Several methodologies for computing plant (un)avail-abilities and system initiator frequencies are compared to results from a Markov model of a simplified cooling-water system. All methodologies are found to be acceptable for small ratios of failure rates to repair rates, but some degenerate more rapidly than others as these ratios increase. Some, but not all, of this degeneration seems to stem from inappropriate use of various forms of a rare-event approximation. The larger errors tend to be conservative in nature, perhaps excessively so.

This content is only available via PDF.
You do not currently have access to this content.