In this research, different optimization models are developed to solve the preventive maintenance (PM) optimization problem in a maintainable multi-state series–parallel system. The objective is to determine for each component in the system the maintenance period minimizing a cost function under the constraint of required availability and for a specified horizon of time. Four genetic models based on the cost associated with maintenance schedule and availability characteristic parameters are constructed and analyzed. They are genetic algorithm (GA), hybridization GA and local search (GA-LS), fuzzy logic controlled GA (FLC-GA), and hybridization FLC-GA and LS. The experiment analyzes and compares the efficiency between them. These experiments investigate the effect of the parameters of the GA on the structure of optimal PM schedules in multi-state multi-component series–parallel systems. Results show that the hybridization FLC-GA and LS outperform the other algorithms.

The paper presents a method for determining an optimal loading in series-parallel systems. The optimal loading is aimed at achieving the greatest possible expected system availability subject to required demand constraint. Then the corresponding system cost is deduced. We consider that system cost is a combination of downtime cost (loss of productivity), and repair cost (supposed proportional to repair time). The former is affected by a penalty value which reflects the importance of downtime cost with respect to repair cost. The model takes into account the relationship between the element failure rate and its corresponding load (element capacity). The universal generating function model is used to assess the performance distribution of the entire system and the system availability (knowing the probability of each performance level). Then, the unavailability and the repair time are estimated in order to study the system cost. The optimization is done for different values of required demand. The effect of required demand on the system availability and system cost is studied. The optimization technique is based on the genetic algorithm in order to determine the optimal load distribution. An illustrative example is presented.

No abstract

No abstract