Check the tr(Uprime2Mean)/(tr(Uprime2Mean)+kSGS)., is it close to one in the domain? The period of calculation would be where your UMean value achieves its statistically steady behaviour... Either symmetric flow, or slow oscillation ( you can use the kOmega sim with time averaging to get an idea, or even better, calculate the integral time scale as an indicator, and then in your LES sim, do your averaging over a few time periods)...