Discrete sensitivity-based evolutionary design optimization

Publisher Summary This chapter describes a paper that elucidates the range of individual design sensitivities that have been used in an evolutionary design optimization algorithm and presents some applications using these design sensitivities, which involve multioptimality criteria and multiphysics. The process, the evolutionary structural optimization (ESO), has been developed over the last 1010 years. Recent developments have significantly extended the original structural concept to most engineering mechanics situations. This paper demonstrates that the method can now be regarded as a gradient-based semianalytic sensitivity approach whereby the sensitivities are element-based and all design modifications take place at the element level. It shows how such sensitivities can be derived jsut like any physical aspect of a design. Accordingly, the decision regarding element modification can be based on its fitness when subjected to a range of physical and structural behaviors. For multiobjective or/and multiconstraint cases, a composite sensitivity can be formed that accounts for a balance in all the individual objectives/constraints in a Pareto manner with equal weights or penalized weights dependent on the amount of constraint violation.