[peterhess]

Hello Saleh!

Try those setups.

https://drive.google.com/file/d/1PnR...ew?usp=sharing

I had a good convergence for more than 10000 iterations.

Regards

Peter

[fertinaz]

Hello Peter and Saleh

I checked this case as well if you don't mind.

Since both solvers fail during the pressure correction evaluation, I focused on the pressure first. I applied Robin BC at the inlet:

    Inlet
    {
        type            zeroGradient;
        refValue        uniform 116800;
        refGradient     uniform 0;
        valueFraction   uniform 0.1;

        // type            fixedValue;
        // value           uniform 116800;
    }

It initially worked fine but then I realized that blow-up was just delayed.

Then I took a look at the "system/fvSolution" and modified:

    rhoMin        0.1;
    rhoMax        2;

and also reduced the pressure relaxation to 0.7. Not sure if 1 is a good value which will chop-off the deviations for stability but may produce unphysical results.

These modifications let simulations run even further but could not prevent blow-up which this time occurred at "Time = 0.4597".

Therefore I changed velocity outlet to kill potential inward reflections on the outlet patch:

    Outlet
    {
        type            inletOutlet;
        inletValue      uniform (150.6 0 0);
        value           $internalField;

        // type            zeroGradient;
    }

It now seems to run fine. It already passed "Time = 3.3" with "deltaT 1e-04;".

Please test it and see if it works for you as well.

I'd highly recommend you to check the last few steps prior to the blow-ups and try to see why they occur. Visualising may help you out.

// Fatih

[Saleh Abuhanieh]

Gunaydin Fatih,


Thanks for your contribution, I'll run with your suggestions and get back to you.


Saleh

Regards,

[peterhess]

Hello Fatih and Saleh!

I tested the suggestions you made (on the case I uploaded lately with the turbulence activated) and they works!

You get convergence very nice and the results are for me have physical meaning...

-------------------------

The last sugesstion is helping the simulation getting convergence. The simulation does not have back flow, anyway, It increase the stability.

I would also do it this way...

-------------------------

The secound suggetion is already applied cause the density at the inlet is higher than 1.05...

-------------------------

The first suggestion is solving the problem!

Anyway, The pressure at the inlet is not fixed anymore as Saleh already confirmed to be!

The problem here is that Saleh is fixing the pressure at the inlet and the outlet and at the same time fixing the velocity at the inlet...

That is physicaly (in my opinion and I may be wrong) not applicable...

Eather you fix the "statical" pressure at the inlet and the outlet and look for the velocity driven via pressure difference.

Or you fix the velocity at the inlet and the "statical" pressure at the outlet and then see which pressure is needed at the inlet to force this pre defined velocity.

The suggestion of Fatih is somehow in this direction.

The simulation as it has been setuped by Saleh is "over defined" in my opinion.

I would define the pressure at the inlet as totalPressure and let the simulation divide it (calculate) between velocity and static pressure.

----------------------------

Saleh: I looked to the paper you mentioned earlier.

The pressure at the inlet is NOT statical.

It is totalPressure

See page 57 (68/100) - table 4-2 - inflow!

Benjamin actualy mentioned on the same page at the bottom that the pressure is fixed at the inlet, but he did not say it is statical! He means fixed totalPressure

Cause on the next page - Figure 4.11 - a, the pressure everywhere is less than the pressure you defined to be "statical" fixed at the inlet.

Regards

Peter

[Saleh Abuhanieh]

Dear Peter and Fatih,


Thank you very much for your support!


I've been running the latest case (the uploaded cases from Peter with Fatih's comments). The residual is quite smooth and in the range of 1e0-5 now and the flow is not changing any more. I attached the velocity profile at 183k iteration, the result still different from the original case (Benjamin's) and there is big difference in the max velocity (270 m/s comparing 498 m/s). I stopped at 205k iteration, complete convergence.







Yes Peter, I agree with you that the case became over-defined!
Thank you for referring to the original work and your valuable comments.

Actually as I highlighted earlier in this thread, I think the problem some where in the BC. I tried before using totalPressure at the inlet, but I got FPE as well, this is why I left it static (as per my wrong interpretation for Benjamen's words).


Starting from the initial case uploaded by me with changing only the inlet BC to totalPressure, I got FPE at around 14k iteration and the residual is fluctuating.



Time = 13801

smoothSolver: Solving for Ux, Initial residual = 0.000258627, Final residual = 2.31127e-05, No Iterations 10
smoothSolver: Solving for Uy, Initial residual = 0.000332803, Final residual = 3.06775e-05, No Iterations 8
smoothSolver: Solving for e, Initial residual = 0.000239535, Final residual = 1.70034e-05, No Iterations 6
GAMG: Solving for p, Initial residual = 0.000125527, Final residual = 1.57558e-06, No Iterations 1
time step continuity errors : sum local = 620.309, global = -309.237, cumulative = -4.36703e+06
ExecutionTime = 964.21 s ClockTime = 965 s

Time = 13802

smoothSolver: Solving for Ux, Initial residual = 0.000301733, Final residual = 2.6563e-05, No Iterations 28
smoothSolver: Solving for Uy, Initial residual = 0.000354937, Final residual = 3.09186e-05, No Iterations 18
smoothSolver: Solving for e, Initial residual = 0.0118356, Final residual = 0.000707792, No Iterations 10
GAMG: Solving for p, Initial residual = 0.000130643, Final residual = 1.77223e-06, No Iterations 1
time step continuity errors : sum local = 637.933, global = -309.221, cumulative = -4.36734e+06
ExecutionTime = 964.3 s ClockTime = 965 s

Time = 13803

smoothSolver: Solving for Ux, Initial residual = 0.00349629, Final residual = 0.000234674, No Iterations 4
smoothSolver: Solving for Uy, Initial residual = 0.00126557, Final residual = 0.000109374, No Iterations 2
smoothSolver: Solving for e, Initial residual = 0.0243903, Final residual = 0.00210812, No Iterations 2



I'll try now the uploaded case by peter with Fatih's comments with changing the inlet BC and post the results.


Regards,

Saleh

[Saleh Abuhanieh]

Hi Foamers,


I just wanted to give a feedback about the case for reference.
Using the totalPressure BC at the inlet produces a stable results but not accurate and far from the reference study which was based on sonicTurboFoam and experimental resluts.


With the same setup, only reducing the time step to 1e-06, I got a good results using sonicFoam.

Then I tried the same for the other two cases in the reference thesis, namely: the weak and strong shock, however I didn't get acceptable results (the velocity is too lower). When I changed the inlet BC to fixedValue (@peter ), The results were acceptable; more accurate in case of weak shock and less accurate in case of strong shock.


Regards,

[Ale22pic]

Hi everybody, I know it is been a while. But it is possible to have the transonic diffuser test case. I'm struggling to solve it with rhoCentralFoam in Openfoam, unfortunately the links in this thread don't work anymore. Thank you in advance.