Optimisation of system resources in reliability availability & maintainability problems using genetic algorithms

Abstract

We propose an application of Genetic Algorithms to the output from Reliability Availability & Maintainability models of complex (mainly industrial) systems to reduce computational costs while maintaining a desired level of fidelity of the model. The paper notes the fundamental difficulty of applying resource optimisation to models of complex systems given the dimensional and computational costs involved in realising it. To date, several analytic approaches have been suggested to provide solutions, which are very fast, but fundamental question marks have arisen regarding their fidelity. By their nature they are limited in the complexity of problems they can tackle. A hybrid approach, which utilises Monte Carlo Simulations together with an analytic metric for the search process has also been proposed. This hybrid approach preserves the model's accuracy and fidelity, but as it involves a study of the model's physics is limited to spare parts and repair teams as the resources, and to the availability as the objective function in the optimisation. The approach proposed here also uses Monte Carlo Simulations, but the difference is that the Genetic Algorithms approach does not require any prior study of the model's underlying structure, instead referring to the Monte Carlo Simulations as a black box engine, which yields an accurate set of outputs for a given set of inputs. Hence, the Genetic Algorithms approach can be applied for the optimisation of a wider resources space and for any desired objective function defined and calculated in the model. A the same time, the application of the Genetic Algorithms is found to be an efficient and relatively cheap search method for optimisation

More information

Identifiers

ORCID for A.J. Keane: orcid.org/0000-0001-7993-1569

Catalogue record

Export record