61 / 2025-04-24 12:55:27

Comparative Study on Predicting Local Scour Depth Using Machine Learning Models

Local scour, ANFIS, GEP, ANN

Problem of local scour depth around bridge piers is a critical issue that is to be considered for
ensuring structural safety and mitigating risks associated with scouring. This study focuses on
predicting local scour depth of unsteady flow under clear water condition using advanced machine
learning methods including Adaptive Neuro-Fuzzy Inference System (ANFIS), Gene Expression
Programming (GEP), and Artificial Neural Networks (ANN). A total of 353 input datasets were
obtained from previous literature data and were divided in 70/30 ratio in which 70% (247) of
datasets were used for training and 30% (106) of datasets were used for testing models. The
performance of the developed models was evaluated using statistical indices such as Root Mean
Square Error (RMSE), Coefficient of Determination (R²), and Mean Absolute Percentage Error
(MAPE). It was observed that ANN shows better results than GEP and ANFIS with RMSE of
0.05, R2 of 0.97, and MAPE of 12%. Thus, ANN can be used as an effective model for predicting
scour depth of unsteady flow under clear water condition. This study contributes to advancing
data-driven approaches for addressing challenges in hydraulic engineering.