A mixed evolutionary/heuristic approach to shape optimization

The problem of nding the optimal shape of a continuous structure is addressed using, alternatively, heuristic, evolutionary and mixed evolutionary and heuristic optimization strategies. Boundaries are represented by B-splines. Two heuristics for minimizing the weight of a structure subject to limits on von Mises stresses and geometrical constraints are implemented: `generalized biological growthʹ and `penalized biological growthʹ. Penalized biological growth adds to generalized biological growth a control for shape changes. This control is based on the overall state of constraints satisfaction in the structure. The two heuristics are very e cient at improving the designs, but they do not yield globally optimal shapes. Therefore, they are interfaced with an evolutionary optimizer. Di erent strategies for mixing evolutionary search and biological growth are compared. Results are obtained for fan disk shape problems. They show that mixing evolutionary search with biological growth improves the e ciency of the optimization. The method o ers to the designer new paths for a better component determination