Non convergence of grdvec

[Alec]

Hello,

On a case I am running with CS 3.0 (stable version), I get the following error message in my listing.

Code: Select all

 
 _iterative_vector_gradient: isweep = 100, residu norm: 1.669612e-02, norm: 3.704496e+04, var_num = 2
@ @@ warning: non convergence of grdvec
 _iterative_vector_gradient: isweep = 100, residu norm: 1.767031e-02, norm: 3.209838e+04, var_num = 2
@ @@ warning: non convergence of grdvec
 _iterative_vector_gradient: isweep = 100, residu norm: 1.867183e-02, norm: 3.370381e+04, var_num = 2
@ @@ warning: non convergence of grdvec
 _iterative_vector_gradient: isweep = 100, residu norm: 1.669612e-02, norm: 3.704496e+04, var_num = 2
@ @@ warning: non convergence of grdvec
 Warning:
 --------
   _iterative_scalar_gradient; variable:  5; sweeps: 100
                               normed residual:  1.4664e-03; norm:  1.5040e-04
 Warning:
 --------
   _iterative_scalar_gradient; variable:  5; sweeps: 100
                               normed residual:  3.4510e-04; norm:  1.7126e-04

What does it mean and how can I solve it ? I also can see that the solution oscillates badly.

Best regards,

A.

[Alec]

It seems to have disappeared by using upwind instead of SOLU for advected variables as recommanded by the BPG.

[Brian Angel]

Hello Alec,

It seems that the number of sweeps that is being used per equation is 100. If this is the case, it would seem that Code_Saturne is not reaching a converged solution with the solver precision that has been specified prior to reaching 100 sweeps. What you can do is increase the number of sweeps in "Maximum Iteration Number" in the "Solver" panel which is located in "Numerical Parameters -> Equation Parameters".

Can you also let me know what has been set for the "Maximum Iteration Number" and "Solver Precision" in the Solver panel.

Regards,

Brian Angel.

[Alec]

So, it is set for all variables to :

Maximum Iteration Number : 10000
Solver precision : 1e-8

Regards,

Alec

[Jacques Fontaine]

You can not change maximum iteration number of gradient computation with the GUI (iterative solver /= gradient).

Do you have an idea of your mesh quality?
You can try to use a centered scheme for convection (or SOLU) with least square gradients over partial extended neighborhood computation.

[Alec]

Hello,

Yes actually the mesh was not very good so I had already chosen to work on neighboring cells with least square method to avoid a divergence.

Checking the mesh, I get :

Code: Select all

  Criterion 1: Orthogonality:
    Number of bad cells detected: 88 -->   0 %

  Criterion 2: Offset:
    Number of bad cells detected: 0 -->   0 %

  Criterion 3: Least-Squares Gradient Quality:
    Number of bad cells detected: 5366 -->   4 %

  Criterion 4: Cells Volume Ratio:
    Number of bad cells detected: 2 -->   0 %

  Criterion 5: Guilt by Association:
    Number of bad cells detected: 0 -->   0 %

The fact is I rarely get better meshes to submit to Code Saturne and usually it behaves well. (Hexa/Prisms layer on the walls, tetra elsewhere).

I did not try to change the number of steps for the iterative solver or the gradient.

The warnings disappear when I use SOLU with any choice of extended neighborhood.

[Brian Angel]

Hello Alec,

Can you upload some images of your flow domain and volume mesh. Can you also upload the listing file of the calculations using SOLU or UD.

Regards,

Brian Angel.