Direct remaining useful life prediction based on multi-sensor information integrations by using evidence theory

Estimation of remaining useful life (RUL) is one of most interesting subjects in prognostic and health management. Performing an analysis of the results of such estimation can increase the reliability and the safety of the system, and reduce the unnecessary costs. In this paper, a similarity-based combination method is proposed to combine several run-to-failure historical datasets in order to directly estimate the RUL. In this method, reference datasets are clustered and the initial RUL is calculated based on the artificial neural networks trained by the reference datasets. By using the extended Dempster-Shafer, the similarity between the initial RUL and the average RUL for each dataset is obtained. The proposed methodology is tested and validated on Commercial Modular Aero-Propulsion System Simulation (C-MAPSS), test-bed developed by NASA. The results of the evaluation show that the proposed method outperforms other methods in the literature.