PGA.
Since the seismic hazard curve is not an analytic function, the derivative dH/da and the integration are carried out numerically.
Given many hazard and fragility curves from which to choose, and there being ne a priori basis for choosing a particular pair, the convolution ex.
pressed in Eq. 2.2 can be carried out for each pair of curves with weighting "
factors assigned to each of the curves in each set.
The resultant collection of Pj,j gives a probability distribution which expresses the uncertainties in the analysis.
The probability density distribution obtained for the Millstone site is shown in Fig. 2.7.
At least in principle, the various hazard and fragility curves (sets i and j) do not have an equal likelihood of being correct. Therefore, a weight-ing factor (wi or ej) should be assigned to each curve which reflects an
" engineering judgement" of its relative validity.
The mean probability for failure is then derived from the following expression, E
- Iwj wj g,) / {w$
P (2.3) t h
1.
(The weighting factors where {wj 1,
{uj 1 and Juj,j
[wj wj
=
=
=
=
.d' assigned for the Millstone case are given in Table 2.5.\\ As can be seen from the table, the "best estimate" hazard curve has been assigned a weighting fac- {lj./ 'AM d
ter of 0.5 with the remaining 0.5 distributed among the median,15 and 85 per-
)
centile curves.
The " Kennedy" set of fragility curves were assigned a totap p f weighting f actor of 0.75 with the remaining 0.25 distributed among the " Zion" j h>[ #(.
set. Assuming an upper limit cutoff of 1.09, the mean probability of failure, Tr. derived from the 24 sets of Pj,j, using the weighting factors listed (v' '
in Teble 2.5 and Equation 2.3, was Ty = 2.2x10-s/ year (Millstone).
In the case of Ginna, only a single hazard curve (Fig. 2.5) was used, there being insufficient data to generate median,15 and 85 percentile curves for this site.
Because of the structure of the Ginna spent fuel pool, the
" Zion" fragility curves are more appropriate, than the " Kennedy" curves.
2-12 4
inerefore, higher weighting factors were assigned to the " Zion" curves as shown in Table 2.5.
Based on an upper limit PGA cut-of f of 1.25, the mean 9
probability resulting from the convolution of the single hazard curve with the six weighted fragility curves was F = 1.6x10-5/ year (Ginna).
f The dif ference between the estimates for Millstone and Ginna, 2.2x10-5 vs.1.6x10-5, should not be regarded as highly significant, but more as an in-dhetion of the sensitivity of the results to the weighting factors assigned to the fragility curves.
i 2.2.2 Structural Failures of Pool Due to Missiles Missiles generated by tornadoes, aircraft crashes or turbine failure could penetrate the pool structure and result in structural failure.
The probability of tornado missiles depends on the frequency of tornadoes at the site, the target area presented to the missile and the angle of in-
{.
pact.
An analysis made by Orvis et al.25 for an average U.S. site derives a probability of <1x10-e/ year for structural loss of pool integrity due to a tornado missile (Ref. 25, pg. 4-44).
Similerly, the analysis for structural failure of a pool from an aircraft crash yielded a probability of <1x10-10/ year (Ref. 25, pg. 4-58).
The damage caused by Missiles generated by turbine failure depends on the orientation of the turbine axis relative to the structure, as well as the fre-of turbine failure.
An quency analysis by Bash yields a probability of
-4x10-7/ year for spent fuel pool damage from a turbine failure miss11e.26 In the case of Ginna, the probability would be several orders of magnitude saaller (i.e., essentially zero) because the spent fuel pool is shielded from turbine missiles by the primary containment.
2.3 Partial Draindown of Pool Due to Refuelino Cavity Seal Failures On August 21, 1984, the Haddam Neck Plant experienced a failure of the refueling cavity water seal, while preparing for refueling.
The water level
A ~l it% CkA M E tJ T
'2
'T)
E *J C. LJ.5 U $C"-
1.0 i
i i
i i
i A = 1.1 g d" = 0.12 0.8 d" = 0.20 E
5% confidence Median f0.6 o
.,=
0.4 o
ct
- - 95% confidence 0.2 limit 1
l 1
1 1
0 0
0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 Fragility parameter (acceleration north-south direction; control room floor stab, g) rigure 22.
Failure of auxiliary shear walls due to N-S ground motion.
diesel generator building shear walls f rom north-south excitation is shown in
(,
Tig. 6.
The median f ragility parameter for this mode of f ailure is expected to be an east-west acceleration of the control room floor slab of approxi-mately 1.1 9 The detaf1s of the analysis of this f ailure mode were presented in Sec. 2.2.3.
A number of concrete block pasonry walls are located throughout the auxiliary building. For the most part, these walls are not load-bearing; they may support an unloaded concrete slab.
The walls are typically constructed of 1-f t-thick concrete blocks, vet tically reinf orced and grouted. The evaluation of these walls was condacted using in-structure respor.se spectra generated in the original design analysis scaled up to the response acceleration level required to cause f ailure.
Failure of these walls may be expected to result in loss of function of any attached conduit or equipment but will be quite localized and will not affect any other structural eerder.
The fragility curves associated with masonry walls at elevation 592 f t are shown in s
Fig. 23.
The median fragility parameter associated with failure of the walls is approximately 1.7 g at the control room floor slab. Walls at lower elevations may be expected to have higher equivalent capacity.
3.4.4 Shear Wall failure for East-West Excitation The auxiliary building, including the diesel generator rooms and the f uel storage building, has higher seismic capacity to withstand east-west excitations than excitations in the north-south direction.
This is because the structure is essentially symmetric about the east-west axis and very little torsional response results for east-west excitations.
3-35
(
ATTAQN(:4T 3 To ErJ Clof V$b TABLE C-10 SEISMIC CAPACITIES OF SIIEAR WAILS plant component Failure A O
O HCLPF Generic or Remarks m
R U
Name Mode (g)
(9)
Specific Important A
Aux Bldg Shear 0.9 0.30 0.28 0.3 S
Risk Contributor Important l
B Control Shear 1.5 0.8 S
Risk Bld9 Contributor Primary Shear Shear 2.8 0.20 0.32 1.2 S
Walls Important C
Aux Bldg Shear 0.9 0.21 0.25 0.4 5
Risk Contributor Standby Shutdown Shear 2.6 0.28 0.30 1.0 S
Facility Standby Gas Shear 1.8 0.20 0.31 1.0 S
D Treatment
.r.
Turbine Non-Li Bldg Shear 0.5 0.23 0.29
- 0. 2 S
Cate' gory I Structure Reactor Building Shear 1.8 0.31 0.26 0.7 S
Aux L
Building Shear 1.5 0.31 0.26 0.6 S
Diesel Building Shear 2.1 0.31 0.26 0.8 S
Turbine Non-ca tegor y Building Shear 0.7 0.30 0.29
- 0. 3 S, G I Structure Reactor Important Building Flexural 1.3 0.31 0.25 0.5 S
Risk Contributor Diesel F
Building Flexural 1.6 0.24 0.24 0.7 5
Spray Pond Pu mp S t r.
Flexural 3,2 0.22 0.24 1.5 S
C-43 N
f 4
TABLE C-10 SEISMIC CAPACITIES OF SHEAR WALLS (Continued)
Plant Component Failure A O
m p
o HCLPF Generic or Remarks g
Mame Mode (g)
(g)
Specific G
Cont Bldg Flexural 7.6 0.31 0.32 2.7 5
Cont Enc 1 Shear 8.2 0.30 0.37 2.7 S
Bldg Primary Flexural 2.6 0.30 0.33
- 0. 9 S
Aux Bldg SW Pump Flexural 2.1 0.17 0.29 1.0 S
House SW Cooling Flexural 2.4 0.30 0.33 1.0 S
Tower 4
I Control & Flexural
- 3. 0 0.29 0.33 1.0 S
Diesel Gen l
Bldg.
Fuel Stor-Flexural 5.0 0.24- 0.32 1.9 S
age Bldg..
]
Secondary S h'e a r H
Cont.
Failure 3.3 0.56 0.51
- 0. 6 S
Primary Shear 9.2 0.49 0.48 1.9 S
Cont.
Failure l
Control Failure 1.7 0.42 0.49 0.4 S
Bldg.
at Base i
Aux Bldg N/S Shear Collapse 1.4 0.19 0.61 0.4 S
SSMRP-Zion Wall I
Aux Bldg E/W Shear Collapse 2.5 0.17 0.63
- 0. 9 S
SSMRP-Zion Wall Diesel Gen Bldg Collapse 1.4 0.16 0.64
- 0. 4 S
SSMRP-Zion l
l C-44
-