Explicit data-driven models for prediction of pressure fluctuations occur during turbulent flows on sloping channels

摘要

Pressure fluctuations are among the favorite topics for hydraulic engineers due to their critical role in the design and safe operation of hydraulic structures. In this study, three explicit data-driven models, including Classification and Regression Trees (CART), Gene Expression Programming (GEP), and Multivariate Adaptive Regression Splines (MARS), are used. These methods are applied to predict the pressure fluctuation intensity oC 0 PTHORN, which occurs beneath a hydraulic jump on sloping channels using reliable experimental observations. The main features of the suggested data-driven models are their capability to provide practical formulas for prediction of C0 P based on the dimensionless form of the effective parameters. The application of these models leads to the extraction of knowledge from data in the form of predictive formulas. In addition, to evaluate the accuracy and performance of the proposed models, some statistical indices and several validation criteria are employed. The obtained results confirm that MARS predicts C 0 P better than CART and GEP. Based on the effective independent variables, MARS presents a simple predictive formula for the computation of C 0 P. GEP-based equations use mathematical trigonometric functions that cause a more complicated and time-consuming process to predict C0 P. However, the application of simple rules generated by CART enables hydraulic engineers to predict C0 P, easily and quickly.