AnsweredAssumed Answered

Mixed mesh simulation convergence / error check

Question asked by Asher Brolsma on Aug 26, 2019
Latest reply on Aug 29, 2019 by Asher Brolsma

Like • Show 0 Likes0
Comment • 4

Hi guys,

I have a mixed mesh structure that I am analysing, the model is relatively large and has a large amount of contacts and fasteners in it (200+ fasteners, 35+ contact sets). the mesh is currently sitting around the 400,000 node mark and has refinements in critical (high stress) areas. The studies take several hours to run and the model has to be analysed in 5 separate load case scenarios. There are ~10 areas in each scenario that will show areas of high stress. The model is being analysed using Linear analysis, I don't have access to Non-linear.

Being a mixed mesh study with fasteners, the p and h-type adaptive meshes are not available, so I cannot automate the mesh convergence, although to convert the shell section to a solid mesh and run them would probably go to a point where the mesh would get too large for my computer to handle, mainly due to the mesh being refined in areas that are not worth considering. To do a manual convergence check could require running 20 or more studies each time a design change is made - I.e this could take weeks if a single change is made and the structure needs to be re-analysed.

Has anyone found any guides on stress error plots (energy norm error) and whether they can be used as an acceptable justification of results and what the level of stress error should be as a maximum in those critical areas? Or is there any other methods out there that could save me some time?

Thanks,

Ash.

No one else has this question

Outcomes

Jay Seaglar

Aug 28, 2019 2:54 PM

The Stress Hot Spot diagnostic tool could be helpful here. Another potentially time-saving approach could be to try using a Design Study, for example by varying the element size in mesh controls applied to critical areas. And finally, just in case Load Case Manager has not been considered for the various load cases, it is worth a look because it could drastically reduce the total solve time.

Beyond that, the following SOLIDWORKS Knowledge Base (KB) solutions about node vs. element stress and associated error may be of interest:

S-020573
S-017218

You can also want to search the KB for the following SPR number and select the 'notify me' option under 'Find what you were looking for' to vote for this enhancement:

533074 - "Adaptive meshing for Shells and Mixed Mesh (both Solid elements and Shell elements present)"

And don't forget to reach out to your SOLIDWORKS Value Added Reseller (VAR) if you need further assistance.

Actions

Asher Brolsma @ Jay Seaglar on Aug 29, 2019 12:29 AM

Thanks for that, definitely useful info.

Actions

Ryan Navarro Aug 28, 2019 3:46 PM

PV-Eng has some very helpful Simulation Tutorials / Blogs:

Mesh Refinement at Discontinuities – Pressure Vessel Engineering

Mesh Refinements Near Discontinuities – Pressure Vessel Engineering

It appears they use < 5% energy norm error as a guideline for mesh refinement

Definitely an area to exercise caution though as 5% energy norm error does not equal 5% error in stress. The discrepancy may vary depending on your study conditions. If I were in your shoes I would run some experiments (as was done in those articles) of convergence tests on a small portion of a model with similar contacts and loading, in order to draw your own conclusions.

Actions

Asher Brolsma @ Ryan Navarro on Aug 29, 2019 12:29 AM

Thanks for those studies, they pretty much show the issues I am having - plenty of sharp corners and therefor plenty of singularities that create mathematical anomalies that cause high stress spikes and high stress errors, so I doubt the criterion would ever be reached. (I.e similar to point 5 in the first study). For the size of my model and the power of computing available I doubt 5% stress error could be achieved at all of the locations.

I do remember about 5-7 years ago I read a study saying keep it all below 100% energy norm error (this mainly applies to areas of high stress and sharp geometry) and the results will typically be within the same region as doing a H-adaptive study set around ~92+% accuracy. I.e stress and displacement values are generally converged to within 8% of a mesh that has half the density. It was a great rule of thumb and I have generally matched the results via hand calcs and they were accurate enough, generally only a couple of percent difference. The areas around the part that are around the 100% mark all fall back to significantly lower energy norm error values (usually around that 5% or lower referenced in the studies above) Unfortunately the geometry in my current model is too complex for that approach this time around and third party verification would like to see some convergence checking or documentation to support he stress error plot being used.

Actions

4 Replies

Jay Seaglar Aug 28, 2019 2:54 PM
Mark Correct
Correct Answer
The Stress Hot Spot diagnostic tool could be helpful here. Another potentially time-saving approach could be to try using a Design Study, for example by varying the element size in mesh controls applied to critical areas. And finally, just in case Load Case Manager has not been considered for the various load cases, it is worth a look because it could drastically reduce the total solve time.

Beyond that, the following SOLIDWORKS Knowledge Base (KB) solutions about node vs. element stress and associated error may be of interest:
S-020573
S-017218

You can also want to search the KB for the following SPR number and select the 'notify me' option under 'Find what you were looking for' to vote for this enhancement:

533074 - "Adaptive meshing for Shells and Mixed Mesh (both Solid elements and Shell elements present)"

And don't forget to reach out to your SOLIDWORKS Value Added Reseller (VAR) if you need further assistance.
Like • Show 0 Likes0
Actions
- Asher Brolsma @ Jay Seaglar on Aug 29, 2019 12:29 AM
  Mark Correct
  Correct Answer
  Thanks for that, definitely useful info.
  Like • Show 0 Likes0
  Actions
Ryan Navarro Aug 28, 2019 3:46 PM
Mark Correct
Correct Answer
PV-Eng has some very helpful Simulation Tutorials / Blogs:
Mesh Refinement at Discontinuities – Pressure Vessel Engineering
Mesh Refinements Near Discontinuities – Pressure Vessel Engineering

It appears they use < 5% energy norm error as a guideline for mesh refinement

Definitely an area to exercise caution though as 5% energy norm error does not equal 5% error in stress. The discrepancy may vary depending on your study conditions. If I were in your shoes I would run some experiments (as was done in those articles) of convergence tests on a small portion of a model with similar contacts and loading, in order to draw your own conclusions.
Like • Show 0 Likes0
Actions
- Asher Brolsma @ Ryan Navarro on Aug 29, 2019 12:29 AM
  Mark Correct
  Correct Answer
  Thanks for those studies, they pretty much show the issues I am having - plenty of sharp corners and therefor plenty of singularities that create mathematical anomalies that cause high stress spikes and high stress errors, so I doubt the criterion would ever be reached. (I.e similar to point 5 in the first study). For the size of my model and the power of computing available I doubt 5% stress error could be achieved at all of the locations.
  
  I do remember about 5-7 years ago I read a study saying keep it all below 100% energy norm error (this mainly applies to areas of high stress and sharp geometry) and the results will typically be within the same region as doing a H-adaptive study set around ~92+% accuracy. I.e stress and displacement values are generally converged to within 8% of a mesh that has half the density. It was a great rule of thumb and I have generally matched the results via hand calcs and they were accurate enough, generally only a couple of percent difference. The areas around the part that are around the 100% mark all fall back to significantly lower energy norm error values (usually around that 5% or lower referenced in the studies above) Unfortunately the geometry in my current model is too complex for that approach this time around and third party verification would like to see some convergence checking or documentation to support he stress error plot being used.
  Like • Show 0 Likes0
  Actions

Mixed mesh simulation convergence / error check

Outcomes

Related Content

Recommended Content