In this post on Computational Fluid Dynamics (CFD) we begin a series of posts dedicated to studies where flow turbulence plays a fundamental role. In this first example we study the flow through a curved channel, where the SST (Shear Stress Transport) turbulence model may be applied.
As shown in the picture, we will use a nearly two-dimensional domain with periodic conditions in the flow. This allows us to represent the behaviour of the fluid in the complete curved channel. In the tutorial available in our Downloads section, you will find all the steps to follow in Hyperworks CFD and Hypermesh to prepare the model and solve it using Acusolve.
SST turbulence model: Curvature correction
The objective will be to compare the convergence and results of the model solved with and without curvature correction of the SST turbulence model.
Symmetry boundary conditions are applied to the side walls, no-slip wall at the ends of the channel and the appropriate conditions for inlet and outlet. In the following picture you can see the final distribution of conditions. It also shows the mesh, which as described in the tutorial is imported using Hypermesh from an Ansys Fluent mesh.
From the physics menu, the options to be simulated are chosen. In both cases of study it will be a stationary analysis with SST turbulence model without transition effects. First it will be used without curvature correction and then with this option activated.
During the Acusolve calculation, the convergence of the two alternatives is found to be similar. Once the resolution has been completed, a series of measurement points are placed to compare the results with the experimental data. These allow the velocity obtained along the centreline of the channel to be exported to a file. To do this, the Point Probes option is used from the Post ribbon in Hyperworks CFD.
These results are exported and after dimensioning, they are plotted on a graph together with the experimental results, as shown in the following image.
The conclusions of this study are that the model achieves a very good fit in the central zone of the channel, and a reasonable fit in the lateral regions, obtaining a good prediction of the overall behaviour of the fluid. The area with the worst results is the area close to the nearest wall, with errors of around 10% in the most unfavourable area.
The other conclusion observed is that for this case, the curvature correction of the SST model has had a practically negligible effect, so it is not necessary. Its effect would be more significant in flows of greater complexity due to the curvature of the surfaces, for which it is convenient to know and activate this characteristic of the turbulence model.