Book suggestions: FEM case studies with experimental validations \ discussions 1

[stru-eng]

Is anybody aware of books that include a selection of FEM analyses or case studies, correlated with discussions and experimental data validations?
The request is mainly focused on structural engineering, but it would be nice to get thermal phenomena as well. I don't know if the question is unusual, but I would personally find such a reference pretty valuable

 

[karachun]

Check this theme

https://www.eng-tips.com/viewthread.cfm?qid=437411

 

[stru-eng]

Well that seems to refer mainly to reference solutions rather than to experimental validaiton, am I wrong?

 

[karachun]

Check documentation of your software - usually there are lot of verification examples, you can check documentation of other solvers like Nastran or ANSYS.

 

[NRP99]

The best resource for "FEM analyses or case studies, correlated with discussions and experimental data validations" are Journal or Conference Papers.

Following is the short list for your reference-

1)Computational Mechanics Springer-Link
2)Elsevier Journals-Link
3)ASME Journals and conference proceedings-Link and Link
4)Computational Structural Engineering-Link
5)ASCE Journal of Structural Engineering-Link

 

[sdm919]

Hi stru-eng,

You touch on an important, but often glossed over topic. Yes, there are many "verification" problems that have been run on FEM software, but almost all of them (probably all of them?) are comparisons of FEM results to theoretical solutions. After all, the Finite Element Method is a numerical method to solve a set of differential equations. So, within the context of linear elasticity, say, you can compare the FEM solution to an "exact" solution from Timoshenko. Almost always, when they say "exact" they are talking about an exact solution to a differential equation, it has nothing to do with test. The publications of NAFEMS fall into this category for the most part, I believe.

The whole comparison of FEM to reality brings up the question of the mathematical theory underlying the FEM. FEM solves mathematical theory to a greater or lesser accuracy. But how well does the chosen mathematical theory (beam theory, shell theory, elasticity theory, etc.) represent your problem is a different question. Off the top of my head, I am not aware of a good place where this full correlation (test-theory-FEM) is rigorously performed. As NR999 suggests, there are many articles where people make a test to compare against their FEM, or vice versa, but these are what I would call "one-offs". If you get a reasonable comparison (perhaps after tuning the model to match the test), for a single case, that does not necessarily translate to "accuracy" of the FEM as a method in and of itself, just that that particular model adequately represents that particular problem. That sort of comparison is done quite often on important hardware, but is usually proprietary to a company.

Another thing comes to mind. There is a British book from the 1950s, where they performed tests on steel building frames, and made comparisons with linear elastic structural analysis calculations (not FEM of course). The results were different enough to cause a revision of structural thinking and lead to the plastic methods of structural design for steel. The book is: The Steel Skeleton, in 2 volumes, by J.F. Baker, published by Cambridge University Press in 1954 & 1956.

I will continue looking, and will post back here if I find something relevant in the near future.

Thanks,
sdm919

 

[stru-eng]

@NRP99
The difference I found between a paper and a book is that the latter is written to *teach* something, while the former is written for expert and researchers in very narrow areas.

@sdm919
test-fem correlation may indeed be a very good keyword to start with. When we model something new we have to do a lot of choices a priori, maybe with o broader view on how different hypotheses may lead us far from actual structural behaviour. I'll keep searching for those books posting updates.
If anybody has suggestions even for reports, feel free to share.

Thank you sdm919 for the excellent resource about plastic design, I'll try to get a copy

 

[NRP99]

stru-eng
IMO the books are just simplified and understandable version of tons of research papers. If there are no research papers, the books will be just information of limited knowledge.

sdm919
To contrary, test-FEA correlation can't be just "one off". Tests are carefully done to get meaningful results. Same way FEA models are matched/tuned with test parameters to get behaviour of the component as close to test results as possible. Does this mean FEA is not accurate or does this translate to just single case of FEA-test correlation for that particular component/case? Can't we extend the same correlation to a different problem but similar phenomena? From results we get the information on real material/component behaviour which is just unpredictable. If we are to use the test FEA correlation for our similar or different problem, we are replicating the actual phenomena but for different problem which confirms our understanding of the component behaviour so as to fix our design strategies or design margins.
It is really good to have some meaningful actual results available to all in the FEA world of approximation and it would not be called as one off.

 

[ShellsPlatesMeshes]

This book contains several examples aimed at comparing analytical and numerical (FEA) solutions for shells:

http://members.ozemail.com.au/~comecau/quad_shell.htm