Enhancing the maintenance strategy and cost in systems with surrogate assisted multiobjective evolutionary algorithms

Abstract

Digital twins need efficient methodologies to design maintenance strategies for decision-making purposes. Recently, a methodology coupling computational simulation and multiobjective evolutionary algorithms has been proposed for developing maintenance strategies consisting in assigning times for preventive maintenance activities and designing the layout of components of a system, minimizing the unavailability of the system and the strategy cost. Here, surrogate assisted evolutionary algorithms (SAEAs) enhance the multiobjective optimization and improve the drawback of the computational cost of the maintenance strategy assessment based on discrete simulation. Several Kriging surrogates were tested. Two industrial test cases are handled in the experimental section, where the methodology succeed in obtaining nondominated designs improving previous benchmarks, and enhancing state-of-the-art multiobjective optimizers, with up to an order of magnitude in terms of the number of fitness function evaluations. Results show that using multiobjective SAEAs in the development of optimal maintenance strategies could foster and improve digital twins operations.