Research on MSC-ANC noise reduction analysis for wire break signal monitoring in prestressed concrete cylinder pipe

Abstract Distributed acoustic sensing systems face significant challenges in monitoring wire breakage within prestressed concrete cylinder pipes (PCCPs) under complex environmental conditions, where raw signals are frequently contaminated by noise interference. To address this issue, an improved adaptive noise cancellation (ANC) system is proposed. Conventional ANC systems exhibit limited denoising efficacy due to discrepancies in noise propagation paths between reference and primary channels. To overcome this limitation, a convolutional autoencoder (CAE) is introduced to reconstruct reference signals. Furthermore, a multi-scale feature-enhanced squeeze-and-excitation network (MSFE-SENet) is incorporated into the CAE architecture to augment its reconstruction capability. The integration of MSFE-SENet-CAE (MSC) with ANC yields the MSC-ANC denoising system. Experimental results demonstrate that MSC-ANC effectively eliminates extraneous interference, outperforming conventional ANC in both time-domain and frequency-domain denoising performance. Quantitative evaluations using signal-to-noise ratio and root mean square error metrics confirm the superior denoising capability of MSC-ANC over existing methods. When applied to denoised signals processed by MSC-ANC, a k-nearest neighbors classifier achieves 100% accuracy in wire breakage signal identification, representing 18.7% and 1.4% improvements over direct signal recognition and ANC-filtered recognition approaches, respectively. These findings conclusively validate the effectiveness of MSC-ANC in processing PCCP wire breakage signals under noisy operational conditions.