Application of unsupervised AI-assisted acoustic wave sound analysis for non-destructive detection of steel corrosion induced deterioration

Reinforced concrete (RC) structures have become increasingly prevalent in infrastructure construction. Ensuring their safety and managing maintenance within limited budgets necessitates effective health monitoring techniques, with non-destructive testing (NDT) methods playing a critical role. Among these, acoustic wave sound emission diagnostics traditionally rely on specialized expertise and sophisticated. This reliance may introduce potential inaccuracies due to interpretive errors and the limited availability of experts. This study focuses on the development of an AI-driven evaluation method for tapping sound analysis using an unsupervised deep learning approach. A laboratory case study was conducted using five RC beams of 2 m length and 0.2 m square cross section, subjected to electrochemical acceleration to simulate random corrosion in steel reinforcements. A total of 6075 tapping sound samples were collected and analyzed using frequency-domain transformations and clustering algorithms. Results showed that the proposed method generated an anomaly index that correlated with both the mass loss of steel reinforcement and surface crack severity. This case study highlights the potential of combining acoustic-based NDT with AI-driven analysis to support early-stage damage detection and improve condition assessment of RC structures.