Abstract
In this paper, we propose a nonparametric estimator of the component lifetime distribution in a binary reliability system, from a type of censored data that arise from the autopsy of black box systems. The observed data consist of the failure time of the system, the set of components that are dead at the instant of the system failure, and the knowledge of the structure function, We study the conditions for the application of this estimator, and we observe its good behaviour in the parallel-series, series-parallel and (tu, a) additive binary reliability systems.