Aerodynamic Modification Doesn't Have to be Expensive or Complicated

Reading through a Youtube post recently on the perceived roadblocks to aerodynamic modification, I was disheartened to see many commenters espouse the idea that effective mods require huge expenditures of money and time, lots of trial and error, or won’t be effective at (legal) road speeds and therefore aren’t worth doing. These attitudes couldn’t be more wrong: aerodynamic modification of road cars can be done very simply and cheaply, and the results can be dramatic.
Myth: To effectively modify your car’s aerodynamics requires a powerful computer and CFD or thousands of dollars’ worth of wind tunnel time.
Reality: CFD and wind tunnels are the two test environments most amateurs are somewhat familiar with and put a lot of faith in. However, there are huge problems with both of these, as I’ve intimated before in several posts.
CFD, or computational fluid dynamics, uses mathematics to model airflows. This can be a very useful tool in the initial development phase of a new car design, for example, since it requires only a digital model of the vehicle rather than a physical mockup or test buck. Lots of iterations can be tested in relatively short time to decide on the best designs for further development.
However (and that’s a huge “however”), CFD is not a 100% accurate simulation of real airflows. In order to reduce the equations of fluid motion to something solvable by a computer in a reasonable amount of time, significant simplifying assumptions have to be made. In low-speed aerodynamics (such as automotive), these assumptions reduce the number of variables from 7 to 4: pressure and three components of velocity. Temperature, density, and viscosity are assumed constant. In reality, these vary. And while they may not vary much, by assuming any change in these variables is negligible, this removes CFD simulations a step back from reality.
Then, we get to the way CFD solves equations. Fluids behave continuously; that is, their properties vary on a continuum that by nature does not allow discrete (step/unit) changes. Computers can only work with discretized information. So, the digital environment around the car must be divided into discrete cells, where the fluid properties are assumed constant within each cell and vary discretely at cell boundaries, in order for a computer to be able to solve for airflow behavior. This is fundamentally different from reality, perhaps even more so than the assumption of temperature, viscosity, and density as constant, and is the primary reason that CFD simulations—even very high-quality ones from reputable software companies or manufacturers—must always be taken as predictors of real-world airflow behavior and not real themselves.

The way computers solve the coupled partial differential equations (known as the "Navier-Stokes equations" for their discoverers) that describe moving fluids also necessitates approximation, as no analytical solutions exist. Instead, these equations are solved numerically or iteratively, over a series of small but discrete timesteps. Over these processes, errors crop up. Comparison of simple aerodynamic problems for which analytical solutions exist, such as the velocity and altitude over time of a missile launched from the ground, show as much as 10% error in predicted range when solved numerically.
Finally, the digital model itself is only an approximation of a real car and, depending how much the geometry has been simplified, the real thing may behave very differently. Surface finishes and imperfections, panel gaps, internal flows and leakages, details such as badging or fasteners—none of these things appear in amateur CFD. Your typical CAD model grabbed online will have perfectly smooth surfaces, a flat underside, no radiator or engine bay, no passenger compartment openings, smooth cylindrical tires, perhaps even no door handles, antenna, or windshield wipers. On a real car in the real world, airflows are affected by parameters as seemingly insignificant as the tread design and sidewall patterning on the tires.
Wind tunnels might seem like a better option, then, but they have drawbacks too. The first problem is dealing with the ground: a real car driving on a road has a speed relative to the ground, and many wind tunnels—especially those affordable enough for amateurs to book time—do not have moving ground planes to simulate this effect. Even those tunnels that do typically use a system of 5 belts, one below each wheel in addition to a large central belt. That might seem like a good representation of a road, except a road extends indefinitely in front of and behind the car and exists between the front and rear wheels, where the belts do not. In stationary floor tunnels, the boundary layer that develops on the floor is typically suctioned off a small distance in front of the car, but since the “ground” isn’t moving at all under the car this still introduces error.
Further, wind tunnels constrict the airflow around the car to some degree by necessity of having walls to contain and direct the moving air. Even in tunnels with a large test section (the area where the car is placed), this removes the simulation another step from real flows which are unconstrained except in certain circumstances such as a highway lined with Jersey barriers and sound walls or a mountain tunnel, for example. Again, wind tunnels available to amateurs are usually worse in this regard; their blockage ratios (the ratio of the car’s frontal area to the tunnel’s) are quite high. Some have movable ceiling panels that can be positioned parallel to the expected streamlines around the car, but these are only guesses and, since they are solid walls, they only approximate streamlines (sort of the opposite of the potential flow problems I explained here).
Then, there’s the question of whether you will test a real car or a model. The same problems with model fidelity apply here as in CFD; additionally, a lot of wind tunnel testing is done on scale models where, unless flow speed is increased commensurately (which brings its own problems with the incompressible assumption as speeds go up), airflows observed in the tunnel do not replicate the actual flow over the real car. This is especially a problem of the popular “desktop” wind tunnels, which frankly shouldn’t be bothered with if you are interested in modifying your real car. The smallest models used by manufacturers for wind tunnel development are about 3/8 scale; any smaller and compressibility will affect results. Desktop tunnels are usually sized for 1/18 or 1/24-scale—so small as to be useless, and they come nowhere close to achieving the flow speed that would be required for similarity of results to a real car (at the speeds required, these small tunnels would be transonic or supersonic anyway, behaving completely different to the flow around a real car at highway or even racing speeds).
Finally, there is the question of turbulence and flow direction. Wind tunnels usually use some straightening matrix that allows for a small degree of unavoidable turbulence in the flow; real flows vary wildly in this regard. Most amateur wind tunnel testing is also done at 0° yaw, that is, airflow approaching the car from directly head on; the smaller affordable tunnels, in fact, don’t have enough room even to turn the car and simulate yawed flow. Try and remember the last time you drove your car on an absolutely windless day, with no other traffic, on a perfectly flat road with no buildings, curbs, light poles, road signs, foliage, etc. in any direction. That’s the environment a wind tunnel simulates. It doesn’t exist in the real world.

Even this road, which is about as perfect a test location as you will find anywhere, differs significantly from the artificial environment of a wind tunnel. Roadside ditches, cornfields, trees in the distance, the influence of passing cars, wind, variations in temperature and density and air pressure: all will affect the flow over your car.

All this is to say: The actual be-all, end-all of aerodynamic measurement, testing, development, and modification is the real car on a real road in real airflow in the real world. Ultimately, it does not matter how your model behaves in a CFD simulation or a wind tunnel unless your car is only going to “drive” in a digital or wind tunnel environment. What matters is how it behaves in real life. And it is ridiculously cheap and easy to measure things on your real car and make effective modifications that change its aerodynamic performance in verifiable ways.
Myth: Making aerodynamic changes to your car will require lots of time, money, and fabrication as you develop shape changes by trial-and-error and finished pieces for testing.
Reality: This is so stupid it’s almost not worth bothering with a rebuttal. But apparently enough people out there believe this—with no evidence—that it must be said: you should not be building a finished product and hope it works. Test pieces should be made of inexpensive, readily available materials such as cardboard, corrugated plastic, or plywood before ever thinking about fabrication of a final body piece, fairing, or spoiler.

For example, when I wondered if an air dam would reduce drag on my truck, I did not manufacture a finished piece and then test it to see if it worked. No, I trimmed a cardboard refrigerator box to size and fixed it to the front bumper with painter’s tape, then trimmed it during testing to see if it had an optimum height, then analyzed the results, and only then built a permanent version out of plastic and aluminum bar.

How did I know that trying out an air dam might be worthwhile? Well, familiarity with the behavior of various aerodynamic devices begins with reading. Read as much as you can about car aerodynamics; books can often be purchased secondhand, borrowed from your local library system, or read at a nearby university library, so it doesn’t even have to be expensive. Of course, if you find a book useful and want to have it as a reference, purchasing books is an option; many can be found on discount bookseller websites for reasonable cost.
When you start to modify your car, there will be some trial-and-error by necessity. Sometimes devices you think should work…won’t! (Which is one reason why you should make them cheap and dirty to test first). I was certain that a large spoiler on my truck’s tailgate would reduce drag, but it turned out to increase it in every position I measured. Fortunately, I wasn’t stupid about it so I was only out a few bucks’ worth of plywood and a couple hours of my time. Now, as I develop a tail on my Prius, I’m doing the same thing: initial theory and conception, specific and measurable requirements so my process is focused, cardboard and coroplast test pieces, and plywood mockups. Cheap, easy, and effective. Once you know how you want to design the real thing, you can make it pretty—after you have optimized it through testing.
Myth: Aerodynamic modifications aren’t effective at road speeds.
Reality: This is almost as stupid as the preceding myth. I’ve written before about people who believe that aerodynamics don’t “work” below some speed, picked at random. Apparently, there are people out there who think that magic speed is well above legal road speeds, as if spoilers won’t change pressures below 100 mph or something.
That is ridiculous. I’ve done most of my pressure testing at just 80 kph/50 mph, and I have been constantly surprised at the magnitude and extent of pressure changes from even small devices such as a tiny lip spoiler. Predicted changes in drag from pressure measurements at these low speeds correlate with measured drag changes from coast down testing at much higher speeds.
Get Out There and Test
These ideas—that effective aerodynamic modification is only possible if you have access to CFD or a wind tunnel, or that modifications will require huge inputs of time and money for something that may not work, or that modification should not even be attempted—are pernicious and idiotic. Don’t fall for them. Testing on your real car, on a real road in real conditions, is cheap and easy and in a lot of ways better than CFD or a wind tunnel. Trial pieces can be mocked up out of very inexpensive materials until you hit on something that works the way you want. And aerodynamic modifications can be effective even at low speeds, with measurable differences in parameters like aerodynamic pressures resulting from even small changes to a car’s shape. Don’t take my word for it; go try it!


  1. Hello, your contact page seems to be bugging out as the "sending..." Just sits in loop. Would you mind fixing this please? I have some questions, thank you.


Post a Comment

Popular Posts

How Spoilers Work

A Practical Guide to Aerodynamic Modification

Optimizing Aerodynamics of a Truck: Part 5