[jnattia]

Hi all,
I'm new to the field and have to do some fluid structure interaction analysis. Can someone please point me to where to start learning about this area (books, references, articles, guides, etc...)?

thanks for the help
Joe

[silvia_petkova]

Simulation Help | Simulation Approach | Special Analysis Topics | Fluid-Structure Interaction (FSI) |
Fluid-Structure Interaction (FSI) - Two Way Transfer



This feature enables you to perform a two way fluid structure interaction problem by setting up the static or flexible dynamic portion of the analysis in Simulation that includes defining faces associated with the fluid-structure interface, continuing the analysis in ANSYS CFX-Solve, and viewing the structural results in Simulation.
Overall Workflow for FSI Analysis with Two Way Transfer

1. Perform all steps for a static structural or flexible dynamic analysis in Simulation but do not solve the analysis in Simulation.
   1. Use the Fluid Solid Interface load to identify faces associated with the fluid-structure interface. You can define multiple interfaces, for example when different types of fluids are present in the ANSYS CFX analysis you may want to individually identify the interface between each fluid and the corresponding parts of the structure.
   2. Specify mesh controls, boundary conditions, and solution settings as you normally would.
2. Highlight the Solution object folder and choose Tools> Write ANSYS Input File... from the main menu.
3. Use ANSYS CFX-Pre to set up the CFD analysis as well as the multi-field analysis controls. The multi-field analysis controls define the loads transferred between ANSYS CFX and Simulation as well as solution settings that define the conditions for an acceptable multi-field solution. More details can be found in “Using ANSYS CFX-Pre in ANSYS Multi-field Mode” with in the ANSYS CFX-Pre Help.
4. Use ANSYS CFX-Solver to solve the analysis. The procedure for using the input file created in step 2 is outlined in the ANSYS CFX-Solver Modeling Guide under “Coupling ANSYS CFX to an External Solver”: ANSYS Multi-field Simulations> Pre-Processing> ANSYS Input File Specification. Typically the displacements from a Simulation static or flexible dynamic solution are passed to ANSYS CFX to change the boundary of the CFX mesh. In turn the surface forces at the fluid-structure interface from the CFD solution are transferred to Simulation as a load.
5. The above solution procedure creates an ANSYS results file (.rst file) for the Simulation portion of the analysis. The directory location of the ANSYS results file is determined by the ANSYS CFX-Solve setup. The directory location of the ANSYS solution is determined by the ANSYS CFX-Solve setup.
6. You can associate the above results file with your Simulation model by highlighting the Solution object folder and choosing Tools> Read ANSYS Results File... from the main menu. Browse to the folder that contains the result file (jobname.rst) and the error file (jobname.err). They will have a jobname of ”ANSYS”.
7. Once associated, you can review the results of the two way FSI analysis on your Simulation model.

Good luck!
Silvi

[jnattia]

Thanks Silvia. This would be of great help when using ANSYS. But i'm also looking for leads to better understand the physics and computational issues involved so i can use either commercial or modify open source tools to suit my problem.

thanks
Joe

[silvia_petkova]

Hi, Joe,
I see what you mean, well you can try to find this book: Fluid-structure interaction, Michael Paidoussi.
I think that it is very usefull,
I am working on my PhD and unfortunatelly I have a problem with my model ( Buried pressured pipeline under seismic excitation) and I don't know how to set base seismic excitation.....but I hope I will find the answer soon!
So, good luck to you and best regards!
Silvi

[jnattia]

thanks. seems like a good starting point and seems to cover the type of problems i have to deal with.

[meb]

Good luck!
We have some experience in this field (using Fluent+custom UDF).
Using weak approach some problems can be handled. It is quite easy to have a stable solution for problems when the structure is immersed in an open flow.

We have faced and solved the following problems:

- the motion of a cantilever immersed in a fluid (2d, contact, including VOF)
- reed valve motion (including contact, 2d and 3d; take special care for closing GAP; linear solver with added contact forces)
- the motion of a paper sheet driven by air jets inside a printing machine (including contact 2d and 3d; large displacement explicit FEM; linear material)
- the motion of a probe immersed in a pipeline (3d linear generic triangular shell)
- vibration of a front wing of a F1 car

For confined flow the convergence of the fluid solver is an issue!
We have faced (and solved after a lot of efforts) the following problems:
- blood flow in vessels for a surgical reconstruction
- filling of board containers

Our current research is addressed to:
- use of explicit for a generic shape shell
- use of modal superposition by mesh morphing

Check this link: http://www.torvergata-karting.it/art...eview/72/1/16/

Take also a look here: http://www.rbf-morph.com/

Regards.

[Jade M]

There is a pretty good tutorial at
http://www.edr.no/blogg/ansys_blogge...cfx_re_meshing

[hosseinhgf]

Tank you.this is a good tutorial.but i think whit cfx and icem we can simulation just STRAIGHT moving.i don t no why it is difficult to have remeshing technic in it.fluent and polyflow can do it.immersed solid i think is not very useful .

[meb]

Hi all,
I have finally implemented FSI within RBF Morph.
I have successfully used the approach for two way coupling in steady simulations.
The method works very well for industrial applications: it has been already used for a F1 front wing and for an aircraft wing (for this application I will soon publish the results).
Meanwhile have a look to this simple FSI test for a steady analysis. At each iteration fluent results are exported to FEM using FSI Mapping tool. FEM nodes and displacements are used as RBF centres to set-up a morphing problem that allows to update the (non conformal) CFD surface and CFD fluid mesh.
The GIF is in the gallery of RBF Morph Web Portal (www.rbf-morph.com) reserved to registered users.

[meb]

Have a look to this stunning results.

[afpierlu]

Mr. Biancolini,

The results are truly impressive. I have been looking at the RBF model slightly, but it seems that you can only do simple transformations to the mesh (rotating or translating geometry features). You claim that you have found a way to incorporate FSI into this? I am very interested to see how you solve the structural problem and feed it back to Fluent for further coupled iterations.

My current research ties directly into this. Please email me and we can discuss.

Regards,

Anthony

[meb]

Hello Anthony,
using RBF Morph you can go well beyond simple transformations. Radial Basis functions allows to prescribe a field on a point by point basis.

RBF Morph was born for shape optimisation and you have plenty of tools to reshape surfaces; and if you need a specific new shape you can define it in your CAD and transfer it as a target surface by means of an STL file.

But it has proven to be very useful also for FSI applications. The attached GIFs are related to a real life problem (an aircraft modelled using 14 millions of cells): the response of a wing accounting for structural deformations.

The FEM mesh (coarser than the CFD one) is used to place RBF sources, FEM displacements are used to prescribe the displacement of each source point; the field is then fitted to have a local solution able to move the CFD mesh; local solution is then combined in a general fit problem. In the case of the aircraft a box is used to limit the morphing action, FEM solution is prescribed on the surfaces of the wing, remaining points on the aircraft are fixed.

The same approach can be used to combine FEM modes in a transient solution, I'm currently working on this feature.

Marco

[afpierlu]

I'm glad to see that the functionality is there. I have a couple of questions that may seem dumb but bear with me...I''m only an undergrad researcher.

1. Was the solid material modeled in fluent to account for heat transfer?
2. How does RBF handle morphing the mesh when it has a very fine boundary layer grid spacing?

My problem is relatively simple. I have a forced lobed mixer geometry from the exhaust of a gas turbine, after getting a good hexa mesh in ICEM CFD, I run the CFD in Fluent 6.3.26 and I get pressures and temperatures on the solid material (heat transfer also modeled in FLUENT). I use those results to run FEA on the mixer (using the exact same mesh). I get a deflected model, and solution files for X, Y and Z, which is all fantastic. 1-way FSI achieved!!!

But 1-way is not enough...I need a way to use these three solution files to update my mesh, presumably by using RBF-Morph. Then I can rerun the CFD and look at performance difference of the lobed mixer due to deflections. I understand that this is a transient process and its not as simple as running steady state CFD, then FEA, then back to CFD but I'm just trying to get through one complete iteration. Then I would look at deforming the mesh after every timestep of an unsteady simulation.

Is RBF-Morph configured to be able to solve my problem? Or is additional script writing/ UDF developing required to adapt the RBF-morph model to my needs? I want to make sure it will work before I make the purchase.

Any help would be great and thanks for replying.

Anthony

[meb]

Hello Anthony,
RBF Morph is the missing link for the analysis of your system because it allows to update your CFD model according to your FEA results. And it works also for non conformal meshes...
And I think that you haven't to concern yourself with mesh spacing. The method has been successfully tested on a fine aircraft mesh (mapped hexa generated by ICEM).

Marco

[phanh]

Originally Posted by meb

Hello Anthony,
RBF Morph is the missing link for the analysis of your system because it allows to update your CFD model according to your FEA results. And it works also for non conformal meshes...
And I think that you haven't to concern yourself with mesh spacing. The method has been successfully tested on a fine aircraft mesh (mapped hexa generated by ICEM).

Marco

Hi Marco,

I am interested in FSI field. Your results look very well. I trying to coupled FLUENT with ABAQUS program, but I still faced some problems with respect to FEM solver. May I asked you some question?
1. What kinds of solver are you using for coupling? How can you make surface mapping?
2. I see your post in this forum. You told that we can make very thin shell that can cover all desired surface. Under thin shell element, Don't you think that there are strong effect on FEM solver?
Thank you for your taking your time.

Mike

[meb]

Originally Posted by phanh

Hi Marco,

I am interested in FSI field. Your results look very well. I trying to coupled FLUENT with ABAQUS program, but I still faced some problems with respect to FEM solver. May I asked you some question?
1. What kinds of solver are you using for coupling? How can you make surface mapping?
2. I see your post in this forum. You told that we can make very thin shell that can cover all desired surface. Under thin shell element, Don't you think that there are strong effect on FEM solver?
Thank you for your taking your time.

Mike

Hi Mike,
1. We have successfully used Nastran (a specific interface is embedded in RBF Morph). Exporting to RBF pts format you can work with the FEA solver of your choice. We did it using ANSYS Mechanical (just completed a test on a turbine blade yesterday!). For a civil application we used a 250 nodes BEAM model (results in excel) to update the CFD mesh (5 millions cells) of a canopy.
2. This is a trick, not mandatory, to flag surfaces. It doesn't affect the solution if the membrane is very thin. Remember that this approach was a post-processing workaround in the past to better recover stresses on surfaces (usually most critical).

Thank you for your feedback!

Marco

[phanh]

Originally Posted by meb

Hi Mike,
1. We have successfully used Nastran (a specific interface is embedded in RBF Morph). Exporting to RBF pts format you can work with the FEA solver of your choice. We did it using ANSYS Mechanical (just completed a test on a turbine blade yesterday!). For a civil application we used a 250 nodes BEAM model (results in excel) to update the CFD mesh (5 millions cells) of a canopy.
2. This is a trick, not mandatory, to flag surfaces. It doesn't affect the solution if the membrane is very thin. Remember that this approach was a post-processing workaround in the past to better recover stresses on surfaces (usually most critical).

Thank you for your feedback!

Marco

Hi Marco,

That sounds great! You are right. Using very thin membrane, it is a trick to flag desired mapping surfaces. Could you introduce some reference paper that mentioned this approach?
In NASTRAN, are you using SOL 600? I have seen OpenFSI (by NASTRAN) that is using SOL 600 or 700?

Thank you for your response

Mike

[meb]

Originally Posted by phanh

Hi Marco,

That sounds great! You are right. Using very thin membrane, it is a trick to flag desired mapping surfaces. Could you introduce some reference paper that mentioned this approach?
In NASTRAN, are you using SOL 600? I have seen OpenFSI (by NASTRAN) that is using SOL 600 or 700?

Thank you for your response

Mike

Hello Mike.
For FSI you have basically two different scenarios.

(1)
If your structural model is non-linear you need a two ways coupling. Very high cost because you have to exchange information back and forth. Aeroelasticity is still feasible (usually 5 iterations are enough). I personally did it using Nastran (sol 101 linear); I hear that RBF Morph customers did it with non linear solutions as well (the FSI interface stays the same and RBF Morph can handle even very large deformations).

However a transient analysis could be not affordable. This is why we have decided to implement structural non linear solvers directly inside the CFD solver Fluent (using UDF). Successful applications of this are sails (computation of flying shape) and thin sheet of paper (study of a rectangular sheet guided by air inside a printing device). This approach, whils effective, is no general and today limited to specific industrial projects or researches.

(2)
If the structure is linear (or almost so). You can go for modal superposition approach that is a way faster and can be easily extended to transietn analyses. In this case you have just an extra cost for initial set-up. Once modes are embedded in the CFD solver (we did it using Fluent), you get a flexible CFD mdoel capable to adapt is shape according to actual pressures. Modal approach works very well both for steady and transient analyses. The cost is similar to rigid simulation (however you have to pay for MDM taht increases the cost of a single iteration).

If I remember right, the skin element as post processing tools are explained by Cook http://www.amazon.com/Concepts-Appli.../dp/0471356050

[phanh]

Originally Posted by meb

Hello Mike.
For FSI you have basically two different scenarios.

(1)
If your structural model is non-linear you need a two ways coupling. Very high cost because you have to exchange information back and forth. Aeroelasticity is still feasible (usually 5 iterations are enough). I personally did it using Nastran (sol 101 linear); I hear that RBF Morph customers did it with non linear solutions as well (the FSI interface stays the same and RBF Morph can handle even very large deformations).

However a transient analysis could be not affordable. This is why we have decided to implement structural non linear solvers directly inside the CFD solver Fluent (using UDF). Successful applications of this are sails (computation of flying shape) and thin sheet of paper (study of a rectangular sheet guided by air inside a printing device). This approach, whils effective, is no general and today limited to specific industrial projects or researches.

(2)
If the structure is linear (or almost so). You can go for modal superposition approach that is a way faster and can be easily extended to transietn analyses. In this case you have just an extra cost for initial set-up. Once modes are embedded in the CFD solver (we did it using Fluent), you get a flexible CFD mdoel capable to adapt is shape according to actual pressures. Modal approach works very well both for steady and transient analyses. The cost is similar to rigid simulation (however you have to pay for MDM taht increases the cost of a single iteration).

If I remember right, the skin element as post processing tools are explained by Cook http://www.amazon.com/Concepts-Appli.../dp/0471356050

Hi Marco,

Thank you for your detailed response.
Have a nice day.

Mike

[meb]

Have a look to this video:
http://youtu.be/1iWS9d2DWnI