[palarcon]

Good morning everybody

I would like to know if it is possible to perform two simple.loop inside the same code. In simplified words, what I want to do is something like this algorithm

int main(...)
{
while(it <= maxIter)
{
     ...
     while (simple.loop())     // first loop
     {
          ...
          #include "UEqn.H"
          #include "pEqn.H"
          ...
     }
     ...
     while (simple.loop())     // second loop
     {
          ...
          #include "UEqn2.H"
          #include "pEqn2.H"
          ...
     }
     ...
}
...
}

The problem is that when I code like that, the solver enter the first loop and solve the flow but then doesn't enter the second loop.

My guess is that happens because simple.loop already have an stop signal recorded, so when I try to enter the second loop by means of while (simple.loop()) for the second loop, simply doesn't enter because the while(simple.loop()) condition was already fulfilled in the first loop.

Is there any way to restart or put to zero the simple.loop in order to be able to enter to the second loop?

Any insight, references to documentation or examples, or threads in the forum I may have missed are welcome.

[pete20r2]

I'm not too strong on c++ and I don't have the source in front of me but I'd say that you are right and the second loop doesn't run because the simple.loop() function is returning 0. I think you should reconsider what exactly you are doing because solving the simple loop again is equivalent to just lowering the solver tolerances so it loops more in the first place. You won't be working on the results of the first simple loop until the timestep (iteration for steady state) is advanced.

[clapointe]

If you aren't using the simple controls (ie just running for a fixed number of iterations) a simple fix is to use your own for or while loops instead of simple.loop().

Caelan

[palarcon]

Actually and in principle, I need the simple controls (or an equivalent one) because is necessary to converge the flow equations in loop 1 before entering loop 2, which, unfortunately solve equations that are really similar. That is the reason why I'm using simple.loop in both.

[Phicau]

Hi Pablo,

how about re-defining the variable simple? You can try adding this line after the first loop has finished:

simpleControl simple(mesh);

I cannot test it right now, but I imagine that in order for it to work, the variable startFrom in controlDict needs to be set to startTime, and startTime to the value of your starting time, usually 0.

Best,

Pablo

[clapointe]

Another option would be to manually set the time with eg. runTime.setTime() -- you'd set it to zero. I think this would reset the simple controls.

Caelan

[pete20r2]

You could have a look at how the chtMultiRegionFoam handles multiple meshes and solver loops. It appears to setup custom loops for the different regions neatly.

[palarcon]

Hello

Everybody, I'm taking notes of all of your comments and I will try them today.

[palarcon]

Hello Phicau

Can you please explain me the effect of

simpleControl simple(mesh);

It will reset the convergence criteria, the number of iteration or something different?

[palarcon]

Originally Posted by clapointe

Another option would be to manually set the time with eg. runTime.setTime() -- you'd set it to zero. I think this would reset the simple controls.

Caelan

I tried that and it "worked" but not perfectly

runTime.setTime(it - 1,it -1);

The idea is that after completing the computation of the loop 1 (with N inner iterations) must enter the second loop, where the inputs are the results of the first loop.
That is the reason why the counter is reset to the value of the global iteration number, because there should be X number of complete solutions of the flow (because loop 2 changes the topology of the design domain).

[Phicau]

Originally Posted by palarcon

Hello Phicau

Can you please explain me the effect of

simpleControl simple(mesh);

It will reset the convergence criteria, the number of iteration or something different?

It will make it have the same value it had when the simulation started. Basically it is the line called by the solver to initialize the object simple.

[palarcon]

Hi Pablo

Unfortunately that option doesn't work

error: redeclaration of ‘Foam::simpleControl simple’
         simpleControl simple(mesh);

Originally Posted by Phicau

Hi Pablo,

how about re-defining the variable simple? You can try adding this line after the first loop has finished:

simpleControl simple(mesh);

I cannot test it right now, but I imagine that in order for it to work, the variable startFrom in controlDict needs to be set to startTime, and startTime to the value of your starting time, usually 0.

Best,

Pablo

[Phicau]

I see... what about using another variable name?

simpleControl simpleSecond(mesh);

while (simpleSecond.loop())     // second loop
     {
...

[clapointe]

Originally Posted by palarcon

I tried that and it "worked" but not perfectly

runTime.setTime(it - 1,it -1);

The idea is that after completing the computation of the loop 1 (with N inner iterations) must enter the second loop, where the inputs are the results of the first loop.
That is the reason why the counter is reset to the value of the global iteration number, because there should be X number of complete solutions of the flow (because loop 2 changes the topology of the design domain).

I see. How many of the outer iterations are you doing? If the secondSimple from above doesn't work, then another option will be to set to zero and then write to zero. Then you'll be at "zero" time but with the data from the previous simple loop.

Caelan

[palarcon]

Do you have any idea on how create that second variable simpleSecond.loop in order to be recognized by OpenFOAM?

Originally Posted by Phicau

I see... what about using another variable name?

simpleControl simpleSecond(mesh);

while (simpleSecond.loop())     // second loop
     {
...

[palarcon]

I can set the time to zero (or any number) but that doesn't change the convergence criteria.
If the first loop is already converged, then there is no way on entering into the second, that is the reason why I have to reset the convergence criteria and not only the time value.

Originally Posted by clapointe

I see. How many of the outer iterations are you doing? If the secondSimple from above doesn't work, then another option will be to set to zero and then write to zero. Then you'll be at "zero" time but with the data from the previous simple loop.

Caelan

[clapointe]

Good point -- I was hoping that writing in addition to changing the time would help. Should have looked at the code first.

What if you just wrote your own conversion criteria? Use two while loops, and when whatever you decide to calculate stops changing, you're converged.

Caelan

[Phicau]

Hi Pablo

the way to define the second variable is easy:

simpleControl simpleSecond(mesh);

I presume that it will need to have the following lines in your controlDict for it to work properly:

startFrom startTime;
startTime 0;

Pablo

[Jesus.M]

Did the last solution work?

[Jansois]

Not sure if it has already been solved but this works for me.


Add an extra constructor for another pimpleControl object in your solver:

pimpleControl pimple(mesh);    //normal line

pimpleControl pimple2(mesh, "PIMPLE2", true);    //added line

Then you can create additional loops in your solver with:

while (pimple2.loop())
{
    //do stuff in outer loop

   

    while (pimple2.correct())
    {
         //do stufff in inner loop

    }
}

Now add a dictionary in your system/fvSolution file like this:

PIMPLE2
{
    nCorrectors         x;
    nOuterCorrectors    y;
}

Hope this helps.