Track acoustic signal compression method based on sparse decomposition and M 2to1 -guided resampling

Abstract

When trains pass over structural discontinuities such as rail gaps or joints, they generate low-frequency acoustic signals characterized by transient impact features. These signals propagate rapidly over long distances and are suitable for track condition monitoring. However, their high sampling rates and volume pose challenges for wireless transmission and real-time processing in structural health monitoring (SHM) systems. To address these issues, this paper proposes a novel compression method tailored for rail impact-induced acoustic signals. The method integrates sparse decomposition with M_2to1.-guided resampling to reduce data dimensionality while preserving fault-relevant transients. The M_2to1. index, a computationally efficient impulsiveness metric, is used to locate pulse-dominant segments, which are then used for k–Singular Value Decomposition (K-SVD) dictionary learning. This approach significantly reduces training time while maintaining high reconstruction fidelity. Experimental validation using field-collected signals from a diesel locomotive passing over a rail gap demonstrates that the proposed method outperforms conventional compression techniques in terms of compression ratio, robustness to noise, and preservation of impact features. These advantages make the method well-suited for deployment in wireless sensor networks for continuous rail SHM.