i am simulating flow behind bluff body and i got Cd v/s time variation as given in figure. i don't know how to calculate avg Cd because it doesn't reach to steady state.
You need to separate your problem out into two parts: Firstly you need to estimate when your start transient is finished. This is not easy, but for separated flow aerodynamic rules of thumb generally look at 50-200 flow lengths (ie Cylinder diameter/freestream velocity) as values where the flow will definitely be established. Maybe you might be able to get away with 10-20, depending on your numerical scheme.
After that you start your averaging and you need enough samples to get a converged average. Here you can plot the convergence of average just like any other convergence variable, and it should asymptotically approach the correct value.
You need to separate your problem out into two parts: Firstly you need to estimate when your start transient is finished. This is not easy, but for separated flow aerodynamic rules of thumb generally look at 50-200 flow lengths (ie Cylinder diameter/freestream velocity) as values where the flow will definitely be established. Maybe you might be able to get away with 10-20, depending on your numerical scheme.
After that you start your averaging and you need enough samples to get a converged average. Here you can plot the convergence of average just like any other convergence variable, and it should asymptotically approach the correct value.
As Anthony has said, take cd values only after the flow is established. Take average of the remaining points and also take standard deviation of the data. Both should represent the type of variation encountered. If needed, you can even present the skewness to represent the distribution of cd around the average.
If you are modeling a high-Re flow using LES, then I will not be surprised that you do not have nice periodic pattern, you are getting what the solution should give. This is in comparison to RANS model where the averaging greatly smooths out the solution.
Other than this, my advice is to wait longer to make sure you are just having an issue of long transience.