Adaptive Signal Reconstruction Based on VMD for Rail Welding Joint Defect Detection

Abstract

Rail welding joint is a vulnerable point in railway, and rail defects often occur at welding joints. To detect these defects, ultrasonic detection is widely used by railway maintenance practice. However, the presence of coarse grains and inclusions in the welding joint leads to numerous backscattering noise in the ultrasonic signal, interfering with defect detection. To address this issue, this paper proposes an adaptive ultrasonic signal reconstruction method VSKR (VMD-SVD-Kurtosis Reconstruction) and introduces a new metric named rail peak signal noise ratio (RPSNR) to measure the effectiveness of this method. This method capitalizes on the distinct frequency characteristics between the noise signals and defect signals, and utilizes variational mode decomposition (VMD) algorithm. VSKR has been successfully applied to signals obtained from both finite element models and real experiments, defect echoes in those signals are highlighted, demonstrating the effectiveness of VSKR. In a specific condition, the RPSNR value has been increased by 8.94 dB. The average increased value of RPSNR is 4.90 dB. These indicates that VSKR can enhance the efficiency of ultrasonic detection of rail welding joint defect by broadening the range of probe positions and directions capable of detecting defects.