Implications of phenomenological uncertainty in engineering reliability assessments

摘要

The consideration of an unknown failure mode is possible whenever expert opinion on the risk associated with the development and application of new technologies is available. Based on the degree of innovation or on the period during which the innovative aspects have been tested, experts may turn in estimates the likelihood L_(n+1) that an unknown failure mode n + 1 exists. It is demonstrated that, mainly for new systems, the reliability should be reduced by a factor (1 - 0.24 L_(n+1)), in order to account for phenomenological uncertainty. This statement is a general result, for it does not depend on the probability distribution of timeto-first-failure in any mode. Common experience seems to indicate that, in general, the maximum uncertainty condition, L_(n+1) = 0.5, is not necessarily pessimistic and may lead to high pronenesses to failure in the case of new technologies. The influence of testing, through the duration of testing or operating time without failure is then quantified. It is shown that high system reliability can only be achieved after successful operation.