Sunteți pe pagina 1din 11

Cyclone tutorial

In the following tutorial you will learn how to set up a simple cyclone with
the Eulerian-Eulerian model.

The geometry is simplified and has one inlet in the positive y-direction
(no inlet pipe, just a surface) and two outlets, one at the bottom where
both the primary and secondary phases will exit, and one at the top
where only the primary phase will exit. This is the main purpose of
cyclones, to concentrate the secondary phase in the lower inlet.

In the current simulation air and solid particles, volume fraction α = 1%,
will enter the cyclone at 5 m/s. The solid particles will be modelled as
having a uniform diameter of 0.5 mm and a density of 1500 kg/m3.
1. Start the double precision (3ddp) version of Fluent 6.3.26 and read
the mesh cyclone.msh. The mesh consists of about 190.000
hexahedral cells.

To familiarize yourself with the geometry, visualize the mesh by


displaying it with GUI: Display → Grid

2. In the GUI, go to Grid → Check

This will check the mesh for common errors like inverted cells and
is appropriate to use when reading in a mesh for the first time.

3. In the GUI, go to Grid → Scale

The grid was constructed using ICEM Hexa which by default uses
millimeters when meshing so scale the grid from millimeters to
meters, i.e. select that the mesh was created using millimeters and
press "Scale"

Forgetting scaling is very common. High turbulence viscosity


ratios when iterating is one of many signs that scaling has
been forgotten.
4. In the GUI, go to Define → Models → Multiphase

As we will use the Eulerian-Eulerian multiphase methodology,


select the Eulerian model with 2 phases.

5. In the GUI, go to Define → Models → Viscous

The standard k-ε model with the default settings will be sufficient
for the current analysis so select that one. Also make sure that
Standard Wall Functions are selected.

We will use the mixture turbulence model in this simulation. This


turbulence model might not be sufficient to capture the full
characteristics of the current system, but as it is less computational
demanding than the other two models it might be wise to try that
model first and evaluate whether it is needed to increase the
accuracy by using one of the other models.
6. In the GUI, go to Define → Materials

Define a new material thast will represent the solid particles with
ρ = 1500 kg/m3 and give it a suitable name (I named it "Solid").

The viscosity does not matter as the particles are modeled as solid
but even though they are solids they have to be of the material type
Fluid as the Eulerian-Eulerian model numerically treats the
secondory phases as fluids .

As we will use air as the primary phase and air is the default
material used in Fluent, you do not have to define any primary
phase.
7. In the GUI, go to Define → Models → Phases

This is normally the most tedious part when setting up an Euler-


Euler simulation as one has to define all models that handles
interaction between phases.

Select phase-1 which we will use as the primary phase

Press the Set-button and check that air is selected as the material

Here you can change the name of "phase-1" to e.g. "air" or


anything else you like, however, I kept the default names.
Close the dialog and select phase-2 and press the Set-button

Set the phase material to the one you defined earlier in step 6.
Enable the granular model which will model the secondary phase
as solid particles and use the following settings:

Diameter: 0.0005 m
Granular Viscosity: Syamlal-Obrien
Granular Bulk Viscosity: Lun et al
Frictional viscosity: None
Granular temperature: Algebraic
Solids pressure: Lun et al
Radial Distribution: Lun et al
Elasticity Module: Derived
Packing Limit: 0.63

The full details regarding the above settings are too lengthy to go
into details here, but there is a lot of material in the Fluent manual
regarding these settings. Additionally, many of the
models/parameters set here are not obvious but vary depending on
the examined system (in many cases one has to use the “least
bad” models).
8. Press the interaction button which will open the panel where one
define how the phases interact with each other:

Select the Schiller-Naumann drag law, as this model is expected to


work well with the current system. We will neglect Virtual Mass and
Lift forces. Under the Collisions tab, enter a value of 0.7 as the
default value is rather high and fairly unlikely.

9. In the GUI, go to Define → Operating Conditions

Enable gravity in the negative z-direction


10. In the GUI, go to Define → Boundary Conditions

Set “Inlet” to a Velocity-Inlet with the settings below

Mixture
Turbulence intensity: 10%
Turbulent length scale: 0.01 m

Phase-1
x-velocity: 0 m/s
y-velocity: 5 m/s
z-velocity: 0 m/s

Phase-2
x-velocity: 0 m/s
y-velocity: 5 m/s
z-velocity: 0 m/s
Granular temperature: 0.0001 m2/s2
Volume fraction: 0.01

Set “loweroutlet” and “upperoutlet” to pressure-outlets with the


settings below

Mixture
Gauge pressure: 0 Pa
Backflow direction specification method: Neighbouring cell
Backflow turbulence intensity: 5%
Backflow turbulent length scale: 0.01 m

Phase-1
-

Phase-2
Backflow granular temperature: 0.0001 m2/s2
Volume fraction: 0
11. Now the case is properly set up. The last parts concern the
solver settings.

Almost all cyclones are transient, meaning that one has to solve
them using the transient solver, even if one is only interested in the
final solution. This is because all cyclones have a center core
having transient oscillations. These oscillations can often make the
solution unstable in steady-state.

12. In the GUI, go to Solve → Controls → Solution

We will solve the Flow, Volume fraction and Turbulence


simultaneously so make sure they are all selected (they are by
default).

Leave the Under-Relaxation factors as default and set the


Discretization schemes according to the figure.

Running the solver with higher order discretization schemes from


start can cause instability, but as we will be running the transient
solver here, where every time step is a “final solution”, we have to
have higher order schemes from t = 0.
13. In the GUI, go to Solve → Initialization → Initialize

As there are no obvious flow direction in the Cartesian frame of


reference (Fluent cannot initialize in cylindrical coordinates!) select
to compute values from the inlet and set the y-velocities and
volume fraction to zero.

14. In the GUI, go to Solve → Monitors → Residiuals

Select whether you would like plot the residuals or not and set
convergence criteria for the variables (I normally use the default
values when running transient simulations).

If you want to monitor other parameters, such as mass flow rate


from the outlets, pressures on boundaries etc, set monitors in
Solve → Monitor → Surface

15. In the GUI, go to Solve → Iterate

Set the Time Step Size to a time small enough so that the temporal
characteristics of the flow are resolved properly. I chosed a value
of 0.001s which represent the time it would take one fluid element
to move 0.005 circumferences at the widest point with the inlet
velocity. However, this number might be increased or one can
enable the adaptive timestepping.

Press Apply (not Iterate)


16. Save your case and data

Now, the simulation would be started and the iterations would go on until
the end-time is reached. If you want you can run some iterations to see
that the case you set up works properly.

However, as we are limited regarding time, there is a file saved at 1.2s


that you can open and post process, Cyclone_Euler_1.2s.cas

This tutorial will not deal with post-processing in detail, but a brief demo
of what can be visualized will be held in front of the class.

S-ar putea să vă placă și