A Novel Cognition-Driven Electromagnetic Optimization Using Impedance Feature-Based Space Mapping for Microwave Filters

This article proposes a novel cognition-driven electromagnetic (EM) optimization using impedance feature-based space mapping (SM) for microwave filters. We propose to transform cognition from the previous <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</i>-parameter features to <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Z</i>-parameter zero and pole features. The impedance feature avoids the issue of being unable to explicitly identify feature information when the starting point of design optimization is far away from the design specifications. Moreover, when the starting point of optimization is worse or the single iteration optimization domain of features is wider, the impedance features are more identifiable than the TF-based <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</i>-parameter features. Besides, the impedance feature space can directly include the transmission zero features without increasing optimization complexity. The issue of the inability to intuitively obtain impedance feature objectives limits the impedance feature-based cognition-driven optimization. We propose an iterative approximation method based on invariant phase factors independent of nonport geometrical parameters for obtaining cognition of impedance features. Furthermore, we derive cognition-driven formulas of impedance feature-based SM, eliminating the need for explicit surrogates and operating in the new impedance feature space. The proposed cognition-driven optimization technique can achieve faster convergence than existing cognition-driven optimization using TF-based feature surrogates while maintaining robustness to worse optimization starting points. The proposed method is demonstrated by applying EM optimizations to three examples of microwave filters.