AnsweredAssumed Answered

Stress Values Not Corresponding to Published Values

Question asked by Cory Nation on Nov 17, 2014
Latest reply on Nov 25, 2014 by Jared Conway

Like • Show 0 Likes0
Comment • 7

I am conducting an analysis on a simple part to determine the stress concentration factor of a shaft with shoulder fillet in torsion. I have referenced Pilkey's Stress Concentration factors and other texts and the values I am achieving from solidworks are much higher than what is published. I have referenced a youtube video and the same issue occurs.

Torsional Static Analysis and Design Study Example Using Solidworks - YouTube

If you go to 5:01 of the video, the author shows a Kts table to relate his model to. When he finishes his analysis, instead of having Kt values of 1.48, 1.25, 1.17, and 1.13 he gets 2.303, 2.235, 2.132, 2.06. Why is the stress level so high? My stress values are charted below for D/d=1.111.

No one else has this question

Outcomes

Jim Riddell Nov 18, 2014 7:10 AM

Cory,

Even your data looks high compared to accepted values in sources such as Machinery's Handbook. I've found that SW is very bad at analysis at 'sharp' contacts and don't trust refined meshes all that much.

Actions

Shaun Densberger @ Jim Riddell on Nov 18, 2014 3:28 PM

I've found that SW is very bad at analysis at 'sharp' contacts and don't trust refined meshes all that much.

What do you mean by, "sharp contacts?" Are you referring to re-entrant corners? If so, then this isn't something specific to SW, but rather the finite element method in general. That being said, the fillet in this model resolve the singularity issue, so this shouldn't be an issue.

Actions

Shaun Densberger Nov 18, 2014 3:26 PM

Couple things I noticed with his model:

At ~2:58 you see the deformed shape and it looks like the rod is bending down. This isn't right and it's hard to say if this is a bug or something else, but we should see the shaft artificially expand due to the linear assumption.
His stress sensor for the fillet is at a point, which is a poor choice. He should have looked at the maximum across the entire fillet.
His stress sensor for the shaft is at a point, which is (again) a poor choice. He should have looked at the maximum around the circular edge of the split line.
He never does a convergence study, so it's possible that he had a poorly constructed mesh.

I went ahead and repeated his work and got the following Kt factors:

Fillet	Kt
0.05 in	1.36
0.10 in	1.23
0.15 in	1.16
0.20 in	1.12

Actions

Cory Nation Nov 18, 2014 3:43 PM

UPDATE: Solidworks outputted is estimated stress using the Von Mises Crtierion. When generating a stress concentration factor, where the stress is in pure torsion, the von mises stress must be divided by sqrt(3). (since sigma(1)=-sigma(2)=tau(y)) tau(yield)=sigma(yield, VMS)/sqrt(3).

When I did this conversion, the stress concentration factor lined up with the predicted values form Peterson.

I did a convergence graph and applied sensors in various regions around the fillet to ensure the stress is accurate and precise. The ERR plot shows error in the energy norm to be less than .0098%. After this aforementioned change, the stress values deviated by .0038 on average from the published values.

Actions

Shaun Densberger @ Cory Nation on Nov 18, 2014 5:12 PM

...the von mises stress must be divided by sqrt(3)...

How were you calculating your stress concentration factor?

Actions

Cory Nation @ Shaun Densberger on Nov 24, 2014 2:11 PM

Kt=max stress/nom stress.

Max stress=VMSmax/sqrt(3)

VMSmax is the maximum von mises stress calculated by solidworks in the region of interest. To convert the VMS to principle stress for a part subjected to pure torsion, you must divide by the square root of 3.

nom stress= 16*Torque/(pi*small diameter^3)

Torque applied to this part was 2500 in-lbf

Actions

Jared Conway @ Cory Nation on Nov 25, 2014 2:17 PM

so in this case the issue was intepretation of the data, not a software/calculation issue

Actions

7 Replies

Jim Riddell Nov 18, 2014 7:10 AM
Mark Correct
Correct Answer
Cory,
Even your data looks high compared to accepted values in sources such as Machinery's Handbook. I've found that SW is very bad at analysis at 'sharp' contacts and don't trust refined meshes all that much.
Like • Show 1 Like1
Actions
- Shaun Densberger @ Jim Riddell on Nov 18, 2014 3:28 PM
  Mark Correct
  Correct Answer
  I've found that SW is very bad at analysis at 'sharp' contacts and don't trust refined meshes all that much.
  
  What do you mean by, "sharp contacts?" Are you referring to re-entrant corners? If so, then this isn't something specific to SW, but rather the finite element method in general. That being said, the fillet in this model resolve the singularity issue, so this shouldn't be an issue.
  Like • Show 1 Like1
  Actions
Shaun Densberger Nov 18, 2014 3:26 PM
Mark Correct
Correct Answer
Couple things I noticed with his model:
1. At ~2:58 you see the deformed shape and it looks like the rod is bending down. This isn't right and it's hard to say if this is a bug or something else, but we should see the shaft artificially expand due to the linear assumption.
2. His stress sensor for the fillet is at a point, which is a poor choice. He should have looked at the maximum across the entire fillet.
3. His stress sensor for the shaft is at a point, which is (again) a poor choice. He should have looked at the maximum around the circular edge of the split line.
4. He never does a convergence study, so it's possible that he had a poorly constructed mesh.
I went ahead and repeated his work and got the following Kt factors:

Fillet Kt
0.05 in 1.36
0.10 in 1.23
0.15 in 1.16
0.20 in 1.12
Like • Show 1 Like1
Actions
Cory Nation Nov 18, 2014 3:43 PM
Mark Correct
Correct Answer
UPDATE: Solidworks outputted is estimated stress using the Von Mises Crtierion. When generating a stress concentration factor, where the stress is in pure torsion, the von mises stress must be divided by sqrt(3). (since sigma(1)=-sigma(2)=tau(y)) tau(yield)=sigma(yield, VMS)/sqrt(3).

When I did this conversion, the stress concentration factor lined up with the predicted values form Peterson.

I did a convergence graph and applied sensors in various regions around the fillet to ensure the stress is accurate and precise. The ERR plot shows error in the energy norm to be less than .0098%. After this aforementioned change, the stress values deviated by .0038 on average from the published values.
Like • Show 1 Like1
Actions
- Shaun Densberger @ Cory Nation on Nov 18, 2014 5:12 PM
  Mark Correct
  Correct Answer
  ...the von mises stress must be divided by sqrt(3)...
  
  How were you calculating your stress concentration factor?
  Like • Show 1 Like1
  Actions
  - Cory Nation @ Shaun Densberger on Nov 24, 2014 2:11 PM
    Mark Correct
    Correct Answer
    Kt=max stress/nom stress.
    
    Max stress=VMSmax/sqrt(3)
    VMSmax is the maximum von mises stress calculated by solidworks in the region of interest. To convert the VMS to principle stress for a part subjected to pure torsion, you must divide by the square root of 3.
    
    nom stress= 16*Torque/(pi*small diameter^3)
    Torque applied to this part was 2500 in-lbf
    Like • Show 0 Likes0
    Actions
    - Jared Conway @ Cory Nation on Nov 25, 2014 2:17 PM
      Mark Correct
      Correct Answer
      so in this case the issue was intepretation of the data, not a software/calculation issue
      Like • Show 0 Likes0
      Actions

Stress Values Not Corresponding to Published Values

Outcomes

Related Content

Recommended Content