Issue of divergence
 

Hello every one,
I m simulating wind flow around rectangular buildings using LES on Ansys/Fluent.the problem of divergence is detected during the simulation after
flow time of 3 sec.
time step size = .0005
Vortex method (no. of vortices = 190) for turbulence generation
Polyhedral mesh is being used
So please guide me how to resolve this issue.


Thanks

Your time step may be too long to capture the vorticies forming at the building edges depending on flow speed and mesh density. Without further info and assuming your mesh is good, its my best idea.

Depending on the SGS model, cell dimension, cfl, the added eddy viscosity can drive to numerical instability for that time-step....

Could you address all the details of your LES setting?

Hello sir.
thanks for reply,

I m using dynamic subgrid scale model.
first cell height is 0.000125m
wind speed at building height=9.5 m/s
vortex method is used to add turbulence and of vortices =190
CFL condition for most of cell around building is <1
no. of polyhedral cells are about 1.9 million

what kind of method are you using? did you set second order in time and space? is the numerical instability onset after many time steps?

Seeing how fluent is, if I have to guess, I would guess that your mesh is too coarse. When the grid is coarse, specially when we are talking of buildings, the disspation that keeps velocity and pressure coupled is too low and equations decouple.

Refine the mesh a bit further and see, how it behaves.

I guess to make things worse times step is too low too (which also reduces dissipation).

spatial discretization for Pressure is second order and for momentum it is bounded central differencing.
time discretization scheme is second order implicit.

Numerical instability occurred after 6880 time steps.

have you tried setting a smaller dt?
you should check some typical variable such as rms and spectra to see if you get energy pile-up.

Just as check of validation, try the same discretization setting the static smagorinsky model but with Cs=0 (no-model) and reduce the Reynolds number. See if you get a stable unsteady solution.

Thank You sir
I will do this..........