Electromagnetic Simulation-Inserted Optimization Method Incorporating Transfer Function Feature-Based Mesh Space Mapping

Simulation-inserted optimization (SIO), integrating the electromagnetic (EM) optimization algorithm into the internal EM simulation, is an effective EM optimization method. However, SIO is computationally expensive, with a large CPU running memory to solve extensive finite-element method (FEM) system equations. To address the additional memory and extra computational time caused by large FEM system equations, this article proposes an efficient SIO method using transfer function (TF) feature-based mesh space mapping (MSM-SIO). The proposed MSM-SIO method incorporates the MSM technique inherently into SIO and transfers most fine mesh optimization time to coarse model optimization and surrogate model development. A parameter extraction technique based on the complex frequency domain (CFD) is introduced to extract the TF features of the fine mesh model. Thus, a new surrogate model utilizing the extracted TF features is developed to accelerate the MSM-SIO method. Frequency mapping (FM) is incorporated into the MSM-SIO to improve the training performance of the surrogate model further and accelerate the overall process. The proposed method effectively reduces computational time while ensuring great convergence. Three examples of EM optimization of microwave components are used to validate the effectiveness of the proposed method.