Components importance ranking considering the effect of epistemic uncertainty

The identification of the importance of basic events or system components is an essential task in system design, which is quantified by reliability importance measure (IM). Traditional reliability IMs are typically calculated based on precisely given parameters (e.g., component failure rate) which is often inevitable when the available data is incomplete or imprecise. This paper proposes a new components importance ranking method concerning the case when the epistemic uncertainties of component parameters are represented by probability distribution (commonly established by Bayesian method). First, a new IM is defined at given working time point based on quantifying the overall difference between the probability distributions of system reliability due to the state change of one component. The larger the proposed IM, the more important the corresponding component is for improving the system reliability. Second, for given system reliability target, another IM is defined based on measuring the overall difference of the probability distribution of system lifetime when the state of one component is changed. A larger value of the second IM indicates a larger contribution of this component for increasing the system lifetime. The proposed method has two merits, i.e., (a) the effect of epistemic uncertainty in component parameters on system reliability performance is considered from the perspective of reliability and system lifetime separately; (b) it provides a simple way for ranking the components considering epistemic uncertainty.