Okay folks, I'm fixing to get in an argument with colleaque. Anyway, I am trying to achieve similar results on a slightly different geometry. We have a molten glass(highly viscous, temperature dependent viscosity and density), with multiple outlets. The glass surface is conditioned by a 'radiation temperature profile' boundary on the glass surface. These profiles come from other models. I run a steady-state solution and it converges smoothly to the fluent default residuals. I then indicate a transient solution with the time step he specifies. At each step, it converge within 2 iterations. It can't be transient! I say to myself. I then look at the transient solution my colleague obtained. His mesh quality is quite a bit poorer. For equi-angle skew, his model has 0.94 while mine is 0.86, and for equi-volume skew, he has 0.99 while I have 0.93. Is the problem with his mesh, or should I be looking for something else? Your feedback is appreciated. Ed

think what you are seeing is a pseudo-steady state solution. If the flow parameters arent changing physically i dont see why you wud expect longer solution times with each time-step! what you are basically doing is tracking time-based solution after it has reached steady-state..so theres not going to be much difference. make sense?

Sai, Possibly, he's getting vortex rolling where I am not. I don't have access to his model, so I can't improve his mesh. All I can do is improve mine. I believe he has numerical dispersion. He's only running 1st order time accuracy on a bad mesh, whereas I'm running 2nd order time accuracy on a reasonably good mesh. I'm going to try increasing my mesh density to make sure I'm not possibly 'missing' the psuedo-steady-state. I don't believe that's the case since I have mesh where he predicts the rolls. I would try reducing the time step size, but I'm just trying to keep a consistency with his model. We'll see what happens. Thanks for your response. Ed