Multi-point response correction for cost-efficient antenna and microwave design optimization

Microwave design optimization using variable-fidelity electromagnetic (EM) simulations and multi-point response correction is presented. In our approach, optimization of the computationally expensive high-fidelity EM model is replaced by iterative design improvement of the low-fidelity model, which is obtained from coarse-discretization EM simulation data. The low-fidelity model is enhanced through a multi-point response correction that utilizes all high-fidelity model data accumulated during the optimization run. The model parameters are obtained analytically by solving an appropriate linear regression problem. Optional frequency scaling is also used to further improve the model alignment. The multi-point correction ensures better generalization than the traditional single-point correction and results in reduction of the computational cost of the design process. Our considerations are illustrated using two examples: a dielectric resonator antenna and a microstrip filter.