Looking for a reference 1

[Worldtraveller]

I have, in the past, read at least a couple of engineering related books (as opposed to scholarly theory) about applying FEA to various structures.

In general, if possible when modelling aircraft structure, it is better to use 'lower dimensional' elements. This is the generally accepted means to model structure, but I'm looking for a good reference to use if/when I get pushback from the customer who seems to want to do everything in 3D (solid elements). An offset beam offers lots of advantages, but if you need nodes to connect other structure, a rod/quad/rod (2D/1D) can be used.

I haven't had much luck looking through my references I have (many of them are currently in storage) and my google-fu seems to be seriously lacking. It's also possible that everything I've read that has this sort of advice is proprietary from previous employers. I don't want to get into the rabbit hole of running a bunch of model comparisons/validations when I just want a general rule for new engineers and guidelines to develop for this program.

Any suggestions?

 

[IRstuff]

" it is better to use 'lower dimensional' elements"

I think that was a matter of expediency, when computers weren't capable of running 3D models in any sort of plausible timeframe. With the current crop of computers, you should run whatever is needed to accurately model your system.

We used to take major chunks of a day to run geometrical and electrical design-rule checks on integrated circuits; one of our divisions was pressed for time and only ran the geometrical checks so that they could deliver photomasks on schedule. Halfway through processing, they finally completed all the design rule checks and found that they shorted power and ground and they wound up with a huge egg on their faces in having to work up a "patch" to the processing masks and having a chunk of non-functional circuitry inside their microprocessor that was "correct by design."

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[GregLocock]

In a previous job I ran a modal analysis lab, generating rather tedious measurements of vehicle bodies and things. These were used to correlate FEA models. Often the highly paid constructor of FEA models was rather irritated when a 26 year old engineer provided data that showed his model was of little use in the real world. Since somewhat prior to that job I had actually constructed FEA models of my own I knew enough to be dangerous when it came to other people's models. and one of the most common errors was the use of simple beam elements tied to more complex representations. In particular RBE3 (I think, rigid anyway) elements seemed to be used with gay abandon, stiffening up the structure locally with no physical justification.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[Mustaine3]

There is no reason to read a book about efficiency. When the model runs in a fraction of a time (and RAM and disk space) and shows nearly the same results, then it is obvious what to use. But you must also consider the time (and so money) it takes to create the lower dimension models instead just meshing everything with solids.
On the other side there are several issues, when you analyze thin walled structures with solids. You can't take linear tets and linear hex can have issues with ourglassing and shear locking. Also composite theory is much more challenging when using solids.

 

[rb1957]

if your customer requires you to use 3D elements, do so, and laugh all the way to the bank. If you really want to help, you could show them how much they'd save if they permitted 1D/2D elements. But then remember it may not be their limitation .. maybe their upstream is creating this restriction.

another day in paradise, or is paradise one day closer ?

 

[ESPcomposites]

How about just giving a list of reasons. For example, if you use a beam element instead of modeling the beam in 3D, some of the advantages are:

- You can very quickly change the beam element's section properties via numerical changes to the inertia and area. Conversely, changing a beam modeled in 3D would require a remesh and potentially a lot of rebuilding of the model. For many cases, this alone should be a sufficient reason.

- Directly extract the shear, moment, and axial loads from the beam element. You then manually apply the Mc/I+P/A. This helps to identify critical load cases the internal load components that tend to drive the sizing. It is also more useful if you want to use plastic bending since that can be done simply with manual calculations (as opposed to a nonlinear material 3D analysis).

- As an add-on to the point above, many customers (or checkers) may insist that you provide them with the critical moment or shear in the beam's section. Depending on your pre/post, this can be a hassle if you are using 3D elements. Conversely, that is the direct output from the beam element. And lets say you want to add a joint to the beam. You do this via the moment, shear, an axial loads on the section...not the stresses from the 3D model.

- Now what happens if you want to perform a crippling analysis on the flange of a beam? This is straightforward via the outputs from the beam element (moment and axial internal loads). However, you can not realistically perform this analysis via a 3D analysis (requires nonlinear geometry, nonlinear material, and possibly a semi-empirical correction factor you won't have). So you end up going through the process of extracting the moment and shear from the 3D model's section. A rather time consuming and inefficient way of getting to the point.

- Convergence issues. It is far more obvious to determine if beam elements have converged than if a 3D model has. This is especially true when you have thin sections that may be in bending since you need several elements through the thickness to reach convergence.

- Obviously 3D models can be time consuming (from meshing, to solving, to processing). There is a lot more information there than you probably need. So if you don't have a justifiable need, you are just wasting time and resources.

- The same arguments apply for shell elements versus 3D elements. Can you imagine having to remesh a model every time you want to change the thickness? And since you need X elements through the thickness, with maximum thresholds for aspect ratio, you could potentially run into a variety of issues with a 3D model.

Rather than thinking of beam/shell elements as "older" or "inferior" to the 3D element (as some might), it is quite the opposite. The beam/shell elements are very efficient elements that give you exactly the I/O that you want without a lot "junk" in the middle. This expedites the entire process and isolates just the data you want. Of course, if you need something that only solid elements can provide then a beam/shell element is not sufficient. And sometimes this is the case. But many aircraft structures are designed to act like beams and shells so you can usually get exactly the data you want from them. That aspect can only be determined by a qualified analyst.


Brian

http://www.structuralfea.com

 

[GregLocock]

That's a good list but completely inapplicable to car bodies -in fact approximating a car body as a series of complex joints connected with beams was exactly what I had been working on. I don't know enough about aircraft to say whether they can be adequately modelled as beam and shell elements. If each 'joint' is small and efficient then I suppose so. In comparison, the joints on cars, say top of the A pillar, were large compared with the surrounding beams and not efficient, as there are no internal bulkheads to prevent oil canning of the large radii shells it is made of.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[ESPcomposites]

Greg,

As I mentioned, beams/shells are only appropriate *if* they capture the physical behavior of the structure. If that is the case, they are usually preferred to solid elements for reasons stated above (the list is longer but I stopped there). The OP did specifically state the application was aircraft structures so I wanted to address the question from that angle.

Aircraft "loads models" almost always use beams/shells to determine the internal loads in the structural elements. This allows the model be flexible as we adjust properties such as inertia and area. The adjustments are numerical as opposed to a total remesh. After we have the internal loads from the FEM, we apply classical solutions to the structural elements. These solutions may account for nonlinear and/or semi-empirical effects not easily (or impossible) to capture via finite element analysis.

In general, for the aircraft industry, we use 3D elements for specific analysis solutions where it is required, but we don't start there. We usually first see if the model is sufficient with beams and shells. There are too many advantages with those elements, provided they can give you the output you desire. Aircraft structures (wings, fuselage, frames, floor beams, shear ties, etc.) tend to act like classical beams and shells so we can usually get away with this approach. However, as you have mentioned, this won't be the case across the board for every industry and application. I have made plenty of solid element models, but I don't go in that direction if it is possible to use beam/shell elements.

Brian

http://www.structuralfea.com