Constrained parameter extraction for microwave design optimisation using implicit space mapping

To date, space mapping remains one of the most efficient design optimisation methodologies in microwave engineering. Still, its performance depends on the underlying surrogate model, in particular, its approximation and generalisation capabilities. By proper selection of the space mapping transformations and their parameters, a trade-off between these can be obtained. Often, this is done by trial and error that may lead to excessive computational overhead and poor quality of the optimisation outcome. In this study, an adaptively constrained parameter extraction is introduced. Based on convergence results for space mapping algorithms, it allows us to automatically find the approximation-generalisation trade-off through the adjustment of the surrogate model parameter space. Improved performance of the space mapping algorithm is obtained both in terms of convergence properties and the quality of the optimised design. Algorithm convergence is additionally improved by constraining the surrogate optimisation process. The authors´ technique is validated using several microwave design problems.