Scour depth prediction around bridge abutments: A comprehensive review of artificial intelligence and hybrid models

Scouring around the bridge structure is a major concern of the globe. Therefore, a precise estimation of the scour depth is essential to minimize bridge failure and provide preventive measures. This review paper aims to analyze the critical review of various artificial intelligence (AI) techniques utilized in the literature to estimate bridge abutment scour depth including artificial neural networks (ANN), adaptive neuro-fuzzy inference systems (ANFIS), gene expression programming (GEP), support vector machines (SVM), and extreme learning machines (ELM). The predictive power of each technique was assessed in terms of different performance indicators, such as correlation coefficient (R), mean square error (MSE), predicted values, Taylor's diagram, sensitivity analysis, and violin plot. This review paper highlights that by comparing different AI techniques, ELM and GEP techniques have superior performance, especially in predicting scour depth and dealing with complex and large datasets. However, various limitations and proposed solutions have been reported for techniques, such as ANN, ANFIS, SVM, and group method of data handling (GMDH). The main challenges in the ANN, ANFIS, SVM, and GMDH techniques were overfitting and hyperparameter tuning. Based on the performance of each technique, the current review paper found the satisfactory performance of the ELM technique because of its computation speed and precise estimation capability. Moreover, the proposed solutions would be helpful to researchers working in the field of hydraulics engineering, particularly scouring around the bridge abutment.