Abstract
This paper outlines research on understanding, characterizing, and managing conservatisms in safety analyses. This research includes a review of national and international approaches for developing and using conservative and best-estimate analyses. A probabilistic approach is discussed to support reducing conservatism while maintaining safety margins. An example of the proposed approach is applied to two case studies for nonreactor nuclear facilities. The objective of this work is to provide a means for better understanding and managing risks associated with nuclear facilities. The results from these examples show that conservative estimates could lead to excessive safety margins when compared to the best-estimate values; the amount of excess margin may be as high as one or more orders of magnitude.
Notes
a Fuel cycle facilities, nuclear waste processing facilities, and nuclear research facilities are typical NRNFs, and safety analyses performed for these types of facilities are examined in this paper.
b A best-estimate model is developed using realistic data (from operations or experiments). Because of limited data, the model parameter values may not be precisely known. Therefore, a best-estimate model associates uncertainties with the form of the model and its parameters, typically in the form of probability distributions.
c The authors note that DOE-STD-5506 is presently being revised by DOE to incorporate additional information regarding energetic reactions, among other changes. This analysis uses the information available in the present standard.
d PE-Ci = 239Pu equivalent curies.
e The generation of 200 realizations of SRS drum arrays, according to the Wilks’ tolerance limits, assures that with more than 95% confidence, the 200 samples will represent 99% of the random variability in the PE-Ci, H2, O2, and ejection ratio. This corresponds to very high confidence that the SRS drum arrays reasonably account for uncertainties in the 705 drums sample. For higher confidence levels than 95%, more realizations would be needed. [See S. S. WILKS, “Determination of Sample Sizes for Setting Tolerance Limits,” Ann. Math. Statist., 12, 1, 91 (1941).]
f For the most centered drums, the adjacent number of neighboring drums was nine (eight at the same level and one at the top, assuming that the bottom drum will not be affected as the INL study reports). For the corner (center row) drums, the number was four (i.e., three at the same level and one on the top), and on the side it was six (i.e., five at the same level and one at the top). For the ones at the top layer, the adjacent corner drums were three, and the side drums were five since no drums existed at the top in this layer.
g See K. DYKES and M. MYER, “TRU Drum Hydrogen Explosion Tests,” WSRC-TR-90-165, Westinghouse Savannah River Co. (1990).