Estimation of Shear Force Distribution in Two-Stage Open Channel Using SVM and ANFIS

Abstract

The prediction of discharge is a challenging task for compound channel due to exchange of momentum at the junction of main channel and floodplain. From the literature, it has been found that, using apparent shear force (ASF) concept at the interface of floodplain and main channel, the accurate discharge can be predicted. ASF is a function of various non-dimensional parameters such as channel width ratio (α), relative flow depth (β), main channel aspect ratio (δ), relative roughness (γ), bed slope (S₀), Froude number (Fr), and side slope (m). In this paper, an attempt has been made to model ASF by considering the aforementioned non-dimensional parameters. A total of 152 datasets have been collected from various literatures related to ASF. In the recent days, artificial intelligence (AI) and machine learning (ML) techniques are widely used to model hydrological and hydraulic flow problems. In this research work, two soft computing techniques such as support vector machine (SVM) and adaptive neuro fuzzy inference system (ANFIS) have been utilized to model apparent shear force at the two-stage channel. First, SVM has been successfully used in classification and then extended for regression analysis. The SVM classification methods are based on the principle of optimal separation of classes. The computation is critically dependent upon the number of training pattern and selection of hyper-parameters for regression. There are four types of SVM named as cubic SVM, fine Gaussian SVM, medium Gaussian SVM, and coarse Gaussian SVM. In ANFIS modeling, different membership functions have been utilized. All types SVMs modelling and ANFIS modelling are performed using MATLAB tool. Sensitivity analysis has also been performed to check the strength of developed model. The present model is compared with other existing models and found to provide satisfactory results with coefficient of determination (R²) value greater than 0.93.