[rv82]

Hi folks!

I have a problem and I need help.

I try to simulate a multisection cyclone unit. And as a result I have plots of pressure and velocity. Velocity values are close to experimental values. But pressure values are too big.

I tryed to simulate this case with this settings:
solver: simpleFoam (is it good choice for twisted flow?).
model: simulation begins with kOmegaSST and then continues with LRR or LaunderGibsonRSTM.
numerical schemes: div(phi,U) are limitedLinear 0.5 or linearUpwind Gauss linear.
under-relaxation factors: 0.3 for pressure and 0.7 for others.

Is it possible to affect to pressure values during simulating? Or may I offer a correction factor to get correct (close to experiment) values for current mesh scale and then multiply new values on it? Or it's nonsense?

A measuring point was marked on figure by red circle.

Thank you!

[taxalian]

Originally Posted by rv82

Hi folks!

I have a problem and I need help.

I try to simulate a multisection cyclone unit. And as a result I have plots of pressure and velocity. Velocity values are close to experimental values. But pressure values are too big.

I tryed to simulate this case with this settings:
solver: simpleFoam (is it good choice for twisted flow?).
model: simulation begins with kOmegaSST and then continues with LRR or LaunderGibsonRSTM.
numerical schemes: div(phi,U) are limitedLinear 0.5 or linearUpwind Gauss linear.
under-relaxation factors: 0.3 for pressure and 0.7 for others.

Is it possible to affect to pressure values during simulating? Or may I offer a correction factor to get correct (close to experiment) values for current mesh scale and then multiply new values on it? Or it's nonsense?

A measuring point was marked on figure by red circle.

Thank you!

Hi i would go for the complete simulation with KOmegaSST and another important thing to know the set of boundary conditions you used for that flow.

Of course you can modify the numerical settings on the fly, if the 'runTimeModifiable' option is turned on in 'controlDict'.

Numerical schemes and under-relaxation factors seems ok.

[rv82]

Hello taxalian! Thank you for your reply!

To my mind, RSTM models are more applicable to cyclone units with complex flows then kOmegaSST, that's based on turbulent viscosity.

Boundary conditions in my case are:
nu 1.5e-5 (air).
for U:
on inlets:
type surfaceNormalFixedVlaue;
value uniform ( 0 0 0 );
refValue uniform -12.6 // for 1st section
refValue uniform -21.2 // for 3rd section
on outlet: type zeroGradient;
on walls: type fixedValue; value uniform (0 0 0);

for p:
all inlets and walls: type zeroGradient;
on outlet: type fixedValue, value uniform 0;

for k:
on inlets:
type fixedValue;
value uniform 1.07; // 1st section
value uniform 3.9; // 3rd section
on outlet:
type zeroGradient;
on walls:
type kqRWallFunction;
value uniform 3.9; // I don't know which value will be correct for this parameter

omega:
on inlets:
type fixedValue:
value uniform 1498; // 1st section
value uniform 13588 // 3rd section
on outlet:
type zeroGradient;
on walls:
type omegaWallFunction;
value uniform 1498; // similarly I don't know which value will be correct

For k and omega I used this material (second formula for omega).

Mesh was built with snappyHexMesh.

After this calculation ends, I calculate values for epsilon and R with createTurbulentFields. Then I change RASModel from kOmegaSST to e.g. LRR, change startTime in controlDict and run the calulation again.

On residual plot initial value so small after computation with kOmegaSST. Then it increases to ~0.017 and then decreases again.

Tell me please how can I use 'runTimeModifiable' to affect to pressure?

Thank you!

[taxalian]

Originally Posted by rv82

Hello taxalian! Thank you for your reply!

To my mind, RSTM models are more applicable to cyclone units with complex flows then kOmegaSST, that's based on turbulent viscosity.

Boundary conditions in my case are:
nu 1.5e-5 (air).
for U:
on inlets:
type surfaceNormalFixedVlaue;
value uniform ( 0 0 0 );
refValue uniform -12.6 // for 1st section
refValue uniform -21.2 // for 3rd section
on outlet: type zeroGradient;
on walls: type fixedValue; value uniform (0 0 0);

for p:
all inlets and walls: type zeroGradient;
on outlet: type fixedValue, value uniform 0;

for k:
on inlets:
type fixedValue;
value uniform 1.07; // 1st section
value uniform 3.9; // 3rd section
on outlet:
type zeroGradient;
on walls:
type kqRWallFunction;
value uniform 3.9; // I don't know which value will be correct for this parameter

omega:
on inlets:
type fixedValue:
value uniform 1498; // 1st section
value uniform 13588 // 3rd section
on outlet:
type zeroGradient;
on walls:
type omegaWallFunction;
value uniform 1498; // similarly I don't know which value will be correct

For k and omega I used this material (second formula for omega).

Mesh was built with snappyHexMesh.

After this calculation ends, I calculate values for epsilon and R with createTurbulentFields. Then I change RASModel from kOmegaSST to e.g. LRR, change startTime in controlDict and run the calulation again.

On residual plot initial value so small after computation with kOmegaSST. Then it increases to ~0.017 and then decreases again.

Tell me please how can I use 'runTimeModifiable' to affect to pressure?

Thank you!

That's true but RSTM model is much more complex and sensitive compared to SST model and as you model less & resolve more, the solution tends to get unstable but more accurate, so you can also try to solve this time dependent.

For the choice of omega i will go for the first formula i.e. the square root of turbulent kinetic energy divided by the turbulent length scale.

well runTimeModifiable option is available in your controlDict file and if you make any changes in your solver settings this will apply the changes to the simulation immediately and you can check the changes in your solution log file

So i would also check the following:

- checkMesh
if check gives some failure or non-orthogonality warnings
then i would go for several non-orthogonal correctors for better
convergence of the pressure equation
- hoping that your mesh is good enough to resolve the near wall effects in terms of having sufficient near wall resolution i.e. appropriate yPlus value

[rv82]

Hello!

That's result of yPlusRAS for my mesh:

Patch 8 named walls y+ : min: 0.0326226 max: 80.6438 average: 10.4437

and that's results of checkMesh:

Checking geometry...
Overall domain bounding box (-0.0725 -0.135 -0.0725) (0.0725 0.252 0.0725)
Mesh (non-empty, non-wedge) directions (1 1 1)
Mesh (non-empty) directions (1 1 1)
Boundary openness (1.01866e-16 -4.41099e-16 -2.10805e-16) OK.
Max cell openness = 4.20861e-16 OK.
Max aspect ratio = 8.92914 OK.
Minimum face area = 1.62543e-07. Maximum face area = 7.37174e-05. Face area magnitudes OK.
Min volume = 1.70347e-10. Max volume = 5.50784e-07. Total volume = 0.00462585. Cell volumes OK.
Mesh non-orthogonality Max: 64.4497 average: 4.44209
Non-orthogonality check OK.
Face pyramids OK.
Max skewness = 2.13567 OK.
Coupled point location match (average 0) OK.

Mesh OK.

Today I tryed to set pressureInletOutletVelocity for outlet and set non-orthogonal correctors to 2. Also I set cellLimited Gauss linear 1 for grad(U) and grad(p) and set div(phi,U) Gauss linearUpwindV cellLimited Gauss linear 1. But pressure is too big anyway