Things I Wish I Had Learned Sooner

Most people are idiots when it comes to aerodynamics.
 
I know; I'm an idiot too. I used to think I "knew" about aerodynamics because I had read a lot of things by posters on the Ecomodder forum, looked at a few cars at auto shows, and used terms like "Kammback" and "pressure recovery"—but without actually knowing the first thing about the science of aerodynamics.

Did any of these modifications reduce drag? Probably, based on the car's long-term fuel economy. How much? I have no way of knowing since it was all done by guess.

In the last couple of years, I've gone back to school to get a degree in aerospace engineering. I've read fluid mechanics and aerodynamics textbooks and journal articles. I've taken coursework in compressible and incompressible flows, flight mechanics, CFD, and applied aerodynamics. And most importantly, I've spent the last several years figuring out how to test things myself on my own cars, building interesting devices that exploit aerodynamic phenomena, coding numerical flow solvers, and even doing some wind tunnel work. It wasn't until I started doing all this that I realized just how much I don't know and still don't.
 
But there are a few things I've learned that I wish I had realized a long time ago, because they would have saved me a lot of wasted effort and many arrogantly misguided comments. In no particular order:
 
1) The behavior of air in motion is not intuitive.
 
Like a lot of things that humans think they "know" intuitively, most of what car enthusiasts "know" about aerodynamics and especially car aerodynamics is complete bunk. Read the comments on any blog post about vehicle aerodynamics (like this one) and you will see recurring comments such as,
 
"The Prius has a Kamm tail precisely so it doesn’t need an unwieldy long tail. Aerodynamically, there's no difference because the rear of the car is specifically designed to disrupt the air in a way that it creates a 'virtual tail.'"
 
This is completely wrong. While there is such thing as a "Kamm tail," it does not create a "virtual tail," and Kamm never claimed it did.

Fix a board to the back of a Prius which continues the slope of the rear window and you will see that the flow stays attached, leading to a smaller wake area and an increase in pressure along the board. Testing showed that this extension reduces the car’s drag about 7%.

A long, tapered tail almost always results in lower drag than a truncated one (I say "almost always" because, again—and I cannot emphasize this enough—the behavior of air in motion is not intuitive).
 
Wolf Heinrich Hucho, the man who almost literally wrote the book on vehicle aerodynamics (Aerodynamics of Road Vehicles, now in its 5^th edition), said in a paper presented at the 1976 conference on bluff body drag at the GM Technical Center in Warren, MI,

 
"As has been shown, most of the knowledge in vehicle aerodynamics is qualitative in nature. Some attempts to transfer results from other fields of aerodynamics to the flow around cars have been quite successful. Nevertheless, what can be deduced from them is guided empiricism rather than systematic design procedures. Unlike other disciplines of fluid dynamics, as for instance turbomachinery or aeronautics, little quantitative information is available on which a rational design procedure for road vehicles can be based."
 
This is still true. We don’t have a set of robust model equations that can adequately predict a car’s aerodynamics based on its geometry the way we can, say, an airplane's or a turbojet engine's. And in the absence of such a library, you can't just intuit your way to a design that manages airflow the way you want it to. Your guesses will probably be wrong because the behavior of air in motion is not intuitive.
 
2) Aerodynamic design of production vehicles does not exist in a vacuum.
 
Here's Hucho again, from the same paper I quoted earlier:
 
"The knowledge generated by the experimental and theoretical investigations that have been suggested would be applied in creating new body shapes. Along with this, two questions arise: how much will these new shapes differ from today's shapes? Is low drag only possible with streamlined bodies of the type seen in Fig. 33, which we already know have almost no chance of being accepted?"

Those are the two shapes at the bottom of the figure; left, Klemperer's streamlined body (1922) and right, one of the VW low-drag models developed in the 1970s.

Now, he leaves off an important addendum, "…by the car-buying public." Because that's the first thing an auto manufacturer must consider, it's an important point: what sorts of cars will consumers actually buy? It does no good to build the most efficient, lowest drag car in the history of the world if no one will buy it.
 
People in auto enthusiast and efficiency forums online often post about this, wondering why the OEMs won't build truly low-drag cars. It often escapes them that the rest of the world doesn't really care about the drag coefficients of their cars and that manufacturers, if they want to stay in business, need to cater to what their base cares about—"infotainment" (how I hate that word, but that’s a subject for another day), style, utility, comfort, price, reliability, safety—and, of course, government regulations specifying everything from how much of the tires' tread must be covered by bodywork to minimum lighted lamp areas. So, it isn't a question of why they won't; it’s a question of why they can't. And they can't largely because the vast majority of people who buy cars don't know anything about aerodynamics, and don't care at least in part because they don't know, but car manufacturers need to sell them cars. It's that simple. That's why you will never see a Prius that is as low-drag as it could be, why there will always be people like that Jalopnik commenter who are convinced that whatever shape a manufacturer offers for sale is hands-down the most aerodynamically efficient form possible and cannot be improved, and why there are simultaneously lots of people out there who think it would be quite easy for manufacturers to build low-drag cars if only they decided to.
 
3) There’s always more than one way to skin a cat.
 
Pardon the macabre expression; I have nothing against cats.

In fact, I'm cat-sitting this one again later this week.

Engineering is a game of tradeoffs, the balancing of competing requirements that necessitate compromise and judgment (this core aspect of engineering is something AI will have a very, very hard time replicating, if it is ever able to do it at all). The Performance group might want a huge battery in an EV, for example, while Structures wants it smaller and Configuration doesn't care so long as it doesn't increase sill height. Or Aero might want a windshield angle of 40° from horizontal while Cost says that the increase in glass area is too expensive and Styling wants a more upright "look."
 
Every car you see on the road that made it to production is the result of thousands of little decisions like this, each one the subject of hours of meetings and argument. I experienced this in microcosm this past semester in senior design: our project was entirely a conceptual design (we didn't actually build or test anything), and fairly basic at that, yet it necessitated hours and hours of meetings throughout the semester, many late nights, and a lot of frustration to generate even a rudimentary design for an aircraft. In my class, nine groups answered the same Request for Proposal (RFP), each with a very different plane because we prioritized different things, made different design decisions, and ultimately approached the same project from varying perspectives and backgrounds.

QF-48 Basilisk. I named it, but not on purpose. Our assigned team names were dragon-themed and I was just spitballing various reptiles during one meeting.

The nice thing about home modification is that, most of the time, you don’t have to please anyone but yourself. When I contemplate building a tail on my car, for instance, I don't have to consider what anyone else thinks about it so long as it is shorter than the maximum 48 inches allowed by law without a red flag or permanent red light.
 
But you still have to make decisions from a variety of possibilities, and every choice you make will constrain your options down the line. That's what makes this fun, at least to me: whatever solution I end up with to address a problem on my car will be different than anyone else's.
 
4) All models are wrong; some models are useful.
 
It is so, so easy to get sucked into the pretty images of CFD models; they're colorful and depict things like streamlines in a visual and easily digestible medium. The same is true of wind tunnel smoke trace images. The problem with both of these is that they are not real, or rather, they do not completely and accurately model a real car in a real environment.
 
CFD is, strictly speaking, an approximation of an approximation of an approximate mathematical description of air. Small wind tunnels don't have flow similarity—the matching of Reynolds and Mach numbers that is required for airflow to behave in the same way—with real cars. Large wind tunnels can achieve similarity but can only artificially approximate things like atmospheric turbulence or a moving ground plane and, especially in smaller enthusiast tunnels, often suffer from high blockage ratios that change the shape of streamlines around the car and restrict flow through the test section (sized properly for a car, tunnels must be very large—we have a tunnel here that is nearly 50 feet long and takes up an entire warehouse. The test section is nowhere near big enough for even a ¼-scale model of a car). All of these are predictive tools: they exist to allow engineers to select a design with a high likelihood of performing as expected in the real world without having to build it first and hope that it works. They are not real themselves.

This tunnel, in the Aerodynamics Research Laboratory at UIUC, takes up an entire warehouse bay. Yet the test section (off in the distance, just ahead of the plywood transom and behind the inlet nozzle) is only big enough for a small model of a wing section—and nowhere near large enough for even a ¼-scale model car.

The only thing that ultimately matters is how the real car performs in the real world. However, real-world measurements are noisy and complicated, which means you must use various mathematical strategies to determine things like averages, uncertainty, and error.
 
Because of the natural uncertainty of testing in the real world, where tested values of just about anything you can measure will fluctuate, you also have to be sure that your test procedure and what you choose to measure will actually tell you if something changed. I see people online (sometimes citing television shows *cough* Mythbusters *cough*) trying to measure drag changes by fuel consumption, for example. This almost never works! Any change in fuel consumption you measure is nearly impossible to ascribe to the thing you altered since myriad other things you have no control over will also change fuel economy. Be smart about it. If you want to use fuel consumption to measure drag change, you will need to compare long-term averages and even then, account for other variables such as the routes you used and how you drove the car.
 
5) Beware assigning cause or explanation to airflow behavior; you’re probably wrong.
 
In one of the first aerodynamics classes I took, we were discussing lift one day when the professor asked a simple question: what causes lift on a wing or airfoil?
 
Lift is proportional to a pressure imbalance between the two surfaces of an airfoil: higher pressure below, lower pressure above (and, in a wind tunnel, taking static pressure measurements along the roof and floor will tell you how much lift an airfoil in the test section makes—this is how early airfoil performance was measured. The same is true of wake pressure measurements and their correlation to drag).

Velocity measurements in the flow behind this cylinder, for example, allowed us to calculate momentum loss in the flow and thus the drag force on the cylinder. In this experiment, we also measured the frequency of vortex shedding in the cylinder's wake.

Lift is associated more fundamentally with a change in momentum of the flow around the airfoil (and the change in pressure distribution over the airfoil's surfaces): as air is "pushed" down by a wing, the wing is "pushed" up by the air. So, our instructor asked, which is it? Is the motion of the wing through air altering the air's momentum such that a lifting reaction force is produced? Or is the lifting force developing on the wing by the air flowing past it resulting in a change of momentum in that flow?
 
The answer to both of those is "yes." Meaning, as she went on to explain, we don't exactly know which is the correct model. Plenty of people argue one way, perhaps vehemently and with lots of "obvious" reasoning—but just as many people argue the opposite, just as vehemently and "obviously."
 
The lesson here is, beware trying to explain the unexplainable or assign a cause that is not directly evident from the information you have. I get a lot of people asking me things about how a certain device on a particular car works or "controls" airflow, and I always have to take a step back when answering, reminding them (and myself) that probably nobody knows the answer to that and if they think they do, they're wrong.
 
"I put this spoiler on my car and it increased static pressure on the backlight an average 10 Pa" is a lot different statement than "this spoiler is trapping a locked vortex over the window, entraining the outer streamline to follow an ideal shape and reducing drag and lift." One of those can be proven with repeated measurement; the other is mumbo-jumbo.
 
6) Most of what you know about aerodynamics comes from marketing, not engineering—so take everything with a huge grain of salt.
 
The senior design class I mentioned above was taught by an industry engineer. On the first day of class, he forbade us from using one word in any of our reports: "optimized."
 
Now, read any article in mainstream automotive media about aerodynamics, like this one, and what word will you encounter, in just about every other sentence? "Optimized." I’m guilty of using it myself on this website in the past but you might notice that over the last year or so I've stopped. Why?
 
"Optimization" is one of those terms marketers like to use, or aerodynamicists engaged in marketing a new model, which is meaningless in an engineering context without a lot more information (or, that professor might argue, meaningless even with a lot of context). "Optimized" in terms of what, exactly? Your fancy aero wheels were not the lowest-drag design tested during development of your new car; they are the result of some compromise between aero, weight, materials, manufacturability, cost, and looks—but marketing will trumpet those same wheels as an "optimized" design to make it sound as if they have the lowest drag possible. The design of the rear end of the Prius that was the subject of the blog post and comment section I referenced above was not the lowest-drag design tested; it was the result, again, of a compromise between the aero, structures, configuration, performance, and styling groups—but marketing's touting their "optimized" design makes the public think it has the lowest drag possible (a result evident in the comment about it). The A6's underfloor is not the lowest-drag design tested; styling probably didn’t care much about it but aero, structures, configuration, and cost probably fought over the design quite a bit—all so marketing (or, in this case, the head of aerodynamics standing in for marketing) could go on and on about how "optimized" it is so that potential customers think it has the lowest drag possible.

"Optimized" might be the most ruthlessly overused word in consumer aerodynamics writing. It's a weasel word. What it means is, "given this set of constraints, we were able to maximize/minimize the value of this parameter." However, when it's used in an article like this or even in technical papers, it doesn't tell you what the constraints were (or, more importantly, why those constraints were put in place) or what the maximum/minimum achieved was and how that compares to the maximum/minimum given a different set of constraints. So it is imprecise at best and deliberately obfuscatory at worst; my professor was right to ban its use.
 
In marketing and promotional materials and press conferences, the aerodynamic design of any new car is sold to consumers as the best thing since sliced bread. Sometimes you can find actual data, as in the drag coefficients for the Audi A6, but these are usually rudimentary. Overall, these interactions with the public are designed to sell cars, and the companies do that efficiently by relaying snippets of digestible and memorable information along with a lot of qualitative descriptions (and sometimes outright falsehood, as in the A6 article—can you spot it? See the end of this post for the answer).
 
Unfortunately, these are by far the largest source of enthusiast knowledge about aerodynamics. This creates a self-reinforcing loop of idiocy (circling back to the subject of this post): consumers who don't know anything about aerodynamics are sold the idea that their new car has the best aerodynamics possible. The consumers don't learn anything about how aerodynamics actually work or the complicated process of design, instead believing that the glossy package of information they have consumed is technically rigorous and correct (one might even say, it’s "optimized" for their consumption). They repeat this information to their friends and other enthusiasts online, all of whom are now influenced by the marketing claims. Rinse and repeat.
 
In the end, consumers believe that their cars are already just about aerodynamically perfect from the factory and cannot be improved upon, especially not by someone building stuff in their garage and testing it on the road. This belief is reinforced by the plethora of efficiency enthusiasts out there who have excitedly modified their cars without measuring anything—like me or the Champrius owner or the builder of Aerocivic or any number of people on Ecomodder.
 
Don't be like that. Use your head. You don't necessarily have to go back to school and get a degree in engineering although, having done exactly that, I highly recommend it if you are truly interested in this subject. But you can still be smart about how you consume information and what information you choose to believe (most of what you come across online, you shouldn't). In short, don't be an idiot.
 
Bonus
 
I mentioned that this article contains at least one outright lie. Highlight the text below to see if you found it:
 
The article makes the claim that the new A6 is "the most aerodynamically efficient production car from any Volkswagen Group brand, ever." Fans of the Volkswagen 1-liter concept cars should recognize the error in that claim. One of those concepts made it to production: the 2014 VW XL1 of which 250 were sold, with a drag coefficient of 0.19—lower than the Audi's 0.21.
 
Notice also that the article makes no mention of reference area. This is an important omission, as aerodynamic (drag) efficiency is determined by drag area, not drag coefficient (the coefficient is just a dimensionless number found by dividing drag area by some reference area based on a vehicle's actual dimensions—which does not have to be its frontal area!). Not just the XL1's drag coefficient but also its drag area were smaller than any other production VW to date—about 40% smaller than the new A6. Designers went to great lengths to reduce the size of the XL1, going so far as to stagger the seats so that the vehicle could be made narrower and its wetted area smaller:

If you didn’t read the text yet, the picture is a giant hint.