Hi all:

I am using GS/SOR to solve the Pressure Equation in SIMPLE algorithm. The code works well for a simple square duct & backward facing step problem. But, when I try to use the same code to simulate flow in a forward-facing step, my code can't take it and diverges...

I found that diagonal dominance is difficult to be achieved particularly on cells near to the forward-step (stagnation region). Can this be the main problem in achieving convergence? What is the remedy? Shall I use a "different" relaxation factor in this region in order to ensure convergence??

Sometimes, backflow may happen at outlet... Any special BC treatment if backflow is detected at outlet? So far, I have set the normal velocity to zero should backflow occurs..(treat it as wall).. However, this does not solve my problem..

Can anyone share your experience?

-khai ching-

Khai Ching wrote: I found that diagonal dominance is difficult to be achieved particularly on cells near to the forward-step (stagnation region). Can this be the main problem in achieving convergence? What is the remedy? Shall I use a "different" relaxation factor in this region in order to ensure conve rgence??

------- A few thoughts:

If you are using a steady-solver, the problem may very well be that you do not have a steady solution to the flow field you have constructed - at least at the velocity you are using.

Try testing this by lowering the inlet velocity & try for convergence. If you obtain a solution, then gradually raise the velocity until your 'lack of diagonal dominance' begins to show itself. Then move over to a 'transient solver', or a steady solver which is 'soft in time', to take you through the transition point.

Take 'snapshots' of your solution at various run intervals & look at where the problem occurs physically. This excercise can provide amazing insights into what the solver is really having to cope with.

You can also try testing your model mesh sensitivity by refining the mesh in the region where you are experiencing problems. Once the 'correct mesh limit' has been reached, then you will most likely have to move towards an unsteady solver.

diaw...

Thanks diaw..

I have reduced the inlet velocities and it works!! Perhaps I should gradually raise the velocity to ensure convergence...

Thanks for your advise..

It is my pleasure... good luck with your work...

diaw...