spinning disk reactor flow behavior using LES
 

I am trying to simulate the flow behavior between two spinning disks using 2d LES model. In the case of one spinning disk and one fixed disk, after 5 seconds of simulation (flow time) convergence is obtained and my desired profiles do not fluctuate. However, in the case that both disks are spinning after 40 seconds of simulation, although all residuals are converged, desired profiles fluctuate severely.
Does anyone know how to solve overcome these fluctuations?
Thanks.