Response Features for Reliability Improvement of Local Optimization Procedures

The focus of this chapter is the employment of the response feature technology for improving reliability of local optimization procedures. This chapter considers two different approaches: adaptively adjusted design specification strategy and frequency-based regularization. Both methodologies involve mechanisms aimed at enhancing the performance of the search process under challenging scenarios, primarily poor initial conditions or the necessity of re-designing the high-frequency structure under study for operating conditions distant from that of the available design. If a decent starting point is not readily available, local search procedures are usually prone to a failure, and, consequently, a designer may resort to the employment of global optimization algorithms, which allow to mitigate the aforementioned issues. However, their computational cost is exorbitant when carried out using full-wave electromagnetic (EM) simulations, otherwise necessary to ensure adequate accuracy. Typically, the computational cost of the global search is at least ten times higher than that of the local procedures. The techniques of this chapter allow to refrain from utilizing prohibitively costly global optimizers by enhancing the performance of the local search routines. Consequently, the computational cost of rendering the optimal design becomes acceptable. In this chapter, we also discuss the incorporation of various acceleration mechanisms into the considered procedures such as variable-fidelity simulations and sparse sensitivity updating schemes. The outline of the mentioned methods is accompanied by illustrative examples including microstrip antennas and passive microwave circuits.