Publications
A DOE-STD-3009 hazard and accident analysis methodology for non-reactor nuclear facilities
This paper demonstrates the use of appropriate consequence evaluation criteria in conjunction with generic likelihood of occurrence data to produce consistent hazard analysis results for nonreactor nuclear facility Safety Analysis Reports (SAR). An additional objective is to demonstrate the use of generic likelihood of occurrence data as a means for deriving defendable accident sequence frequencies, thereby enabling the screening of potentially incredible events (<10{sup {minus}6} per year) from the design basis accident envelope. Generic likelihood of occurrence data has been used successfully in performing SAR hazard and accident analyses for two nonreactor nuclear facilities at Sandia National Laboratories. DOE-STD-3009-94 addresses and even encourages use of a qualitative binning technique for deriving and ranking nonreactor nuclear facility risks. However, qualitative techniques invariably lead to reviewer requests for more details associated with consequence or likelihood of occurrence bin assignments in the test of the SAR. Hazard analysis data displayed in simple worksheet format generally elicits questions about not only the assumptions behind the data, but also the quantitative bases for the assumptions themselves (engineering judgment may not be considered sufficient by some reviewers). This is especially true where the criteria for qualitative binning of likelihood of occurrence involves numerical ranges. Oftentimes reviewers want to see calculations or at least a discussion of event frequencies or failure probabilities to support likelihood of occurrence bin assignments. This may become a significant point of contention for events that have been binned as incredible. This paper will show how the use of readily available generic data can avoid many of the reviewer questions that will inevitably arise from strictly qualitative analyses, while not significantly increasing the overall burden on the analyst.