[backscatter]

Hi,


Recently, I noticed that having high aspect ratio cells away from the wall leads to a very low determinant for those cells. It has absolutely no effect on flow field (UEqn) but when I add a passive scalar transport equation to the simulation, the presence of these underdetermined cells leads to the blow-up of the scalar equation (only) with the momentum equation still converging.

To circumvent this issue, I alter my mesh such that I get rid of these underdetermined cells away from the wall. That helps and doesn't cause scalar field to blow up.

Has anybody noticed this kind of behavior before?
If yes, is there anything we could do than to alter the mesh?
I have tried all kinds of bounded schemes. They don't help.

PS:To get the idea of what I am talking about look at my old thread:
TEqn (passive scalar) diverges all of a sudden

[wildfire230]

Just to double check, once you introduce a scalar transport equation, there are at least two new conditions on the solver in order to have stability and convergence. A new limit on the time step is imposed, and the grid must be sufficiently refined in order to converge. If the time step is too large the velocity and pressure may still converge, while the scalar value could blow up, and if the grid is too coarse the scalar may blow up no matter how small the time step is.

[backscatter]

Originally Posted by wildfire230

Just to double check, once you introduce a scalar transport equation, there are at least two new conditions on the solver in order to have stability and convergence. A new limit on the time step is imposed, and the grid must be sufficiently refined in order to converge. If the time step is too large the velocity and pressure may still converge, while the scalar value could blow up, and if the grid is too coarse the scalar may blow up no matter how small the time step is.

Thanks for the comment. I knew that would be the case but in the back of my mind I was thinking if the undetermined cells might be the cause. It turns out that reducing the time step by an order of magnitude does the job. In the regions where Reynolds Analogy breaks down, the turbulent Prandtl number decreases to a large extent and that increases the turbulent thermal diffusivity, thereby making the time scale of thermal transport small- which, of course, the timestep suited for solving momentum equation could not resolve.

[damon707]

Hello all,

How exactly did you manage to eliminate the cells with low determinant?

Kind regards,
George