Fault identification of bridge expansion joints based on vibration and acoustic signals

Bridge expansion joints (BEJ), a critical component of bridge structures, are often subject to structural anomalies. To address this issue, the study proposes an integrated framework: vibration- guided acoustic denoising and abnormal recognition (VGADR). Firstly, the vibrational and acoustic signals are respectively inputted into the wavelet packet transform (WPT) and the learnable wavelet packet transform (LWPT), through which the acoustic signal is adaptively denoised and feature- enhanced, guided by the vibration signal, followed by signal reconstruction. Subsequently, Filter- Bank (Fbank) features are extracted from the reconstructed acoustic signal and input into the Conformer model to achieve BEJ abnormal recognition. Finally, comparisons with various existing methods indicate that under conditions of strong noise interference, the VGADR can accurately identify abnormal states of BEJs, demonstrating notable robustness and applicability.