Is to simulate a laminar flow over a backward-facing step and
Give some insight into the influence of the grid density and
Order of the spatial discretization.
Laminar flow over backward-facing step (2D) simulation
1. Laminar flow over a backward-
facing step (2D) simulation
By :Hayder Jawad Kadhim
hayder.jawad@uokerbala.edu.iq
Thualfaqir J. Kadhim
Thualfaqir.j@uokerbala.edu.iq
Abdalrazzaq K. Abbas
The University of Karbala
College of engineering
2. Purpose
The aim of the exercise:-
• Is to simulate a laminar flow over a backward-facing step.
• Give some insight into an influence of the grid density and
Order of the spatial discretization.
•Analyze the differences and possible numerical errors.
• Present results.
The University of Karbala
3. Introduction to ANSYS
Workbench ver. 16.2
ANSYS Workbench assists the user in carrying out the tasks
involved in an analysis process; typically the software is
divided in three main parts:
- A pre-processor.
- A Solver.
- A post-processor.
The University of Karbala
5. Project start-up
Before starting a solution of any flow problem using the ANSYS Fluent we will
need to create geometry and make the computational mesh. Next the Fluent
solver has to be properly set-up.
Drag and drop a Fluid Flow (Fluent) component from Analysis Systems into
the Project
The University of Karbala
6. Project start-up
Alternatively, each component can be shifted separately from Component
Systems window into the Project Schematic window. We could start by adding
the Geometry box, later Mesh and finally the Fluent box, as shown below.
The University of Karbala
7. Geometry
Double click on the Geometry cell to start Design Modeller
The geometry of pipe that we create
The University of Karbala
8. Geometry
Select the Draw/Rectangular toolbox. Sketch the arbitrary rectangular
in the Graphics window (it will be used for creating right side of our
geometry – outflow section).
The University of Karbala
9. Mesh
Drag and drop the Mesh cell into the Geometry box in the Project Schematic.
Double click on the Mesh to open the meshing program. Select the Mesh under
the Outline.
The University of Karbala
10. Uniform Mesh
Now we are going to make a finer mesh. This requires a specification of a proper
size of the cell elements on selected edges. Activate the edge selection filter
option in the upper menu panel. Select all vertical edges (use Ctrl to select
multiple edges) right click on the Mesh.
The University of Karbala
11. Non- Uniform Mesh
We can introduce the boundary layer mesh near to walls. The aim is to
make the grid denser in the high velocity gradient flow regions. Right click
on Mesh cell and introduce Insert/Inflation. Activate the Face symbol in
the geometry selection filter in upper menu panel and select all surfaces
under Geometry selection cell. Select all edges corresponding to walls
under Boundary. Use the Ctrl button to select multiple edges. Set the
number of layers to 10 (Maximum Layers) and generate the mesh.
The University of Karbala
12. Solution
Drag and drop the Fluent cell into the Geometry box in the Project Schematic.
Double click on the Fluent to open the program.
The University of Karbala
13. Result and analysis
The simulations have to be performed using the k-ɛ model with the
wall function approach. The Reynolds number is based on the second pipe
diameter D and the averaged velocity at the inlet and treated to be turbulent.
The flow is 2D axisymmetric so only half of the domain is shown.
Solution converged for Coarse mesh and 1st order upwind
The University of Karbala
15. Velocity fluid profile at
x=0cm,x=6cm
Velocity profile at the step length (x=0) and 6 cm. after the step (x=6cm.) is plotted
and compared with experimental results
The University of Karbala
16. Result
Thus the required simulation is successfully done and solution is concluded.
After simulations on the basic mesh with the 1st and 2nd order scheme we
can make a conclusion
1. that numerical results received after modeling in Fluent are quite close
to experimental data.
2. Fluent gives reliable results in such case.
3. After mesh refining (2 times) we can notice that numerical results are
not so close to the experimental data in comparison with not refined
mesh.
The University of Karbala
