[tladd]

I am using simpleFoam from OpenFoam 2.3.0 to solve for laminar flow in a flat channel (boring; but a simple test). The log file when the simpleControl debug flag is on is attached. I deliberately set a weak pressure convergence criterion (1e-2) which is satisfied after about 50 iterations in this case. But the velocity convergence criterion (1e-6) is nowhere close to being satisfied. The velocity field is about 5% off at this point.

So why does OF exit when the U norms are so far off (the segregated solver does not do this). Its because the simpleControl object is not checking for the U field, as shown by the debug output. Only the p field convergence is checked. The problem seems to be in fvMatrixSolve.C - the solver performance is not being added to the solverPerformance dictionary as it is for the segregated solver. Interestingly it seems the appropriate line is commented out; but uncommenting it leads to a compiler error so this was done for a reason.

But its the same in OF 2.2.2 so what gives? Does anyone know?

Tony

PS: there is another issue illustrated by the large relative norms for Ux and Uy (which are zero fields). OF, incorrectly in my view, normalizes the residual component by component instead of by the norm of the vector itself. First, this is inherently coordinate-system dependent, which suggests a mistake, and second if the velocity in a particular direction is zero (or small), the reported residual can dominate the solution control. We have patched this for the segregated solver but we are still puzzling over the coupled solver.

[allett02015]

Hello,

I had some similar problems for the coupled solver using OF 2.3.0. when
giving back the output of solve(...) to a object of type solverPerformance

and output the content of the object solverPerformance ( Info << ....) to the standard output the result is

( 0 0 0 0 (0) ).

It seems that only the standard consturctor is called without assigning any values to the variables of the object.