An evolutionary optimized artificial intelligence model for modeling scouring depth of submerged weir

Abstract

The advancement in computer aid and artificial intelligence (AI) models have received a noticeable progression in several engineering applications. In this research, an investigation for the capacity of a hybrid artificial intelligence model for predicting depth scouring of submerged weir. Scouring phenomena is one of the most complex problems in the field of the river and hydraulic engineering. Accurate and precise prediction for the depth scouring (ds) is one of the essential processes for maintaining a sustainable hydraulic structure. This article introduces a new predictive model called tBPSO-SVR, which is a hybridization of an enhanced binary particle swarm optimization (PSO) algorithm with support vector regression (SVR) model as an efficient predictive model. The roles of the PSO algorithm are tuning the internal hyperparameters of the SVR model in addition to the optimization of the predictors selection “feature selection” for the ds modeling. The prediction matrix is constructed based on several related geometric dimensions, flow information and sediment properties. The proposed model is validated against several well-established machine learning models introduced over the literature. The prediction potential of the proposed tBPSO-SVR model exhibited a superior capability. In quantitative terms, tBPSO-SVR attained minimum mean absolute error (MAE = 0.012 m) and maximum coefficient of determination (R2 = 0.956). Remarkably, the proposed hybrid artificial intelligence demonstrated an efficient prediction model for depth scouring prediction with reducing the input parameters.