Re. Sylvian's response:

http://www.cfd-online.com/Forum/main.cgi?read=21155

"I do believe this is due to the fact that RANS is not derived to solve unsteady flows."

So, is this true? I've seen many application of RANS to unsteady flows, are they all wrong??

Fred.

>> So, is this true?

It depends on how you are performing your average.

>> I've seen many application of RANS to unsteady flows, are they all wrong??

No, not all but I suspect many are wrong.

If the scales of the energy containing eddies are substantially distinct from the unsteadiness then it is reasonable to decompose the solution into a slowly evolving mean flow field with a superimposed rapidly fluctuating component handled by an eddy viscosity based turbulence model. If the scales are not distinct then you need to use an LES type approach.

I would have thought that it's more an issue with the energy contained within periodically moving fluid vs. the energy contained within statistically turbulent (aperiodic) motion.

I wonder how this all relates to coherent strucutres in turbulent flow and the preservation of transient flow features through transition into fully turbulent regimes? This all seems similar to the route to chaos behaviour exhibited by Xn+1=LamdaXn(1-Xn) as Lambda increases. Anyone still looking at non-linear dynamics (chaos) to enable to a better understanding of turbulence?

Fred.

>> I would have thought that it's more an issue with the energy contained within periodically moving fluid vs. the energy contained within statistically turbulent (aperiodic) motion.

In what way?. At high Reynolds numbers, the energy in the coherent motion is going to end up mainly in the turbulence at some point but the simulation does see enhanced mean gradients to account for it. However, I believe the scales are going to control the rate of energy transfer and determine whether the assumptions made in the turbulence model are reasonably maintained or not.

(A) If the scales are adequately separated, the coherent vorticity in the shedding will interact with the turbulence in a manner similar to the mean flow. The vorticity will remain unidirectional, largely coherent and dissipate steadily. Things seem reasonable?

(B) If the scales are not adequately separated then the coherent motion will start to take part directly in the energy cascade of the turbulence. The interaction (energy transfer) will be stronger, the orientation of the coherence motion will break down swiftly as vortex stretching shifts the energy to other orientations. It seems unlikely to me that whatever assumptions the RANS turbulence model is based on will hold up under this extra injection of energy into the energy containing scales. Seems unreasonable to perform quasi-steady turbulent RANS predictions?