Computationally-efficient surrogate-assisted dimension scaling of compact dual-band couplers

摘要

Reusing existing results in order to facilitate optimisation of structures for various sets of performance requirements such as operating frequencies is an attractive concept. When implemented properly, it can lead to considerable speedup of the microwave design process. In general, re-design of microwave structures is a challenging task, particularly for compact components where considerable electromagnetic (EM) cross-couplings make the relationships between geometry parameters and the structure responses complex. In this study, the authors address geometry scaling of miniaturised dual-band couplers by means of inverse surrogate modelling. It allows for fast estimation of the optimum values of structure dimensions corresponding to required operating frequencies, independently for each band. The inverse surrogate is constructed based on a set of reference designs obtained by optimising the equivalent circuit model of the coupler, and subsequently corrected using output space mapping to accommodate the discrepancies between the network model and the EM-simulation one. As demonstrated using a compact dual-band microstrip branch-line coupler example, geometry scaling is possible in a wide range of operating frequencies and at a very low cost of just a single EM simulation of the structure at hand. Numerical results are validated through physical measurements of fabricated coupler prototypes for several test cases.