SCHEDULING OF PREVENTIVE MAINTENANCE FOR PARALLEL, SERIES, AND SINGLE-ITEM REPLACEMENT SYSTEMS BASED ON STATISTICAL ANALYSIS

Preventive maintenance scheduling of generating units is addressed as a crucial issue in power system studies due to its severe impacts on power systems' asset man-agement by reducing operation cost while enhancing re-liability worth. In this paper, we describe an application of statistical analysis for determining the best preventive maintenance strategy in the case of parallel, series, and single-item replacement systems. A key aspect of indus- trial maintenance is the trade-o between cost and time of performing preventive maintenance operations. This article deals with two main issues related to the preven- tive maintenance in order to minimize the cost per unit time: (1) specifying the best time for performing preven- tive maintenance actions in parallel, series, and single- item replacement systems; (2) determining the required number of spare parts and facilities. in single-item replacement and parallel systems, respec- tively. In this proposed maintenance strategy, preventive maintenance operations are regularly performed on the production unit in equal time intervals and maintenance actions return the facilities to a good-as-new condition. In the proposed model, the number of failures for a spec- i ed facility follows the Poisson distribution, and failure rate of the Poisson distribution follows Gamma process. In addition, equipment' failure being earlier than the de- termined starting time for the preventive maintenance operations would impose too much cost on each system. Therefore, to consider this constraint, it is assumed that there is a linear relation between recovering cost from unplanned failures and time to failure. Moreover, all costs related to maintenance are supposed to be known and the cost of a preventive repair is appreciably less than that of a failure and its associated repair. Finally, three examples are presented to demonstrate the e ec- tiveness of the proposed models for each system. We use MATLAB software to obtain the optimal value of parameters in cost models in the proposed approach. To nd the optimal value of decision variables in each system, a search procedure is applied.