Approximating the effects of a roll cage in a production car

[Greg Locock]

"It's not easy though even at the manufacturers they struggle to do FEA on the original body with much accuracy."

True 30 years ago. Not true now. We are very close to being able to set up our crash sensor system using FEA alone, and crash is much harder to get right than static stiffness or linear dynamics.

I'm not sure what FEA program you are describing, but if you don't model the test rig as well then you are not going to get decent results.

Using the 100 node limited version of Femap/nastran it is perfectly possible to build a model of a ladder chassis car that correlates well with real world tests.

[gixxer_drew]

"It's not easy though even at the manufacturers they struggle to do FEA on the original body with much accuracy."

True 30 years ago. Not true now. We are very close to being able to set up our crash sensor system using FEA alone, and crash is much harder to get right than static stiffness or linear dynamics.

I'm not sure what FEA program you are describing, but if you don't model the test rig as well then you are not going to get decent results.

Using the 100 node limited version of Femap/nastran it is perfectly possible to build a model of a ladder chassis car that correlates well with real world tests.

That's interesting. I was having a chat with an engineer at a large Japanese manufacturer and he was telling me they really struggle to get good correlation in absolute values. Good for relative stuff but he was talking about some fairly large inaccuracies. Worse than I usually see with CFD. Or maybe I should rephrase that from "they struggle" to "its not easy to achieve"?

With that said, either way they seem not make that data available to race teams.... at least not when I asked! My eyes are roundish though, that may be the problem.

[Greg Locock]

When i started here 22 years ago we had 3 guys doing FEA, now we have 80. The models are at least 10 times bigger, and are non linear, not linear.

Yes, I'm not suprised you won't get details from any OEM, notice that everyone quotes ratios not absolutes in their PR stuff.

If your deflections are wrong, your stiffnesses are wrong. if your stiffnesses are wrong then your dynamic frequencies (root k/m) won't match. Since we spend a lot of time on modal analysis then i am pretty sure that the frequencies match well these days. FEA is absolutely fundamental to the way cars are designed now, it is used for - noise, vibration, static stiffness (not much of that), durability and crash.

[gixxer_drew]

When i started here 22 years ago we had 3 guys doing FEA, now we have 80. The models are at least 10 times bigger, and are non linear, not linear.

Yes, I'm not suprised you won't get details from any OEM, notice that everyone quotes ratios not absolutes in their PR stuff.

If your deflections are wrong, your stiffnesses are wrong. if your stiffnesses are wrong then your dynamic frequencies (root k/m) won't match. Since we spend a lot of time on modal analysis then i am pretty sure that the frequencies match well these days. FEA is absolutely fundamental to the way cars are designed now, it is used for - noise, vibration, static stiffness (not much of that), durability and crash.

Thanks for the info. I'll concede it sounds like the manufacturer level is ahead of what I thought with FEA. The guy I talked to was in the heavy equipment division of a company that also builds cars.

"Easy" may be a relative term though, but I'm guessing theres a group of specialized people and practices that took some years to develop. I'm sure F1 is doing it but I havent seen much chassis FEA work outside the manufacturers and they usually say the problem is cost/difficulty. Just having an accurate CAD model of a production car chassis sounds like a huge project to me.

[Greg Locock]

80 people/2 programs*2 years*200000$pa= 16 million dollars. yes that is quite a lot. Our group employed a guy who did FEA for an F1 team. For whom he analyse conrods. Every day. All day.