timestep and mesh size vs sensor sensitivity in experimental data

mbn1454 · November 18, 2011, 23:35

I am trying to validate interaction of waves with a bridge structure. I do mesh and time step sensitivity analysis. How ever it turns out that coarser mesh results with bigger time step is closer to experimental data.
in CFD how do we usually validate our simulation? dont we need to make the mesh finer and finer and time step smaller and smaller till the result don't change?

abdul099 · November 19, 2011, 08:49

That's impossible in the most cases. Just imagine a complex system like a F1 car or an underhood case etc. You can't refine the mesh to get a mesh-independent solution, that would be way too much.
Also keep in mind, CFD usually is good in predicting trends, but it's not easy to predict an absolute number. Therefore it doesn't always make sense to keep refining the mesh just to try to hit the absolute number.

Usually (but not necessarily) a finer mesh and a smaller time step gives better results. But when refining the mesh, you might change unintentionally something else with the mesh size. Maybe the ratio of the biggest to the smallest cell. Or the CFL number or whatever. Some side effect can cause the worse result, although the mesh would produce the better result. Or you just don't iterate long enough.

November 18, 2011, 23:35	timestep and mesh size vs sensor sensitivity in experimental data	#1
mbn1454 New Member Mehrdad Join Date: Apr 2009 Posts: 11 Rep Power: 17	I am trying to validate interaction of waves with a bridge structure. I do mesh and time step sensitivity analysis. How ever it turns out that coarser mesh results with bigger time step is closer to experimental data. in CFD how do we usually validate our simulation? dont we need to make the mesh finer and finer and time step smaller and smaller till the result don't change?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Converting Starccm+ mesh	Ladnam	OpenFOAM	0	September 14, 2011 07:30
Moving Mesh Velocity Issue: Mesh velocity does not equal displacement	Doginal	CFX	2	September 8, 2011 13:02
Timestep with mesh deformation	rbarrett	CFX	4	July 21, 2011 20:25
Convergence moving mesh	lr103476	OpenFOAM Running, Solving & CFD	30	November 19, 2007 15:09
Timestep with different size of mesh	Liu	Main CFD Forum	3	June 6, 2004 05:04

November 19, 2011, 08:49		#2
abdul099 Senior Member Join Date: Oct 2009 Location: Germany Posts: 636 Rep Power: 22	That's impossible in the most cases. Just imagine a complex system like a F1 car or an underhood case etc. You can't refine the mesh to get a mesh-independent solution, that would be way too much. Also keep in mind, CFD usually is good in predicting trends, but it's not easy to predict an absolute number. Therefore it doesn't always make sense to keep refining the mesh just to try to hit the absolute number. Usually (but not necessarily) a finer mesh and a smaller time step gives better results. But when refining the mesh, you might change unintentionally something else with the mesh size. Maybe the ratio of the biggest to the smallest cell. Or the CFL number or whatever. Some side effect can cause the worse result, although the mesh would produce the better result. Or you just don't iterate long enough.