Pipeline Defect Detection Using Artificial Intelligence-Based Active Acoustic Sensing

Abstract This study presents an innovative AI-enhanced active acoustic sensing approach for continuous pipeline defect detection, overcoming limitations of conventional in-line inspection methods that include sparse inspection frequencies and extensive data analysis requirements. The proposed method introduces acoustic energy to detect circumferential notch-type defects. Numerical simulations are performed using Finite Element Analysis (FEA), and experiments performed using axial piezoelectric transducers and acoustic emission sensors. Due to the rapid attenuation at typical inspection frequencies, we excite the system close to natural resonance frequencies, improving defect detection accuracy. Our approach incorporates power spectral analysis and root mean square error (RMSE) computations against baseline references, with Z-score normalization addressing signal variations. The neural network model at the core of this AI methodology significantly boosts defect detection accuracy, proving its efficacy and reliability for enhancing pipeline structural health monitoring (SHM) and integrity management.