First of all, you should monitor the residuals for the variables you are simulating (continuity, velocity, temperature...)... The level of convergence depends on your application, but less than 10^-3 or 10^-4 is not recommended. Additionally, you can monitor drag/lift/moment force coefficient. After all this, there is the most important thing: validation! You should compare your data with experimental data in order to provide evidences that your simulation reflects the real physical behavior of your application. I suggest you go to cfdonline.com and look for answers according to your specific application

you must compare your CFD results with the open literature .. diff must not exceed 10%..you must also do grid independence test.

thank every one for you answer please I'm used transient some times the solution converged and in an other time step not that is true ?