View Single Post
      06-19-2008, 02:13 PM   #43
Major General
lucid's Avatar

Drives: E30 M3; Expedition
Join Date: Apr 2007
Location: USA

iTrader: (0)

Originally Posted by jm1234 View Post
Buildings and bridges undergo predictable loads? Google Tacoma Narrows (awesome video) for the most obvious bridge example and think about what happens to a building during a fire, earthquake, hurricane, tornado and/or impact (planes and cranes being recent impact examples). What loads do you think a car undergoes that are as complex in comparison?
Tornados and plane crashes would be difficult to model for a building as well. I was referring to the loading conditions a structure would experience on a daily basis--the way your car would driving around a track without crashing into anything--not the extreme cases.

Originally Posted by jm1234 View Post
Car companies spend years and tens of millions developing a new car design. There is both time and money for modeling. Not as much as in other industries but the models are simpler.
Of course they do. Where did I say, they don't? The issue is at what level of precision? Do you really think the models they build account for every millimeter of movement of the chassis for 99.99% of the behaviors the car will exhibit under different loading conditions? Is that level of precision necessary for a passenger car (it clearly was for the F1 car in the example I gave earlier)? Even if it is, is it even possible? There is cost/benefit trade-off there. And I think you are mixing up the aerospace industry's requirements with the automotive industry's requirements.

Originally Posted by jm1234 View Post
A system controller would not include a model of the controlled system in it's design. A controller would be built based on a model of or even the real controlled system. I never said the ECU would be running a model of the car.
Yes, you got me on this one.

Originally Posted by jm1234 View Post
When an engineer speaks of model validation he is talking about an iterative process not a one time check.
Yes, I said the same thing.

Originally Posted by jm1234 View Post
The sheer number of scenarios is exactly the reason that a model is more helpful than real world testing. You can run a model through the same set of thousands of hours of "real world scenarios" much cheaper and faster than real world testing.
This is obvious. That is why it is cheaper to use models than running a million real life tests. However, the inputs are a part of the simulation together with the model. And virtual inputs are not going to be as representative as the real world conditions. It is perfectly possible that when the model is tested with real world inputs, issues that could not have been found with the model using virtual inputs surface. I don't know, maybe we are actually saying the same thing here.

Originally Posted by jm1234 View Post
There comes a point later in the design process where real world testing is more valuable but at that point you have limited the parameter ranges and are designing for subjective criteria.
I disagree with your use of the term "subjective" criteria. Why won't you simply agree that models can be faulty for various reasons and produce invalid outcomes? This is often caught in testing and corrected. There is nothing subjective about that. Do you really believe BMW would spend thousdans of hours logging data in this car just to make sure it "feels" right? They are also improving the objective performance measures when they are testing. Again, we don't have comprehensive ontological knowledge of all things/systems, and the real world is a complex environment.

That said, I actually agree with you on not messing with the stiffness of the springs. That has been my position from the beginning. Whoever designed this system has spent a lot time in optimizing--by using models and real world data--and I think one would take a chance by assuming everything would work fine by swapping springs. Moreover, a system working fine might be a system far from its potential peak performance.

Last edited by lucid; 06-19-2008 at 04:13 PM.