Recent Advances in Automated Multiphysics Parametric Modeling for Microwave Components

This paper presents an efficient automated multi-physics modeling technique for microwave components using artificial neural network (ANN). Automated model generation (AMG) method is expanded from developing single electromagnetic (EM) domain parametric ANN model to EM-centric multiphysics domain parametric ANN model. This technique integrates all subtasks of multiphysics parametric modeling into a unified algorithm, including automated multiphysics ANN structure adjustment and automated multiphysics data sampling/generation. To further reduce the time comsumption of multiphysics data sampling and multiphysics simulations, parallel computation is utilized to drive multiple EM-centric multiphysics evaluations simultaneously on multiple computers. The presented technique can effectively reduce the time of multiphysics ANN modeling and the consumption of manpower compared to manual multiphysics modeling techniques. Finally, we use a multiphysics parametric modeling example of a microwave filter to demonstrate the advantage of presented technique.