Numerical Study on Sharp Defect Evaluation Using Higher Order Modes Cluster (HOMC) Guided Waves and Machine Learning Models

The inspection of corrosion and pitting-type defects is critical in the petrochemical, marine, and offshore industries. Guided wave inspection is widely used to detect these flaws and control operational costs. Higher order modes cluster (HOMC) guided waves, composed of higher-order Lamb wave modes, offer enhanced resolution compared to low-frequency guided waves. They exhibit minimal dispersion, reduced sensitivity to surface features such as T-joints, and retain most of their energy upon interacting with surface defects. This study employs two-dimensional finite element simulations to investigate the propagation and interaction of HOMC guided waves with defects in a T-joint and an aluminum plate. Both conventional fitting methods and machine learning (ML) models are used to estimate the depth of sharp defects reaching up to half the plate thickness. The results demonstrate that both approaches can utilize data from defects of one width to predict the depth of defects with a different width. The ML model outperforms the fitting method, achieving higher prediction accuracy while reducing dependence on expert knowledge. The developed method shows strong potential for characterizing sharp defects of varying widths, closely resembling real-world pitting corrosion scenarios.