A multi-dictionary fusion method based on improved dynamic time warping for moveable rail damage localization

Traditional structural health monitoring (SHM) of rails relies on a fixed single sensor, limited by detection range and noise interference. Therefore, a multi-dictionary fusion method for movable rail damage localization is proposed based on improved dynamic time warping (DTW). The approach combines onboard acoustic emission sensors with peak detection frames to measure the moving distance of the inspection wheels and monitor a wide range of rails. Aiming to enhance the damage information, an innovative DTW-based multi-dictionary sparse representation algorithm is presented for data fusion. The second-order difference of the Mahalanobis distance is employed to optimize the fusion weights from the global property. A two-feature adaptive threshold is designed to precisely detect and localize damage on rails. The effectiveness of this method is verified at laboratory testing speeds less than 0.75 m s⁻¹. The results demonstrate that it can accurately detect 2 mm deep strip and square damage, providing new inspiration for rail SHM.