[marialhm]

Dear all,

I'm a new Foamer, and currently learning how to use it. And now I have a very confusing problem.

My model is simple, just a planar jet flow. Which is showed in figure 1, the gas inlet flow velocity is higher in the middle, and lower in the sides. But I can only get this correct flow field using icoFoam, when I use other solvers, like pimpleFoam, pisoFoam, reactingparcelfoam, etc., the flow field will be wavering, as is shown in Figure 2. The gravity is zero, and the flow field is wavering from up and down, not just one side.

Can anybody tell me why? and how to fix this? I can't just use icoFoam, because it's only for laminar flow, and cant do anything else.

Many thanks.

Maria

[GerhardHolzinger]

This is likely due to an interaction of the jet and the walls. If you want to simulate a free jet, consider making the domain wider. You could also compare the results of pimpleFoam, when using a RAS-type turbulence model and an LES-type turbulence model.

[piu58]

The second result seems to be unplausible, but it is realistic. If a wall comes near a flow, the wall creates some kind of suctions which moves the flow toward the wall. A tiny irregularity starts that process.

[marialhm]

Originally Posted by GerhardHolzinger

This is likely due to an interaction of the jet and the walls. If you want to simulate a free jet, consider making the domain wider. You could also compare the results of pimpleFoam, when using a RAS-type turbulence model and an LES-type turbulence model.

Thanks, I will try to make the domain wider.

[marialhm]

Originally Posted by piu58

The second result seems to be unplausible, but it is realistic. If a wall comes near a flow, the wall creates some kind of suctions which moves the flow toward the wall. A tiny irregularity starts that process.

Thanks for your explanation, it makes sense to me. But the result from Fluent doesn't have this phenomenon, do you know why?

Maria

[arjun]

Originally Posted by marialhm

Thanks for your explanation, it makes sense to me. But the result from Fluent doesn't have this phenomenon, do you know why?

Maria

What pressure interpolation you have for Fluent?

[marialhm]

Originally Posted by arjun

What pressure interpolation you have for Fluent?

Hi, I chose PISO.

[arjun]

Originally Posted by marialhm

Hi, I chose PISO.

PISO is not pressure interpolation. It seems you did not make any changes to it, so that means you have standard pressure interpolation if it was old version of fluent and second order central if it is new fluent.

The pressure interpolation differences could cause what you are seeing.

[marialhm]

Originally Posted by arjun

PISO is not pressure interpolation. It seems you did not make any changes to it, so that means you have standard pressure interpolation if it was old version of fluent and second order central if it is new fluent.

The pressure interpolation differences could cause what you are seeing.

Oh, is that the reason? but I think in Openfoam, I also use the default setup, and the second order.

[alexeym]

Hi all,

@marialhm

I think it would be MUCH easier if you disclose your geometry, mesh, boundary conditions, schemes, linear solvers, convergence criteria used. Right now, people are trying to explain why the second picture is possible, while simpler answer could be: "your simulation does not converge". And for icoFoam it could be just lucky coincidence.

[marialhm]

Originally Posted by alexeym

Hi all,

@marialhm

I think it would be MUCH easier if you disclose your geometry, mesh, boundary conditions, schemes, linear solvers, convergence criteria used. Right now, people are trying to explain why the second picture is possible, while simpler answer could be: "your simulation does not converge". And for icoFoam it could be just lucky coincidence.

Please see figure 3 for the model, and the files of fvScheme and fvSolution are in figure 4 and 5. Please give your suggestions, thanks!

Regards,

Maria

[marialhm]

Originally Posted by alexeym

Hi all,

@marialhm

I think it would be MUCH easier if you disclose your geometry, mesh, boundary conditions, schemes, linear solvers, convergence criteria used. Right now, people are trying to explain why the second picture is possible, while simpler answer could be: "your simulation does not converge". And for icoFoam it could be just lucky coincidence.

following is figure 5.

[alexeym]

You have hidden the most interesting thing: PIMPLE dictionary in fvSolutions (can't you just post gzipped files?). Also boundary conditions are missing or should we assume, you use "standard" ones?.

[marialhm]

Originally Posted by alexeym

You have hidden the most interesting thing: PIMPLE dictionary in fvSolutions (can't you just post gzipped files?). Also boundary conditions are missing or should we assume, you use "standard" ones?.

Pimple dictionary is,

nNonOrthogonalCorrectors 0;
nCorrectors 2;

Boundary conditions for U is in figure 3. For P, it's fixed value in outlet, and others are zeroGradient.

Acturally I tried several cases, different cases may have small differences. but almost all of them have wavering flow field.

Thanks again.

Maria

[alexeym]

So, basically, you do not check convergence, use GAMG solver and limitedLinear scheme. Do you adjust time step? What is your Courant number? Did you try with PCG instead of GAMG? Did you try to check for convergence (using residualControl)? Did you try other discretisation schemes?

[marialhm]

Originally Posted by alexeym

So, basically, you do not check convergence, use GAMG solver and limitedLinear scheme. Do you adjust time step? What is your Courant number? Did you try with PCG instead of GAMG? Did you try to check for convergence (using residualControl)? Did you try other discretisation schemes?

The maximum Co is 0.5~1, I changed limitedlinear scheme into upwind, but didnt change GAMG, maybe I should.

[alexeym]

Maybe you could also post domain sizes? So people can at least try to reproduce your behaviour.

I do not believe single GAMG -> PCG change will solve your problem. Though it can postpone start of the oscillations (like it does in case of von Karman vortex street).

[marialhm]

Originally Posted by alexeym

Maybe you could also post domain sizes? So people can at least try to reproduce your behaviour.

I do not believe single GAMG -> PCG change will solve your problem. Though it can postpone start of the oscillations (like it does in case of von Karman vortex street).

domain size:
x length: 2m
y:0.45m
injection width: 0.01m

Thanks.

[marialhm]

Hi all,

I think it's not the problem of settings, but it's the reality. @Uwe Pilz is right.

But the problem is, how can I set up to avoid this phenomenon? I tried outlet boundaries for the bottom and top, but the results are still not satisfactory.

Thanks in advance.

Maria

[dferrando]

You could try with symmetry boundary condition.