Measurement and Modeling of Magnetostriction in Transformer Core Based on a BPNN Method Assisted with Levenberg-Marquardt Algorithm

摘要

The reasons causing the vabriations and noises of transformer cores include the magnetostrictive force in the core material and electromagnetic forces acting on lapped seams of cores and among silicon steel sheets. This paper carried out the measurement of magnetostrictive strains in a grain-oriented steel sheet under alternating magnetizations and dicussed the effect of a dc bias present in the exciting current on the magnetostriction. Based on the experimental data, the magnetostrictive characteristics are modeled by a back propagation neural network (BPNN) method. In order to achive an acceptable convergence condition, the Levenberg-Marquardt (LM) algorithm instead of the traditional gradient descent algorithm was employed to assist the BPNN.