AT07.1b - Diffuser
AT07.1b - Diffuser
1b:
Diffuser
Master’s
Master’s DegreeDegree
ininNumerical
Numerical Simulation in Engineering
Simulation with Ansys with Ansys
in Engineering
© 2023 Ansys, Inc.
1Universidad Politécnica de Madrid - Ansys 12th EDITION
Introduction
• Workshop Description:
– Turbulent Flow Through a Planar Asymmetric Diffuser is simulated.
• Learning Aims:
– The purpose of this tutorial is to provide guidelines and recommendations for solving a CFD problem
which includes:
– using the pseudo-transient solver and high order term relaxation.
– comparing the results with experimental data.
– Performing a precursor calculation in a small periodic domain to obtain fully developed inlet flow and
turbulence profiles.
• Learning Objectives:
– To understand how to the mesh refinement and the turbulence modelling could influence the CFD results
in term of accuracy of the solver.
At that stage of the Module, this tutorial assumes that you are familiar
with the Fluent interface and you have a good understanding of the basic
setup and solution procedures. Some steps will not be shown explicitly.
• The problem is to simulate the flow through an asymmetric plane diffuser with a Reynolds number Re =
20000. The Reynolds number is based on the centerline velocity and the channel height at the inlet.
• The complete experimental results were obtained by Buice and Eaton [1]. This is a classical test case for
flows dominated by adverse pressure gradient and boundary-layer separation.
• The profiles computed by the periodic model will be used for the inlet boundary condition in the full
geometry. Take the latter approach in this tutorial to minimize the size of the domain and CPU time.
• Materials
• The fluid is standard air with constant density,
so there is no need to visit the materials panel.
• Boundary Conditions
• Set the periodic conditions by clicking the
button reading “Periodic Conditions…”
below the list of
boundary zones
• Enter 0.314 kg/s
• Solution
• Solution > Solution Methods
– The panel will be populated
as shown here
• Monitors
• In the Residual Monitors panel, use the
drop-down list to change Convergence
Criterion to None.
• Set up a surface monitor for shear stress at
walls.
• Initialization
• Select “Standard Initialization”
• Press Initialize
• Run calculations
• Set the reporting interval to 100
• This is a very small case and therefore reporting
every iteration will significantly increase the
calculation time
• Run 1500 iterations
Master’s Degree in Numerical Simulation in Engineering with Ansys
© 2023 Ansys, Inc.
8
Write profiles
Generating Inlet Profiles with a Precursor Calculation
• Profiles
• Save the case and data file
• Select File > Write > Profile. Choose “periodic.1”
from the list of boundaries and x-velocity, turbulent
kinetic energy and turbulent dissipation rate.
• Click “Write…” and save the file as “channel-rke-
profile.prof”.
• Models
• Activate the realizable k-epsilon turbulence model with Enhanced Wall Treatment, as shown before.
• Materials
The fluid is standard air with constant density, so there is no need to visit the materials panel.
• Boundary Conditions
• Click on Boundary Conditions in the navigation pane on the left of the GUI.
• Below the boundary conditions panel is a button labeled “Profiles….”. Click this button, then click
“Read…” on the Profiles panel and select the file “channel-rke-profile.prof” created before. The panel
will be populated as shown in the figure.
• Apply boundary conditions for inlet_v
• Boundary Conditions
• Apply boundary conditions for outlet
• Solutions Methods
• Same as before
• Monitors
• In the Residual Monitors panel, change the value of
“Absolute Criteria” from .001 to 10-6 for all equations.
• Set up a surface monitor for shear stress at walls
• Initialization
• Select “Hybrid Initialization”
• Click “More Settings”
• Click the “Turbulence Settings” tab
• Unselect “Averaged Turbulent Parameters” and enter the
values shown in the panel (4 and 40)
– They are approximately the same as the average values
computed before for the channel upstream of the
diffuser section.
• Click “Initialize”
• Request another 400 iterations. The solution will reach the residual criteria. Write Case&Data (as asd-
rke-diffuser-final).
• Plot the skin friction on the top and bottom wall for the SST results
Remember the realizable k-epsilon Cf predictions along the top wall are substantially different (lower) than the
experimental. The main reason for the failure is due the fact that it does not correctly predict the size of the
separation/recirculation zone along the inclined wall.
On the other hand, SST k-ω is the only turbulence model among all the two-equation turbulence models which
can successfully capture the recirculation zone. SST k-ω model’s prediction of Cf on the top wall is good, but
along the bottom wall it predicts the flow separates slightly upstream of the actual separation point.
• To test whether the converged results (from the SST k-ω model) obtained so far are independent of the grid
resolution, you can either uniformly double the total cell count, or use the grid adaption feature of the
solver to achieve the objective more efficiently.
• Grid independence is attained when further mesh refinement yields only small and insignificant changes in
the solution fields. Different mesh adaption criteria could be considered (region, gradient,…), but here we
choose to adapt a region extending slightly upstream and slightly downstream of the separated flow region.
• First, save the case and data before attempting any adaption since any change is irreversible.
• Now, we will identify the separated flow region by creating an iso-surface where the x-velocity is equal to
zero.
• Run calculation
• Lower the timescale factor (0.1) and continue the
iterating until convergence.
• Save the case and data files
(asd-sst-diffuser-adapt-final).
• Plot the results
• Between the two-equation turbulence models, only the SST k-ω model gives reasonable predictions of the
skin friction and the recirculation zone.
• You have also learned how to use Fluent´s regional grid adaption feature to test whether or not the
calculation is grid independent, without having to uniformly double up the cell count in the whole flow
domain.