Infinitely Long Bearing - Linear Wedge

Adam Riese · July 7, 2011, 05:52

Dear all,

I try to simulate some principal Bearing Problems and am quite new to CFX.
As a test case I chose the Infinitely Linear Wedge, i.e. a moving plate at the bottom and an inclined upper plate with an inlet 2 microns high and an outlet 1 micron high. For this case an analytical solution to the reynolds equation exisits, resulting in a maximum pressure of 50 MPa.
I modeled a thin 3D structure 0.1 micron thick and meshed it with hex elements.
My problem is that the solution does not fit to the analytical one. I started a parameter study. The only influence I recognized is the fluid timescale control. Choosing an automatic timescale with a specified lenght scale of 1 micron and a timescale factor of 0.2 produced the best results (37 MPa, velocity profile qualitatively correct) . Further reducing the length scale or timescale factor results in an error. I tried to define a physical timescale but also with very small values the results were insufficient.
I recognized that it is important to have a slow convergence rate as then the results get closer to the analytical solution. But the convergence is still decreasing significantly when reaching the RMS residual target of 1e-6.
Has anyone some hints for me in solving such a thin film flow problem?
Thanks in Advance.

ghorrocks · July 7, 2011, 19:03

http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

Adam Riese · July 8, 2011, 05:02

Thanks for the link. But I already tried the points mentioned there.
I now additionally raised the mesh density in flow direction much more and then the solution gets a bit better. Is it possible that CFX can only cope with cube like elements? The other meshes have long element sides in flow direction. Is Fluent the better choice for such elements?

ghorrocks · July 8, 2011, 07:01

It is likely your domain has very high aspect ratio elements as the wedge is so thin. In this case, yes, hexahedral elements will be significantly better than tets. Also double precision might help.

I know of no difference between CFX and Fluent in being able to handle high aspect ratio elements.

Adam Riese · July 8, 2011, 07:22

The mesh contains only hex-Elements (20 over the gap, 200 or more in flow direction and 1 element in symmetry direction). I tried double precission showing nearly no effect.
Thanks for your help.

July 7, 2011, 05:52	Infinitely Long Bearing - Linear Wedge	#1
Adam Riese New Member Join Date: Jul 2011 Posts: 3 Rep Power: 14	Dear all, I try to simulate some principal Bearing Problems and am quite new to CFX. As a test case I chose the Infinitely Linear Wedge, i.e. a moving plate at the bottom and an inclined upper plate with an inlet 2 microns high and an outlet 1 micron high. For this case an analytical solution to the reynolds equation exisits, resulting in a maximum pressure of 50 MPa. I modeled a thin 3D structure 0.1 micron thick and meshed it with hex elements. My problem is that the solution does not fit to the analytical one. I started a parameter study. The only influence I recognized is the fluid timescale control. Choosing an automatic timescale with a specified lenght scale of 1 micron and a timescale factor of 0.2 produced the best results (37 MPa, velocity profile qualitatively correct) . Further reducing the length scale or timescale factor results in an error. I tried to define a physical timescale but also with very small values the results were insufficient. I recognized that it is important to have a slow convergence rate as then the results get closer to the analytical solution. But the convergence is still decreasing significantly when reaching the RMS residual target of 1e-6. Has anyone some hints for me in solving such a thin film flow problem? Thanks in Advance.

July 8, 2011, 07:22		#5
Adam Riese New Member Join Date: Jul 2011 Posts: 3 Rep Power: 14	The mesh contains only hex-Elements (20 over the gap, 200 or more in flow direction and 1 element in symmetry direction). I tried double precission showing nearly no effect. Thanks for your help. Last edited by Adam Riese; July 8, 2011 at 07:44.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
solution diverges when linear upwind interpolation scheme is used	subash	OpenFOAM	0	May 29, 2010 01:23
Problems in compiling paraview in Suse 10.3 platform	chiven	OpenFOAM Installation	3	December 1, 2009 07:21
IdeasUnvToFoam Bug amp Fix	benru	OpenFOAM Bugs	42	November 13, 2009 07:59
OpenFOAM-1.5-dev svn revision 1438: libOpenFOAM does not compile in SP	4xF	OpenFOAM Bugs	3	October 16, 2009 05:35
Missing math.h header	Travis	FLUENT	4	January 15, 2009 11:48

July 7, 2011, 19:03		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

July 8, 2011, 05:02		#3
Adam Riese New Member Join Date: Jul 2011 Posts: 3 Rep Power: 14	Thanks for the link. But I already tried the points mentioned there. I now additionally raised the mesh density in flow direction much more and then the solution gets a bit better. Is it possible that CFX can only cope with cube like elements? The other meshes have long element sides in flow direction. Is Fluent the better choice for such elements?

July 8, 2011, 07:01		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	It is likely your domain has very high aspect ratio elements as the wedge is so thin. In this case, yes, hexahedral elements will be significantly better than tets. Also double precision might help. I know of no difference between CFX and Fluent in being able to handle high aspect ratio elements.