Complex Frequency Domain-Based Multifeature-Assisted Electromagnetic Surrogate Model Optimization for Microwave Filters

This article proposes a novel multifeature-assisted electromagnetic (EM) surrogate model optimization method based on the complex frequency domain (CFD) EM parameter extraction technique. By introducing a CFD-based fast frequency sweep, the method enables efficient extraction of internal features from EM parameters, thereby facilitating the construction of high-accuracy parametric surrogate models and supporting advanced design of microwave filters. In scenarios where the initial design response is far from the target or the EM response lacks clarity, a feature-assisted optimization strategy is adopted to guide the design process. By exploiting multiple feature parameters and introducing a novel objective function, the proposed method effectively relocates the passband into the desired design specifications. The novel feature parameter extraction technique based on CFD can solve the problem of poor optimization performance in traditional surrogate models caused by the multiple-solution issue in vector-fitting (VF) parameter extraction. Compared to conventional surrogate model optimization methods, the proposed method achieves faster convergence for practical high-level design applications.