Mask Reconstruction via Kernel-Based Surrogate Modeling for Automated Circulator Design

This letter introduces a mask-based surrogate modeling framework that generalizes the feature-based paradigm for efficient microwave design exploration using a limited number of full-wave simulations. Unlike feature-based approaches, which extract scalar performance metrics, the proposed method reconstructs the complete S -parameter boundary contour within the operating band. This preserves each design's parametric signature and enables on-demand evaluation of multiple performance specifications without model retraining. The framework employs two vector-valued kernel ridge regression models to predict boundary coordinates from geometric parameters. Validation on a three-port circulator demonstrates accurate reconstruction and rapid exploration of thousands of candidate designs.