An Efficient Stochastic Search with Minimal Initial Population for Structural Optimization

Genetic Algorithms are best suited for unconstrained problems; however, most of the practical cases have constraints. As a common approach, modifying initial population due to problem-specific information has not yet come to an end. This is due to the generalization challenges and also the lack of diversity and effectiveness regarding relatively narrow size of the feasible subspace of the entire search space. In this article, a new type of expanding genetic population is presented starting from its minimal size. Suitable ideas from ant colony and simulated annealing approaches are utilized for an adaptive efficient search which is also tuneable by the developed extra control parameters. Effectiveness and efficiency of the proposed method are illustrated by capturing the global optimum in a number of well-known structural size and layout optimization examples in a considerably less fitness evaluations compared to the other standard methods