My suggestion will deviate slightly from that of Aris Iturbe. In context to vortex shedding simulations, the shedding phenomenon itself might not occur at all if turbulence boundary conditions (k and epsilon) are inappropriately applied....I am speaking here from experience.
I would hence suggest that you begin your simulations from a laminar model and then move on to two-equation turbulence models. Whilst using a turbulence model, care needs to be taken in specifying values of turbulence quantities at inlets as they are susceptible to excessive turbulent viscosity generation.
The added viscosity would induce strong damping in the wake of the cylinder preventing amplification of vertical asymmetries (in momentum) whose growth is crucial for shedding to initiate.
My answer might have added to the confusion....but in my experience, both laminar and LES (for 3-D) models are better suited at capturing the shedding phenomenon in unsteady simulations.
thank you all for helping me in this project. I use the k-epsilon model in this project and vortex shedding occurred in downstream flow but it called wrong values of sound pressure level (SPL) (my goal) . because it has contrast with experimental reports!
Given the values of the velocity, your Re number is O(10^4) and the flow is turbulent.
Using RANS formulation you cannot observe the unsteady shedding. Assuming that DNS is too computationlly expensive, I suggest to adopt the LES formulation.