An adaptive predictive maintenance model for repairable deteriorating systems using inverse Gaussian degradation process

Predictive maintenance is a promising solution to keep the long-run operation of industrial systems at high reliability and low cost. In this spirit, we aim to develop an adaptive predictive maintenance model for continuously deteriorating single-unit systems subject to periodic inspection, imperfect repair and perfect replacement. The development consists of four steps: degradation modeling, maintenance effect modeling, maintenance policy elaboration, and performance evaluation. Compared with existing models, ours differs in three main aspects. Firstly, we take into account the past dependency of maintenance actions in the degradation modeling via the random effect of an inverse Gaussian process. Secondly, we use both the system remaining useful life and maintenance duration to enable dynamic maintenance decision-making. Finally, we take advantage of the semi-regenerative theory to analytically evaluate the long-run cost rate of maintenance policies whose decision variables are of different nature. We validate and illustrate the developed adaptive predictive maintenance model by various numerical experiments. Comparative studies with benchmarks under different maintenance costs and degradation characteristics confirm the flexibility and cost-effectiveness of the model.