Masonry arch collapse loads and mechanisms by heuristically seeded genetic algorithm Original Research Article

A novel method of initialising and applying genetic algorithms to the limit analysis problem of masonry arch collapse mode identification is described. The method of encoding an individual uses half the resources (with no accompanying loss of information) in comparison with a simple genetic algorithm (SGA) implementation. The new genetic algorithm (GA) uses a set of heuristic rules to seed the initial population and control subsequent genetic operations (mutation and reproduction). The heuristics developed are seen to be applicable to different sized problems. The heuristic GA (HGA) results are compared with simple GA results and are shown to be significantly more efficient. The heuristic GA finds better solutions than the simple GA after far less iteration and with smaller populations. Several conventionally challenging problems illustrate the solving efficiency of the heuristically initialised genetic algorithm formulation.