Dynamic dispatching and preventive maintenance for parallel machines with dispatching-dependent deterioration

Abstract

A dynamic decision model that coordinates dispatching and preventive maintenance decisions for failure-prone parallel machines in make-to-order (MTO) production environments is developed in this research. The primary objective is to minimize the weighted long-run average waiting costs of MTO systems. Two common but seldom studied stochastic factors, namely, the dispatching-dependent deterioration of machines and machine-health-dependent production rates, are explicitly modeled in the proposed dynamic dispatching and preventive maintenance (DDPM) model. Although the DDPM model is developed using Markov decision processes, it is equally effective in non-Markovian production environments. The performance of the DDPM model is validated in Markovian and non-Markovian production environments. Compared with several methods from the literature, simulation results show an improvement of at least 45.2% in average job waiting times and a minimum reduction of 48.9% in average machine downtimes. The comparison results between the optimal dynamic dispatching policies with and without coordinated preventive maintenance show that performance improvement can be mostly attributed to the coordination between preventive maintenance and dispatching decisions.