Blasphemy! Sacrilege!
Apologies for the click bait-y subject line, but I couldn’t resist.
To be fair, that’s not the full story; I do care about drag. But I don’t make it my biggest concern in a project. I do what I can to reduce drag overall without severely compromising other factors. Then I let the chips fall where they may and see how close my predictions are to the real physics.
I have a handful of reasons for this mindset. Primarily, when you really think about all the forces and moments on a vehicle, drag is honestly the weirdest and most complex.
Lift is easily predicted since it’s based on how much pressure air applies as it moves over a lifting surface. Pitch, roll, and yaw can be predicted using the forces that each lifting surface generates, since they capture the balance of moments about the aircraft’s center of gravity. (Side force is more or less ignored in design, it’s just the force in the Y axis.) We have great, accessible methods for predicting all these, and they’re pretty straightforward to calculate by hand in a pinch.
Drag is on a whole different level. It’s the force that acts in the same direction as the airstream, and therefore it encompasses all the various ways that can air move over and around the vehicle. And because there are multiple ways air can move when it encounters something in its path, there are multiple types of drag:
- Induced drag, which is created as a byproduct of lift—more specifically, the high pressure air under the wing meets low pressure air at the wingtip and generates vortices that add drag (which is why long skinny wings are more efficient than stubby ones)
- Skin friction drag, which as the name implies comes from the friction of air molecules moving across the aircraft’s surfaces
- Form drag, which encompasses the effects of separation bubbles, turbulence, and places where airstreams mix (like where the wing joins the fuselage) and is determined by the actual shape of the aircraft
- Wave drag, generated from shock waves present on wings, fuselages, and propeller blade tips; thankfully this only matters at transonic and supersonic speeds, so most UAVs will never have to deal with it
Calculating the drag from these factors is much more complex than calculating the lift of a wing. Both skin friction and form drag are specifically generated by viscous effects. But most lower-order tools don’t model viscous effects, so they only provide the induced drag of any given configuration. In fact, most panel codes actually operate better if you entirely exclude fuselages and other non-lifting surfaces. So everything else—fuselage lift and moment contributions, plus skin friction and form drag—needs to be manually added to the tool’s output.
To get a great prediction of drag you’d need detailed, rigorously checked CFD simulations of the vehicle, run by an engineer proficient in the tool. And even then you’d want some sort of wind tunnel data to validate the results. A lot of smaller companies just don’t have the resources in time, money, or people for this.
And even if you have wind tunnel data to validate your CFD outputs, you still wouldn’t have the most accurate drag prediction. A good chunk of total drag comes from skin friction. There’s a big difference between the friction coefficient of a super smooth, all-aluminum wind tunnel model versus a production aircraft with all its different materials and fasteners and post-mission grime. That reality is going to add even more drag to your total—one of my colleagues regularly adds an extra 25% drag to his estimates, based on what he’s seen when comparing designs against flight test results.
When there are so many factors that go into a drag prediction, and practically no way to verify the truth until you fly your airplane, I make it a point to not fret about getting perfectly accurate predictions.
A second reason I don’t stress is that drag often ends up being mostly a propulsion and performance problem.
If you get lift or pitch stability wrong in your design, you’re going to have a bad time: either your aircraft will struggle to get off the ground to begin with, or it’ll turn into a lawn dart when it finally does. Yaw and roll stability issues are slightly more manageable, but they can make developing and tuning the flight control software more difficult than it needs to be.
Meanwhile, if you underestimate drag, you’ll still have an airplane. At absolute worst, you won’t be able to take off or launch successfully. Get it a bit wrong, and you’ll just need more throttle for forward motion and your maximum altitude and airspeeds will shrink. Often these problems can be rectified with a different propeller or some adjusted contours on the next aircraft iteration. And if you happened to overpredict drag, you get the unexpected bonus of better performance.
Finally, there’s my philosophy of doing what best serves the mission, not just the aircraft. The payload is called the payload for a reason: it’s the part you get paid for. If integrating a payload onto your aircraft results in some funky retrofitted geometry, then that’s the price of admission. Many payloads have orientation or exposure requirements, so though we try our best we are often limited in where we can install them—for example, you probably want your Starlink dish on the top of the airplane, where it can actually see the satellite constellations it uses.
We can’t perfectly quantify drag until something flies. But we often still need to have reasonable performance predictions partway through development. What can we do?
When in doubt, overestimate. Expect your aircraft to be the draggiest beast ever made, so take advantage of any drag reduction opportunities you find, like reducing airfoil thickness or smoothing out geometry.
Understand the limits of your analysis tools, and be fully aware of what they do and do not model. Always augment your panel code, and perhaps your CFD results, with a drag buildup. Textbook methods of drag prediction are often based on real-world flight test and wind tunnel data. Even if those predictions turn out to be less accurate, they’ll be more defensible than using exclusively what your computer programs tell you.
Predict by assuming the worst, but design by envisioning the best. That way you’re pleasantly surprised if your aircraft performs beyond your expectations.