عنوان مقاله [English]
نویسندگان [English]چکیده [English]
Most of the rivers in nature are meandering and they have constantly erosion at outer bank and sedimentation at inner bank with strong secondary current. The water intake should be sited where there is the maximum strength of the secondary flow which causes sediment movement from the inner bank towards the outer bank and the lowest levels of sediment enter into the intake. So, recognizing the patterns and locations of erosion and sedimentation on bends is really important. Moreover, changes in the meanders and their bed and bank erosion causes river bend to move and destruction of the surrounding structures, agricultural farms and adjacent pumping stations; indicating there is a need to understand the flow patterns, maximum secondary flow position and shear stress in river bend using mathematical models. In this research, mathematical models of turbulence were studied using FLUENT software to introduce the models in good agreement with experimental data to use in different bends.
The experiments were performed in a 1.3 m wide laboratory Plexiglas’s flume consisting of a 193° bend with a constant centerline radius of curvature of R = 1.7 m, preceded and followed by straight reaches 9 m and 5 m long, respectively. The flume was located in the hydraulic laboratory of EPFL, Lausanne, Switzerland. The bed was covered by a quasi-uniform sand with a diameter d = 0.002 m. The curvature ratio R/B = 1.3 which is representative for sharp natural meander bends. The bed was frozen for future examinations after sediment injection for three weeks and forming a developed topography. Fluent software was used for three-dimensional simulation of flow pattern. K-?, K-? and RSM Models and LES technique were used to simulate flow patterns. To evaluate the turbulence models, depth averaged parameters were used. Also for Quantitative study of models, some predicted longitudinal velocity profiles were selected and compared with measured velocity profiles. Before performing a calculation using Fluent, a computational domain must be generated to define the geometry of problem. The Gambit software was used for mesh generation. This study comprised of 910,000 computing nodes in the Gambit. This computational grid was obtained finally after size reduction as much as possible and testing several networks in Gambit and then FLUENT software.
The results of depth averaged velocities showed that K-ε, K-ω and RSM did not have the ability to determine important points of flow and the models showed maximum flow separation zone at 90-degree angle whereas measurements determined the region identified at the angle of 75 degrees. So, there was 15-degree different between RANS model and measurements in flow separations' zone. But the LES model had good agreement with measurements to locate the separation zone. Also, LES model determined longitudinal depth averaged velocity location better than RANS models. Furthermore, about other sections of flow pattern, there were better predictions in LES than others. Between 30 to 60 degree cross sections, LES model had good agreement with experimental measurements and showed maximum velocity moves toward outer bank. Between 60 to 90 degree cross sections, maximum of separation zone predicted by LES places in 75 degree that is near measurements whereas the maximum of separation zone in K-?, K-? and RSM models based on Boussinesq equation were in 90 degree cross sections which did not have agreement with the measurements. From 90 to 120 degree cross sections, predictions with LES method was the same as measured flow pattern on channel and maximum velocity places between central line of channel and outer bank. Also, Between 120 to 150 degrees, predicted and measured flow patterns had good agreements with each other. In 120 to 150 degree cross sections, maximum velocity was near central line of channel and moved toward outer bank. From 150 degree cross section toward end of bend, maximum velocity moved and placed in the side of outer bank and had good agreement with measurements. Investigation of LES model showd there was good agreements between predictions and measurements flow patterns whereas K-?, K-? and RSM models, named RANS models, did not have good agreement with measurements flow pattern and had more different. Also, Quantitative comparison of depth averaged velocity profiles between experimental results and predictions of models showed that the LES model with 7.9 percent error had the best results than other models. Then K-? model with 9.6 percent error, RSM Model with 10.4 percent error and K-? with 10.9 percent error are placed. So, LES model acts better than RANS models on sharp bend and it is usable to predict flow pattern on sharp bend.