Shape-preserving response prediction with adjoint sensitivities for microwave design optimization

A new development of the shape-preserving response prediction (SPRP) technique for microwave design optimization is presented. The original SPRP method is enhanced by employing low-cost derivative information obtained through adjoint sensitivities. Sensitivity ensures first-order consistency between the SPRP surrogate and the high-fidelity electromagnetic (EM) model under optimization and to speed up the surrogate optimization process. Fast surrogate optimization allows us to use coarse-discretization EM simulations as an underlying low-fidelity model and, therefore, efficiently apply SPRP to cases where reliable circuit models are not available. Our approach is demonstrated using a dielectric resonator filter example.