Modeling of Anisotropic Magnetostriction Under DC Bias Based on an Optimized BP Neural Network
Zhen Wang, Yanli Zhang, Ziyan Ren, Chang-Seop Koh, Osama A. Mohammed
Abstract
The existence of direct current (dc) bias magnetic field changes the magnetostrictive properties of electrical steel sheets (ESSs) and intensifies the vibration and noise of transformers. In this article, the dynamic magnetostrictive properties of the ESSs under dc bias are measured and analyzed, and it is shown that the magnetostriction of the ESSs under dc bias has a complex dynamic anisotropy. An optimized backpropagation neural network (BPNN) is proposed to model the dynamic magnetostriction curves under the dc bias. In the construction of the BPNN, the topology is optimized by using the genetic algorithm and the parameters are optimally decided by using particle swarm optimization (PSO), respectively, to get a fast convergence and avoid the local optimal solution. The comparison of the proposed model with experimental measurements shows the accuracy of the proposed model is improved by 6%.