steady state fsi-convergence problem

mortazavi · July 21, 2016, 00:05

Hi,
I am simulating steady-state 3D fsi on a NACA66 hydrofoil in the 5m/s flow of water. A steady state flow-only result is provided as the initial condition. The problem is that errors doesnot converge below 1e-4.
even for the flow-only simulation i had convergence issues. So i provided a coarse mesh result as the initial condition for the ine mesh, and then started with a very small time scale (1e-6) and edited the run in progress by increasing the time scale.
this strategy now does not work for fsi simulation.
Any idea please?
where is my mistake?
thanks

ghorrocks · July 21, 2016, 00:37

Have you read the FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Mesh quality is almost certainly a major factor - it always is.

mortazavi · July 21, 2016, 09:58

Thanks for your reply,
Of course I have performed lots of tests on mesh. Convergence has no problem with a coarse mesh. but all the issues are with a fine mesh.
That is the reason a course-mesh-result is set as the initial solution for the fine-mesh simulation in flow-olnly part to converge. For low AOAs like 0-2degree it converges fast.
thanks

ghorrocks · July 21, 2016, 19:44

It is common for a coarse mesh to converge quickly but convergence gets harder as the mesh gets finer. There is less dissipation in a finer mesh so the same level of mesh quality becomes more of an issue.

So as you refine the mesh you need to improve mesh quality to keep the solution stable.

mortazavi · July 27, 2016, 12:06

Thank you so much for your helpful answers.
Besides the mesh quality, on which I have worked a lot, timescale has also a key role. I am now trying to investigate its effects on my fsi simulation. previous ly I have done "timescale convergence study" on the fluid-only part.
Now do you think if fsi problems also follow the same way as flow-only?
I mean small timesacles should be used then gradually change to bigger one to have faster convergence?
thanks

ghorrocks · July 27, 2016, 18:26

Yes, the same principles apply.

mortazavi · August 2, 2016, 01:25

Hi Glenn,

I found out that my simulation does not converge because it has to be simulated as a transient problem. for the transient fsi simulation the cfx part, the solid part and the interface load converges in each time step and this must be true. however, my question: why does the lift or displacement at the interface (user points) does not converge to a constant value within a time step? what parameter should be modified to achieve that?
thanks,

July 21, 2016, 00:05	steady state fsi-convergence problem	#1
mortazavi Member SMN Join Date: Jun 2009 Location: CANADA Posts: 71 Rep Power: 17	Hi, I am simulating steady-state 3D fsi on a NACA66 hydrofoil in the 5m/s flow of water. A steady state flow-only result is provided as the initial condition. The problem is that errors doesnot converge below 1e-4. even for the flow-only simulation i had convergence issues. So i provided a coarse mesh result as the initial condition for the ine mesh, and then started with a very small time scale (1e-6) and edited the run in progress by increasing the time scale. this strategy now does not work for fsi simulation. Any idea please? where is my mistake? thanks

July 21, 2016, 19:44		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,890 Rep Power: 144	It is common for a coarse mesh to converge quickly but convergence gets harder as the mesh gets finer. There is less dissipation in a finer mesh so the same level of mesh quality becomes more of an issue. So as you refine the mesh you need to improve mesh quality to keep the solution stable. mortazavi likes this.

July 27, 2016, 18:26		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,890 Rep Power: 144	Yes, the same principles apply. mortazavi likes this.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Steady state VOF problem	Fatima.k	FLUENT	0	February 4, 2014 09:24
Time step dependence of convergence behavior of steady state simulations in CFX	Chander	Main CFD Forum	5	December 23, 2013 06:31
FSI convergence problem	YY	CFX	0	April 13, 2008 11:54
Steady State Natural Convection problem with PISO	jerry	Siemens	3	August 12, 2002 07:47
Psedo-time steps in Steady state problem	Blackdraft	FLUENT	0	July 14, 2002 06:23

July 21, 2016, 00:37		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,890 Rep Power: 144	Have you read the FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria Mesh quality is almost certainly a major factor - it always is.

July 21, 2016, 09:58		#3
mortazavi Member SMN Join Date: Jun 2009 Location: CANADA Posts: 71 Rep Power: 17	Thanks for your reply, Of course I have performed lots of tests on mesh. Convergence has no problem with a coarse mesh. but all the issues are with a fine mesh. That is the reason a course-mesh-result is set as the initial solution for the fine-mesh simulation in flow-olnly part to converge. For low AOAs like 0-2degree it converges fast. thanks

July 27, 2016, 12:06		#5
mortazavi Member SMN Join Date: Jun 2009 Location: CANADA Posts: 71 Rep Power: 17	Thank you so much for your helpful answers. Besides the mesh quality, on which I have worked a lot, timescale has also a key role. I am now trying to investigate its effects on my fsi simulation. previous ly I have done "timescale convergence study" on the fluid-only part. Now do you think if fsi problems also follow the same way as flow-only? I mean small timesacles should be used then gradually change to bigger one to have faster convergence? thanks

August 2, 2016, 01:25		#7
mortazavi Member SMN Join Date: Jun 2009 Location: CANADA Posts: 71 Rep Power: 17	Hi Glenn, I found out that my simulation does not converge because it has to be simulated as a transient problem. for the transient fsi simulation the cfx part, the solid part and the interface load converges in each time step and this must be true. however, my question: why does the lift or displacement at the interface (user points) does not converge to a constant value within a time step? what parameter should be modified to achieve that? thanks,