Citation

Keey RB, Smith CH. Reliab. Eng. 1985; 10(2): 105-127.

Copyright

(Copyright © 1985, Elsevier Publishing)

DOI

10.1016/0143-8174(85)90004-6

PMID

unavailable

Abstract

Hazard analysis is based on computing the incidence of undesired outcomes by considering the structure of fault propagation from lesser mishaps. These events are not single-valued, but are described by random-variable distributions. The authors consider how the input-variable distributions combine into higher-order events, with particular reference to the changes in variable range and uncertainty, as indicated by the coefficient of variation. By evaluating the combination of binary, normally-distributed inputs, it is shown that the coefficient of variation of the output is lower than the larger coefficient of the inputs in the case of an OR-gate, but that the coefficient for the output from an AND-gate is always higher than the input coefficients. A numerical method for combining inputs of any kind of distribution is described, and the method is applied to an example of the reliability of a water-deluge system. The propagation of uncertainty is discussed, as well as the effect of the proof-testing period on the coefficient of variation of the system.