I have calculated the laminar flow around a circular, two-dimensional cylinder.

I got different reults for the same problem when different initial conditions were used.

The max. residuum were lower than 1 * 10^{-4} in all simulations.

Could someone tell me the reason ? Thanks !

Are these results grid-independent? Problem with cylinder is generally speaking one of the most difficult. With best regards

There are three possibilities1). there is something wrong with your program. it is fairly common, everybody makes a mistake from time to time, (2). the solution is not fully converged, (3). the boundary conditions have been changed. So, for the two solutions, check whether the boundary conditions are identical or not, before and after the iteration. run more iterations to see whether the solution remains the same.

yes I have similar experiences with flows in 3D ducts. I looks like the solution has not converged fully. A good indicator is to check your varible history for the scalars at some point in flow field. If these do not change over a number of iterations then the solution hase reach convergence.

First, it's not clear, are you computing a transient or steady-state solution. One have to remember, that for large Re a steady-state solution may not exist at all. For a time-dependent problem it's not surprising to get different results, but asymptotically they should look the same.

Are you talking about different boundary conditions (BC)or different initial conditions(IC)? Whereas different BC will give you different solution, different IC (i.e. merely a first guess) should (in principle) converge to a unique solution for a set of BC. However this may not happen due to many reasons errors in the program etc.. But 2 main numerical causes are if Re is very high (there will be more than 1 solutions) and if you are using seggregated solver soution is highly influenced by the initial conditions. Hope this helps!