PREDICTION OF ENERGY DISSIPATION OVER TRAPEZOIDAL STEPPED SPILLWAY WITH BAFFLES

Abstract

The stepped spillway of a dam holds significant importance in the field of river engineering and can be utilized in diverse applications. Conducting research on flood control is imperative to examine the mechanism by which energy dissipation occurs across stepped spillways. Numerous research endeavours have been conducted in the past to investigate rectangular stepped spillways without baffles, employing diverse research methodologies. The present research was carried out on a tilting flume in order to compute the energy loss through the trapezoidal stepped spillway at various channel slopes for different baffle arrangements. Additionally, it demonstrates the application of a machine learning method, specifically Support Vector Machine (SVM), for predicting the energy dissipation in terms of nondimensional parameters. The models were constructed using laboratory datasets of an exceptionally high quality, which were gathered from both recent and earlier experiments. The experimental findings suggest that the energy dissipation rises with an increase in the number of baffles from zero to five, and is comparatively greater in channels with zero degree slope as opposed to those with one degree slope. The research has determined that the trapezoidal stepped spillway experiences a twenty percent increase in energy dissipation compared to the rectangular stepped spillway. The statistical indices developed during the experimental research are utilized to validate the suggested models and assess their efficiency and usefulness. The proposed models appear to be influenced by a diverse range of factors, including but not limited to the dimensions of the spillway in terms of width and height, the positioning of the baffle, the upstream and downstream heads, the slope of the channel, the inclination of steps, and the Froude number. A nonlinear relationship has been identified between the aforementioned factors. The results demonstrate that the SVM model proposed, which employs a quadratic kernel function, yields the most accurate predictions for energy dissipation across a trapezoidal stepped spillway. This is evidenced by its superior R2 value and lower RMSE, MSE, and MAE values. The observed phenomenon of overtraining is not exhibited by it. The present research validates the application of machine learning methodologies in this field, and it is noteworthy for its ability to predict energy dissipation across trapezoidal stepped spillways, incorporating diverse baffle designs and configurations.