help, solution blows up!
 

Hello,

I am trying to do a finite volume LES calculation using a commercial code for flow across a blunt body. I am using a steady inflow BC with 2% turbulence intensity and fixed pressure outflow BC. I am using Crank Nicholson for time advancement.

After about 600 time steps, I see that the code blows up. (Kinetic energy of the flow tends to infinity). I think there could be several reasons, listed below. Please try to eliminate them or add to them..

1. Could the instability be due to crank nicholson? I know that the method is unconditionally stable, but only in the weak sense. I know my CFL numbers (U*delta_t/delta_x) in many parts of the grid are way above unity. Could this be the problem? I don't want to use a first order implicit method to avoid it's diffusion, and I cannot really reduce the time step as it would make the computation impractically expensive.

2. Could waves be reflecting off my outflow BC and causing instability? How do I check for this? How do I remedy this?

3. I am using central differencing for my advective terms (to avoid the well known diffusion due to upwinding). Based on the cell Peclet number, I know that I cannot satisfy boundedness at every point on my grid, and this may well be the cause of instability. Does anyone have a similar exeprience? How do I get around it? Does anyone have any experience with second order upwind methods? How do they perform in an LES?

Re: help, solution blows up!
 

CFL numbers need to be ideally below 0.3 in LES. Reduce your time step significantly.

Re: help, solution blows up!
 

crank nicolson should be an implicit scheme, but it doesn't guarantee stability.

Is there a way you can look at the residuals in your computational domain, i.e. where are the locations which have the highest residuals? It might have to do with extremely fine/highly unstructured grid.

Are you using a pressure-based or a density-based formulation?

Tell us about the problem you're running, the Reynolds number, the time-step, the grid resolution etc.. and we may have a better shot at convergence.

Re: help, solution blows up!
 

I never did LES but some suggesations : Can you limit the kinetic energy till u have passed trasient stage and then may be remove this limit. amol