numerical collapse

I’m implementing a D3Q19 model, with bounce back at walls and wrap around in the direction of flow using body force forcing.

My problem is, that after running a couple(about 10000) simulations, i have found that the highest reynold’s number at which the velocity profile converged is very small- about 300(cylindrical pipe flow). At higher reynolds no., and a lot of lower ones as well, the parabola collapsed into random fluctuations.

My question:
What (if any) among these: tao,body force, lattice dimensions, reynolds no., max velocity or perhaps some combination, is the parameter which can definitively predict numerical collapse?(from my data i’m certain it is not Re, as i was led to believe)
Is any research available on the subject?


You need to be explicit enough. In varying your Reynolds number, did you vary your resolution? What was your Mach number as your maximum velocity in the domain should not be greater than 0.1.

even though (strangely) i do not have the details of the parameter variations with me right now(it would take some work to access them, on a different computer), what i really wanted to ask doesn’t depend on them.

All I wish to know is why, and in what parameter ranges does numerical collapse happen. If this is widely known, a short explanation would be great, otherwise perhaps a paper referral?

Thanks a lot