Selection and Identification of a Shape Function for Modeling Degradation as a Gamma Process

Abstract

This paper proposes an approach for estimation of Remaining Useful Life (RUL), in which data from earlier lifetimes is utilized in the identification of the functional shape of degradation. A set of candidate shape functions are identified, and one of them is selected for RUL estimation, and consequently for inference, based on information theory. The feasibility of the approach is studied using functional forms of degradation acquired through air filter loading experiments and air quality measurements representing degrading factors. The study indicates the usefulness of utilizing data from previous lifetimes when reliable enough a priori information of the degradation phenomena is not available. The observed improvement in accuracy was considerable at long prediction horizons, which is beneficial especially in ConditionBased Maintenance applications (not limited to air filters) with infrequent maintenance shutdowns.