Résumé: Handling constraints is a common challenge to all optimisation methods. To no exception is the planning and optimisation of manufacturing processes that often involves a number of constraints reflecting the complicated reality of manufacturing to which the pursuit of the best operation condition is subject. Mathematical models describing today's manufacturing processes are generally discontinuous, non-explicit, and not analytically differentiable; all of which renders traditional optimisation methods difficult to apply. Genetic Algorithm (GA) is known to provide an optimisation platform method capable of treating highly nonlinear and ill-behaved complex problems, thereby making it an appealing candidate. However, several issues in regard to the handling constraints must be rigorously addressed in order for GA to become a viable and effective method for manufacturing optimisation. In this paper, a new constraint handling strategy combined with (?,?)- population initialisation is proposed. Twelve numerical test cases and one surface grinding process optimisation are presented to evaluate its optimisation performance.

Langue: Anglais