
Re: 2d vs 3d analysis results discrepency
David Anderson Jul 8, 2011 2:19 PM (in response to Steve Fiacchino)assuming mat props are equal, some things to check
1. is the model scale/geometry/constraints exactly the same for both? my guess is the geomtry/bc's/loads are the issue.
other things to try...
2. can you replace the triangular shell with a revolve or swept higher order quad shell?
3. are the plane strain and triangular shell elements higher order, if not try higher order?
4. are the stresses the same in the same areas or different.
5. in both models, try decreasing the element size until the stress no longers changes. then compare values
if you want to attach you model i can take a look at it...

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 12, 2011 11:46 AM (in response to David Anderson)I have not had much luck reconciling the results of the 2d and 3d models yet. Have you have any luck with the models I provided? Thanks in advance...

Re: 2d vs 3d analysis results discrepency
David Anderson Jul 14, 2011 12:35 PM (in response to Steve Fiacchino)how did you provide the models? i do not have any.

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 14, 2011 12:44 PM (in response to David Anderson)I sent them to the email notice I received from your post. I guess that did not work. Attached is the model. Thanks for looking at it.

bellows_shared.zip 570.0 KB

Re: 2d vs 3d analysis results discrepency
David Anderson Jul 14, 2011 1:19 PM (in response to Steve Fiacchino)and your loads and constraints?
looks like the closed endcap is fixed, and axial compression load of .01" and and internal pressure of 340psi correct?

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 14, 2011 1:27 PM (in response to David Anderson)That is correct.

Re: 2d vs 3d analysis results discrepency
Igor Isayev Jul 15, 2011 11:26 AM (in response to Steve Fiacchino)I've ran both studies and received much closer values .
Comparing results for 3D shell & 2D Models , I got Max.VonMises stress as 64.8 ksi vs.74.4 ksi for Top face or 57.5 ksi vs. 47.3 ksi on Bottom face .
Here is what I did.
1. Mesh of 2D model surface was refined at the fillets area where the part is deformed , at least 34 nodes between top & bottom faces. You can generate even more  it won't affect the time to solve the study
2. Uncheck "Draft mesh quality" in 2D model.

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 15, 2011 11:36 AM (in response to Igor Isayev)Thats an improvement but still 2x the max stress on the bottom surface of my 2d model which was at 32.5 KSI (see first attched powerpoint). I'm suprised your 2d model had a much higher stress then mine (47.3 vs. 32.5) I tend to think the 2d model is more accurate (vs 3d shelled) becasue I find it hard to believe a bellows material would go beyond the yeild (55 KSI) with only a .010 defelction. Any ideas why there is such a large discrepency between the 2d and 3d shelled models?

Re: 2d vs 3d analysis results discrepency
David Anderson Jul 15, 2011 7:08 PM (in response to Steve Fiacchino)some modeling issues
1. your surface model is not a midsurface model. so there is a geometry mismatch from the get go. lever arms and radii are not the same.
2. the surface area of surface model is larger than area of 2d model by about 5% thus surface model will have 5% higher load and stress from pressure
for a 2d axisym model, stress for
displacement only:26973
displacement and pressure:75803

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 18, 2011 10:50 AM (in response to David Anderson)I understand your comments on the shelled surface model not being a midplane model and having some effect on the resutls due to radii and lever arm difference. But I believe that should be minimal when analyzing results at the top and bottom shelled surface resulting in a 510% difference comapred to the 2d model.
Regaring your results on the 2d model, I do not understand why they are so much different than the results I shared at the beginning of the discussion:
displacement only 16.3 KSI
displacemen and pressure 32.5KSI
In the end, I believe the resutls of both the 2d and 3d shelled models should have similar values (under 15 precent delta) and be similar in magnitude to the 2d modeled results. This could be wishful thinking but I need to understand WHY there is such a signifiant difference... Thanks for your assistance.

Re: 2d vs 3d analysis results discrepency
David Anderson Jul 18, 2011 11:13 AM (in response to Steve Fiacchino)i am using ANSY and not SW sim so i cannot interrogate you model for discrepancies.
what are your exact material properities? specifically youngs and poissions.
what is the element type and order?
have you tried to continually refinin your mesh until the stress results do not change significantly? this can play a huge role in stress levels. give this a shot and let me know your results.
see my mesh below. regardless whether is it SW sim or ANSYS, we should be getting the same answer

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 18, 2011 11:32 AM (in response to David Anderson)
Re: 2d vs 3d analysis results discrepency
David Anderson Jul 18, 2011 12:28 PM (in response to Steve Fiacchino)are you using higher order elements? also you are using triangular elements which can understimate stresses.can you try using high order quads?
if you refine your mesh further, does the stress change?
are you using a load combination of pressure and displacement or pressure only?

Re: 2d vs 3d analysis results discrepency
Joe Galliera Jul 18, 2011 8:56 PM (in response to Steve Fiacchino)Hi Steve, I took a look at the model that you had attached. The mesh is insufficient in crucial areas and the element type is firstorder "Draft" quality, which are not suited for bending applications. PDF attached with comments.
I made some modifications to both the shell model and the 2D axisymmetric one, attached. For the shell model, I created a midsurface by offsetting one of the sides, and cut the model down to only 30° to take advantage of symmetry so that one can add more elements without adding too much more solver time. The result that works well here is the 3rd principal stress on the bottom, since the inside is under compression.
The 2D problem, I decreased mesh size globally and added mesh controls around the bends like I did for shell model. For the results, I used von Mises and focused on the inside, because that's what the other posts showed. Bonus, the 2D results are shown with a 3D visualization... see below.

bellows_shared_jg.zip 791.1 KB

analysisvariationshared.pdf 589.1 KB

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 20, 2011 4:27 PM (in response to Joe Galliera)Thank you Joe. I will see if I can replicate your solution. I was really hoping my original 2D analysis showed the accurate results since it was much lower then the 3D model. I still find it hard to believe that a .010 deflection of a bellows would yield the bellows material. But then again, thats what the FEA is for, right? Thanks again.

Re: 2d vs 3d analysis results discrepency
David Anderson Jul 20, 2011 4:44 PM (in response to Steve Fiacchino)if 55k is your yield, then the .01 displ alone does not yield it. its the pressure.

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 20, 2011 4:56 PM (in response to David Anderson)Yup. I see it. Thanks. And thanks to everybody who answered this post. Your feedback was very helpful.

Re: 2d vs 3d analysis results discrepency
David Anderson Jul 20, 2011 5:06 PM (in response to Steve Fiacchino)that being said, .0204 inches will yield it!

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 20, 2011 5:16 PM (in response to David Anderson)Yes. Not good, something will need to change. Most likely a change in the material specs, then wall thickness. I think I will need to get to about a .040 deflection with a FOS of at least 2.
Design, Thats what its all about...




Re: 2d vs 3d analysis results discrepency
Wayne Matus Jul 20, 2011 5:22 PM (in response to Steve Fiacchino)You left out one major step. You validate your FEA with physical tests, strain gages. Once you validate your FEA techniques then you can apply those techniques to other similar products.
When I was doing this for a living, I designed products per API 8C. Basically that spec said that we had to validate our product and calculation methods by strain gage testing loading the product to 2.25 to 3 times load rating depending on rated load. Strain gage stress values had to be less than yield strength of material. If they were, that model was stamped as meeting this code. Also if our calculated stresses were greater than the actual stresses, then we could use these calculation methods on the same product with a load rating one size larger and one size smaller than the one we tested as long as we did not change the design.

Re: 2d vs 3d analysis results discrepency
Steve Fiacchino Jul 20, 2011 5:32 PM (in response to Wayne Matus)I agree, I usually tell our customers that the results aren't much more then pretty picutes unless the models (or something similar) have been tested and the results correlate.














