Crack characterization in ferromagnetic steels by pulsed eddy current technique based on GA-BP neural network model

Abstract

Ferromagnetic steels are widely used in engineering structures such as rail track, oil/gas pipeline and steel hanging bridge. Cracks resulted from manufacturing processes or previous loading will seriously undermine the safety of the engineering structures and even lead to catastrophic industrial accidents. Accurate and quantitative characterization the cracks in ferromagnetic steels are therefore of vital importance. In this paper, the cracks in ferromagnetic steels are detected by the pulsed eddy current (PEC) technique. Firstly, the physical mechanism of the relative magnetic permeability of the ferromagnetic steel on the detection signal of PEC is interpreted from a microscopic level of magnetic domain wall movement. The relationship of the crack width/depth and the detection signal of PEC is then investigated and verified by numerical simulations and experimental study. Finally, the cracks are inversely characterized by using Genetic Algorithm (GA) based Back-Propagation (BP) neural network (NN) considering the nonlinearity of the crack geometric parameters with the detection signal of PEC. The prediction results indicated that the proposed algorithm can characterize the crack depth and width within the relative error of 10%. The proposed approach combining PEC and GA based BPNN has been verified to quantitatively detect cracks in ferromagnetic steel.