Weibull Component Reliability Prediction with Masked Data

Abstract

Under the conditions that the exact component causing system failure may be unknown or masked, we present an iterative approach to estimate the component lifetime parameters of a system consisting of two Weibull components. The time and cause of system failure are assumed to follow a competing risks model, and the masking probability of minimum random subsets are not subjected to the symmetry assumption. On the basis of considering the effect of failure time and component, we redefine the computation of masking probability and the likelihood function is derived for the masked data. An iterative procedure for finding the maximum likelihood estimates is presented via an EM algorithm. The developed approach is illustrated with a simple numerical example.