With the aid of direct numerical simulation, this paper presents a detailed investigation on the flow around a finite square cylinder at a fixed aspect ratio (AR) of 4 and six Reynolds numbers (Re = 50, 100, 150, 250, 500, and 1000). It is found that the mean streamwise vortex structure is also affected by Re, apart from the AR value. Three types of mean streamwise vortices have been identified and analyzed in detail, namely, “Quadrupole Type” at Re = 50 and Re = 100, “Six-Vortices Type” at Re = 150 and Re = 250, and “Dipole Type” at Re = 500 and Re = 1000. It is the first time that the “Six-Vortices Type” mean streamwise vortices are reported, which is considered as a transitional structure between the other two types. Besides, three kinds of spanwise vortex-shedding models have been observed in this study, namely, “Hairpin VortexModel” at Re = 150, “C and Reverse-C and Hairpin VortexModel (Symmetric Shedding)” at Re = 250, and “C and Reverse-C and Hairpin VortexModel (Symmetric/Antisymmetric Shedding)” at Re = 500 and Re = 1000. The newly proposed “C and Reverse-C and Hairpin VortexModel” shares some similarities with “Wang’s Model” [H. F. Wang and Y. Zhou, “The finite-length square cylinder near wake,” J. Fluid Mech. 638, 453–490 (2009)] but differs in aspects such as the absence of the connection line near the free-end and the “C-Shape” vortex structure in the early stage of the formation of the spanwise vortex.
27 June 2023 in Journal of Water Process Engineering
Gates and weirs are frequently used hydraulic structures employed for controlling water flow rates in irrigation and drainage networks. Therefore, accurately estimating the discharge coefficient (Cd) is important for precise flow measurement. The present study used intelligent predictive models for modeling Cd in labyrinth sluice gates. For this purpose, key dimensionless parameters and reliable experimental datasets were used. The support vector regression (SVR) model was hybridized with particle swarm optimization (PSO) and genetic algorithms (GA). The statistical metrics and graphical plots evaluated the performance of the generated models. Three commonly used statistical indicators, namely root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2), were used for quantitatively evaluating the performance of the proposed models. The SVR-PSO model achieved the lowest values of RMSE (0.0287) and MAE (0.0209) and the highest value of R2 (0.9732), indicating that it was more accurate than SVR-GA (RMSE = 0.0324, MAE = 0.0257, R2 = 0.9685) and SVR (RMSE = 0.0575, MAE = 0.0468, R2 = 0.8958) on the testing data. The findings revealed that the hybrid SVR methods were more accurate than the standalone SVR model. In addition, regarding the value of the objective function criterion (OBF), the SVR-PSO (OBF = 0.0245) and SVR-GA (OBF = 0.0273) had lower OBF values and provided more precise estimates of the Cd compared to existing nonlinear regression-based formulas and existing data-driven approaches. Finally, sensitivity and SHapley Additive exPlanations (SHAP) analyses determined the relative importance of each input variable for the prediction of Cd.