View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0003046 | OpenFOAM | Bug | public | 2018-08-16 14:41 | 2018-08-21 18:58 |
| Reporter | janGaertner | Assigned To | henry | ||
| Priority | normal | Severity | minor | Reproducibility | sometimes |
| Status | closed | Resolution | suspended | ||
| Platform | GNU/Linux | OS | Ubuntu | OS Version | 14.04 LTS |
| Summary | 0003046: Initialization error of dgdt in createFields.H in compressibleInterFoam | ||||
| Description | In createFields.H the dgdt field in compressibleInterFoam is initialized with dgdt = alpha1*fvc::div(phi) However, for the first iteratrion the source term should be zero in the alpha equation: (Sp*alpha1 + Su + alpha1*fvc::div(phi)) With, dgdt = Sp*alpha1 + Su This is the case when dgdt is initialized with: dgdt = -alpha1*fvc::div(phi); If the field is initialized incorrectly it leads to an error and to a mass imbalance of the solver. | ||||
| Steps To Reproduce | Run a case with a nozzle flow and a backfacing step and set the pressure as a fixed value boundary condition at inlet and outlet. The mass will not be conserved and errors are in the range above 5%. | ||||
| Tags | compressibleInterFoam | ||||
|
|
> However, for the first iteratrion the source term should be zero in the alpha equation: Why? If div(phi) is non-zero then dgdt should be non-zero. Or do you think that dgdt should always be zero, only zero at the beginning of each time-step or zero only at the start of the run (time 0)? |
|
|
Is the issue that the initial phi field you have in your case is not conservative? What do you get if you initialize your case with potentialFoam? |
|
|
The case is that if we start with a field of zero divergence, div(phi)=0. The terms on the RHS of the alpha equation should equal to zero as well. If dgdt is initalized with dgdt = alpha1*fvc::div(phi) then this adds together with the divU field and the Su expression in the alpha equation. As a result the RHS is never zero, even for a divergence free field. |
|
|
Take a very simple test and run with one iteration and one outer loop with the compressibleInterFoam solver and see that the alpha field changes. Even if the initial divergence field is zero. |
|
|
If div(phi)=0 then surely dgdt = alpha1*fvc::div(phi) is also 0? It is not clear what change you want, can you clarify? Do you think that dgdt should always be zero, only zero at the beginning of each time-step or zero only at the start of the run (time 0)? Is the issue that the initial phi field you have in your case is not conservative? What do you get if you initialize your case with potentialFoam? |
|
|
Ah yes, of course it is a bad example. Case: div(phi) not equal 0 in intial condition AlphaEqn: (as implemented in VoF/alphaEqn.H) ddt(alpha) + div(alpha*phi) = dgdt + alpha*div(phi) Maybe I have an error how I think about this. Should it not be, that the RHS equals to zero in the first iteration and alpha only changes due to the divergence field on the LHS? Currently it is implemented as: ddt(alpha) + div(alpha*phi) = alpha*div(phi) + alpha*div(phi) => ddt(alpha) + alpha*div(phi) + phi*grad(alpha) = 2*alpha*div(phi) => ddt(alpha) + phi*grad(alpha) = alpha*div(phi) => DDt(alpha) = alpha*div(phi) |
|
|
One Note to add: Assume a liquid vapor mix and liauid is represented with alpha1 and is nearly incompressible, so that rho1 approx. constant. Then the alpha transport equation is: ddt(rho1*alpha1) + div(rho1*alpha1*phi) = 0 or for nearly constant rho1: => ddt(alpha1) + div(alpha1*phi) = 0 This equation is only solved in alphaEqn.H if dgdt is initialized with: dgdt = -alpha1*div(phi) |
|
|
What do you get if you initialize your case with potentialFoam? |
|
|
not if I initialize with potential Foam. But I will have cases with phase change and in these the divergence does not have to be zero. Generally, I believe the descritpion as it is implemented is ok for most cases. But probably is not if the initial field is not well defined. So changing the inital value is a small thing to guarantee that the equation system makes sense in these cases too. |
|
|
> not if I initialize with potential Foam. That is what I get. > But I will have cases with phase change and in these the divergence does not have to be zero. OK, in which case dgdt should not be 0 either. > Generally, I believe the descritpion as it is implemented is ok for most cases. But probably is not if the initial field is not well defined. Right, and for these cases I would recommend using potentialFoam to make the initial field well defined. > So changing the inital value is a small thing to guarantee that the equation system makes sense in these cases too. What about restart? |
|
|
I have run some tests and confirmed that dgdt converges to alpha1*fvc::div(phi) as it should and this is correct initialization for dgdt. If your initial phi field is very non-conservative then there is a problem with this initialization for dgdt but this is not the only problem with the start-up. It would be best to ensure the initial phi field is consistent and appropriate, failing that you could use a large number of outer correctors at the beginning of the run to obtain convergence and conservation. Have you tried increasing nOuterCorrectors? For my cases this resolves the problem. |
|
|
I do not understand why dgdt converges to alpha1*fvc::div(phi). Expecially if the phase 1 is nearly incompressible. As the RHS of the alpha equation is then always: ddt(alpha1) + div(alpha*phi) = 2* alpha1 * fvc::div(phi) |
|
|
The proposed work-around will cause incorrect restart. It is not clear that any change is necessary and some important questions remain unanswered by the reporter. The general solution to the problem is to apply better initialization and/or more outer iterations early in the run to converge the initial fields. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2018-08-16 14:41 | janGaertner | New Issue | |
| 2018-08-16 14:41 | janGaertner | Tag Attached: compressibleInterFoam | |
| 2018-08-16 15:35 | henry | Note Added: 0009961 | |
| 2018-08-16 15:49 | henry | Note Added: 0009962 | |
| 2018-08-17 08:34 | janGaertner | Note Added: 0009966 | |
| 2018-08-17 08:36 | janGaertner | Note Added: 0009967 | |
| 2018-08-17 08:52 | henry | Note Added: 0009968 | |
| 2018-08-17 08:53 | henry | Note Edited: 0009968 | |
| 2018-08-17 09:33 | janGaertner | Note Added: 0009969 | |
| 2018-08-17 09:52 | janGaertner | Note Added: 0009970 | |
| 2018-08-17 10:19 | henry | Note Added: 0009971 | |
| 2018-08-17 10:35 | janGaertner | Note Added: 0009972 | |
| 2018-08-17 11:03 | henry | Note Added: 0009973 | |
| 2018-08-17 11:46 | henry | Note Added: 0009974 | |
| 2018-08-17 13:37 | janGaertner | Note Added: 0009975 | |
| 2018-08-21 18:58 | henry | Assigned To | => henry |
| 2018-08-21 18:58 | henry | Status | new => closed |
| 2018-08-21 18:58 | henry | Resolution | open => suspended |
| 2018-08-21 18:58 | henry | Note Added: 0009984 |
