[yashjo]

Hello,

I am trying to use under relaxation with rhoCentralFoam.

I added the following in the fvSolution but it made no difference in the solution.

relaxationFactors 
  { 
      fields 
	{
		//p 0.9;
		//rho 0.9;
		".*"	1;
		//U 0.7;
		//k 0.7;
	//	omega 0.7;
	}
      equations 
      { 
          ".*"               1; 
          //"(rho|U|k|omega|epsilon).*" 0.7; 
      } 
  }

This did not change the solver output at all (the output log files are identical with and without relaxation factors).

Is it not possible to use relaxation factors with rhoCentralFoam? Or is there some naive mistake that I am committing.

[deepbandivadekar]

Originally Posted by yashjo

Hello,

I am trying to use under relaxation with rhoCentralFoam.

I added the following in the fvSolution but it made no difference in the solution.

relaxationFactors 
  { 
      fields 
    {
        //p 0.9;
        //rho 0.9;
        ".*"    1;
        //U 0.7;
        //k 0.7;
    //    omega 0.7;
    }
      equations 
      { 
          ".*"               1; 
          //"(rho|U|k|omega|epsilon).*" 0.7; 
      } 
  }

This did not change the solver output at all (the output log files are identical with and without relaxation factors).

Is it not possible to use relaxation factors with rhoCentralFoam? Or is there some naive mistake that I am committing.

I felt the same with the tutorial case. However a working case for my intended problem still remains illusive so I am not sure if under relaxation factor indeed makes any difference in rhoCentralFoam

[jnanabrao]

rhoCentralFoam is an explicit solver. It does not use the PIMPLE algorithm to solve for the properties for the next time step. It evaluates the current spatial derivatives (fvc) and equates them with the time derivative in the conservation equation, to march forward in time. I believe the only term that is discretised in the traditional implicit manner in this solver is the viscous dissipation term in the energy equation. The under-relaxation method for equations is used to stabilize the implicit solvers by increasing the explicitly calculated terms in the equations. To maintain accuracy in traditional solvers this usually means use of higher number of innerCorectors and outerCorrectors.


If your rhoCentralFoam simulation diverges, you should set all gradSchemes to faceLimited Gauss linear 1 and the reconstruct variable to either vanLeer 1, vanAlbada 1, upwind or Gamma 1. If you don't get a solution even with these schemes, the issue must be in the gird or boundary conditions. Not to mention you must go down to at least a Courant number of 0.2 before any of these changes.