BuildingsLife - The use of genetic algorithms for maintenance plan optimization

Whilst systems have been developed for bridge management, when it comes to buildings a much larger variety of maintenance problems and a small number of buildings per owner means that building management systems are still quite rare. Indeed, efficient approaches to the management of building maintenance still need to be developed. This study addresses the application of a recently developed Building Management System (BdMS) BuildingsLife. This software uses a genetic algorithm applied to Markov Chains to estimate the best maintenance plan. This simulation compares different maintenance plans actions. Each one considers different material properties in a building facade and consequently the building's performance varies during its service life. The varying durability is given by the transition probability of the Markov Chain method. This method proved to be very accurate in the description of the uncertainty of degradation laws, leading to good results in the service life estimation of the facades analysed. The best maintenance plan can be characterized as the plan offering the lowest global cost over a, certain analysis period which allows an acceptable degradation level, as established by the building manager. The genetic algorithm was used to generate multiple combinations of adequately-performing maintenance actions in order to obtain the best result. The application of the proposed method is demonstrated with a case study, leading to coherent results. (C) 2016 Elsevier Ltd. All rights reserved.