Low-cost data-driven modelling of microwave components using domain confinement and PCA-based dimensionality reduction

Fast data-driven surrogate models can be employed as replacements of computationally demanding full-wave electromagnetic simulations to facilitate the microwave design procedures. Unfortunately, practical application of surrogate modelling is often hindered by the curse of dimensionality and/or considerable nonlinearity of the component characteristics. This paper proposes a simple yet reliable approach to cost-efficient modelling of miniaturized microwave components which adopts two fundamental mechanisms to improve the computational efficiency of setting up the surrogate. Firstly, the model domain is confined-using a set of pre-optimized reference designs-to the region of the parameter space containing high-quality designs with respect to the relevant performance figures. Secondly, the domain is spanned by the selected principal components of the reference set for dimensionality reduction. Consequently, the surrogate model, covering wide ranges of the device parameters and operating conditions, can be established using a fraction of training data samples required by conventional approaches, without compromising its predictive power. The proposed technique is illustrated using two miniaturized microstrip components: a rat-race coupler (RRC) and an impedance matching transformer. The following accuracies of the PCA-based surrogates have been obtained: 0.9% for RRC and 2.1% for the transformer (for 800 training data samples).