Slashdot Mirror
CAE Tools for Car Performance Modifications?
RevHead asks: "Although after-market performance modification of cars is a discipline which claims a significant following all over the world, most of the information available on the topic tends to be more of anecdotal nature. To add to this and the plethora of conflicting information out there,
most of the tips and techniques tend to be of 'do it and see if it works' type of experimentation. I am interested in the simulation approach prior to actual experimentation to get a decent picture of what to expect during the experimentation phase, which IMHO should be safer and more cost-effective. Has anyone resorted to this approach (successfully) when it comes to engine modification, suspension design, aerodynamic performance and emission control? If so what software is available for these tasks? Which are the most popular/most effective? Does anyone know of any public-domain automotive engine models available for CAE applications such as Catia and ADAMS?"
Instead of fixing up the car, modify your life by getting rid of it entirely. You will learn patience and calm as you wait for the bus. Seriously! You will also be able to read more. Less stress, more thought, all good! No simulation needed!
I guess one really needs to look at the problem and figure out if your really need all those and other variables in the model.
There was a great example of this in economics, matmaticians spent years attempting to model market performance, so they could properly price options. The equation took everything in to account, and was pages long.
Finally a few guys figured out that most of the variables effectively cancelled each other out, and came up with a five variable or su equation that could accurately plot market performance and stock option prices in real time at any resolution (continuous time).
The auto manufacturers do use super computers to model the fluid dynamics through the intake, compustion and exaust systems, but starting with the basic functioning and building a rough model would probably allow an modder to determing if putting a larger turbo on would add any torque, or if a larger intake and exaust system would be necessary first.
What I do find humourous many times is all these tweeked Toyotas with all the laughable stickersand emblems that smoke their tires off the line. They whole point of the engine mods is to get the powet to the pavement to move the car, or so I thought. Making pretty white smoke and noise I could do with a small windshield pump and brake fluid installed on the drive wheels of any vehicle.
Article X: The powers not delegated... by the Constitution...are reserved...to the people
I'm an aero engineer who also happens to be a car nut. It seems like you would like to apply engineering rigor and the computational capabilities commonly available today to get a grip on the real physics behind various car modifications.
Unfortunately, the answer is that the availability of tools for these purposes to an engineering-minded enthusiast who is not directly affiliated with professional automotive engineering is fairly low. I think some people have posted links to various "desktop dyno" kinds of programs, and that is certainly a step in the right direction, but the fundamental problem is that fields such as aerodynamics are really hard to rigorously model, period.
To support this assertion, I'm going to ramble for a good while about the aerodynamics aspect of the answer to your question.
Even in the professional realm (i.e. Aerospace and Automotive Engineering Industries) you have tools such as CFD (Computational Fluid Dynamics), but very few good aerodynamicists in this world will trust CFD solutions on a magnitude-for-magnitude basis. There are a lot of specific cases or realms in which CFD folks can point to their solutions and say "for this aspect of the vehicle engineering, we trust the results of this CFD code" but CFD is never a substitute for wind tunnel or physical testing kind of work in conjunction with expertise & experience. You may say, "well duh, that's obvious, the same is true of computers, etc." and you'd be right, but that "phenonmenon" applies to a significantly higher degree to CFD. Moreover, low-speed, "low compressibility" kinds of flows (IOW "car aerodynamics") are some of the most difficult to model computationally.
Setting aside all these kinds of "how much do you trust the results" questions, you should also realize that high-fidelity solutions require mighty close attention to the way you model (e.g. physical features -- small features can make a big difference in the flowfield or stress analysis) and "discretize" (grid) your geometry. To put it another way, it requires a *lot* of time and experience to get the most (or even worthwhile) results out of the tools we do have. There are people whose full-time job it is to generate *just the grids* for CFD cases for *one* part of an airplane or aircraft engine, etc.
After all this, some will correctly point out the "analysis paralysis" problem, and that's definitely the case. If you want your model to capture the vast majority of the physical effects going on, then you'll have to (for the case of an engine for example) get down 'n dirty with seal/friction calculations (e.g. rings, crank seals), structural dynamics (e.g. crank torsion and vibration), reacting and non-reacting flows, and the list goes on... You can probably neglect a lot of these in order to get an answer to a specific topic or modifcation you're exploring, or to get "trends" as you change relevant variables, but if you want to conclusively prove in an engineering sense (via modeling) that some exhaust headers are going to be superior to another design, you'd better be prepared to spend a crapload of time with CFD, solid models, and actual engine hardware to measure the hell out of the relevant parts. Though parenthetically, it's still my personal plan to (once I get the right project car) design my own optimized turbo exhaust manifold with whatever CFD-type tools I can come up with or get access to. I predict the need to do a lot of straight-up control volume analysis in solving that problem as well.
Or you take the approach that the vast majority of car modifiers do: use experience and intuition (the latter grows with the former) plus any sort of engineering/science background you do have to come up with candidate designs, and try them out. Unfortunately, the results from these kinds of tests frequently lead to theories created by people without the formal background to know whether or not what they postulate really makes sense. That leads to what you describe with conflicting, incomplete, vague, and sometimes logic-defying "conclusions" that are presented to the auto-modifying enthusiast crowd (see the persistent fallacy that you "need backpressure" for an engine to run or run well or "make torque"). Make no mistake, while the analytical/modeling tools we have today may have a hell of cost (in terms of time, money, knowledge) to use, these tools in conjunction with a good engineering understanding of the principles involved will get you a long way there. Which is obviously a part of how groups like automotive race teams with big money make significant advances in car performance.
So unless you have: the appropriate engineering background, lots of time, access to the software tools and significant computational power, access to experts on the various "subtools", don't expect to have an enthusiast-level system of CAE-type modeling that can supplant large chunks of the "system" that we use today. However, you can get a huge leg up on the majority of people particpating in this society of car modification, by getting an appropriate engineering-level background in the aspect(s) you are interested in. Any good engineering degree+"hard" textbooks on the specific aspects of auto engineering (e.g. engine, suspension) you're interested in will be best complement to a great deal of testing-based experience and will get you very close to your goal of not just "trying stuff" but having a good idea of where the solution space lies and what its bounds are.
If you get that far, you will (as I do) undoubtedly continue to wonder about the fate of the human race seeing the sheer number of ways the moron crowd with cut springs, egregious quantities of gaudy vinyl graphics, big, heavy-ass wheels, and aerodynamically "retarded" (for lack of a better word) "body kits" and "wings" violate good engineering judgment when it comes to anything that can affect objective metrics of automotive performance.
Simulation is nothing in the autmotive world, particularly the aftermarket. You have to test and test and re-test. Sure, I suppose a competent mechanical engineer could come up with some formulas to reasonably estimate what modification X might produce. In fact, a lot of drag guys will run simulations that reasonably calculate, based on their current weight, what kind of power they need to do a 1/4 mile at a certain time. That's a reasonably simple calculation.
The catch is going to be getting the actual specs for a given car.
Let's say I want to bolt on a simple turbocharger and intercooler setup. It's a reasonably good quality turbo with minimal lag. Now let's say I bolt the same turbocharger onto the following three cars:
Older model Honda ('98) Honda Civic LX: My turbo and the accompanying modifications make, say, 400 hp at the wheels. My car is now close to undriveable because torque steer is ridiculous when the turbo starts to pull, and the necessary suspension mods will either make the car fast but unable to steer (too stiff a rear) or negate most of the power gain (rear is too soft and the front wheels hop).
97' Camaro SS: My turbo add-in is basically worthless unless I do a lot of engine work. The higher compression pistons on the Camaro mean that unless I knock the boost down to about 6 PSI, I'm going to be knocking my engine apart with detonation. For the same $5,000, I could've stroked the engine, ported and milled the heads, and upgraded the intake and exhaust.
'97 Toyota Supra: My intercooler is too big to work with the factory radiator, and thus, the car constantly overheats (not good). Either I need to get a bigger radiator or a smaller intercooler.
Now, I could probably have predicted that the Honda would torque-steer like a mother, given a basic knowledge of its drivetrain. I would have selected a supercharger and been content to put 250 or 300 horsepower to the wheels and call it a day. Any reasonably well-written program would predict the same torque-steer. But it might not have been able to tell you about the suspension issues, as it would likely assume a purely physics approach to the suspension, while I would recognize that things like gravel on the road, short inclined exits from fast-food places, and crappy roads would mean that my car would lose traction all the time.
As for the Camaro, any reasonably intelligent program could tell you that the compression ratio was too high too feed heavy boost to. It would likely recommend the same Natural Aspiration tuning.
Now for the Supra, that would be the hardest to predict. Without a lot of raw data on engine bay heat buildup in various areas, I couldn't see any program reasonably predicting such an occurence. So, while the '97 Supra would've been the ideal recipient for such an upgrade, the computer would probably fail to recognize the terrible amounts of heat such an upgrade might generate.
Blah Blah... let me shorten the long story: Tuning a car is not an easy process, and it certainly isn't like a lot of tuning mags will have you believe ("Joe had this and this and this done. He rolls on these type of tires and posts this timeslip. Joe hopes to add these modifications in order to acheive this lower time"). There's a reason professional shops charge about 3X what reasonable part and labor costs for a given complete upgrade might be. If you're running straight lines all day long and rebuilding at the end of the day (as in drag), a pick-a-part approach is somewhat suitable as long as you have a good idea of what you're doing. For turning a reliable, daily-driven car into a much faster, reliable, daily-driven car, it's a hell of a lot harder. Open source or no, there's a lot more that a computer needs to know beyond basic physics and math calculations in order to accurately predict what a given upgrade might do.