A new approximate method for uncertainty propagation in system reliability analysis

Abstract All existing methods of uncertainty propagation for system reliability analysis follow a similar approach: estimate the distribution function f S (X S ) of the system parameter X S (whatever is estimated, e.g. unavailability or failure frequency) and then, given a confidence, calculate the corresponding upper and lower bounds on X S from f S (X S ) . This paper presents a new approximate method that calculates the confidence given an upper or lower bound on X x . This method does not estimate f S (X S ) so that the calculation effort is much less than that of existing methods. The accuracy of this method is illustrated in an example. Based on this method, a new uncertainty measure for the element importance is presented.