7 replies to this topic

[desmo]

In vehicle dynamics the bicycle model is well trodden ground right? I mean it's all pretty cut and dried Newtonian physics. Except it apparently isn't. It appears there are significant holes in our understanding of the vehicle dynamics of even this simple seeming device.

http://www.newstates...w-bicycles-work

I should warn you need to register to access the full paper linked to in the article, although it is free.

I'm profoundly skeptical of the underlying soundness of much common mathematical modeling, if not its utility--reality is seldom amenable to neat mathematical analysis--and this illustrates why. Models, even well designed ones, almost inevitably omit significant variables because any gaps in knowledge or understanding in the people constructing the models get built into them and omniscience isn't a known human trait. At least physics models generally have some pretty fundamental grounding in empirical reality and predictive utility, unlike say economic models...

[Greg Locock]

whoa thar. The (often discussed here) bicycle model is a model of a car not a bicycle!

And the fact that a gyro cancelled bike is still stable has been known for ages.

The authors of the paper found a whole bunch of other stability mechanisms and seem to get good correlation. So, I think what the author of that rather breathless NS article meant was that the common perception of bicycle stability was incomplete.

Also, if you build a mathematical model of something that predicts the same behavior as the real system do you 'understand' it? If I build an ADAMS model of something, do I 'understand' it? (Quick answer to the latter, no not really, not in the same way I understand my bank balance, or a pendulum. I have a model that I can exercise, modify and measure, but the effect of the precise interactions are still masked by complexity. Every interaction is calculable, as is the overall net effect, but there may be no intuitively obvious link).

[NotAPineapple]

In vehicle dynamics the bicycle model is well trodden ground right? I mean it's all pretty cut and dried Newtonian physics. Except it apparently isn't. It appears there are significant holes in our understanding of the vehicle dynamics of even this simple seeming device.

http://www.newstates...w-bicycles-work

I should warn you need to register to access the full paper linked to in the article, although it is free.

I'm profoundly skeptical of the underlying soundness of much common mathematical modeling, if not its utility--reality is seldom amenable to neat mathematical analysis--and this illustrates why. Models, even well designed ones, almost inevitably omit significant variables because any gaps in knowledge or understanding in the people constructing the models get built into them and omniscience isn't a known human trait. At least physics models generally have some pretty fundamental grounding in empirical reality and predictive utility, unlike say economic models...

What Greg said. The bicycle model of a car isn't a representation of a push bike for obvious reasons.

Engineers construct models (all of which are simplified and all of which are wrong) as a "representation" of a real system. Most handling phenomona can be explained with a simple 2 dof bicycle model. Nothing more is needed.

This is why we don't use a FEM model of the full vehicle running co-sim with a CFD solver to find the understeer gradient of a car.

[desmo]

Thanks for the replies guys. I found a free link to the full paper:

http://ruina.tam.cor...1959SOMtext.pdf

I didn't mean to imply that mathematical models aren't useful tools for analyzing and predicting real world behaviors, only that almost inevitably they will be to some degree compromised by simplifying assumptions built in to make them practical. And it isn't by any means always either obvious or explicit exactly what those simplifying assumptions are and how they might affect model outputs.

[NotAPineapple]

Thats where the engineering comes into it. Models are always wrong but a good engineer engineer/mathmatician/physicist can ensure that:
The errors are small enough to be neglected
OR
The errors are not affecting the particular point you are looking at.

If these are done correctly, then there is no need for the skepticism.

[Greg Locock]

Ferzackly. That's why correlation is the essential part of the model building process. Over on the bloody awful LinkedIn you see software proponents committing 3 deadly sins: using a complex sim that uses the hand calc assumptions, and expecting a better answer, encouraging people to do sims when they couldn't do the hand calc, and of course not emphasizing the necessity for correlation.

[Canuck]

I work with a physicist who has recently been "made redundant" and sent off to an early retirement. However, before he's gone they want him to take two weeks to teach his newly-graduated, newly hired replacement how to interpret the complex radioactive simulations he runs. Should be a piece of cake.

[meb58]

Unfortunately, with time - experience - we begin to understand what's important and whats not...what to correlate and what not to. That's the gift of experience...or age

I work with a physicist who has recently been "made redundant" and sent off to an early retirement. However, before he's gone they want him to take two weeks to teach his newly-graduated, newly hired replacement how to interpret the complex radioactive simulations he runs. Should be a piece of cake.