# Finer mesh, worse result?

 Register Blogs Members List Search Today's Posts Mark Forums Read

March 17, 2016, 12:51
Finer mesh, worse result?
#1
New Member

ryan
Join Date: Jun 2011
Location: Jülich,Germany
Posts: 19
Rep Power: 15
Hi, everyone,

I was trying to model a transient Couette flow with FLUENT (cf.Geometry). The walls at Y-direction are defined as translational periodic, the walls at Z-direction are defined as symmetric. Walls at X=H/2 and X=-H/2 started to move at t=0 with U=U_wall. The flow is laminar and calculation was performed for only one time step.

I found that with finer mesh, it is more difficult to obtain the converged result. Take the default convergence criterion 1e-3 for example (cf. Result), velocity profile along the +X-axis gets steeper with finer mesh( higher H/delta_X value), meanwhile it takes more iterations. However, with smaller convergence criterion (depending on how fine is the mesh), the result of finer mesh will eventually approach to the result of coarser mesh.

This blows my mind, does anyone knows why? Does it mean that I have to choose a specific convergence criterion for each mesh size individually?

thanks!

Ryan
Attached Images
 Geometry.jpg (15.0 KB, 38 views) Result.jpg (39.2 KB, 53 views)

 March 17, 2016, 14:52 #2 Senior Member   Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,685 Rep Power: 66 The velocity profile tends to be steeper on finer grids because there's less numerical dissipation/damping. Convergence doesn't mean the solution is accurate. All else being equal (same problem, same initial guess, all settings equal), increasing the mesh resolution (finer mesh) will take more iterations to converge to the same solution. It's a result of the implicit discretization schemes which results in a sparse linear system. That is, changes in cell properties only affect their immediate neighbors. Hence, it takes many many iterations for adjustments of the solution to slowly propagate cell by cell throughout the entire domain to reduce the global error. This behavior can be overcome/accelerated by using a multigrid algorithm to improve the speed at which global errors are reduced, but when you go to a finer grid (you need to make the multigrid method more aggressive). If you don't change these settings in the multigrid solver, your finer grid would still take more iterations (because the coarse grid has more aggressive settings relative to the finer grid). However, I don't recommend changing the AMG parameters unless you are an expert. Merely, this is to explain why the behavior is normal. Also, I recommend defining convergence criteria based on solution values (like a sane person would) rather than residuals. Residuals do not measure convergence. Residuals are good stopping criteria, but are terribly convergence criteria. Last edited by LuckyTran; March 18, 2016 at 09:42.

 March 18, 2016, 03:48 #3 New Member   ryan Join Date: Jun 2011 Location: Jülich,Germany Posts: 19 Rep Power: 15 Hi, dude, Thanks for your reply! I will try to improve my calculation by your suggestion. cheers!

 Tags couette flow, fluent, mesh dependency