Skip to Main Content
Space mapping (SM) is probably one of the most efficient simulation-driven design methodologies used in microwave engineering to date. Nevertheless, its performance heavily depends on the proper selection of the surrogate model, a computationally cheap representation of the microwave structure under consideration, which is the key component of the SM algorithm. Despite some attempts of automating the set-up of the surrogate model and its parameters, the successful application of SM still requires some user experience. In this work, an efficient method to alleviate this problem is discussed that exploits adaptively constrained parameter extraction and surrogate optimisation processes. As a result, improved convergence properties and overall performance of the SM algorithm are observed. The proposed technique is verified through the design of several microstrip filters. It is also compared with standard SM as well as SM enhanced by a trust-region methodology.