[hotboy]

hello friends !I was modeling natral convection.I want to know weather the initial value is important or not?If the initial value is not goog it ,the result will not converged ?

[dreamz]

Initial value is important for Natural convection to give direction to the solver. However I don't think it will directly affect convergence of your problem. It might take a larger number of iterations to achieve convergence

[LuckyTran]

Initial guesses always help, but you can certainly get results even without the best initial guesses.

For natural convection, taking zero initial velocities has always worked well for me. The hardest part has always been starting with a decent initial temperature profile, especially in complex geometries. For simple geometries I just use a uniform temperature (say 300 K).

For complex geometries, what I've adopted is to freeze the flow solver and run only the energy equation for say 30 iterations. This gives realistic temperature distributions in the solid regions. Then you can enable the flow solver again. To get the natural convection right, you (only) need to get the exposed surface temperatures right, and this step helps. But for simple cases, and causes where you're not having divergence problems, it's probably not worth the hassle and you can just skip.

[hotboy]

Originally Posted by LuckyTran

Initial guesses always help, but you can certainly get results even without the best initial guesses.

For natural convection, taking zero initial velocities has always worked well for me. The hardest part has always been starting with a decent initial temperature profile, especially in complex geometries. For simple geometries I just use a uniform temperature (say 300 K).

For complex geometries, what I've adopted is to freeze the flow solver and run only the energy equation for say 30 iterations. This gives realistic temperature distributions in the solid regions. Then you can enable the flow solver again. To get the natural convection right, you (only) need to get the exposed surface temperatures right, and this step helps. But for simple cases, and causes where you're not having divergence problems, it's probably not worth the hassle and you can just skip.

Thank you very much!Why do you use a uniform temperature(say 300K)?If we use a uniform temperature 400K,500K or 600k ,what's the result ?

Can I comprehend your meaning as this ? When modeling complex geometries ,we can freeze the flow solver and run only the energy equation for 30 iterations,then enable the flow solver . But when modeing simple geometries we can solve the flow and equations at the same time?

What is you meaning of "To get the natural convection right, you (only) need to get the exposed surface temperatures right", the temperature is the exposed surface temperature of solid or the gas?

[LuckyTran]

Originally Posted by hotboy

Thank you very much!Why do you use a uniform temperature(say 300K)?If we use a uniform temperature 400K,500K or 600k ,what's the result ?

Can I comprehend your meaning as this ? When modeling complex geometries ,we can freeze the flow solver and run only the energy equation for 30 iterations,then enable the flow solver . But when modeing simple geometries we can solve the flow and equations at the same time?

What is you meaning of "To get the natural convection right, you (only) need to get the exposed surface temperatures right", the temperature is the exposed surface temperature of solid or the gas?

300 K is only an example. Use whatever you like.

You only need a few iterations ~30 to get temperature fields that are more consistent with the boundary condition (and not the initial conditions). For simple geometry (usually small mesh) the time it takes to compute these 30 or so iterations (with energy + flow) is not very long so you can just keep the energy + flow enabled from the start.

Supposed you were doing a conjugate heat transfer simulation. The temperature boundary conditions for the natural convection flow field is the exposed solid surface temperature. If you have a reasonable temperature distribution for the exposed solid surface, you could more quickly establish the convective flow. In the extreme case, if you knew exactly what the surface temperature distribution was, you could avoid a conjugate CFD model altogether.