-
1
- #1
Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations cowski on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
Basics of FEA 20
- Thread starter cwmullion
- Start date
-
1
- #2
Twoballcane
Mechanical
ANSYS is the respectable one among analysis people, but if your company uses Pro E, than Pro Mechanica may be the way to go. If you search on Amazon.com, there are many great workbooks and books on theory to choose. You can go by the reader’s ratings.
Tobalcane
"If you avoid failure, you also avoid success."
Tobalcane
"If you avoid failure, you also avoid success."
-
1
- #3
You could certainly start with the Wikipedia article:
TTFN
FAQ731-376
Chinese prisoner wins Nobel Peace Prize
TTFN
FAQ731-376
Chinese prisoner wins Nobel Peace Prize
-
5
- #4
ASME offers some classes that might help you out.
As far as books go, Finite Element Method - Its Basis and Fundamentals by Zienkiewicz & Taylor seems quite good and is available for free on the ASME e-library. I used Concepts and Applications of Finite Element Analysis by Cook in my graduate classes. Knowledge of continuum mechanics, energy methods, and the theory of elasticity will make these books much more digestible.
Before you get too carried away, though, the most important thing is to understand the fundamentals of the analyses that you're trying to perform. In the case of structural analyses, you should have a good grasp of stress, strain, thermal strain, material science, thermally dependent material properties, fracture and fatigue, and of how materials fail. You should be able to determine acceptability limits for your analyses. You should understand stress concentrations and their impact on designs. Strictly speaking, the same concepts that are used in analytical calculations are often used in finite element analyses (if nothing else, for head-checking what the computer churns out).
A few pointers for structural FE analyses:
-Before you begin an analysis, you need to have a clearly defined goal. You need to make judgments about where stresses matter and where they don't before you begin (usually supported by hand calcs). Things tend to crop up during the analysis, so be prepared to make sub-models and use hand calculations to address them.
-Sharp interior corners are stress/strain singularities in an elastic FE model (i.e. stress will keep rising as the mesh is refined). Stresses in interior fillets must often be dealt with through hand calculations or through sub-models.
-Thermal stresses and temperature dependent material properties should be considered for anything that's not operating at room temperature.
-FE Models become quite unwieldy when you incorporate nonlinearities (a-la contact, plasticity, radiation, etc.). You have to use engineering judgment and hand calculations to determine whether these are necessary.
-Welds suck to analyze. Period. Especially socket welds, fillet welds, or any weld that might result in a buried crack. The best practice is to take the loading on the weld from your FE model and do hand calculations to determine whether it is acceptable. Blodgett's Design of Welded Structures book (from Lincoln Electric) is pretty good for that. The other approach is to model the weld. A weld with a buried crack requires an ugly elastic-plastic analysis to determine the stress distribution. Crack growth and fatigue failure is another ugly thing to analyze.
-When in doubt, refine your mesh. An FE mesh can never be too fine (other than that it might take weeks to solve). So the best approach is to start with a courser mesh and refine it several times to ensure that your stress numbers aren't changing. This is called checking for "grid-independence". If you're concerned about a particular feature, plot the stress in that feature at each refinement. You'll either find that it approaches some value or heads to the moon. If it heads to the moon, your model's wrong.
-Use brick elements over tets whenever possible. Higher order tets aren't too bad when it comes to accuracy, but they're a killer when it comes to problem size. Just consider that one brick element necessarily equals several tet elements (sketch it out). If you can built a model with brick elements, you can refine the mesh further while keeping your node count reasonable.
-2D models are your friend. If you can build something as an axisymmetric or planar, you'll be able to achieve a finer grid far more easily. If you can build something out of shell elements, do it. However, you need to be mindful of their limitations as well.
-2D Plane stress = thin; 2D Plane strain = thick. Think of a model of a pipe cross section (thick, into the page) vs a model of a plate with a hole (thin, into the page).
Well, I hope that helps. Good luck.
As far as books go, Finite Element Method - Its Basis and Fundamentals by Zienkiewicz & Taylor seems quite good and is available for free on the ASME e-library. I used Concepts and Applications of Finite Element Analysis by Cook in my graduate classes. Knowledge of continuum mechanics, energy methods, and the theory of elasticity will make these books much more digestible.
Before you get too carried away, though, the most important thing is to understand the fundamentals of the analyses that you're trying to perform. In the case of structural analyses, you should have a good grasp of stress, strain, thermal strain, material science, thermally dependent material properties, fracture and fatigue, and of how materials fail. You should be able to determine acceptability limits for your analyses. You should understand stress concentrations and their impact on designs. Strictly speaking, the same concepts that are used in analytical calculations are often used in finite element analyses (if nothing else, for head-checking what the computer churns out).
A few pointers for structural FE analyses:
-Before you begin an analysis, you need to have a clearly defined goal. You need to make judgments about where stresses matter and where they don't before you begin (usually supported by hand calcs). Things tend to crop up during the analysis, so be prepared to make sub-models and use hand calculations to address them.
-Sharp interior corners are stress/strain singularities in an elastic FE model (i.e. stress will keep rising as the mesh is refined). Stresses in interior fillets must often be dealt with through hand calculations or through sub-models.
-Thermal stresses and temperature dependent material properties should be considered for anything that's not operating at room temperature.
-FE Models become quite unwieldy when you incorporate nonlinearities (a-la contact, plasticity, radiation, etc.). You have to use engineering judgment and hand calculations to determine whether these are necessary.
-Welds suck to analyze. Period. Especially socket welds, fillet welds, or any weld that might result in a buried crack. The best practice is to take the loading on the weld from your FE model and do hand calculations to determine whether it is acceptable. Blodgett's Design of Welded Structures book (from Lincoln Electric) is pretty good for that. The other approach is to model the weld. A weld with a buried crack requires an ugly elastic-plastic analysis to determine the stress distribution. Crack growth and fatigue failure is another ugly thing to analyze.
-When in doubt, refine your mesh. An FE mesh can never be too fine (other than that it might take weeks to solve). So the best approach is to start with a courser mesh and refine it several times to ensure that your stress numbers aren't changing. This is called checking for "grid-independence". If you're concerned about a particular feature, plot the stress in that feature at each refinement. You'll either find that it approaches some value or heads to the moon. If it heads to the moon, your model's wrong.
-Use brick elements over tets whenever possible. Higher order tets aren't too bad when it comes to accuracy, but they're a killer when it comes to problem size. Just consider that one brick element necessarily equals several tet elements (sketch it out). If you can built a model with brick elements, you can refine the mesh further while keeping your node count reasonable.
-2D models are your friend. If you can build something as an axisymmetric or planar, you'll be able to achieve a finer grid far more easily. If you can build something out of shell elements, do it. However, you need to be mindful of their limitations as well.
-2D Plane stress = thin; 2D Plane strain = thick. Think of a model of a pipe cross section (thick, into the page) vs a model of a plate with a hole (thin, into the page).
Well, I hope that helps. Good luck.
-
1
- #5
GregLocock
Automotive
+Boundary conditions
+load application
+stupid use of rigid links
Those account for more errors than anything fancy
Cheers
Greg Locock
New here? Try reading these, they might help FAQ731-376
+load application
+stupid use of rigid links
Those account for more errors than anything fancy
Cheers
Greg Locock
New here? Try reading these, they might help FAQ731-376
-
3
- #6
cwmullion, FEA is just a tool, yes a very powerful tool, but still just a tool, and as with many other powerful tools (nail gun, chainsaw....) in the wrong hands it has the capacity to cause great harm (I'm guessing, but consider the possible underlying reasons for engine failures on passenger aircraft, who would bet against FEA being in there?).
As flash said, you need to understand the problem you wish to solve before you delve into FEA. FEA is then merely a refinement of your hand calc. Don't accept the argument that some problems can only be solved with FEA (anyone seen any FEA of Concorde parts when it was designed?).
Some software vendors make sweeping judgements and claims about their products that mislead gullible managers into thinking that the software does all the work and that they no longer require FEA experts to run it, as if the software possessed artificial intelligence.
The reality is exactly as Greg has pointed out and no software on the market is capable of questioning incorrect boundary conditions being applied.
As flash said, you need to understand the problem you wish to solve before you delve into FEA. FEA is then merely a refinement of your hand calc. Don't accept the argument that some problems can only be solved with FEA (anyone seen any FEA of Concorde parts when it was designed?).
Some software vendors make sweeping judgements and claims about their products that mislead gullible managers into thinking that the software does all the work and that they no longer require FEA experts to run it, as if the software possessed artificial intelligence.
The reality is exactly as Greg has pointed out and no software on the market is capable of questioning incorrect boundary conditions being applied.
Twoballcane
Mechanical
FEA should be thought of as correlation to your hand calcs and presenting pretty pics to management.
Tobalcane
"If you avoid failure, you also avoid success."
Tobalcane
"If you avoid failure, you also avoid success."
-
1
- #8
rowingengineer
Structural
FEA analysis varies with materials used, what are you planning to model, for example if you modelling concrete the FEA tool needs to be adjusted for many different cracking situation’s.
ANY FOOL CAN DESIGN A STRUCTURE. IT TAKES AN ENGINEER TO DESIGN A CONNECTION.”
ANY FOOL CAN DESIGN A STRUCTURE. IT TAKES AN ENGINEER TO DESIGN A CONNECTION.”
-
2
- #9
I found that starting with simple elements like beam elements helped understand the concept initially. I did a few calcs by hand (using stiffness matrices) and checked them against Nastran results for the same basic problem. Its fiddly but I learned a lot this way about the importance of a few things like boundary conditions. I also appreciated the importance of using hand calcs to check your results, to make sure you're in the right ballpark. And if there are differences, it helps to try understand why.
- Thread starter
- #10
-
2
- #11
-
1
- #12
Nileo2005,
I cannot help but smile. It would not be the first time that the best technical solution would have been to give the local kindergarten a box of colored pencils and a few prints. Even with $50 in the Christmas party fund, it would have been a steal to decorate the walls of the PMs office.
"But, but .... we have paid $500k, we have thousands of pages of printout, and scores of coloured drawings, it must be correct". You are now in the real world where RI=RO. You are now $500k lighter. More that twice that because you need to have it done again - properly by somebody who know what he is doing.
Even more if you have built the equipment and it has already failed at a cost of +$10m.
I cannot help but smile. It would not be the first time that the best technical solution would have been to give the local kindergarten a box of colored pencils and a few prints. Even with $50 in the Christmas party fund, it would have been a steal to decorate the walls of the PMs office.
"But, but .... we have paid $500k, we have thousands of pages of printout, and scores of coloured drawings, it must be correct". You are now in the real world where RI=RO. You are now $500k lighter. More that twice that because you need to have it done again - properly by somebody who know what he is doing.
Even more if you have built the equipment and it has already failed at a cost of +$10m.
-
1
- #13
you can study in user manual of program (getting start ) and general finite element method book (J.T Reddy and R.D. Cook are introduced (Finite element analysis))
thank you
Rubber engineering
thank you
Rubber engineering
-
1
- #14
- Status
Similar threads
- Question
- Question
- Locked
- Question