Fea questionable reliability

[chipwitch]

Attached are images that I believe may indicate a glitch in the FEA plugin.  The first image shows the rig, which should be self-explanatory.  Red are grounded parts, yellow part is bonded to red part bilaterally where indicated (only one side indicated).  Please forgive the colors in the first image.  The Autodesk uploader changed them.  I've also attached the image.  It will be clearer.  <edit> the uploader isn't changing it... it was my browser.  If the colors look wrong to you, please let me know.. </edit>

The three images that follow are variations of the same simulation.  The only thing that I changed were the contacts at pins 1 and 2.   Everything else is constant.  I am not attempting to analyse the pin contacts themselves, rather the forces acting on the parts.  The pins merely act to constrain the assembly and transfer forces.

Constant among all 3 images: 

Pin 3 = Sliding, no separation contact

Applied Force => magnitude and direction

Frictionless constraint on the top face of the Top Link (to prevent rotation)

All simulations converged very near 24 ksi.

The color bar is set to a very low 1 ksi to indicate stress uniformly.

Image 1:  Contacts at pins 1 and 2 are sliding/no separation.  This was my initial and preferred simulation.  Note the Stinger pin 2 contact.  Virtually no stress is being translated to the flanges.  Also note the lower stress between pins 2-3 vs pins 1-2.   This cannot be correct and the reason I began to question the results.

Image 2:  The contacts at pins 1 and 2 were changed to bonded.  I'm only including this image to compare to image 3.  Please read on.

[chipwitch]

(Please forgive the OP being broken into 3 different posts.  The forum kept saying there was a problem with the HTML in my code, though didn't use any.  This was the only way I could get it all posted.)

[chipwitch]

Image 3:  This is my "control" simulation.  Here, I replaced the contacts at pins 1 and 2 with pin constraints.  These results are more believable than those in image 1.  However, they obviously exclude the yellow part from the assembly image I posted first.  For this reason, using pin constraints are unacceptable.  Further, there are more links and pins to be added to the assembly to form a complex linkage sytstem.  It was my intention to add those and apply loads at various points in the system to analyse the resultant forces on the individual parts.  If I must use the pin constraint at each contact, I won't be able to determine how the loads transfer from one part to the next.

 

What puzzles me most is how the ears in the first image at pin location 2 appear to have almost no stress.  Regardless of the contact type at pins 1 and 2, the stresses around the pins should be similar to one another (given that the cross sections of their corresponding ears are similar) and the forces are equal and opposite (moment).  If anything, the ears at pin 2 should be subject to greater force as indicated in Image 3, the opposite of image 1.

 

When it comes to FEA, the results you get are heavily dependent on the data input.  Checking and double checking are a given, as I've done here.  But, it seems to me that the results in the Image 1 are so far off base, I can't help wonder if it is a glitch in the software.  Perhaps I'm missing something here.  I would be very grateful, if someone would be so kind as to point it out.

 

Best Regards.

[chipwitch]

Looks like the color problem I'm having with my browser of the assembly is related to PNG files.  Here is the file in JPG format.

[raviburla]

Hi,

I am assuming the model is same as in the post here. I ran the model and found that the "region near pin 2" is undergoing a near rigid-body motion and hence that portion is not undergoing significant stresses. I am attaching the animation with this message. Also, when the contacts are set to sliding/no separation, the pins are undergoing  a rotational motion, so perhaps this is the reason why the results are different when bonded contacts are used.

Please let us know if you have more questions.

Thanks,
Ravi Burla

[chipwitch]

Thank you Ravi.  I did make some changes to the assembly to simplify the simulation, but yes, it is essentially the same as the model you referenced.

I'm looking at this simulation as a fulcrum.  If we assume the distance between pins 1 and 2 to be 1/3 the distance between pins 2 and 3, then the 600 lb force would imply a force of 1800 in the same direction at pin 1.  The result on pin 2 is 600 + 1800, or 2400 lbs in the opposite direction.  Regardless of flexibility at pin 2, am I wrong in thinking the stress will be greatest there?

I realize the fulcrum isn't a perfect model to apply to this situation and the 3:1 ratio is a slight exaggeration, but isn't it close?

[raviburla]

Hi, 

You are correct, the model would behave as a fulcrum if there is no motion for the portions where we expect high stresses. I ran an experiment (similar to the one you had done with pin constraints) - for that I had excluded everything except stringer and applied bearing load on pin-3 and applied pin constraints on pins 1 and 2. Looking at the reaction forces, I got the values to be 1226 lbf  at pin2 and -622 at pin-1.

However for the actual model - the rigid motion does influence the amount of force transfer and hence the stresses. I ran another experiment by fixing portions of the axle mount part so that this motion can be stopped. With this change, we can see some stress taken up the the "pins". However, due to some sharp geometry changes, there is stress-concentration as can be seen in the attached image. Note that in the attached image, I have made all parts excepting Stringer invisible, but the only change to simulation is the addition of fixed constraints. Hope this helps.

Thanks!
Ravi

[chipwitch]

Ravi, I appreciate the time you've spent helping me.

I don't understand the "rigid motion" reference.  It was my understanding that the FEA is a STATIC analysis and that any "motion" illustrated by the animation is merely a representation of the displacement vs applied force, not time.  Be that as it may, even the image you posted, appears to indicate far more significant stress at pin 3 than pin 2, which doesn't make sense.

I understand what you're trying to say about the flex in the axle mount, but since this is a STATIC analysis, motion SHOULD be removed as a factor from the calculation.  No?

[raviburla]

Hi,

You are correct, I had used the term rigid motion rather loosely. What I meant was rigid deformations. Even in Linear Static Analysis, finite elements undergo rigid deformations. A rigid deformation is characterized by the deformations without inducing any stresses. For example, consider a long cantilever beam - on the end that is  loaded the elements undergo large deformations but these deformations include rigid deformation + actual deformation. The actual deformations would be relatively very small. The actual deformations lead to strains and hence stresses. (Otherwise, the loaded end should have very large stresses).

I was trying to say that there is no relative deformation between the pin and Stringer part, therefore stresses were not induced. Also, for the modified model (with fixed constraint), the forces would be transferred properly.

As you have mentioned, we should factor out the rigid motion - this is correct. In fact, after removing the rigid deformations - the actual deformations are so small that they are not inducing the stresses. The model is simple but if we can simplify it further and still keep the issue that is of concern, the analysis/debugging the model would be much more simpler. 

Thanks!

Ravi

[chipwitch]

So, taken to its logical conclusion, the FEA is only reliable when used to analyze the stresses on a single part?  That is as simplified as we can make it and I'd argue not far from the model I currently have here.

I'm having a hard time understanding the relevance of the deformation as it pertains to my model.  It sounds as if you're suggesting, so I can only assume I am misunderstanding you, that the deformation occurring in one part somehow relieves the force acting on an adjacent part.  THAT I would find very difficult to believe is how the FEA plugin calculates the stress unless it were a bug or glitch.  I apologize if I'm not understanding you correctly.  I can be a little dense at times.

As I've said before, I wish to analyze an assembly, not just a part.  Calculating the stress and deformation in a single part isn't all that difficult to approximate by hand.  If that is all the FEA plugin is capable of, then I'd have to say it isn't very useful.  However, I don't believe it was ever intended to suffer such limitation.  There must be something else at play here.  Only one of three things is possible.  There is a glitch in the software, the plugin is only good for single parts, or I'm setting up my model incorrectly.  I would be pleased to discover which one.

[swalton]

Based on the animation posted by raviburla,  I see the following.

1. The stinger has a very robust construction.  It is very ridged.

2. The yellow support structure that captures Pin 2 is not.

3. The horizontal plate that captures pin 2 flexes when the load is applied.

A key assumption about a free body diagram is that the members are so rigid that they do not flex, ever.

The FEA animation shows components in the load path between Pin 2 and the fixed constraints that flex.  The flex in those components prevents the lugs around Pin 2 from developing any stress.  Instead, that load is carried by the horizontal plate and should show as stress at the joint between the horizontal plate and the two vertical supports.  I expect that the stress level there is very low because there is a large amount of area.  Stress=force/area.

Try adding a fixed constraint to the yellow plate.  I expect that will show stress in the lugs at the Pin 2 mount.

A question that you might ask yourself is this:  Does the deflection caused at the end of the stinger because of the flex in the yellow plate exceed the allowable limits? If not, your design might be ok (depending on any other criteria).  If so, stiffen the yellow plate to prevent deflection at the end of the stinger.  As you start to do that, the yellow plate should behave more like a fixed constraint on Pin 2 and you should see more stress in the lugs.

[chipwitch]

Ravi,

If you please, I re-examined the simulation this morning.  Your suggestion that deformation may be playing a role prompted me to focus my attention on pin 2.  The pin is seeing less than .001" of displacement.  However, what I noticed was that after isolating pin 2, I was able to probe the maximum stress in the pin.  It was 0.108 ksi!  100 psi!?  That would indicate a total force being transferred through the part of approximately 35 lbs, by my rough estimation.  (There is approximately 0.31 in^2 of contact area normal to the force). This, at a point that should be experiencing the MOST force.  Common sense dictates that this figure should be at the very least > 600 lb, the applied force.

Thank you

Sherri

[raviburla]

Hi,

Thanks for the detailed investigation, and I understand your concerns. If it is OK with you, I would like to take this issue and log in our internal system as a defect. This allows our internal experts to take a detailed look into the model and we should also be able to debug into the solver and look for any discrepancies.

Thanks!

Ravi Burla

[chipwitch]

swalton, thank you for the input.  The animation is an exaggeration.  The tip of the stinger is only seeing a displacement of <0.003". Rigidity is relative and I'd say this system is pretty rigid.

Firstly, FBD's and FEA are two entirely different things.  Yet, they should not be in contradiction.  My earlier comment regarding FBD's was merely meant to identify the force that MUST be transmitted through the contacts at pin 2.

Like I mentioned in my previous post, by my calculation, pin 2 is indicating a total load of about 35 lbs when an applied load at pin 3 is 600.  The force at pin 2 must exceed the applied load, as the applied load is cantilevered with it's primary support coming from pin 2.  Every link in a chain must bear the load applied to one.  Thus, pin 2 MUST exceed 600 lbs.  An FBD can be used to determine the resultant forces at pin 2.

Clearly, there is SOMETHING wrong.  Either my model, the FEA plugin or my understanding of the limitations of FEA.  I'm trying to identify which.  Why would a mere .001 inches of deflection (or any deflection for that matter) at the yellow part somehow absorb 95% (or any for that matter) of the applied load?  In a dynamic simulation, I would agree that flex can create moments of seeming discrepancies such as this, but there ARE other forces acting aside from the applied force, inertia and centripetal to name two, that can account for those "discrepancies."  What can account for them here?  For all intents and purposes any static system is "rigid."  And, the FEA illustrates just that.  Even though there may be some flex, we still have to account for the forces acting on a system.  If you hang a 600 lb weight from a chain, does the link at the top miraculously somehow absorb 580 lbs? Certainly not.

All I'm saying is that my model isn't making any sense.  If I'm expecting more from the FEA plugin than it can deliver. Fine. Can you help me understand why?  If I've modelled something incorrectly, can you point out what it is and why?

I did try as you suggested and applied a fixed constraint on the yellow part.  The results showed slight improvement.

[chipwitch]

Ravi,

Thank you.  Please keep my aprised of their findings.  Again, I really appreciate all your time and help in this matter.

Sherri

[chipwitch · Accepted solution]

Ravi, Nathan Chandler found a problem in my model in a previous thread I started, here.

Thank you for your assistance.  Please suspend further investigation.  Consider the matter resolved.

[raviburla]

Thank you for the update. I am glad that the issue has been resolved to your satisfaction. We will close internally logged issue.

Thanks,

Ravi

[raviburla]

Hi,

We have logged the issue as Bug 40589. We will keep you updated on the progress of this issue.

Thanks,
Ravi