Estimating component reliabilities from incomplete system failure data

Abstract

Engineers often need to estimate component reliabilities from system failure data that are obtained either from in-house tests or field operations. The components that caused the system failures usually can be identified by checking the failed systems; however, in some cases, the exact cause of failure is very difficult or impossible to identify. For the latter case, the information on the system failures is not complete, and such data are usually called masked failure data. In this paper, we propose a method for estimating component reliabilities from masked system failure data. Solutions for systems with serial and parallel configurations are provided. The failure distribution of each component is obtained from the proposed method, and is then used to calculate the probability that a system failure is caused by the given component when the exact time or an interval time of the system failure is known. This probability is very useful since it can be used to decide which component should be analyzed first when a system failure occurs, depending on the failure time.