Solving Multi-objective Optimal Design and Maintenance for Systems Based on Calendar Times Using GDE3

Abstract

Industries that expect to optimize the performance of physical assets have to contemplate consider design alternatives and maintenance strategies from the phases of project and construction. The problem to be solved presents two objectives in conflict: maximising times in which the system is available and minimising costs due to both maintenance activities and recoveries after failure. From the system functionability profile, information in relation to the system availability and operation costs arises. It allows the joint optimisation of system design and maintenance strategy using Multi-objective Evolutionary Algorithms (MOEA). The system functionability profile is generated and modified in an iterative process, which uses both (a), Discrete Simulation to characterise the randomness of the process until the failure and (b), the maintenance strategy (optimum period to perform preventive maintenance activities to the devices included in the system design) obtained along the evolutionary process. An application case is presented and successfully solved. Moreover, several configurations of an MOEA with selection criteria based on Pareto Dominance with Differential Evolution as operator (GDE3) are compared to solve the multi-objective problem with Hypervolume indicator and statistical significance analysis. A similar performance is observed, bringing to the light the robust behaviour of the GDE3 method, while a balanced solutions set is successfully found.