ML022400370

Request for Additional Information on Davis-Besse Response to the Bulletin

ML022400370
Person / Time
Site:	Davis Besse
Issue date:	08/27/2002
From:	Office of Nuclear Reactor Regulation
To:
References
BL-01-001, FOIA/PA-2002-0229
Download: ML022400370 (2)

Text

RAI on Davis-Besse Response to The Bulletin In your response letter (page 13 of an attachment) to the Bulletin stated that multiple failures of CRDM would not occur and, apparently, the bounding analysis of a single failure for LOCA and non-LOCA would be applicable to Davis-Besse. Could you explain the rationale and assumptions of such statement.

0.1 The initiating event frequency evaluated by the Monte Carlo simulation did not provide the basis, data, or bench marking using the available data. The result of the Monte Carlo simulation, 1.3E-5 (probability of having an OD flaw propagate in one fuel cycle to be large enough to cause catastrophic failure),

did not provide detailed information including assumptions and reference documents. The analysis did not provide nor discuss the uncertainty. PRA does not create uncertainty, but uncertainty is derived by knowledge limitations or lack of data.

1.2 Actual calculations, equations, and assumptions were not provided.

1.3 The probability of having a leaking nozzle with boric acid crystals present, but not identifying the leak as a result of human error (either failing to conduct the test of failing to detect evidence of a leak during an inspection) is estimated to be 6.OE-2 or 6%.

The NRC staff needs the supporting data and the basis for this number, and an explanation of why this 6% human error factor was either not included anywhere in the risk assessment or clarified.

1.4 The number of flaws found by inspection that resulted in leaking nozzles experienced at Oconee and Arkansas Nuclear One Unit 1 (ANO-1) was 14 and 1, respectively. Conservatively assuming these flaws initiated over the last two operating cycles, an initiating frequency of 1.25 CRDM leaks can be estimated from fifteen leaks identified in twelve reactor-years.

The NRC staff notes that ANO-1 is an outlier with regard to the calculation, and those 15 may not represent the state of Davis-Besse. Furthermore, the staff's calculation using just Oconee, and adding the 6% human error factor discussed above, results in an increase in the initiating event frequency. The assumption of two year initiation appears to be non-conservative since there may be cracks developed but identified.

1.5 Discussion on the licensee's Monte Carlo simulation stated that a conservative estimate of one failure (out of 80,000 Monte Carlo simulations) was used to estimate the 1.3E-5 value."

The NRC staff notes that a Monte Carlo simulation is a computational method, and is not a mathematical model for describing catastrophic failure. The staff requests the behch mark data points and the basis for the Monte Cado simulation which resulted in the 1.3E-5 probability value. The staff needs the initiating event frequency to complete the review of the response.

1.6 The NRC staff needs the CCDP for a Medium LOCA, and if that value is different than the value presented in the IPE, the staff needs an explanation of the discrepancy. In addition, the CCDP of a medium break LOCA is not conservative, it may be a bounding or limiting case based on post break configuration.

1.7 The NRC staff would like to have Davis-Besse plant specific conditional population dose, if available, and the supporting data and the uncertainty used to get the value.

2. We are interested in treatment of the recirculation blockage after gross failure(s) of CRDM(s) and other risk assessment details. For example, how the human errors were factored into the risk assessment during the initiation and mitigation phases of the postulated bounding accident analyses?

It would be desirable to provide us the core damage probability (as opposed to conditional core damage probability, given event initiation) and frequency of the bounding LOCAs as well as their dependency with time since the probability of the event initiation would be depend on the duration of operation as postulated in the susceptibility model. How does the cumulative core damage probability increase for three months, six months ?