FEA's a math model that's been around for most of a century, its simply gone digital in recent decades. Like any tool, in the hands of someone that doesnt know what they're doing its very easy to input the wrong value and get a garbage answer that looks like a pretty possibility (the old "garbage in, garbage out" mentality). In the hands of someone knowledgeable however its a very powerful, very accurate tool.
I'm real curious how its results compare to real world tests. Like say bolt clamping distribution of pressure across a surface, such as flanges and blocks and heads etc. I would think it would be very difficult with part shapes and materials loaded with so many variables.
If the model is accurate, the results will be accurate.
You can be darn sure that FEA models of a connecting rod, let's say, will account for the clamping load of the rod bolts.
If something needs to be finely analysed then the mesh size also has to be fine, to capture the details of the shape and the intricacies of the load distribution.
It still all just calculation though, not true test, so in a sense kinda like fancy guess work? I've never been exposed to FEA, is there an inexpensive learners version some place?
There have always been "calculations". That's what engineering is - a bunch of mathematical methods for predicting reality. New techniques for mathematical modelling of any kind are always validated before general acceptance. This done by comparing to actual experimental results (true test). That's what a large percentage of University research does.
MIT has an online course** describing how FEA methods work. Basically, if you believe that stress = force/area and that strain = stress * some point on a graph of stress vs strain, then FEA is pretty good.
Where FEA usually gets odd answers is in cases that real parts cannot have, such as perfectly sharp inside corners where stress = force/0. However, making the mesh smaller in those areas can limit the distance those reach and otherwise can be ignored. In real parts, were they made to such sharp changes, the parts will plastically yield some tiny amount to spread the load over a non-zero area.
All models are wrong, some models are useful. Currently the FEA model of a car hitting a concrete block is sufficiently accurate that the airbag calibration can be (and is) done on the model, rather than in real crash tests. Those are very good models, that run on supercomputers. OTOH we don't believe the predicted fatigue life from an FEA model to better than about a factor of 2 on stresses, partly because it is difficult to measure the exact input loads during a durability test.
Cheers
Greg Locock
New here? Try reading these, they might help FAQ731-376
FEA is just another tool and, like any tool, it can be misused. As far as I know, nobody designs to the limit indicated by FEA, they apply a safety factor. The gain from FEA is that it reduces uncertainty and thus allows use of smaller safety factors than the old rule of thumb methods yielding lighter and less costly assemblies. I read a great article a long time ago about the lifespan of the last fighter designed using old-school rules, the F-15, and the first designed using FEA, the F-16. The F-15 is proving to have a lot longer life span than specified even when loaded far beyond the original specification (i.e. the Strike Eagle) while the F-16 structures are failing from fatigue stress very near the specified life span. The article did similar comparisons to old school bridges and modern bridges, and the result was the same. Some say that extra strength resulting from use of larger safety margins driven by greater uncertainty is a good thing while others say it's over-design. No doubt the Air Force is happy the F-15 lasts far beyond their original requirement, but in my mind, that implies their original requirement didn't accurately reflect their true desires.
Yes and no. Within single part and simple assembly design its easy enough to validate models to trust them to a very high degree. Where "safety factors" typically come into play IME is at the engine assembly level where the decision is often purposely made to overbuild an engine to account for future performance upgrades, lousy fuel overseas, and other potential issues. I purposely use quotes around "safety factors" as its not really a FoS so much as it is often a CYA business decision to prevent poor planning by the marketing folks translating into executive blaming of engineering.
"The old ways of analyzing" gets you a Ford Model T. Or maybe the 1950s-era iteration of the Chevrolet small block V8, if you wish.
"The new ways of analyzing" gets you a Ferrari F1 engine. Or a MotoGP engine. Or a compact 4 cylinder engine for a normal daily-driver car that beats the 1950s-era Chevy small block in every meaningful way.
Safety factors are indeed "factors of ignorance".... ignorance of material variance across suppliers and lots, ignorance of workforce assembly quality, ignorance of consumer misuse and abuse, ignorance of whether FEA using a given mesh size and generalized material yield at estimated temperature are all accurate, etc. The magnitude of the safety factor depends on how well controlled all the factors encompassing the real-word fabrication of components combined with the consequence of failure; money can be thrown at improving control over design and fabrication factors (model fidelity, material variance, assembly quality, and so forth), and the safety factor can be relaxed if the consequence of failure isn't severe. Folks keep mentioning F1 engines which are modern marvels, but those engines are inspected and hand rebuilt by experienced professionals at intervals the common consumer would never tolerate. Furthermore, the drivers wear a lot of safety gear and race on a controlled course with rescue personnel nearby, so the consequence of failure is low. Contrast this with a consumer engine which is manufactured on an assembly line, must operate for hundreds of thousands of miles between rebuilds, operated in a real world environment with unconstrained stresses, and is maintained by less qualified personal. Professionals know there's a lot they don't know and use safety factors to ensure a product will yield customer experiences leading to a strong reputation and increased sales.
I'm always careful when I use the term "safety factor" as I don't care to have something which I view as good and ethical to be confused with laziness or ignorance. A safety factor to me accounts for the unknown rare event - 100 year flood, unforeseeable perfectly wrong combination of driver inputs, etc, and in many cases are a necessity but somewhat of a rarity in engine design. Things that are not only known but well within the designer's control like variation in tolerancing and material specs shouldn't be a sole cause for a FoS being applied as most every FEA code has the ability to quickly analyze this manufacturing variation.
