سیاست دو مرحله ای تعمیر و نگهداری پیشگیرانه برای یک سیستم خراب چند حالته

This paper is to propose a two-stage preventive maintenance (PM) policy for the multi-state deterioration system under periodic inspection and with multiple candidate actions for PM. Such policy is mainly to schedule PM optimally and also on how to choose the action optimally for each PM. The scheduling includes two tasks: to determine after completing each PM when to make the decision and then at it to schedule the next PM optimally. We assume that: (1) such actions except replacement are imperfect, (2) the inspection and action times can be ignored, (3) the system can be modeled by a multi-state discrete time Markov chain whose transition probabilities will change and be updated only at the instant after completing each PM, and (4) the risks of such imperfect actions will be updated only at the instant after completing each PM. Finally, an example through simulation is presented to illustrate how such policy can be carried out.

This paper is to propose a two-stage preventive maintenance (PM) policy for the multi-state deterioration system under periodic inspection and with multiple candidate actions for PM. Such policy is mainly to schedule PM optimally and also on how to choose the action optimally for each PM. The scheduling includes two tasks: to determine after completing each PM when to make the decision and then at it to schedule the next PM optimally. We assume that: (1) such actions except replacement are imperfect, (2) the inspection and action times can be ignored, (3) the system can be modeled by a multi-state discrete time Markov chain whose transition probabilities will change and be updated only at the instant after completing each PM, and (4) the risks of such imperfect actions will be updated only at the instant after completing each PM. Finally, an example through simulation is presented to illustrate how such policy can be carried out.

A two-stage PM policy is proposed herein for a deterioration system under scheduled inspection with period d. Such policy is mainly to derive how to schedule the PM optimally and how to choose the action optimally for each PM. The action for each PM can be chosen from multiple candidate imperfect actions plus replacement. The system can be modeled into a multi-state discrete time Markov chain whose transition probabilities change only at the instant after completing each PM. Both transition probabilities and action risks are updated at the instant after completing each PM in terms of maintenance data. The scheduling of the next PM and the choice of the optimal actions for the next PM will be both time and state dependent. Due to inspection taking time, our results will also hold to the system under continuous inspection if we take d either as inspection time or small enough. The future study may go on along the lines that: (1) choose better objective functions for determining the action for the rth PM optimally, (2) take the nonzero action time into consideration, (3) take the time-dependent candidate action set into consideration, and (4) take the time-dependent action cost into consideration.