I want to assign free outflow as boundary condition. I know that free outflow means stresses + pressure null. But i cannot specify pressure, I must leave it free to develop. Is it enough if I set only stresses equal to zero?
what model equations and what kind of discretization are you using?
For (compressible) Navier-Stokes equations and finite element discretizations, you might consider the work of Rannacher, who is an expert in this field. Or, for finite volume discretizations and large eddy simulation, surely Thierry Poinsot as indicated above by Heiko is an expert.
I have myself been working a bit in CFD using Navier-Stokes and finite elements. In this context, you might do a quick search for the "do nothing" outflow boundary conditions, i.e. gradient of velocity - pressure times outer normal = 0 at the outflow boundary.
The answer depends on the equation of state of the fluid inside. If it's compressible, but the flow is subsonic, and you've specified outflow velocity, then you've already imposed a lot of conditions on the normal derivative of the pressure. (For a moving boundary problem, of course, you can give the pressure, also.) The tangential components of stress do not present computational problems.