RUL-centered postponed maintenance planning under random failure threshold

State-driven asset management, which fully considers the health state evolution pattern, is essential to improve the operation performance and reduce the downtime loss of industrial systems such as railway vehicles, wind power equipment and nuclear equipment. A common phenomenon in health management is the uncertainty (improvisation and/or cognition) of degradation-based failure threshold. This paper solves this problem by scheduling a new postponed replacement policy to capture the distribution characteristics of fault threshold. The health evolution is described by a generalized stochastic process, and its lifetime model is established by combining the randomness of both deterioration increment and threshold. A reliability threshold is set to determine whether preventive replacement (PR) is available. If the predicted reliability is lower than the threshold, PR will be postponed following the expectation of the residual service lifetime adjusted by a postponement factor, so as to enhance the asset availability. The objective is to minimize the long-run operational cost by jointly optimizing the inspection interval, reliability threshold and postponement factor. A case study of bearing health management demonstrates that the proposed policies and models outperform some conventional maintenance policies in availability enhancement and cost reduction.