LuckyTran 
April 25, 2012 23:22 
Quote:
Originally Posted by ryanmoser
(Post 357082)
Thank you so much.
I have tried, but a new problem comes out. The thermal conductivity of fluid is very low causing big energy convergence difficulty. To eliminate the influence of flow, I let the fluid be solid temporarily to compute pure thermal conduct problem. With low thermal conductivity, only 1e01 of energy residuals can be reached, which leads to a bad temperature fied.
If thermal conductivity is increased to more than 70, this model can be easily converged.
How can I improve convergence performance at this stage?

You're doing it all wrong. You should try to do some tutorials before attempting a problem on your own. You clearly lack basic knowledge of CFD in general.
You don't need to turn the fluid into a solid, you can simply disable solving the equations for fluid flow off temporarily. Also, the energy equations are by far the easiest to solve. You should have no trouble getting a converged solution.
It also makes absolutely no sense to change your thermal conductivity. If you already understand that with poor convergence leads to a bad temperature field, how is changing the thermal conductivity going to help!? It can only make it worse!
Next, your energy equation solution will not make any sense until there is a well converged flow solution. If anything you need to solve the flow well enough before trying energy and not the other way around. You cannot eliminate the influence of flow as you say, if you could, why are you doing CFD in the first place? The whole point is to be able to capture the flow influence. It doesn't matter how well converged your temperature distribution is, it is meaningless and outright incorrect until the flow is correct.
I would not say that a new problem came out, just that you really have no clue what is going on. You need to correct this first and not just turn random knobs. You are shooting a bullet through a car engine in hopes of making it run better!
