FSI Analysis (Opristruct - Acusolve) - 3D analysis with structural shell mesh

I would like to perform a fluid-structure interaction (FSI) analysis on a thin metal plate subjected to fluid flow. Since the plate has a thin geometry, I plan to model it using a 2D shell mesh to reduce computational complexity, especially in terms of resolving the plate's thickness. My goal is to maintain the fluid domain as a 3D volume, ensuring that the plate is represented by a 3D hole within the fluid field to preserve the correct interaction between the flow and the plate, as a full 3D model would.

The reasoning behind this approach is to avoid the need for excessively small elements in modeling the plate's thickness, which would otherwise significantly increase computational demands. However, when attempting this method, I encountered the following error during the first increment:

*** ERROR # 5905 ***
Severe element distortion detected

Could you provide insights into whether this approach is feasible?

Find more posts tagged with

AcuSolve

English

HyperMesh

HyperMesh CFD

OptiStruct

v2022.3

Accepted answers

All comments

acupro

Can you please attach the AcuSolve input (.inp) and Log files?

You may want to add a multiplier to the EXTERNAL_CODE command to ramp the passing of the fluid forces over the first several time steps, to allow the flow to converge nicely before passing full forces. This tutorial includes such a multiplier:

https://help.altair.com/hwcfdsolvers/acusolve/topics/tutorials/acu/acu_5403_intro_cfd_r.htm
https://help.altair.com/hwcfdsolvers/acusolve/topics/tutorials/acu/acu_5403_intro_sl_r.htm

THEODOROS BALSIS_22459

I am working on the same model that I mentioned in my previous question here: FSI Analysis (Opristruct - Acusolve) - Error.

I am attaching the relevant files in this reply for your reference.

Additionally, I applied the multiplier, but it doesn't appear to have a significant impact on the analysis.

Test_model_benchmark_ACU_2D_2.1.Log

Test_model_benchmark_ACU_2D_2.inp

Test_model_benchmark_OPT_2D.fem

acupro

I am working on the same model that I mentioned in my previous question here: FSI Analysis (Opristruct - Acusolve) - Error.

I am attaching the relevant files in this reply for your reference.

Additionally, I applied the multiplier, but it doesn't appear to have a significant impact on the analysis.

Try mesh-displacement-type = tied rather than slip on the AcuSolve side External-Code-Surface "Beam". Also try increasing the multiplier steps to 50 instead of 20.

The AcuSolve mesh seems very coarse with only 15 to 16 thousand nodes. Have you tried refining the mesh to better resolve the flow and thus the forces passed to OptiStruct?

acupro

Try mesh-displacement-type = tied rather than slip on the AcuSolve side External-Code-Surface "Beam". Also try increasing the multiplier steps to 50 instead of 20.

The AcuSolve mesh seems very coarse with only 15 to 16 thousand nodes. Have you tried refining the mesh to better resolve the flow and thus the forces passed to OptiStruct?

I believe in the other ticket you had mentioned the DC-FSI tutorial:

https://help.altair.com/hwcfdsolvers/acusolve/topics/tutorials/acu/acu_5403_intro_cfd_r.htm

I'm attaching here a shell-only OptiStruct model that runs with the AcuSolve deck from that tutorial. Towards the end, there's some distortion error from OS - but in general it does work. I'm not an expert in OS - but the person that put this together mentioned also creating 'surface' definitions on the edges, to account for how the 'thickness' surfaces in AcuSolve connect to those edges in OS.

Maybe this helps - for comparison at least

pshell_single_surf.fem

THEODOROS BALSIS_22459

I believe in the other ticket you had mentioned the DC-FSI tutorial:

https://help.altair.com/hwcfdsolvers/acusolve/topics/tutorials/acu/acu_5403_intro_cfd_r.htm

I'm attaching here a shell-only OptiStruct model that runs with the AcuSolve deck from that tutorial. Towards the end, there's some distortion error from OS - but in general it does work. I'm not an expert in OS - but the person that put this together mentioned also creating 'surface' definitions on the edges, to account for how the 'thickness' surfaces in AcuSolve connect to those edges in OS.

Maybe this helps - for comparison at least

Thank you so much for your help, I really appreciate it!

Just to keep you in the loop, I’ve implemented all the suggestions from your previous reply. Specifically, I refined the mesh for both the structural and CFD models, with over 1 million nodes for the CFD and a suitably small element size for both models. I also adjusted the step multiplier from 20 to 50 and switched the mesh-displacement type from "tied" to "slip." Unfortunately, the analysis still diverges, stopping after a few iterations due to severe element distortion.

Additionally, I wanted to ask for your thoughts on the time steps I'm using: 1e-03 for the structural model and 1e-06 for the CFD model. Do you think these are suitable for this application? I’d really appreciate your feedback. I'll also try the .fem model you provided as a reference and will get back to you with any new results.

acupro

Thank you so much for your help, I really appreciate it!

Just to keep you in the loop, I’ve implemented all the suggestions from your previous reply. Specifically, I refined the mesh for both the structural and CFD models, with over 1 million nodes for the CFD and a suitably small element size for both models. I also adjusted the step multiplier from 20 to 50 and switched the mesh-displacement type from "tied" to "slip." Unfortunately, the analysis still diverges, stopping after a few iterations due to severe element distortion.

Additionally, I wanted to ask for your thoughts on the time steps I'm using: 1e-03 for the structural model and 1e-06 for the CFD model. Do you think these are suitable for this application? I’d really appreciate your feedback. I'll also try the .fem model you provided as a reference and will get back to you with any new results.

I believe the time step size needs to be the same for both the OS and AcuSolve models. The time increment 1.e-6 seems to be quite small. What frequencies are you trying to capture? You would want 20 to 50 time steps per cycle, for something that is generally repeating.