Difference in directional displacement for the same mesh and model settings

  • 39 Views
  • Last Post 31 January 2020
Tayyaba posted this 14 January 2020

Hi all,

I hope you are doing well,

I am doing FSI analysis of a thin flap, I repeated my simulations for the same geometry,mesh, boundary conditions and Fluent model but I found a difference in directional deformation of the flap although bending of the flap is the same. I am confused why difference is occurring in the directional deformation although everything is same as the previous simulations.

Can anybody help in this regard?

Tayyaba

Order By: Standard | Newest | Votes
rwoolhou posted this 14 January 2020

Please post some images. 

Tayyaba posted this 15 January 2020

Here I am attaching the total deformation and directional deformation of the flap. FSI 1 is the initial simulation I have done for 10sec and FSI2 is the repeated

simulation just completed for 6sec. but there is a clear difference in directional deformation. I do not know why is it so , therefore a kind response is needed.

 

Regards

Tayyaba

 

 

l

rwoolhou posted this 15 January 2020

That looks very similar to what I'd expect for differing levels of convergence or solver settings (ie first to second order). Check the solution has converged properly and run again with a refined mesh & smaller time steps (plus more iterations per step). 

Tayyaba posted this 15 January 2020

Thank you very much for your reply, But I have done the both simulations for the second order. Do you mean to change the time step and mesh for the second simulation?But I need results at the same time step and mesh as i have done but getting difference in z direction deformation which shouldn't be happen.

Regards

Tayyaba

Tayyaba posted this 15 January 2020

I just realized that the difference is in coupling iteration for the previous and this case,initially it was 10 and now it is5. So I think this is a factor which is changing the results.

Tayyaba

rwoolhou posted this 15 January 2020

So you're not updating tools as frequently? Means the mesh/flow will be slightly different, and in a transient model the error will accumulate. 

Tayyaba posted this 17 January 2020

Thank you very much for your reply, could you please explain your point a little bit further, I didn't understand that.

 

Regards

Tayyaba

rwoolhou posted this 17 January 2020

If we run a transient solution and slightly fail to converge a step we get an error. As time progresses the error slowly builds until it's large. At no point do we necessarily fail the convergence check, because each step is nearly OK. 

Simple equivalent. You carry a sack of grain down the road. Every 10 paces I add an extra grain. You won't notice the difference but after a mile or so the sack will be far heavier than it was at the beginning. 

Tayyaba posted this 20 January 2020

Thank you very much for your explanation. Yes you are right, but in my transient simulation, solution is converging at each coupling iteration. And my understanding is because of difference in coupling iteration, there is a difference between directional deformation of the two simulations performed. Is this a matter?

and sorry but I did not understand the last statement of you "At no point do we necessarily fail the convergence check, because each step is nearly OK"

Regards

Tayyaba

 

rwoolhou posted this 20 January 2020

You've not failed the convergence checks, but have you fully converged the calculation?

Effectively by adding more sub steps or by reducing the time step we force the solver to calculate a more accurate solution: similar to using a finer mesh for a mesh dependency study. 

Tayyaba posted this 21 January 2020

Thank you very much for your explanation,

did you mean the following convergence check for system coupling?

Regards

Tayyaba

rwoolhou posted this 21 January 2020

I don't do much (any) FSI as most things I deal with break if bent very far!  You've tried (I think) 5 & 10 substeps so try 20 and see how the result comes out. 

Tayyaba posted this 22 January 2020

Yes thanks for your valuable replies, I am working with 15 and 20 coupling iterations and then see how it works.

Regards

Tayyaba

abenhadj posted this 22 January 2020

Also monitor the forces applied on the flap in Fluent along the number of iterations.

Best regards, Amine

Tayyaba posted this 23 January 2020

Yes, thank you for your reply. I am working with drag coefficient also.

Tayyaba

Tayyaba posted this 30 January 2020

I hope you are fine, my question is about the coupling iteration, if I set it as 20 then sometimes its take 2  , sometimes 5 and sometimes 20 to complete a coupling steps. I mean its not same in each coupling step and also it is not same for the same two cases I run for same coupling iterations lets suppose 20.Some times its vary from step to step although the cases are identical.

 

Regards

Tayyaba

abenhadj posted this 30 January 2020

That it does not do always 20 that is ok but same case with same settings and same initialization same machine and it differs that is strange but also might mean that on Amg side still not deeply converged. Sorry to say now that at this stage no other help can be provided in this regarding on this platform.

Best regards, Amine

Tayyaba posted this 31 January 2020

Thank you very much for your reply.

 

Regards

Tayyaba

Close