Vol. 115

Front:[PDF file] Back:[PDF file]
Latest Volume
All Volumes
All Issues
2021-09-26

Neural-Network-Based Source Reconstruction for Estimating Linear Synchronous Motor Radiation

By Lu Xing, Yinghong Wen, David W. P. Thomas, Jinbao Zhang, and Dan Zhang
Progress In Electromagnetics Research C, Vol. 115, 219-232, 2021
doi:10.2528/PIERC21071205

Abstract

An equivalent source model based on neural network is proposed to rapidly estimate the magnetic radiation characteristics of linear synchronous motor (LSM) in electromagnetic suspension (EMS) maglev system. The equivalent source is composed of electric dipoles and a closed three-dimensional (3-D) surface, and is developed in terms of source reconstruction method. A few sampling data of magnetic field simulation results serve as the input information to determine the unknown current distribution on equivalent source model. To solve the inverse radiation problem and characterize the whole radiation pattern with high efficiency, the current distribution signature of equivalent model is fitted into artificial neural network models. Separate neural network models are fitted under different phases of winding excitation, which enables the low-frequency magnetic field estimation under both 3-phase balanced operation and unbalanced operation. The equivalent source model is extended to estimate LSM radiation in multi-source environment, and the comparison with numerical simulation verifies its accuracy and efficiency.

Citation


Lu Xing, Yinghong Wen, David W. P. Thomas, Jinbao Zhang, and Dan Zhang, "Neural-Network-Based Source Reconstruction for Estimating Linear Synchronous Motor Radiation," Progress In Electromagnetics Research C, Vol. 115, 219-232, 2021.
doi:10.2528/PIERC21071205
http://www.jpier.org/PIERC/pier.php?paper=21071205

References