Introduction of a New Selection Parameter in Genetic Algorithm for Constrained Reliability Design Problems

In this article we propose to introduce a new selection parameter in Genetic Algorithms (GAs) for a class of constrained reliability design problems. Our work demonstrates two major points. The first one is that the populations are quickly included in the space of the feasible solutions for a sufficiently large selection of parameter value. The second one is that the value of the selection parameter controls the exploration strategy of the feasible space. These two properties illustrate that an adapted choice of the selection parameter value allows to improve the performance of GA. Furthermore, our numerical examples tend to show that, with an adapted choice of the selection parameter, these GAs are in practice more efficient than previously proposed GAs for this class of problems.