Asymptotic availability allocation with genetic algorithms
Résumé
This paper proposes a genetic algorithm to allocate the availability goal to a parallel-serial system components. The problem is formulated as a multi-objective optimization problem in continue and discrete variables. It is a combinatorial optimization that is NP-hard, so genetics algorithms are suitable for attacking such a problem. We use the weighing approach and exact penalty method to relax the problem constraints. An experimental factorial plan is proposed to analyze each parameter's influence on the genetic algorithm's results. From this factorial plan, we have indications on how to choose the parameters of our algorithm. The approach used is relevant for genetics algorithms users. The results that we obtain are suitable and enable us to show how powerful are genetics algorithms. A numerical example is given to assess the algorithm.