Now i have found a question in CFX11.0 for all type of heat transfer problem. It will occur the very low temperature in some area(below the enviroment). Anybody have some solution about this?

Hi,

Is your solution properly converged? Probably not I guess.

Glenn

You can test this in the CFX tutorial 18. But it is not only this case, there are a lot of this type of CHT questions occure in the Version10.0 and 11.0. I also test this in Version 5.7.1,it is okey, So i think it is caused by the numeric scheme change from Version5.7 to 10.0 or 11.0.

Hi,

Yes, there have been changes to the numerics over the years. The default differencing scheme in CFX is High Resolution which has some second order in it so it will potentially lead to undershoots/overshoots. The high res scheme tries to minimise these under/over shoots - have a look at the documentation.

For further reading on the effects of differencing scheme have a look at the book "Computational Fluid Dynamics" by Roache. This book explores accuracy and errors in CFD in detail.

Regards, Glenn Horrocks

Thanks,Glenn Horrocks

Now in my practicle CHT question,I will change the energy sheme to up wind,thus it will some time avoid the low temperature. But it will not always useful.

Hi,

Yes upwind will also eliminate the problem but will reduce the accuracy of your simulation. You will need a very fine mesh to get accurate results with first order differencing. Do a mesh refinement study to check your accuracy.

Glenn Horrocks

Hi, The following is the explain from the headquater:

1) The "Freestream Damping Option" for the High Resolution advection scheme was changed from 1st order (5.7.1) to 2nd order (10.0 and 11.0). To see the difference, look at the "Energy and Mass Fractions.Beta" variable in Post 2) By default energy is excluded from calculation of the common Beta for energy and mass fractions.

Because of 2) the HR Beta for energy is determined from the component mass fractions alone, and because of 1) in 10.0/11.0 the advection scheme will be second order in regions of constant composition (in 5.7.1 it used to be blended to 1st order). However, if temperature gradients are present in such a region, then the second order scheme may cause oscillations/over-/under-shots as observed in the current case. In 5.7.1 such regions were blended to 1st order, which did avoid the oscillations, but at the cost of reduced accuracy in regions where we could afford the higher order.