Anyone good with Finite Element Analysis (FEA)?

[Triumph]

Need to get into this stuff for work (engineer). Used it some in school, done some tutorials with Pro/Engineer's Mechanica package. It seems that either this software is very limited and doesn't work very well, or I'm not using it right. The former may be true compared to other FEA packages, from what I've read on a few websites. The latter is definitely true, which I'm hoping to cure through guided training on how to use it, instead of just messing around with settings.

Anyone know any good books for teaching FEA? I plan on taking a class sometime, but I want some things to do on my own time, as well. Anyone with experience with other software?

 

[TecHNooB]

PM me your AIM for a book you might find useful

 

[CycloWizard]

There are many different finite element methods (h, p, hp, others), and most textbooks focus almost exclusively on one or the other. The only one that I know of that covers just about everything, though more as a survey, is the Finite Element Handbook. It's a fairly old book, but the nice thing about math is that not many things change. If you have this book and a manual for your particular software, you should be able to solve most problems.

 

[PsiStar]

Do you have a package now?

 

[Triumph]

Do you have a package now?

Yes, Pro/Engineer. I've done some tutorials with it but I think that more guided tutorials than the 1 book I have, would help me understand better how to use it.

 

[PsiStar]

What kind of problems are you interested in.

BTW I am a sparky, so I will not pretend to get too deep, but I am/was a killer with a very specific type of problem. :twisted::twisted::twisted:

 

[bobsmith1492]

What kind of problems are you interested in.

BTW I am a sparky, so I will not pretend to get too deep, but I am/was a killer with a very specific type of problem. :twisted::twisted::twisted:

Thermal analysis, by any chance?

 

[PsiStar]

I have a static thermal solver. What I was referring was nonlinear mechanical spring analysis ... but was too expensive for me to keep versus the amount of work I had of that type. I used COSMOS/M then SolidWorks when it was all incorporated into it.

For Pro/E they used to have some pretty good classes offered. But really it depends on the type of problem you want to work on. The software has become so refined & inexpensive (the deceptive part), that it can solve most anything. So I suggest that you figure out the most common type of problem that you would need to solve & investigate FEA for working on that. Maybe find a consultant to give you a good kick start?

 

[Born2bwire]

It would be helpful if we knew more about the type of problems that you want to solve as FEM can be applied to a large variety of situations. I know a good book but it is for electromagnetic simulations which I doubt is what you are interested in.

 

[Triumph]

It would be helpful if we knew more about the type of problems that you want to solve as FEM can be applied to a large variety of situations. I know a good book but it is for electromagnetic simulations which I doubt is what you are interested in.

Very true. Looking mainly at static stress analysis, can this structure support this weight. Simple stuff, but most of the time I can't get the results to converge. So I'm not doing something right, either with the model, or the settings, or I'm just expecting too much from it.

 

[CycloWizard]

Very true. Looking mainly at static stress analysis, can this structure support this weight. Simple stuff, but most of the time I can't get the results to converge. So I'm not doing something right, either with the model, or the settings, or I'm just expecting too much from it.

Do you have any background in how FEA works? I'm not familiar with the software you're using, so I can't speak to your particular problem, but I could give pointers if you have a basic understanding of the process. Convergence is a huge issue in FEA - perhaps the least understood major issue. There is solver convergence (i.e. where an iterative solver finds a solution to the working model on a given mesh), then there is solution convergence (i.e. where the solution of the working model is independent of the mesh parameters for the quantities of interest). The two are very different and ways of addressing them are completely different.

 

[Born2bwire]

Do you have any background in how FEA works? I'm not familiar with the software you're using, so I can't speak to your particular problem, but I could give pointers if you have a basic understanding of the process. Convergence is a huge issue in FEA - perhaps the least understood major issue. There is solver convergence (i.e. where an iterative solver finds a solution to the working model on a given mesh), then there is solution convergence (i.e. where the solution of the working model is independent of the mesh parameters for the quantities of interest). The two are very different and ways of addressing them are completely different.

I agree. Even with commercial software, it is best to have an understanding of how the algorithms work. Software that use finite element, finite difference, moment method, etc. are not foolproof and require a level of understanding of the techniques themselves to be used effectively. Most commercial solutions, however, are pretty basic in their implementation. They often do not incorporate a large number of advanced techniques even if they offer appreciable improvement because the background knowledge to effectively use these techniques is increased. So the general theory and basic examples that you can read up on are probably not too far removed from what the program is actually implementing.

The kinds of things that a good background can provide is being able to determine what you can and cannot use the software to do. Other examples are generating a good mesh, validating results, and deciding what kind of compromises you can make to your model and still obtain reasonable results (which can mean the difference between a simulation taking hours or minutes). These examples can greatly help the efficacy and speed of your simulation. Another thing to consider is that many of the companies that make such software have engineers on staff to answer technical questions. I have used them myself when I did not know if my ideas could be properly implemented with their solver.

 

[Triumph]

Do you have any background in how FEA works? I'm not familiar with the software you're using, so I can't speak to your particular problem, but I could give pointers if you have a basic understanding of the process. Convergence is a huge issue in FEA - perhaps the least understood major issue. There is solver convergence (i.e. where an iterative solver finds a solution to the working model on a given mesh), then there is solution convergence (i.e. where the solution of the working model is independent of the mesh parameters for the quantities of interest). The two are very different and ways of addressing them are completely different.

Mainly just running through the examples in this book. That's the extent of it, really.

 

[CycloWizard]

Mainly just running through the examples in this book. That's the extent of it, really.

If you bought a textbook with examples for one particular FEA program and the program can't even find a solution when you follow the instructions he gives, I would contact the author and ask why he's selling a defective product. I can't see any of the examples in the book, but I've never had a problem solving any tutorial problems provided by other FEA publishers (ABAQUS, COMSOL, StressCheck, etc.).

 

[CP5670]

Anyone know any good books for teaching FEA? I plan on taking a class sometime, but I want some things to do on my own time, as well. Anyone with experience with other software?

I used this book to learn about finite elements and other such methods, although its emphasis is more on the theory than any actual software packages.

 

[Triumph]

If you bought a textbook with examples for one particular FEA program and the program can't even find a solution when you follow the instructions he gives, I would contact the author and ask why he's selling a defective product. I can't see any of the examples in the book, but I've never had a problem solving any tutorial problems provided by other FEA publishers (ABAQUS, COMSOL, StressCheck, etc.).

No I mean for my own designs, I have trouble getting convergence. Not the examples in the book. Although you bring up a good point, something I wanted to mention. The author includes in his tutorial an example of a simple part that looks something like a crankshaft connecting rod, with an extra bend in the middle. Even in the tutorial, the results never converge, and the author simply states (paraphrasing), "As you can see, not all models will converge easily, and this is the type of problems you'll have with FEA." This is a simple lever with a fixed constraint on one end and a load on the other! Meanwhile I want to analyze entire assemblies. If this is the limits of FEA, no wonder I'm having trouble.

 

[CycloWizard]

No I mean for my own designs, I have trouble getting convergence. Not the examples in the book. Although you bring up a good point, something I wanted to mention. The author includes in his tutorial an example of a simple part that looks something like a crankshaft connecting rod, with an extra bend in the middle. Even in the tutorial, the results never converge, and the author simply states (paraphrasing), "As you can see, not all models will converge easily, and this is the type of problems you'll have with FEA." This is a simple lever with a fixed constraint on one end and a load on the other! Meanwhile I want to analyze entire assemblies. If this is the limits of FEA, no wonder I'm having trouble.

How is the load applied? Is it a point load? If so, that is not valid input in FEA, as it induces a strong singularity (though it results in largely local errors). If the mesh is refined around the point of application, then convergence of the solution should be improved. If you add in contact between multiple bodies, then things become more complex still and the best method for addressing them depends largely on the software implementation of contact detection. Sorry I can't be more helpful, but all of these things are software-specific so I can't even guess how you might address them. The good news is that there are a few companies working to essentially standardize FEA tools using hierarchical hp-extension-based methods which are more efficient and easily implemented. The bad news is that all of the more popular softwares have invested a lot in their package already and are very reluctant to change, especially since it would open the black box that they now have.

 

[tritonofg]

I have used PTC Pro/Engineer and Pro-Mechanica for years and Solidworks Simulation for 17 months and can tell you that taking a class by the vendor is worth it. pro/mechanca uses a p element solver and requires that the use set up the boundary conditons correctly ... These are after all ordinary differential equations. If your bcs are not correctly set up you will always have a displacement which it outside of the specified convergence...or you are running out of resources before it can converge. Lastly...make sure that you have the correct ptc lic for what you are trying to do...basic mech will not have all of the opt ins that the adv lic has

 

[radioouman]

CATIA FEA and Abaqus here.