Response Features for Global and Multi-objective Optimization

Previous chapters have discussed utilization of response feature technology for solving local optimization problems. Specifically, Chap. 6 addressed local feature-based optimization of high-frequency components for various types of structures and responses, whereas in Chap. 9 , the employment of the response feature technology for reliability improvement of local optimization procedures has been considered. The examples described in Chaps. 6 and 9 corroborate that the incorporation of the response features technology into local search procedures enable quasi-global search capabilities. This is because the employment of feature-related reliability enhancements inherently regularizes the objective function landscape, thereby improving the immunity of the search process to poor initial conditions and permitting redesign of a given device for operational parameters significantly different than those at the starting point. Nevertheless, in some cases, the enhancement of local search procedures may not be sufficient, and a truly global search may be indispensable. Still, computational cost of simulation-driven global design optimization of high-frequency components is usually exorbitant. A workaround is offered by surrogate-based procedures; yet, the curse of dimensionality strongly limits their applicability to the devices described by few design parameters within the limited ranges thereof. This chapter outlines two feature-based surrogate-assisted frameworks for global optimization of antennas and microwave devices. The first one employs feature-based inverse surrogate models, whereas in the second, simplex-based regression surrogates have been exploited. The remaining part of this chapter outlines feature-based multi-objective design optimization. All the discussed methods have been extensively illustrated using examples of microwave and antenna structures.