Optimizing maintenance and repair policies via a combination of genetic algorithms and Monte Carlo simulation

In this paper we present an optimization approach based on the combination of a Genetic Algorithms maximization procedure with a Monte Carlo simulatio n. The approach is applied within the context of plant logistic management for what concerns the choice of maintenance and repair strategies. A stocha stic model of plant operation is developed from the standpoint of its relia bility/availability behavior, i.e. of the failure/repair/maintenance proces ses of its components. The model is evaluated by Monte Carlo simulation in terms of economic costs and revenues of operation. The flexibility of the M onte Carlo method allows us to include several practical aspects such as st and-by operation modes, deteriorating repairs, aging, sequences of periodic maintenances, number of repair teams available for different kinds of repa ir interventions (mechanical, electronic, hydraulic, etc.), components prio rity rankings. A genetic algorithm is then utilized to optimize the compone nts maintenance periods and number of repair teams. The fitness function ob ject of the optimization is a profit function which inherently accounts for the safety and economic performance of the plant and whose Value is comput ed by the above Monte Carlo simulation model. For an efficient combination of Genetic Algorithms and Monte Carlo simulation, only few hundreds Monte C arlo histories are performed for each potential solution proposed by the ge netic algorithm. Statistical significance of the results of the solutions o f interest (i.e. the best ones) is then attained exploiting the fact that d uring the population evolution the fit chromosomes appear repeatedly many t imes. The proposed optimization approach is applied on two case studies of increasing complexity. (C) 2000 Elsevier Science Ltd. All rights reserved.