Need an Expert to Validate a NL Static Model with Multi-Cyclic B.C's

hooshsim

Attached is a model file is for a 9-bladed fan assembly with rivet connection. Due to cyclic B.C,. only one segment of the fan is modeled.

I need an expert in Mecway (CCX) to review the attached model and validate the setup to make sure multi cyclic B.C's are applied correctly. Please note that solution is cleared to make the file size small for uploading...the run time for converged solution is about 5-6 minutes.

Thanks in advance.

Hooshsim

Sergio

Hi, in your model (on the righ side) you forgot to define an element set where the cyclic simmetry is applied. This leads to some displacement out of the simmetry plane (see the second picture) and slighty different results.

By the way, have someone noted that in CCX 2.15 the CHECK parameter in CYCLIC SYMMETRY MODEL card is not working? I have tried with CHECK=YES, and CHECK only but it doesn´t seem to work.

Atached is the new fixed model.

Best regards

hooshsim

Hi Sergio,

Thank you for your review. I downloaded your model and ran it over without any changes in the file and got the exact results as my original model (figures on the left). What version of the Mecway did you use? My install is at 12.0 64 bit.

Thanks-Hooshim

hooshsim

Sergio,
Thank you for your review. I downloaded your model and ran it without any changes and got the exact results as my original model. What version of the Mecway did you run My install is at 12.0 64 bit.

Sergio

I'm using Mecway 12 with CCX 2.15 with Pardiso. As far as I remember Mecway came with CCX 2.14 and that could be the difference

hooshsim

Sergi,

Initially from your reply, I thought the first model (on the left) was my model. That's why I raised a question of software version. Please discard my first reply.

However, after further investigation of both models I realized that my model have identical results (stress and displacement) on all cyclic surfaces (top slide in the attached plot) while your model has similar results on displacement only and has some discrepancies on the stresses (bottom slide). Your thought?

Do you think my symmetry card defined in the CCX card is incorrectly defined? What is the proper way of defining symmetry condition for multi cyclic surfaces? I like your approach but I'm afraid of losing accuracy.

Thanks again - Hooshsim

Victor

I can't see anything wrong with the cyclic symmetry definition in your original model, Hooshsim. According to the CCX manual, ELSET is only used for generating the output segments so I think it's harmless to omit.

Combining the 3 pairs of surfaces into one seems to cause some nodes from one part to get tied to faces of the wrong other part along their common edges. That probably accounts for the stress differences. This is something I hadn't thought of before and is a limitation of the built-in cyclic symmetry constraint too. Better to keep 3 separate cyclic symmetries.

The CCX manuals says 3 are OK "Several *CYCLIC SYMMETRY MODEL cards within one input deck defining several cyclic symmetries within one and the same model are allowed. This, however, always is an approximation, since several cyclic symmetries within one model cannot really exist." Since you have identical axes and numbers of segments, I don't think it's an approximation.

I prefer to set 2 output segments (NGRAPH=2) so you can easily see from the deformed view, that the displacements as well as curvature agree. With NGRAPH=2, you do need ELSET like in Sergio's model otherwise CCX seems to ignore it.

hooshsim

Thank you Sergio and Victor. Much appreciated.

Sergio

And what abou the mid plate moving outside the simmetry plane? Should not be in the plane to attend the simmetry boundary condition?

Victor

That would be true for mirror symmetry, but static cyclic symmetry requires displacement (in cylindrical coordinates) to be equal on both sides, but not necessarily zero. Here, you can see the opposite side has the same circumferential displacement to match it.

Though, since the geometry and load seems to be symmetric about the middle of the segment, it should behave the same as mirror symmetry. I wonder if that small shift is just random rigid body motion or caused by some accidental asymmetry in the mesh.

hooshsim

Sergi,

In your fixed model posted earlier I realized that both symm_1 and symm_2 surfaces in the Named Selections are assigned to 230 faces. Could you please explain how you came up with these numbers and how would you handle a more complex 3D cyclic surfaces. See attached snapshot from your fixed model.

Thank you - Hooshsim

Sergio

The 230 faces are those of the simmetry plane, for using only one cyclic simmetry condifion.

Have checked and using only one pair of TIE for the cyclic simmetry leads to nodes of different components being artificially stitched, even if I had used very small tolerance for the tie (0.1mm). I have used again three pairs of TIE/CYCLIC SYMM bc and now works ok (stress are the same on both sides of the simmetry).

Be aware that in the initial model, the contact of the shaft against the hole in the nylon part was not achived, has to increase the contact stiffnes to get it work. Didn't check this contact in the other components.

EDIT: I would improve (reduce element size) the meshes of the components near the hole, and also include some rivet pretension in the assembly.

Regards

Sergio

I would apply the simmetry at the middle of the bolt (so, half bolt in each simmetry plane) in order to check easily the behavieur in this area. Have exported the results to Paraview for a clear view, and we can see that the other plates will not be afected by the contact as they are first glued by the TIE on the upper/lower surfaces.

hooshsim

Sergio,

Thanks for the update. Solution now makes more sense.
Would you please add pretension to the assembly as you suggested. I'm unsure how to add the pretension.

Thanks again.

Sergio

Please upload the geometry CAD files needed in order to create a partition on the rivet shaft, and improve the mesh at the hole area.

hooshsim

Sergio,

Requested CAD files are attached.

Thanks again

Sergio

Hi, here is the final work, took a few iterations but now the preload, centrifugal force and pressure on blade are all applied. In case of using the middle of the rivet as the simmetry plane, there will be some nodes that will participate in contact and simmetry bc, that is not allowed by ccx, that's why I back to the initial model.

I did some animations on Paraview in order to capture the middle plane of the bolt. The loads were applied every one in different step (so one can see the effect separetely), so at time 1 there is the rivet preload fully applied, at time 2 the centrifugal force and time 3 the blade pressure.

Maybe for a complete analysis the central hole must be modeled with it bolt and contact the same as the rivet and not just "fixed" as it now.

Regards

hooshsim

Thank you so much Sergio...great job. I will rerun your final model and get back with you for any questions that I may have.
Hooshsim

hooshsim

Sergio,

I'm not too familiar with the CXX cards. It appears that some parameters have been added to the original one. Can you please add a brief explanation to each one of the cards describing the parameters? For example what are the values under CLOAD, etc.?

Thanks again. Hooshsim

Victor

I'll quickly try to answer this, hope I don't step on your reply Sergio:


*CLOAD
48393,1,-1000.

• CLOAD = concentrated (point) load
  
• Applied to node 48393
  
• DOF 1 (X direction)
  
• -1000 N.
  
• This load has the special meaning of pretension force because it's applied to the *PRE-TENSION SECTION's NODE.
  

*DLOAD
Aluminum,CENTRIF,45000.,0,0,0,0,0,1

• DLOAD = distributed load
  
• Applied to ELSET (component in Mecway) named "Aluminum"
  
• Centrifugal force
  
• Angular velocity squared=45000 (rad/s)^2
  
• Axis is through points (0,0,0) and (0,0,1) which is the Z axis.
  

*DSLOAD
pressure_faces,P,3447.378646584

• DSLOAD = distributed load on surface
  
• Applied to surface (face selection in Mecway) named "pressure_faces"
  
• P = normal pressure
  
• 3447.378646584 Pa
  

*NODE FILE,GLOBAL=YES
U,RF
*EL FILE
S

• Output displacement (U), external force (RF), and stress (S) in the global coordinate system.
  

hooshsim

Victor,

This is great. Much appreciated.

Hooshsim

hooshsim

Victor,

In your previous reply to this thread, should't preload (-1000 N) be applied in the Z-direction instead of X?

Thanks again - Hooshsim

Victor

Oh, I was a bit wrong on that. Because of the special meaning of that load, DOF 1 here means the pretension direction, not the global X direction. It can be oriented any way and it's still DOF 1.

disla

Hello,

I'm running Sergio's liml file.

¿How is it possible that being the time period for the Analysis settings menu 1s the calculation is at 1.4s and still running?

Sergio

Hello Disla, that is because there are two more steps defined in the CCX custom cards.