"The SST model does not always converge to the solution quickly, so the k-epsilon or k-omega models are often solved first to give good initial conditions" [from comsol.com]. This confirms Carlo's answer above.

https://www.comsol.com/blogs/which-turbulence-model-should-choose-cfd-application/

You may try to first obtain convergence using the standard k-epsilon and next shift to SST model. I tried successfully this in the past when I was using the standard k-omega model in Multiphysics.

"The SST model does not always converge to the solution quickly, so the k-epsilon or k-omega models are often solved first to give good initial conditions" [from comsol.com]. This confirms Carlo's answer above.

https://www.comsol.com/blogs/which-turbulence-model-should-choose-cfd-application/

I appreciate your advice , already successfully with k-epsilon and k-omega model. SST ongoing. how we can play around the tolerance and also the meshing size if we have very fine edges . Thanks in advance.

Generally speaking, a CFD solver fails to return a converged solution when the mesh is too coarse to resolve steep scalar gradients near walls. So, you may want to refine the mesh "locally" where spatial variations in the solution (e.g., around edges in your specific case) are expected.

Note that numerical solutions are always approximate ones whose relative error could be adjusted through convergence criterion. So, if the solution is converged, but still there is a large scalar imbalance, you may reduce convergence criterion to improve relative error/tolerance.

Good luck with your simulations! 

The k-ω SST turbulence model uses k-ω model at walls and k-ε model away from walls by using blending function to ensure a smooth transition between the two models. So,  if  k-ω and  k-ε models successfully work, then SST should be work as well. 

Check: turbulence quantities at BCs and under-relaxation factors. 

for more details about SST model>>

Article Prediction of swirling cold flow in a solid-fuel ramjet engi...