The Study of Electromagnetic Finite Element Method and Bayesian Deep Learning for SWS Analysis Based on Python Programming

In this paper, Python programming is employed to study the electromagnetic finite element method (FEM) and Bayesian deep learning. Rectangular cavity and folded waveguide (FWG) slow-wave structure (SWS) are taken as examples. Compared with the theoretical calculation and HFSS simulation, Python-based FEM (Py_FEM) has considerable calculation accuracy. Additionally, multi- process computing is studied to parallelize the calculation process to reduce the computational time, and adaptive mesh refinement (AMR) is studied to enhance the calculation accuracy. Finally, based on the Py_FEM, a Bayesian neural network (BNN) is trained to provide the prediction value and prediction uncertainty of the dispersion characteristics.