This paper is concerned with proposing a fuzzy set of unreliability based on 'likelihood of failure' which is expressed by a fuzzy set instead of failure rate in the conventional fault tree analysis, and presenting a method of fault tree analysis. In this analysis we consider the fault caused by a component, human error, environment and common mode failure. A causal relation of common mode failure is expressed by a fuzzy relation. Standby redundant systems including series systems and parallel redundant systems are analyzed. Especially we define an operator for unreliability of a standby redundant system.
Five criteria are defined to evaluate the results of the present analysis,
i) the numerical unreliability which gives the greatest grade of the membership of the fuzzy set
ii) 0.9-cut of the fuzzy set
iii) the fuzziness of the first criterion
iv) the potentiality of failure
v) the importance of a basic event.
As an illustrative example, a circuit diagram of a power supply in a nuclear power plant is shown and its fault tree is analyzed, where the top event is a loss of electric power supply to engineering safety features. It is shown that the present method gives us more useful informations than a probabilistic method does.