The determination of maintenance strategies for plants subject to breakdown

When plant components fail the opportunity may be taken during the ensuing shut down to carry out maintenance on other components which have not yet failed. The choice of components for this opportunity maintenance depends on the probability distributions of their residual operating lives which may, in turn, be influenced by the process operating conditions. Three methods are described for determining the strategy of opportunity maintenance which will maximize the overall performance of the plant. The problem posed in the form of a stochastic decision tree is treated by a branch and bound procedure using first a Monte Carlo simulation and secondly multidimensional numerical integration formulae to account for the stochastic variables. The third method treats the problem as a Markov decision process. This proves to have computational advantages particularly when treating broad component life time distributions and long planning periods. Furthermore, the dynamic programming nature of the Markov procedure yields maintenance strategies for all defined states of the system from a single computational run.