Research on microcrack algorithm based on eddy current pulsed thermography

With the increase of fatigue rolling cycles between the wheel and rail, it's easy to generate small angle microcracks on the surface or shallow surface of the rail, which is located in the blind spot of ultrasonic detection. Therefore, pulsed eddy current thermography (ECPT) is selected for detection. In this paper, the finite element software was used to simulate the thermal characteristics of microcracks, and the data was extracted for simulation data analysis. Then principal component analysis(PCA), tensor CANDECOMP/PARAFAC de- composition(CP), and Tucker decomposition were compared horizontally, and the signal-to-noise ratios (SNR) of the three algorithms were obtained by experimental data. Experimental analysis shows that the characteristics of microcracks are very weak, the optimal extraction time is in the beginning of the heating stage, and the tensor algorithm can improve the SNR ratio of the microcrack effectively.