A genetic optimization design methodology based on support vector regression

Aiming at addressing the optimization design problems with implicit objective performance functions, a genetic optimization design methodology based on the support vector regression (SVR) response surface is proposed. First appropriate design parameter samples are selected by experimental design theories, then the response samples are obtained from the experiments or numerical simulations. Applying the genetic algorithm (GA) to optimize the parameters of SVR, the response surface is constructed and treated as the objective performance functions. Combing other constraints, the optimization model is formed and ready to be solved by GA. The structure optimization of the microwave power divider is adopted as an example to illustrate this methodology. The learning samples are obtained from uniform design theory and the high frequency electromagnetic field finite element analysis codes (HFSS). Three response surface objective functions for the magnitude, phase and VSWR of the microwave power divider model are obtained and the multi-objective optimization problem is solved. The results show that this methodology is feasible and highly effective, and thus can be used in the optimum design of engineering fields.