Sensitivity analysis and optimization utilizing an approximate auxiliary problem

We propose an optimal shape design technique for full-wave structured-grid solvers. The technique provides superior accuracy over other approximate adjoint techniques. We develop a new adjoint problem where perturbations in the design parameters take place. This is in contrast to existing techniques which exploit perturbation dependent original problem solutions. The new approach improves the accuracy of the estimated derivatives while preserving the computational efficiency typical for the adjoint technique. We illustrate the technique through the computation of response sensitivities and the optimization of a waveguide filter and a printed antenna.