Prediction of local scour around bridge piers: artificial-intelligence-based modeling versus conventional regression methods

This paper presents the use of two artificial intelligence modeling methods, namely genetic programming (GP) and adaptive neuro-fuzzy inference system (ANFIS), to predict pier scour depth based on clear water conditions of 320 data sets of laboratory and field data measurements. The scour depth was modeled as a function of five main dimensionless parameters: pier width, approaching flow depth, Froude number, standard deviation of grain size distribution, and channel open ratio. A functional relationship was established using the trained GP model, and its performance was verified by comparing the results with those obtained by the ANFIS model and seven conventional regression-based formulas. Numerical tests indicated that the GP model yielded much superior agreement than the ANFIS model or any other empirical equation. The advantage of the GP model was confirmed by applying the derived GP equation to predict the scour depth around the piers of Imbaba Bridge, Egypt.