A method for scour depth identification of single pile foundations based on adaptive neuro-fuzzy inference system

Scour around pile foundations is a major threat to structural safety and durability. While vibration-based methods have shown promise in scour identification, most existing approaches rely solely on frequency changes, neglecting other modal parameters. This study proposes a scour depth identification model based on adaptive neuro-fuzzy inference system (ANFIS), which integrates various modal parameters, including frequency change ratio (FCR), modal assurance criterion (MAC), mode curvature (MC), and mode curvature difference (MCD). Flume experiments and numerical simulations are employed to systematically evaluate the sensitivity and accuracy of these modal indicators under both scour depth and pre-existing local damage. The results indicate that FCR, MC, and MCD are strongly correlated with scour depth, whereas MAC remains limited sensitivity. Furthermore, pre-existing local damage has a negligible effect on predication accuracy. The proposed ANFIS model using three modal indicators achieves a high prediction accuracy (R² = 0.95), with majority predictions falling within the 95 % prediction interval. These findings demonstrate a novel and accuracy approach that enhances the accuracy and reliability of scour identification.