Transient 1-way FSI with transient structural and external data modules

  • Last Post 21 December 2018
Siva Teja Golla posted this 19 December 2018


      I am trying to do Transient 1-way FSI of Sloshing in ANSYS 18.

- For this, I am using Fluent to solve for the fluid domain. I have solved for 0.525 sec with a time-step of 0.005 sec. I have written the data      files for every 5 time-steps i.e., 0.025 sec.

- Then using the CFX-macros 'CFX_FSI_IOWrite.cse", "CFX_FSI_IORead.cse" and "FSI_Transient_Export_Surf.cse", I have retrieved the      fluid pressures that are acting on the fluid and tank interface at all nodes, at all the time steps. This data is stored in the form of ".sfe" and    ".csv" format.

- Then I have added the Transient structural module to the project to identify the response of the tank walls for the fluid pressure acting        upon them during sloshing. I am interested in accelerations of the tank wall. These accelerations I cannot identify through static structural    analysis as shown in various tutorials.

- I used the External data module to give the fluid pressure loads at different time steps as input to the Transient structural module. I        have followed the exact procedure as mentioned in ANSYS tutorials to add the data to the External data module and connecting it to the        Transient structural module.




Analysis settings

- I wanted to know the variation of tank wall acceleration for the total time i.e, 0.525 sec.  The fluid pressure acting on the tank walls varies      harmonically w.r.t time. But the accelerations are decaying to '0' w.r.t time, which should not be. Moreover, the magnitudes of acceleration are        also coming to be very less, order 10^-3. Expected order is between 10^0 and10^1.

- I am thinking that the transient structural module is taking only the load at first time-step and solving for the entire time duration. Though it      is showing that the load at all the time steps are loaded.

- I am encountering the following error while the simulation is going on..


I need help in resolving this issue. 

Thanks in advance

 PS: I have tried the Transient-FSI extension also and the scenario is the same with that too....







Order By: Standard | Newest | Votes
RD2016 posted this 20 December 2018

First off, thank you for including all the pics. They really help paint the picture of your setup, and you wouldn't believe how rare it is for someone to be this detailed.

I have a little experience in FSI and the External Data module myself, and with that being said I'm a little confused by your setup. You exported your pressures and nodal points correctly, but you did not take into consideration how the nodal points change with respect to time. Again, I don't know the specifics of your setup so I might be out of line with this, but let me try to walk through it with you.

Let's say this box is your fluid domain.

So, at time step 1, the pressure acting on the right hand-side might be greater than the initial pressure. And then at time step 2, it might be less. As you say, it oscillates. Keep this in mind for later.

Time step 1. Time Step 2.

Since this is purely a fluids-based simulation, I'm assuming that the walls do not move. This is an important distinction, because if the walls do move, then the resulting pressure force would change to adjust to the altered geometry. As they do not, your nodal points remain fixed in space, with the pressure varying as a function of time. I'm also assuming that the harmonic pressure attenuates. If it remains constant, I'd need to think about this a little more.

Now, you feed this information to Mechanical, and attempt to perform a 1-way FSI. Allow me to say that, while I certainly don't know everything about FSI, I'm not quite sure you can do this. I mean, you can physically do this, but I don't think you will get a right solution. Typically, 1-way FSI is performed by transferring displacement, calculated by Mechanical, to CFX with the resulting force on the walls being neglected (I am ignoring the vast majority of 1-way FSI to focus on this one case). If you attempt to go the other way with the simulation, I think you will run into serious problems. Just as CFX will calculate the resulting forces from a moving boundary, Mechanical will calculate the resulting displacement from a force.

Back to the box diagram. Within Mechanical, during the first time step, the increased force on the wall will cause the box to expand. This means that the nodal points will move from their initial positions. This answers your last question: what is going on with the mapping? Essentially, at the second time step, Mechanical is having difficulty finding the node that was previously in the center of the box, because it moved. To fix this and continue mapping, it adjusts its mapping algorithm, which is why you see the message of "The program controlled pinball grew to more than 10%...". It's trying to find the node that the External Data module says should be there. Eventually, it will find it and continue solving.

However, this introduces a flaw in your simulation. Think back to the CFX sim: at time step 2, the pressures were calculated with respect to a specific fluid domain. Now, that domain is changed in Mechanical. So, the forces experienced at time step 2 are no longer representative of your CFX simulation, meaning that the resulting displacements in Mechanical are not correct. Explained out a bit: CFX said that there is a lower pressure response on the walls than at initialization (in our example) in the second time step. In Mechanical, however, this lower pressure might not be enough to cause the shape to change much, so it will stay bulged out. What does this mean for your acceleration? It basically decreases/goes to 0. Not to mention the fact that, were this altered geometry imported into CFX you would experience drastically different results from what you previously calculated. In short, you might be doing everything correctly, but just set up your problem in such a way as to not give you the desired output.

This simulation might be better suited for a two-way FSI. Have you considered that?


Siva Teja Golla posted this 21 December 2018

Hi RD2016,

             Thank you very much for your detailed response. Your explanation really helped us in understanding where we are going wrong. I will try to go with 2-way FSI and get back to you soon.


                                                                      Thank you once again.