Stream is sticking to the wall
 

Hello,
I have a problem with swirl coal burners calculations. Burner design provides 3 air steams. Main air stream is swirled.

I am using steady state.
I have tried various methods to converge these burners:
- different mesh (coarse, fine, with cells directed along and across the stream, with and without boundary layer)
- with and without: energy, combustion
- calculations start with all 3 streams, calculations start with 1 stream and after obtain convergent swirl stream I added second one and third one.
- different wall functions for k-epsilon
- another turbulence models like: RSM
- first order/ second order discretization methods, and pseudo transient options.
Unfortunately, results are almost always the same.
At the beginning streams are formed (Fig. 1)

Later stream starts to oscillate (Fig. 2)
At the end stream is sticking to the wall (Fig. 3)
I am also attaching animated .gif file (simulation inside combustion chamber) - http://imgflip.com/i/499sp


If you have and ideas how to fix this problem, or what should I check, please help.

Best regards,
Pawel