ML20054B728
| ML20054B728 | |
| Person / Time | |
|---|---|
| Site: | Waterford  |
| Issue date: | 04/02/1982 |
| From: | Hanauer S Office of Nuclear Reactor Regulation |
| To: | Mattson R Office of Nuclear Reactor Regulation |
| Shared Package | |
| ML20054B726 | List: |
| References | |
| NUDOCS 8204190143 | |
| Download: ML20054B728 (19) | |
Text
.
v.
,: n.
APR 2 1932 MEMORANDUM FOR:
R. Mattson, Director Division of Systems Integration FROM:
S. Hanauer. Director i
Division of Safety Technology
SUBJECT:
FEED AND BLEED ISSUE FOR CE APPLICANTS Per your request, we have reviewed the F. Rowsome memo of January 29, 1982, Feed and Bleed Issue for CE Applicants and our assessment is in the enclosure.
We do not believe that there is any need for imediate action as noted in the summary of the enclosure.
\\
l M-x,. t. 9;
.uw,<
s.
.3
' Stephen H. Hanauer, Director Division of Safety Technology
Enclosure:
As stated cc:
H. Denton E. Case R. Vollmer H. Thompson R. Bernero F. Rowsome T. Speis l
L. Rubenstein J. Knight M. Ernst F. Schroeder K. Kniel A. Marchese A. Thadani S. Israel O. Parr I
R. Tedesco 7 8204190143 820412 PDR ADOCK 05000382 G
0 6
o EVALUATION OF FEED AND BLEED AS BACEUP TO AFV AND VERY SMALL LOCA CORE MELT FREQUENCY
- ~
l
)
TABLE OF CONTENTS 1.
Introduction 2.
Suma ry 3.
Discussion i
3.1 Transients and Failure of All Feedwater 3.2 Loss of Offsite Power and The Availability of Two Motor Driven AR1 Pumps 3.3 Very Small LOCAs and Failure of HPSI
^
4.
Re ferences i
1 i
i I
I
(
t l
I..
1.
Introduction Based on a quick and dirty analysis, core melt estim.ites pre,cr:ed in Ref. 9 of loss of main feedwater, loss of offsite power, and very snull LOCAs suggested that improvements could be achieved by implementing certain recommendations.
In particular, it was recommended that feed and bleed capability te installed on CE plants (which do not have PORVs) as a backup to AN to cope with loss of main feedwater events; that the non-safety grade AN pump be capable of being powered from an emergency bus at CE plants; and that HPS) reliability be re-viewed and the frequency of very small LOCAs be reviewed and reduced for all PW Rs. These issues fall within the unresolved safety issue of decay heat re-moval (USl A-45). We have reviewed the sequences of interest and compared the estimated frequencies with those calculated in WASH-1400.
2.
Sumha ry The frequency of core melt for loss of main feedwater events (including loss of offsite power with emergency power available) was estimated for Palo Verde using an AN system unreliability (7x10-5/D) based on the meth,-dology in Ref. 3.
-5 The estimated frequency of core melt is about 10 /RY which is similar to the contribution for such events in WASH-1400, therefore, we do not believe these results, by themselves, support a need for feed and bleed capatility in CE plants without PORVs, provided the AN reliability is in the range of 10 to
-5 10 /D and procedures for restoring main feedwater for decay hc a renoval are in place.
T i
l
The licensee for Palo Verde has corr 1* ted to provih capability of pt a" ring the non-safety grade AFW pump from an emergency bus to upgrade the reliability of the AFWS.
The frequency of core melt for very small LOCAs in PWRs was estimated to be
-5 about 6s8x10 /RY which is about a factor of 2s3 higher than the estir.ates of small LOCA induced core melt frequencies in WASH-1400.
This difference is due to our current understanding that the small LOCA frequency is higher than that assumed in WASH-1400.
The generic issue of high core melt frequency of very small LOCAs was raised previously with regard to reactor coolant pump seal failure (Ref.12) which was given a high priority rating by DST (Ref.13).
The issue of very small LOCAs is also part of Task Action Plan A-45 " Shutdown Decay Heat Removal Requirements" which is scheduled for completion in 0:tober These current estimates of small LOCA related core melt frequencies do 1985.
not seem to require immediate action; however, we have identified several intermediate term actions that can be coordinated with USI A-45 to determine what measures, if any, could be taken to reduce overall risk.
This study brought forward previously published information on feednater systems that should be reviewed by the relevant technical groups in the intermediate term and incorporated in their reviews as appropriate.
I I
l t
3.
DISCU5510).
3cl Transients and Failure of All Feedwater i
In general, Combustion Engineering and Westinghouse plants respond to reactor trips in the same fashion; namely, the main feedwater is isolated, auxiliary feedwater is actuated, and the operator is responsible for modu-lating the flow rate to maintain water level in the steam generators.
There is an assumption in all PRA's that the use of the main feedwater system provides an alternate success path for removing decay heat for all feedwater transients, even those events involved with the direct loss of main (including MSIV closures, loss of offsite power, etc.).
At these low heat levels the control of the main feedwater is manual and very touchy; however, the operators should be familiar with the idiosyncrasies of their ;,lants,
since they perform this procedure at power levels less than 15% during power escalation.
Of more concern are those events where loss of main feedwater is part of the initiating transient.
In an EPRI survey (Ref. 1), the frequency c' transient.
involved in the loss (total or partial) of main feedwater (loss of.ain feed-Of water, condensate pumps, condenser vacuum) is about 1.3/RY for PG's.
these, the frequency of events that affect both main feedwater purt.s is only 0.13/RY which includes closure of all M51V's with a frequency of 0.05/RY.
Thus, the frequency of total loss of MFW events for which the MFa' N.Ds poten-tially are not recoverable in a short time is 0.08/RY.
Otner survejs of feed-1976 to 1975 water related events in Westinghouse and CE plants for a period fr:r:
The combined results of these surveys indica,'te a are presented in Ref. 2 and 3.
f requency of about 1.0/RY of partial or tot;l loss of noin feedwater event.. a*.2 a f requency of about 0.48/RY for events which affect both main feedwater pu ps.
(
A breakdown of the causes of these events is presented in Table 1.
The first nine items identified in Table 1 should be recoverable in the short term (less than 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br />) considering the low flow rate required to remove decay heat.
- Thus, the frequency of potentially irrecoverable events is about 0.12/RY which is con-sistent with the EPRI results.
The main feedsater systems of interest consist of electrically driven condensate pumps, which have a discharge pressure of 400 to 600 psi, in series with stear driven main feedwater pumps.
If the main feedwater pumps are not available, it is possible to supply feedwater with the condensate pumps if the steam generators are depressurized to 400 to 600 psi using turbine bypass or atmospheric dump valves. The saturation temperature at 400 psi is 444 F which is about 110 F decrease in the normal SG operating temperature, but well above nil ductility temperatures for the pressure vessel.
Thus, assuming procedures and operator training are in place for using the condensate pumps, it is reasonable to assure that the frequency of irrecoverable losses of main feedwater could be reduced to the value of 0.03/RY used in WASH-1400.
The estimated demand failure probability of the auxiliary feedwater system for the
~4 Palo Verde plant is about 10 /D (Ref. 4); however, if the recommendations inv:lving valve position confirmation are implemented, it is estimated that the system w:;1d
~
have a demand failure probability of about 7x10 /D based on the methodology defined in Ref. 3.
The precursor program (Ref. 5) has identified eight events where the total AFW system was failed in operating reactors and, based on thest
~
results, they estimated the demand failure rate for AFM to be about 10 /D.
l t
i
A discussion of these event. is pre.,ented in Talde 2.
- k. noted, the utnhilving causes of these events are recoverable because of subsequent actions or iin y are not relevant to Palo Verde.
The human error estimates are based on those con-sidered in the AN studies in Ref. 2.
Thus, although the AFW events cited in Table 2 are important and should be considered in the design and operation of AN systems, these events are not considered to impact significantly the low demand failure estimated for Palo Verde.
There are two sequences related to loss of main feedwater and utilization of feed and bleed that can be evaluated quickly for Palo Verde. The first is a lo.s of main feedwater transient (with offsite power available). The estimated frequency of extended loss of main feedwater is 0.03/RY assuming adequate procedures and operator training for restoring MFW under various degraded conditions including use of the condensate pumps alone.
The demand failure probability of the AN
-5 for the Palo Verde plant is estimated to be 7x10 /D assuming implementation of the recomendations. Thus, the frequency of core melt (assuming no other means available or removing decay heat) due to this sequence is:
-6 Acm = 0.03x7x10-5 Rr 2x10 /RY This estimate assumes independence between the loss of main feedwater and loss of AN, thus it is probably a lower bound for a point estimate of the frequency of core melt. As pointed out in Ref. 9, all potential common mode failures are not considered in an isolated system reliability analysis such as performed in Ref. 3.
The types of comon mode failure considerations include support systems, equipment layout and external events; however, failure in support systems and extern.sl events.'
could also adversely affect feed and bleed. The quantitative impact of these ef fects is unknown. Uncertainty in this estimate due to statistical vari.itions in the failure rates is probably about a factor of ten (both up and down) con.idering the uncertainty bounds presented in Ref. 3.
i i
l A second event is the loss of offsite power with on-site power available. T h.-
frequency of loss of offsite power is estimated te be 0.27/RY from Reference t..>
based on operating experience at nuclear power plants. The probability of non-recovery of offsite power in ih hours is estimated to be 0.3/D based on operating experience reviewed in Reference 7.
The recovery of offsite power implies recovery of MFW. Thus, the core melt frequency for this sequence is estimated to be:
-5 Acm = 0.27x0.3x7x10 6x10-6/RY The uncertainties in this estimate are the same as those discussed above.
The combined estimate of core melt frequency from these two sequences is about
-5 10 /RY and is similar to their contribution in WASH-1400.
In both of these sequences, the containment cooling systems are operable so there is only a small effect on offsite consequences compared to other sequences which involve contatri-ment failure. Thus, from a risk standpoint, the potential effect of these sequences at Palo Verde (or other CE plants without PORV's) is in the same ranqe-of estimates for other plants.
Thus, we do not recommend that feed and bleed be a licensing requirement for Palo Verde because of potential concern for the two sequences considered above provided that the recommended rrodifications to the AN system are implemented and viable procedures are available for restoring MN to remove decay heat.
In the course of this evaluation, it became obvious that recovery of MN as a back up to AFV is an important contribution to reducing risk and that there are potential comon mode failures in the AN system which should receive additional ((
attention in the longer term. We consider it prudent to review the available information and where feasible develop cost-beneficial recommendations that can be implemented generically. These actions should be coordinated with Task Actir..
Plan A-45.
i A.
Loss of Mrn' l.oss of MFW surveys pre,ented in References 2 and 3 indicate a number of generic causes that result in the loss of both MFW pumps which has a fre-quency in the range of 0.1 to 0.5/RY.
The risk associated with this event (although small) is directly proportional to the frequency of the loss of M FM. On cursory examination, it appears that some of the causes may be amenable to simple corrections, e.g., more frequent cleaning of the conden-sate strainers and the use of two lubricating oil pumps for the MFW pumps.
DSI should review these feedsater events on a low priority basis to determine what modifications, if any, may be cost-beneficial to reducing the frequency of loss of all MFW and/or reducing the loss to a single train of MN.
B.
Restoration of MN in all our evaluations, significant credit (conditional probabilities ranging
-2 from 0.3/D to 10 /D) is given to the restoration or use of MFW as a backup to AFV whenever there is a reactor trip.
For most transients, the MFW is completely operational; however, for transients initiated by disturbances in the secondary system, the MFW system may be degraded. The restoration or use of the MFW for shutdown heat removal is a completely manual operation.
Because of the potential impact of this function on risk DSI/DHFS, within a reasonable time frame, should develop scenarios and procedure requirements for the restoration of MFV under various degraded conditions.
This effort should be complementary to work being pursued on procedures for inadequate core cooling and recovery from significant events with multiple failures.
.z
~
1 C.
AFW Connon Mode Failures The precursor program di* cussed in Ref. 5 indicated potential common mode failures of the AFW systems which have a potential impact on risk. DSI should review these potential failures in the intermediate term and develop cost-beneficial recommendations to minimize the situations if feasible.
3.2.
Loss of Offsite Power and the Availability of Two Motor Driven AFW Pumps The impact of AFW during a loss of offsite power was evaluated in Item 1 above.
The staff has previously required that Palo Verde modify its design to include operation of the non-safety grade AFW pump with emergency power.
3.3.
Very Small LOCAs and failure of HPSI PWRs are designed so that the chemical volume and control system (CVCS) and the high pressure injection system (HPSI) provide acceptable protection for any effective pipe break size up to about six inches. The composition of these syster.s varies among the three PWR vendors.
Generally, B & W plants use the same high head centrifugal pumps for the charging and ECCS function. One pump is normally running and on receipt of a safety injection signal, the remaining pumps are actuated and the suction and discharge lines are realigned to accommodate the higher ECCS flow rates.
Generally, CE plants use constant displacement pumps for the charging function and separate centrifugal pumps (shut-off head about 1400 psi) for the ECCS function.
Upon receipt of a safety injection signal, the charging pumps are stopped and the HPSI pumps are actuated using the ECCS piping arrangemer.t.:
Westinghouse plants use either of the above functional arrangements except for some plants which use charging and HPSI pumps for ECCS.
The maximum charging flow rate is about 150 gpm which is usually sufficient to accommodate breaks in instrument lines (about 3/4 inch).
For systems with constar:
displacement pumps this would require opening modulating valves.
for larger breaks, HPSI must be utilized to mitigate the consequences.
Evaluation of core melt frequency in WASH-1400 for very small break sequences
-3 was based on a break frequency estimate of 10 /RY for breaks less than 2" and
~4 a frequency of 3x10 /RY for intermediate size breaks 2 to 6 inches.
DSR, in Reference 9 has estimated the very small LOCA frequency (presumably less than
-2 2 inches) to be 3x10 /RY because of concern for instrument line breaks, steam generator tube ruptures, charging pump line breaks, and gross reactor coolant pump seal failures.
Significant reactor coolant pump seal failures (>l50 gpm) have occurred at Robinson and ANO-1; however, there have been four RCP seal failures with 50 to 150 gpm flow rates in PWRs based on the supporting information for Reference 10.
Assuming 300 years of PWR operation, this would result in point estir.ates of
-3
-2 8x10 /RY for RCP seal failures requiring HPSI and 1.3x10 /RY for R P seal failures requiring only charging pumps.
Cracks have occurred in the charging system mainly in small instrume'.tation lines upstream of the check valves which isolate the charging system from the reactor coolant system as indicated in Reference 11.
A break in this syster would not constitute a LOCA since any sustained loss of primary coolant would result in a safety injection signal which would automatically isolate the syste ar.d terminate the loss.
Similarly, a steam generator tube failure does not constitute a LOCA :.y itself unless subsequent operator actions result in a sustained loss of flo. cut of the The frequency of SG tube rupture is about 2x10~2/RY a-d we would primary system.
estimate the conditional probability of an additional failure and/or P.. an error
leading to sustained lo.. of coolant (and thereby requiring 111'51 and/or LPSI) to be less than 10-I /D, which would make this event a small contributor to very small breaks; however, it could lead to a LOCA outside of containment, which is a separate consideration outside of this evaluation.
Open PORV's have been identified as contributors to small LOCAs in the precursor review (Reference 5).
Because of the procedures implemented following THI-2, this event would be tenninated by closing the PORV block valve.
Therefore, the estimated frequency of a LOCA through the PORV is judged to be not greater than 10-3/RY. This value is being confirmed by RRAB for all plants under the TMI-2 Action Plan Item II.K3.2.
As noted above, instrument line breaks can be acconsnodated by the charging system alone. HPSI is also a success path for mitigating this event.
Reference 9 assumed a frequency of 10~ /RY for an instrument line break. Cracks in welds at small tees and nipples were noted on 12 occurrences in Reference 11. These types of cracks could potentially become complete tube failures with a conservative frequency estimate of 4x10~2/RY.
-3 The demand failure probability for HPSI is estimated to be about 4x10 /D based on
-3
-3 previous PRAs. Oconee point estimate is 1.4x10 /D; Surry, 3.7x10 /D; Sequoyah,
-3 3.5x10-3/D; and Crystal River (exclusive of human error), 3.2x10 /D.
The human error estimates for Crystal River were estimated to be 5.4x10-2/D associated with the operator mistakenly terminating flow and switching to the recirculation mode prema turely. For the very small breaks of interest herein, these human errors are.','
not appropriate because of the long time frames involved.
In fact, the two sig-nificant RCP seal failures were mitinated and the plant depressurized using less i
l l
l
e A
a than one-half of the RT.! inventorj.
it.e conditional probabili:j
. tem c.ombi rimi loss of both charging and HPSI for LOCAs less than 150 gpm i'. estirined to be le..
than 10 /D regardless of whether the functions are shared or separate, because
>f the AND gates involved in the valve and pump combinations.
The core melt frequencies for very small LOCAs (with flow less than 150 gpm)
-5
-2x10- 3 2 to 4 x10 /RY Acm = (2 to 4) x 10 which require either charging pumps or HPSI for success For larger breaks which require HPSI for success, the estimate is
-3
-5 Acm = 10-2x4x10 :::: 4x10 /RY Although there are large uncertainties in these estimates, they indicate that the core melt frequency from small LOCA initiators is about a factor of 2.3 higher than the WASH-1400 estimate.
The core melt estimates are in the range of previously accepted v'alues, and there-fore, we have no imediate concern.
However, these sequences represer.: dominant contributors to core melt as previously discussed in Ref.12 and has Men given high priority (Ref.13).
The initiating frequency for very small LOMs is in the same range as the probability of extended loss of main feedwater even.s.
We have previously recomended that the demand failure probability for AN sys, ens be in the 10' to 10 /D range because of their high challenge frequency due to loss of MW.
Although' nost HPSI systems have three pumps similar to the better Ari systems, they have a lower reliability because of the general use of two-train pipir p-valve arrangements.
Thus, any improvement in reliability in these systems m uld probably require modifications associated with the valves.
Another potential i n rovement
,1 would be the capability to depressurize the primary system with the 5t:cndary syste' so the low head safety injection system could be used as a redundant cooling system.
8 s
Becaun th:.e event, repre,ent dominant contributors to cor.: relt, ernerit s tuih e.
should be performed in a reasonable time frame to examine potential cost-benrfiti.il I
modifications that would reduce the frequency for these events.
These studies should be coordinated with Task Action Plan A-45.
A.
Improvement in HPSI Reliability D51 should review available PRAs to identify HPSI weaknesses and investigate valve arrangements and sources of valve unreliability in HPSI systems to identify potential cost-beneficial modi fications.
A longer term program would be developed on the basis of these studies and consistent with pre-vious assessment of RCP seal failure in Ref.12.
B.
Alternate ECCS D51 should investigate in the intermediate term whether PWRs can be de-pressurized to LPSI pressures for very small LOCAs by using existing secondary systems.
If this process is feasible, procedure requirements should be developed in conjunction with DHFS.
C.
Reduction of Initiating Frequency Prior studies cited in Reference 11 have addressed the frequency of potential small breaks.
DE should review these studies in the intertnediate term to determine if cost-beneficial modifications can be recomended to reduce the frequency of these events.
O
TABLE 1 - CAUSES OF LOSS OF ALL MAIN FEEDWATER L,per of Events 1.
Loss of Suction Pressure Due to Disturbances it in Heater / Drain System 2.
Loss of Suction Pressure Due to Clogged 5
Condensate Strainers 3.
Loss of Suction Pressure Due to Other Causes 2
4.
Disturbances in Feedwater Control 7
5.
Loss of Condenser Level 2
6.
7.
KilV Closure 3
8.
Condensate Recirculation Control 3
9.
Isolation of Suction Pressure Sensor 1
- 10. Loss of Lubrication Oil to Both Main Feedwater 5
Pumps
- 11. Loss of Steam to Feedwater Punps 2
- 12. Disturbances in Onsite Power Distribution 2
~ 13. Loss of All Condensate Pur.ps 2
- 14. Unknown 1
48 l
s.
l I
Table 2 Total Loss of AFW Events
?
1.
Turkey Point 4 June 18,1973 - Failure of the pumps to auto-start due to failure to install fuses.
The pumps were still capable of being started manually. The immediate operator actions following any reactor trip is to ascertain that the AFW system is operating.
In addition, AFW flow indicators have been installed since TMI-2. Thus, this event is considered 7
totally recoverable with a human error probability of less than 10 /D.
2.
Point Beach 1. April 7,1974 - Failure to deliver flow due to clogged suction strainers.
3.
Kewaunee, November 5,1975 - Failure to deliver flow due to clogged suction strainers.
There are no strainers in the Palo Verde AFW system so this potential common mode failure is not applicable. However, it is recommended that AFW installations at operating plants be examined to correct this potential problem.
4.
Turkey Point 3, May 8,1979 - Failure of pumps to start due to over-tightened packings and controller malfunction.
5.
Davis-Besse, December 11, 1977 - Loss of AFW pump control due to mechanical binding and blown control power fuses.
These plants have only turbine driven AFW pumps which are suscep-tible to control problems.
Palo Verde has 2 motor and 1 turbine driven pump, so this failure mode is not applicable.
6.
Farley, March 25, 1978 - Failure of turbine driven pumps to start plus open by-pass valves, i
recoverable with a human error probability of less than 10- is event The modifications instituted because of TMI-2 should make tp/D.
l l
7.
Rancho Seco, March 20, 1978 - Failure of AFW to deliver flow due to NNI failure.
This event is probably peculiar only to B&W plants which had a sig-7 nificant amount of control and readout instrumentation on a single
[
bus. This situation was corrected following the Crystal River event of February,1980, when all plants were required to have redundant readout in the control of vital plant parameters.
This requirement l
l
e Tal,le 2 - continued Total Loss of AN Events coupled with the THI-2 modifications and the operator's imediate action following a scram should make this type f event recoverable with a human error probability of less than 10'g/D.
8.
TMI-2, March 28,1979 - Failure of AN to deliver flow to closed valves.
This type of event is recoverable as discussed in Items 1 and 7 above.
4.
Re f erences 1.
ATW5: A Reappraisal Part III Frequency of Anticipated Transients.
EPRI NP-801, July 1978 2.
Generic Evaluation of Feedwater Transients and Small Break Loss-of-Coolant Accidents in Westinghouse-Designed Operating Plants, NUREG-0611, January,1980 3.
Generic Evaluation of Feedwater Transients and Small Break Loss-of-Coolant Accidents in Combustion Engineering Designed Plants, NUREG-0635, January,1980 4.
Palo Verde Nuclear Generating Station Units 1, 2 and 3, Auxiliary Feedwater System Reliability Study Evaluation, NUREG/CR-2322, December,1981 5.
Precursors to Potential Severe Core Damage Accidents: 1969 - 1979, Draft NUREG/CR-2497, January,1982 6.
Loss of Offsite Power, Survey Status Report, Revision 3, R. Scholl, November,1980 7.
Letter from J. Anderson (ORNL) to P. Baranowsky (NRC), dated August 24, 1981 8.
Data Summaries of Licensee Event Reports of Pumps at U.S.
Commercial Nuclear Power Plants, NUREG/CR-1205, January,1980 9.
Feed and Bleed Issue for CE Applicants, Memorandum from F. Rowsome and J. t1urphy to R. Tedesco, dated January 29, 1982 10.
RRN3 Preliminary Assessment of the Reactor Coolant Pump Seal Failure Problem, Memorandum from A. Thadani to R. Baer, dated December 12, 1980 11.
Investigation and Evaluation of Cracking Incidents in Piping in Pressurized Water Reactors, NUREG-0691, September,1980 12.
Reactor Coolant Pump Seal Failure, Memorandum from T. Marley to D. Eisenhut, dated March 27, 1981 13.
NRR Prioritization of Generic Issues,11emorandum from S. Hanauer to D. Eisenhut, et al, dated March 4,1982 a
w