ABSTRACT
Topological optimization has been widely tackled by algorithms with stochastic principles. This kind of algorithm is rather simple to implement and avoids local minima. Also, energy functions can be used when the gradient does not exist or is difficult to approximate, and an initialization procedure is not needed. Genetic algorithms (GAs) have been used to solve shape optimization problems, providing feasible solutions with acceptable objective values [1-5]. Nonetheless, GA solutions would be difficult to manufacture since small holes are present or unconnected pieces appear in the evolved design. This behavior can be explained by population diversity issues, which favor premature convergence and reduce exploration of the search space [1, 2, 5]. In this chapter we present a multi-objective optimizer and a post-process which improves the exploration and solves the problem of small holes and connectedness of the structure.
