Condition based maintenance policy for series-parallel systems through Proportional Hazards Model: A multi-stage stochastic programming approach

Condition based maintenance for series-parallel systems is studied in this paper. Due to the effect of covariate values on the component's deterioration, proportional hazard model would be adopted to model hazard rate of each component in the whole system. A control limit is determined at each inspection point for each component to minimize a total expected cost during planning horizon subject to reliability constraint of the whole series-parallel system. Because, covariates play a stochastic role in the proportional hazard model and the maintenance planning has a sequential nature, we would employ a multi-stage stochastic programming to model CBM for series-parallel systems. Different from other studies that researchers attempt to present a fixed control limit at the start point maintenance planning, this paper presents an optimal control limit per each inspection point and provides flexible dynamic control limits. Due to curse of dimensionality, a novel hybrid meta-heuristic algorithm constructed by Parallel Genetic Algorithm and Invasive weed optimization is proposed to find efficient control limits for each component over planning horizon. Its efficiency would be compared with some other classical meta-heuristic algorithms such as Genetic Algorithm, Particle Swarm Optimization and Invasive Weed Optimization. The results of the computational experiments are statistically discussed and indicate that the proposed hybrid algorithm outperforms the other mentioned algorithms.