Bayesian Network for Decision Aid in Maintenance

Most companies register a large quantity of information in their maintenance data bases, but this is often poorly exploited. An efficient valorization of a maintenance history can improve future maintenance strategies. In maintenance optimisation, four general types of maintenance strategies are taken into account such as: a) corrective, b) preventive, c) predictive, and d) reliability centered maintenance. In this paper, a decision aid tool and a model for preventive maintenance are proposed. Preventive maintenance (PM) is an important component of the maintenance activity. Within a maintenance organization it usually accounts for a major proportion of the total maintenance effort. PM may be described as the care and servicing activities carried-out by maintenance people to keep equipment/facilities in satisfactory operational state by providing for systematic inspection, detection, and correction of incipient failures either prior to their occurrence or prior to their development into major failure. To develop an effective PM program, the availability of a number of machines is necessary Corrective maintenance (CM) may be defined as the set of remedial actions carried out due to failure or deficiencies discovered during preventive maintenance, to repair an equipment/item to restore its operational state. Usually, corrective maintenance is an unscheduled maintenance activity, basically composed of unpredictable maintenance actions that cannot be pre-planned or programmed. The aim of present work is to give a solution in developing probabilistic models for decision aid in preventive maintenance strategies starting from corrective maintenance data. The paper is organized as it follows: first, the main maintenance indicators and are reviewed. Having made a survey of the significant results presented in the literature, probabilistic models are proposed and Bayesian dynamic networks are used in a case study to illustrate the methodology.