External Validation Criteria and Uncertainty Analysis of Maximum Scour Depth at Downstream of Stilling Basins Based on EPR and MT Approaches

Abstract

In the present research, external validation criteria and uncertainty analysis are conducted to predict maximum scour depth at downstream of the stilling basins using evolutionary polynomial regression (EPR) and model tree (MT) and nonlinear regression (NLR). Data sets used to develop models were collected from experimental models of stilling basins. Effective parameters on the scour depth were included those of discharge, approaching flow depth, mean flow velocity, geometric characteristics of stilling basins and physical properties of bed sediments at downstream of stilling basins. At first, MT and EPR models are implemented based on input–output data sets and result of performances for training and testing stages are evaluated. In addition, external validations and uncertainty analyses on the formulations given by EPR and MT are carried out using statistical errors functions and have absolute mean prediction errors less than that NLR technique. Finally, results of uncertainty and the prediction errors are compared with the GEP, MT, EPR and NLR models to select GEP model as a reliable model with the highest accuracy and smaller errors than other models.