A Novel Method Using DS-MCM for Equipment Health Prognosis with Partially Observed Information

Abstract

Health prognosis for equipment is considered as a key process of the condition-based maintenance strategy. However, in many cases, due to the performance and working environment of the sensor, working condition of equipment, and monitoring error, the monitored data can unavoidably exist uncertain data information, for example noisy data and noisy data. The accurate health state of equipment is difficult to be obtained. To address it, this paper presents a joint optimization model for equipment health prognosis by combining Dempster-Shafer evidence theory (DS) with Markov chain model (MCM). First, based on Markov model, DS is used to develop the state recognition framework of equipment. Then, the uncertain data is allowed to be handled in the form of interval number, and the basic probability assignments (BPA) can be generated based on the distance and similarity among interval numbers. BPA is transformed into the probability distribution of basic health states by Pignistic probability conversion in order for increasing the reliability. Finally, a case study is used to show the effectiveness and rationality of the proposed model. The results show that the proposed model has an efficient ability in dealing with uncertain data information, and can effectively solve the problem of equipment health prognosis with partially observed information.