A Novel Mobile Onboard Multi-Sensor Rail Damage Detection Method Based on Dictionary Enhancement Fusion With Acoustic Emission

With the perfection of high-speed railroad networks, maintaining the structural integrity of rails is crucial for safe transportation. However, the contact friction between wheels and rails yields a noisy background that hinders nondestructive testing of the rails. Aiming to accurately detect the emerging defects, based on dictionary enhancement fusion, a novel mobile onboard multi-sensor rail damage detection method with acoustic emission (AE) is proposed for the structural health monitoring (SHM) of rails. In this method, the mirror extension-based adaptive local mean decomposition (ME-ALMD) algorithm is developed to avoid the endpoint effect and reduce the random component of wheel-rail rolling noise (WRRN). Aiming to dramatize the defect characteristics, an enhanced dictionary fusion with relevance constraints (EDF-RCCs) based on Cramer's V coefficient is innovated to fuse the redundant information from multi-channel data and further eliminate the noise. Adaptive thresholding based on sampling entropy precisely determines the rail damage situation. A customized experimental platform with strip-deep and square damage validates the presented approach. The results show that, based on the proposed method, the signal-to-noise ratio (SNR) of the fused signal is at least 1.81 dB higher than other denoising methods. The average damage detection accuracy reaches 93.75% under experimental conditions. This method provides a guidance for AE-based SHM practical applications of rails.