Creative Approaches Towards Development of Computing and Multidisciplinary IT Solutions for Society

11Performance Comparison of Prediction of a Hydraulic Jump Depth in a Channel Using Various Machine Learning Models

Nguyen Minh Ngoc^* and Bui Hai Phong

Hanoi Architectural University, Hanoi, Vietnam

Abstract

A sequent depth of a hydraulic jump depends on various factors, such as Froude number, type of cross section and characteristics of a roller zone. In fact, theoretical or experimental equations cannot describe all factors affecting the depth of the jump. First, this study uses the theory of the momentum equation to establish the correlation function between the factors, thereby building the research data field and the target data field. Then, we apply and fine-tune various machine learning models to predict the sequent depth of the jump. A wide range of machine learning models including neural networks, decision trees, random forests, and support vector machines, have been applied to predict the depth of the jump in the trapezoidal channel. Performance evaluation evaluation was carried out on two datasets, and a comparison with existing methods showed that the highest error according to neural networks was 3.3%. Moreover, the other statistical metrics showed high prediction accuracy (R² = 0.99, MSE = 0.011, RMSE = 0.104, MEA = 0.088, and MAPE = 1.49%). The obtained results illustrate the efficiency of the prediction of jump depth using machine learning models.

Keywords: Machine learning, sequent depth, hydraulic jump, trapezoidal channel, statistical indicator

Nomenclature ...