I have a 3D fluid flow problem that converges but not to the criteria I set. I set convergence for all directions and continuity to 1e-06. The flow is for a tubular SOFC. Air enters a 10mm diameter air inducing tube which is 1450mm long. The air emptys into a larger closed end tube with a 24mm diameter and 1500mm long. The air reverses directions against the closed end of the outer tube and flows out of the domain in the annular region between the air inducing tube and the outer tube. The air inducing tube has a .5mm thickness. There is also a fuel flow on the outside of the outer tube. which enters the domain at the closed end of the outer tube. Also the outer tubes walls are porous.

When I run the simulation, it begins to converge and then levels off around 1e-04 for x,y,z, and 1e-02 for continuity. I refined the mesh but that did not improve the convergence.

Is there a way to improve convergence? Is it even important to get it to 1e-06 or are the current values good enough? Also when I visualize the flow it looks correct. It is not perfectly symmetrical but pretty close.

Thanks Emily

From your problem setup, it appears that the vast range of dimensions 1000mm to 0.5 mm could be why you are facing some of this complexity. A general rule of thumb that many practitioners use that if you get a reduction of about 4 orders of magnitude in your residuals, then you can assume that your flow has converged even if the absolute convergence is not 1e-6. You can also try to reduce the under relaxation factors. This will improve the damping of oscillations/stability in the code. -Phil

Are you using single precision or double?? i mean 3d or 3ddp. I remember facing the same problem sometime ago when the truncated errors were rounded off to 1e-2! only and the solution dint coverge. I switched to 3ddp, then i got the solution converged in no time. May I suggest/advice that mesh refining has nothing to do with convergence. u can try 3ddp on coarce mesh too.

Hope this helps Regards Rizwan