Learning Forum

[tandinhtamky]

Hi there,

I'm working for this simulation. This model include 2 cylinder (battery cells) with heat source in side. Air domain with velocity inlet = 2m/s, temperature = 300K.

I have already control the y+ at wall of cylinder below <1.

As I know, I use K-w SST, and SIMPLEC, second order upwind, but the residual not descent below 10e-6 (even 10e-3).

I'm quite sure about quality of mesh. (min orthogonal quality >0.16)

So, what should I use turbulent model ?

Thank you so much.

[DrAmine]

Sst model is the model of choice. Coupled solver is recommended.

[tandinhtamky]

Here is my new setup.
I'm nervous about time step size, is it good ? because now I want to simulate the discharge process of battery in 3600 second (1 hour).
Coupled scheme with second order upwind.
Solution control here:
But, the residul (continuity) just about 1e-2 (orther ~1e-3). And, I also check velocity inlet plot -> value become constant. And I also check in the fluxes in the results, about mass imbalance, total heat tranfer rate -> that's ok.
So, my simulation now is convergence or not, Sir ?

[Rob]

We typically set a time step so the flow crosses a cell in roughly 10 time steps and then adjust to converge each time step in 10-15 iterations. There are a few transient tutorials in the Help system so have a look through those for some more guidance.

[DrAmine]

Your time step is very enthusiastic and can understand why you are so nervous about it. A good rule of thumb to have a time step corresponding to delta_t= (minimum over all cells of cell volume of 10*CellVolume^0.333/maximum Velocity)*10 (using her CFL = 10 for quick runs. You can increase it to even higher values if things are stable).

[tandinhtamky]

Thanks for replying me !
But, I don't quite understand what you mean about delta_t formula. What is "minimum over all cells of cell volume of 10" ? And Cell volume is max or min or average value ?
Also, I switch viscous model to RNG k-e enhance wall function, still coupled method but flow courant number I descent from 200 to 50. Time-step-size still 10 second. Below is residual.
Is it convergence, right ?

[tandinhtamky]

Just one more question.
I see this mesh on internet, this model quite similar to my model, but I don't know how to create this mesh.
Mesh in the cylinder is OK, but mesh in domain not.

[DrAmine]

Just look after CFL number and you will understand my reasoning, ­ƒÿÇ.

[DrAmine]

Judging convergence will also need to consider behavior of couple of monitored quantities and imbalances. Only residuals looking is not enough: you any hmway have now low global scaled continuity residual for example which does not automatically 100% confirm that solution towards end if time step is not changing but it tells that likely it won't change: u should verify that for this very simple exercise.

[DrAmine]

Mesh can be created with ICEM CFD or Ansys Meshing.