Residual in steady-state simulation
 

Hi,

I have a 2D steady-state simulation in laplacianFoam (simple unit square made with blockMesh).

I have set the SIMPLE convergence criteria to be 1e-9 and the solver tolerance to 1e-11. The solver is set to go from 1 to 10s [for steady state simulation this can be understood has an iteration counter].

In this case, it will solve in step 1 to a final residual of something to e-12. When the solver enters the next step, the final residual from time 1 is not the same as the initial residual of time 2 but it is still some value to e-12 and as such the simulation is considered converged.

Why is this happening? I am not using any relaxation, so I would expect the residual value to be the same. Is it the normalization of the residual that is recalculated and slightly different?


Now, I change the mesh to have an arch [now with some non-orthogonality and skewness]. I have adjusted the numerical schemes to leastSquares for gradient and Gauss linear corrected for the laplacian. The simulation requires far more time steps but the solver convergence is always achieved (1e-11) at each time step. Why is it not converging in a single step as the previous one?

Are the gradients not correctly calculated in a single loop iteration and so I need to keep on solving the system to get it right?

I am new to OpenFOAM and CFD so I am still struggling with some concepts. Sorry if the question is too basic. I would appreciate some insight on this.