Replacement Scheme for Lubricating Oil Based on Bayesian Control Chart

• Published in: IEEE transactions on instrumentation and measurement Vol. 70; pp. 1 - 10
• Main Authors: Du, Ying, Duan, Chaoqun, Wu, Tonghai
• Format: Journal Article
• Language: English
• Published: New York IEEE 2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Availability; Bayes methods; Bayesian analysis; Bayesian control chart; Computational modeling; Control charts; Fault detection; fault prognosis; Four ball testers; hidden Markov model (HMM); Hidden Markov models; History; Lubricating oils; Maintenance engineering; Markov chains; Monitoring; oil deterioration modeling; oil replacement; Optimization; Statistical models; Stochastic processes; Wear particles
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2020.3042231

Lubricating oil carries important health information of operating machines, and oil replacement scheme is crucial for ensuring machine health, reducing operation costs, and improving machine availability. However, few works have been done on the determination of reasonable replacement time for lubricating oil in the industry. Therefore, the main motivation of this article is to present a replacement scheme based on the Bayesian approach to detect and prevent the potential failures of lubricating oil. A three-state statistical model based on the hidden Markov chain is applied to characterize oil deterioration, which contains partially observable healthy and unhealthy states, and an observable failure state. A novel Bayesian control scheme for oil replacement based on the hidden stochastic process is proposed under the objective of long-term expected average availability maximization. A computational algorithm in a semi-Markov decision process is presented to estimate the optimal decision variable of the Bayesian control chart. The 2-D oil data based on wear debris collected at regular time epochs from a four-ball tester are adopted to validate the effectiveness of the proposed replacement approach. Given comparisons with the age-based scheme and the failure-based scheme, the proposed Bayesian replacement approach for lubricating oil is demonstrated to achieve better fault detection performance and the longer average availability.