Vibroacoustic Diagnostics of Degradation and Wall Thinning in Pressurized Pipelines: From Physical Wave Propagation Models to a Hybrid DSP-Deep Learning Pipeline

This paper presents the development and field validation of the RIVIXI Diagnose module for continuous structural integrity monitoring, degradation assessment, and localized wall thinning detection in pressurized heat and water distribution networks. The proposed system combines physical digital signal processing (DSP) of vibroacoustic signals acquired from MEMS accelerometers and hydrophones with deep convolutional neural networks (1D-CNN and 2D-CNN). The methodology integrates signal filtering, cross-correlation analysis, phase coherence estimation, ultrasonic thickness gauge (UT) calibration, Mel-spectrogram processing, CNN-based classification, and decision fusion. The system was validated on a municipal district heating pipeline operating at 1.6 MPa. The AI classifier successfully ruled out active leakage while the diagnostic pipeline reconstructed the wall-thickness profile and identified uniform corrosive degradation. Estimated residual wall thickness ranged from 5.70 mm to 6.40 mm, corresponding to 10.0–18.5% wall loss, with strong agreement to independent ultrasonic thickness measurements. Additional validation on the Mendeley Water Distribution Network dataset achieved 85.0% accuracy and 100.0% precision on hydrophone recordings. The study demonstrates that hybrid DSP–Deep Learning architectures can provide non-intrusive diagnostics of wall thinning, degradation mechanisms, leak detection, and Remaining Useful Life (RUL) estimation for industrial pipeline infrastructure.