I am solving natural convection problem using CFX 5.5.1. Since it seems I could not get a converged steady state results, I am trying transient simulation. I have the following questions about setting and solving my problem, where all BCs are constant.

1)Should I run the transient simulation based only on steady state results (not converged, of course)? Can I run transient simulation based on transient simulation results (converged better than steady state results as seen from the residual plot)? If I can, should I set automatic as the initial time? I think I still can set a value of 0 since all the BCs are constant. Does it matter?

2)Can I choose timestep based on the residual plot? A simple time step dependence study is recommended in the menu, is it OK to take the effect of the timestep on residuals as such a study for my problem?

3)How can I decide at what time steps the transient results are valid? Should I exclude the time steps when residuals are above my target values (for example, 1 e-4)? What should I do if the residuals are above 1e-4 for all time steps (I think I have chosen a very short timestep)?

Thanks in advance.

Hi, sounds like an interesting problem. Firstly I'm supprised that you cant get a converged Steady state solution. Low or zero ambient velocity flows are more difficult to converge but it is possible. Can you describe your present steady state setup and model. Are you using inflation on your model surfaces, what are your boundary conditions ?? Dave

Thank you, Dave. My problem is not pure natural convection. You are right, I can get a converged solution for my model at some BCs, but not always. Do you think a converged steady state solution can always be got with proper mesh and other settings for natural convection problem? My model is about kitchen hood. To make it easy to be understood, it just like a wind (at different temperatures and different flow rates) over a very hot surface. The hot plume over the hot surface is likely not steady. BTW, if I put a moving block (moving in direction crossing the wind) between the wind and hot surface, do you have any idea how to simulate the effect on the plume? Does CFX 5.5.1 have such ability? Best Regards.

Hi, I don't see why you shouldn't be able to get a Steady State solution out from all your models. Obviously the higher the wind over the hot surface the easier the convergence will be. Do you output your residual values ? if not try doing this and have a look at them in Post. They can be output by creating an advanced variable using the edit definition file tool in the solver manager window. The problem will be unsteady in real life, and I'm not sure how CFX would cope with that ?? May be Neal or Robin could answer that question ?? Dave