Method of measuring the mechanical properties of ferromagnetic materials based on magnetostriction EMAT and sound velocity

Abstract

• A NDT method to measure the mechanical properties of ferromagnetic materials based on the magnetostriction EMAT and sound velocity . • Extract reliable pattern parameters (including five magnetostrictive parameters and sound velocity parameter) from the EMAT signal. • Parameter estimation of mechanical properties based on generalized regression neural network. The measurement of mechanical properties parameters has been found as the standard process to predict the quality of ferromagnetic materials, which is a critical step in the production of ferromagnetic materials. The conventional testing methods of mechanical properties are dependent on Destructive tests. In this study, a non-destructive method was proposed based on magnetostriction EMAT and sound velocity to measure the mechanical properties of ferromagnetic materials imprinted by the materials’ microstructure since the microstructure can have an effect on the pattern of the magnetostrictive effect and sound velocity of ferromagnetic materials. For experimental verification, the ultrasonic propagation time was calculated by filtering the orthogonal demodulation of the electromagnetic ultrasonic signal. Based on BP (back propagation) neural network, the Mean Impact Value (MIV) of each characteristic value was investigated to predict the effect arising from sound velocity on the network output. Moreover, with GRNN (general regression neural network), this study established the mapping relationship between magnetostrictive parameters (with or without combined sound velocity) and mechanical properties of the material, and the mechanical properties of the materials were non-destructively measured. The testing samples of cold-rolled steel specimens synthesized by Bao-steel Inc. were verified. As indicated by the results, the method combined with sound velocity achieved higher prediction accuracy, and the passing rate of relative error less than 10% could reach 95%.