Application of finite state models for system design and reliability assessment

In order to discuss reliability, particularly software reliability, a conceptual model of a computing system is developed based on finite state automaton and associated state set transition diagrams. The problem of software unreliability is an example of systematic failure that all protection systems are prone to but on which a greater emphasis should be placed as the complexity attainable with computer technology is exploited. The reliability criteria for such systems are quantitative so that the assessment of the software reliability must contain quantitative arguments. Procedures for achieving this are discussed together with the assumptions on which such estimates are based. The scope for statistical inference is considered given both the theoretical problems of the underlying failure processes and the probable paucity of data from integrity systems. A Bayesian method, common to other type of Probabilistic Risk Assessment is suggested for the reliability estimation. This method takes into account some of the uncertainties discussed. (author)