I am reviewing an FEA performed by a third party using Autodesk Simulation Mechanical. Within the program I am most familiar with (ANSYS) convergence is used to ensure an accurate result of the von Mises stress. Simulation Mechanical does not seem to have a similar tool so I would like to ask how the results can be assured to be accurate. The autodesk website states that mesh convergence is partially automated but this doesn't confirm to me that an FEA will be accurate automatically using this program. Can anyone shed light on the automatic convergence or accuracy of the program?
Eng-Tips is the largest engineering community on the Internet
Intelligent Work Forums for Engineering Professionals
Autodesk Simulation Mechanical Accuracy Check 2
- Thread starter PJ18
- Start date
-
1
- #2
-
1
- #3
IceBreakerSours
Bioengineer
Accuracy without context is meaningless. What might be accurate in one model can easily be inaccurate in another even "slightly" different model. Also, convergence is a mathematical/numerical concept. It is not a sufficient condition for accuracy so you have to check whether the physical quantities make sense to be sure the converged solution actually is meaningful. Quite often that ends up being the case but there are no guarantees.
Now, you could ask the third party for an order of convergence computation or, at least, a mesh convergence analysis. That said, as important as they might be (and they are, to be sure), numerical matters are secondary to physics.
*********************************************************
Are you new to this forum? If so, please read these FAQs:
Now, you could ask the third party for an order of convergence computation or, at least, a mesh convergence analysis. That said, as important as they might be (and they are, to be sure), numerical matters are secondary to physics.
*********************************************************
Are you new to this forum? If so, please read these FAQs:
BlasMolero
Mechanical
Dear PJ18,
If you are revising a FEA project not matter the FEA code brand used the important question here is if the task of stress convergence was performed or not as part of the project, this is up the project author and in my opinion this should be included in any FEA project, as well as the corresponding element quality check, a task that seems that most FEA engineers forgot at all (Jacobian check?, what´s Jacobian?...): the target is to demostrate that the solution is mesh independent, basically run your model, show your stress solution, clone your study, repeat the meshing task increasing the mesh density (the target is doubling the total DOF) and run the analysis: if your new stress solution is in the range of say 5% to 8% then you can say that the stress solution is convergent, the mesh density is correct, the model stress solution is mesh independent, OK?.
In displacements the solution is practically directly convergent, not need to refine mesh, unless the mesh is very coarse!!. Of course, a quality mesh check is required as well, element quality is critical in the accuracy of the stress solution, and the rules of meshing should be followed, for instance: when meshing walls with solid elements a minimum of two elements in the thickness are required in order to capture stress gradient, etc...
In summary, today not any FEA code of the market is wrong, the name of the FEA solver is not the most important, the key is the engineer to be a FEA professional: it cost years, is simply a question of time ...
Best regards,
Blas.
~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director
IBERISA
48004 BILBAO (SPAIN)
WEB: Blog de FEMAP & NX Nastran:
If you are revising a FEA project not matter the FEA code brand used the important question here is if the task of stress convergence was performed or not as part of the project, this is up the project author and in my opinion this should be included in any FEA project, as well as the corresponding element quality check, a task that seems that most FEA engineers forgot at all (Jacobian check?, what´s Jacobian?...): the target is to demostrate that the solution is mesh independent, basically run your model, show your stress solution, clone your study, repeat the meshing task increasing the mesh density (the target is doubling the total DOF) and run the analysis: if your new stress solution is in the range of say 5% to 8% then you can say that the stress solution is convergent, the mesh density is correct, the model stress solution is mesh independent, OK?.
In displacements the solution is practically directly convergent, not need to refine mesh, unless the mesh is very coarse!!. Of course, a quality mesh check is required as well, element quality is critical in the accuracy of the stress solution, and the rules of meshing should be followed, for instance: when meshing walls with solid elements a minimum of two elements in the thickness are required in order to capture stress gradient, etc...
In summary, today not any FEA code of the market is wrong, the name of the FEA solver is not the most important, the key is the engineer to be a FEA professional: it cost years, is simply a question of time ...
Best regards,
Blas.
~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director
IBERISA
48004 BILBAO (SPAIN)
WEB: Blog de FEMAP & NX Nastran:
- Status
Similar threads
- Locked
- Question
- Locked
- Question