EXPERIMENTAL USE ONLY!!! -- [What?] -- Login to add public annotations [How? >>YouTube].

It turns out that my teachers in college, a ton of engineering textbooks, and the internet in general all seem to understand what wings do. Also my airplane has wings, and those wings are designed to interact with the air as much as possible, so I can flight-test my airplane at any time (and I constantly do) to collect information about what the wings in said airplane do. And then I use the information from the several sources listed above to really dial in the flight dynamics in X-Plane. Without question, on my death-bed I will look back on my many flights flown while frantically scribbling down notes and flying the airplane at the same time fondly. This is a challenge that not enough people get to enjoy, and then turning that knowledge into a simulator that then turns into money for me… well, let’s just say I have very little to complain about.

ANYHOO, a ton of people including me know what sort of forces wings put out.

But what about airplane FUSELAGES? Who knows THAT?

My teachers never really told me: they just could not give a specific answer in school.

Aero textbooks never tell me: They can’t because every fuselage is different. They can’t give a single good answer.

The internet doesn’t seem to tell me: All the references say that it varies with fuselage shape, and some give forces but not WHERE those forces are centered, which is surely important!

Flight test can’t really tell me: My airplane is controlled by its WINGS so much that I can’t really find the forces of the FUSELAGE!

Now, in X-Plane, the WINGS are handled extremely well… but what about the FUSELAGE forces? Those are not nearly as accurate for the reasons stated above.

Now, we DO have references online for the drag of a cylinder going thru the air sideways, and we DO enter the fuselage coefficient of drag in Plane Maker for when the fuselage is aimed perfectly into the air, but what about all the cases in between? What about when the fuselage is SORT OF slipped in the air, as happens every slip or imperfect turn, crosswind take-off, or landing? What about THOSE cases? While we can find decent estimates for the FORCES online, I can’t find a good estimate of WHERE those forces are centered. THAT is what I need to know to get the pitching and yawing of the airplane correct in X-Plane, especially in crosswinds.

So how do I find where these forces are centered?

The normal stability-test in cases like this is to SWING A MODEL AROUND ON A STRING, and see if it is stable.

But that is no green here because:

  1. The location of the center of gravity will alter the test results.
  2. When swinging the thing around, you can tell if it is pointing up or down, but cannot really tell if it is pointing towards you or away from you, which is surely important in determining what the thing is trying to do.
  3. I don’t need to see IF a model is stable: I need to see WHERE its stability-point is, requiring me to experiment with a large number of models or attach-points to find that neutral-point about which the model is just BARELY stable.

So clearly I need a generic fuselage model that has a variable-length body, and a wind tunnel where I can attach the model at a precisely-defined point to see about what point, exactly, the body wants to rotate.
Since I don’t have a wind tunnel, how about a ROLLING ‘wind tunnel’? A machine that will move my generic fuselage model through the air so that I can find it’s aerodynamic center of pressure? Such a machine must allow the test article to be mounted ABOVE any airflow around the rolling wind tunnel, in the CENTER so the flow is symmetrical, and the rolling wind tunnel should be able to drive itself on a deserted road outside of town so I can observe the test article while the wind tunnel drives itself.

What would such a setup look like?

Well, how about like… THIS?!?!

Austin's car "wind tunnel"

So here is the interesting thing about this:

I just used parts from my usual suite of model rocket parts to build this generic fuselage, and I made it so that the nose and tail of the rocket can be easily lengthened or shortened to adjust the body-length both fore and aft of the pivot-point. I set the autopilot to hold a steady low speed on a straight deserted road while holding the model well clear of the streamlines around the car and bang: Tesla wind tunnel.

So here is the question: Where is the aerodynamic center of a generic fuselage?

Put another way: When there are side forces from a side-slip, where is the center of where those forces act?

What you cannot tell from the picture above is that the rockets swings like a weathervane: The steel rod coming out of the bottom of the rocket is going into a SLEEVE like a shower-curtain rod, so the rocket swings left and right with no friction at all: It is like a weathervane with no fins. (Again: X-Plane already knows all about how the fins would work…it is the BODY I am working on here).

So, if we EXTEND the NOSE of the fuselage so there is MORE fuselage in FRONT of the pivot than BEHIND, then the fuselage should swing around backwards! UNSTABLE! So, if we EXTEND the TAIL of the fuselage so there is MORE fuselage BEHIND the pivot than IN FRONT, then the fuselage should point like a weathervane! STABLE!
So, if we SET the NOSE AND TAIL of the fuselage so we neither weathervane nor flip around, then we should have found our aero center! NEUTRALLY STABLE! THE POINT ABOUT WHICH THE FORCES ACT IS FOUND! YIPPEE!

So we just have the car drive while we see how the rocket wants to pivot with various-length fore and aft sections until we find the perfect lengths that have the thing be neither stable nor unstable: The neutral-point. And apply the forces in X-Plane there.

Sound like a plan? Well, it ain’t.

As it turns out from many ‘flight’ tests in the ‘wind tunnel’, the neutral-point MOVES! It MOVES depending on the sideslip! HAR!

So what I see are various different AMOUNTS of side-slip that the body wants to generate depending on how far back from the nose the mounting-point is!

And HERE is the data, scribbled down after the most-reliable test-sessions (least wind, stablest speed, etc)
How MANY degrees the rocket wobbled circled on the LEFT, WHERE the rocket pivots circled on the RIGHT (0.0% mounted right at the nose, 50% mounted in the center, 100% mounted right on the tail… positions tried here are 41.4%, 39.0%, 37.3%, 28.7%, and 28.3%):

notes from Austin's tests

So what are we seeing here?

The farther-aft the mounting point, the more the nose would yaw left and right (45 degrees either side of center when mounted 41.4% of the way back along the body!). The farther-forward the mounting point, the less the nose would yaw left and right (5 degrees either side of center when mounted 28.3% of the way back along the body!).

So what is this data trying to tell us?

When you know one or two really basic things, it suddenly becomes obvious:
WINGS (which we know VERY well) are VERY well known to pivot about their 25% chords. In other words, if you mounted the wing about the 25% chord, it would be neutrally stable: No wandering or weather-vaning. As we clearly see here, the side-slip is running down to just 5 degrees when the mount-point is 28.3% of the way back… very close to that magic 25% mounting-point.

BUT, if the body were mounted right at the MIDDLE (50%) then the body would simply flip to 90 degrees.. the FRONT half of the body, in its fresh, un-disturbed air, would ALWAYS have more aero force than the BACK half of the body with its turbulent, boundary-layer are (from the front!) so the front would always push to the back. And then the back would BECOME the front. Then IT would push back more. Whichever part of the body was in FRONT of the centrally-mounted pivot would push back MORE than the part BEHIND the pivot, since the FRONT part is catching all the air, so the result would INEVITABLY be that the rocket runs at a 90-degree side-slip: Exactly sideways.

So, we KNOW from all WINGS that at ZERO sideslip we will pivot about the 25% chord. So, we KNOW from the thought-experiment above that at 90 degrees sideslip we will pivot about the 50% chord.

Put another way:
So, we KNOW from all WINGS that mounting at the 25% chord will give us ZERO sideslip.
So, we KNOW from the thought-experiment above that mounting at the 50% chord will give us 90-degrees sideslip.

Put another way:
So, we KNOW from all WINGS that at ZERO sideslip the aerodynamic center is at the 25% chord.
So, we KNOW from the thought-experiment above that at 90 degrees sideslip the aerodynamic center is at the 50% chord.

Do my experiments bear this out?

Let’s see: (the 2 known endpoint cases shown in bold red, the experimental data in black)

Angular Offset Pivot Percent chord
90 degrees 50.0
45 degrees 41.4
20 degrees 39.0
15 degrees 37.3
15 degrees 32.9
10 degrees 28.7
5 degrees 28.3
0 degrees 25.0

Well, well, well.

My experimental data fills in that area in the middle perfectly. Using both the:

  1. absolutely inevitable fact that mounting the body in the center will cause it to flip to 90 degrees, (WHICHEVER part is in FRONT will catch MORE wind and flip BACK!), so the aero center is at 50% of the body-length for 90-degree side-slip cases
  2. absolutely documented a million-times over fact that bodies hold an aero center right around the 25% chord for small angular deflections, as we see with all wings constantly, so the aero center is at 25% of the body-length for low side-slip cases

We have now experimentally connected the dots to show:

  1. That this happens and
  2. What happens in between these two cases!

SOOOOOO, now it is off to X-Plane to tweak the fuselage forces to use this new information!

The MAGNITUDE of the fuselage forces will remain un-changed… that was taken from much more careful studies.
But the LOCATION at which it is applied will be updated today in code by me, to roll out in an X-Plane update, and this will cause the airplanes to be more accurate in their SIDESLIP performance, as those SIDE-forces on the slipped fuselage will be applied… AT THE RIGHT PLACE!

How I got lucky enough to get this job I’ll never know.

29 comments on “Better Fuselage Dynamics Through Science”

  1. > How I got lucky enough to get this job I’ll never know.

    Like all entrepreneurs, you created the job. And subsequently we all benefited from your wise decision to create yourself a job! Long live X-Plane!

  2. Genius. Burt Rutan and Elon Musk would be proud. Now all you need to do is text someone while on autopilot, hanging the test fuselage out the window with your favorite aviation movie blaring on the DVD player. Or “Ride of the Valkyries.”

    Nice write up, Austin. That ought to let the non-VR enthusiasts know what the heck is going on while LR is soooooooo focused on making us all look silly with HMD’s on our faces.

    You do have a great job, which makes possible lots of jobs for designers all over the world. Thanks for that. 😉

  3. I never had an issue with slipping and skidding with X-Plane, but I guess it was an area that needed improvement. Looking forward to trying it out!

  4. Still not going to fix the prop forces being decimated by crosswinds and causing totally unrealistic prop wash off the opposite wing and zero over the tail.

    1. That’s quite abruptly said but I agree with you, Ham Solo.
      And thank you Austin for this delightful article. I can’t wait to test the différence!

    2. +1

      What surprises me is that the twitch videos from aerosimgaming are not even mentioned or acknowledged here. In my opinion, a big thank you to these guys would be appropriate and a confirmation that the problem is being worked on.

  5. Some random thoughts:
    First, nice to hear something from Austin in his usual style.
    Second thought is a typical body with little wings is a rocket; maybe we need more rockets in X-Plane. The X-15 is somewhat close 😉
    Finally, it’s quite cool to stickout your misc bodies from the top of your car while driving, but there are alternatives:
    A friend of mine (the physics guy) one had mounted some force sensors to a board to calibarate them and read the measurements at a high speed. He used that to measure the force of various model rocket engines (fireworks fall in that category, too). The sum of all those forces over time (the integral) is the total power the engine delivers…
    Now if you mount your bodies to such force sensors while blowing the wind at it, you can measure the forces (and even record the numers automatically). Would be more reliable results, I guess.
    Finally: Did you perform measure the performance of one of the most important bodies, a marmelade bread? Is it true that it will prefer to land with the marmelade-side down? Obviously as the marmelade side is smoother, the opposite side will cause more drag, so the marmelade-side will turn downwards… 😉

  7. meanwhile, several years passed since X Plane moved on to x64 architecture, the “vatsim” and generally flying on-line multi-sim fauna and flora has been left out in a corner. erm.. -That multi platform / multi-sim “swift” client .. where it is at, how deep LR is involved in it and what should we expect out of it? Because, the current “on line” situation and experience we have in X Plane v11 is rather dull … we have to deal with a number of different incompatibility layers and outcomings of all kind when flying online. Is there a project in order to ‘harmonize’ the on-line flying experience for all users at once (P3D & X Plane), so no matter of what flight sim we are using on line, we have to get the best of immersive experience while flying online, meaning by that, we must see the ‘ohter flying aircraft’ deploying flaps, turning nav lights.. etc etc.. it’s been a while since we haven’t heard anything plausible about this matter… what’s LR’s road map? i would love to hear an elaborated statement in a separated blog post about this. thanks in advance.

    1. Hi Rudy,

      If you want your inputs on vatsim and squawkbox, you have come to completely the wrong forum to discuss/voice your opnion. Whilst Ben Supnik did used to have a hand in xsquawkbox, that work has now been delegated to a seperate entity. Please refer your questions to Chris Collins from xsquawkbox…or better yet vatsim. Swift, email those guys. Laminar is in n way responsible for their individual developmental efforts.

      re roadmap > The roadmap has always been openly discussed by Laminar on multiple occasions (the latest, being the Aero discussion by Philip Munzel.) Laminar are continuoly maintaining the sim, working on shipping 11.20, implementing further features and flightmodel changes for 11.30, and working on implementing vulkan into X-Plane.

    2. Lockheed Martin (Prepar3D) and Laminar Research (X-Plane) are not involved in the development of Swift, VatSim, Ivao, PilotEdge, or any other multi platform client/network.

      If you are asking for improvements in these clients I suggest contacting the maintainers of these projects.

    3. PilotEdge already handles most of what you are asking for. It’s a subscription service but it’s definitely the top of the ‘online’ clients. Other aircraft are shown (if a model exists, it will show the exact model), drone aircraft are shown, lights are shown and change depending on what switches are on, flap positions and gear are modeled, etc. With that existing, I can understand why LR would focus on core development of the sim.

  8. I would not be surprised to see you with a ‘land-shark’ rig on the front of the Tesla next! 😉
    It is stuff like this that really makes X-plane stand-out. Keep it up!

  10. You don’t have to know the neutral point if you know the moments. In fact, non-symetrical airfoils’ neutral point isn’t exactly at the 0.25c point, but we apply forces there and account for the difference in location using pitching moment. The fuselage should be modeled as a map (lookup table) of forces an moments at/about some representative point as a function of alpha and beta (aoa and sideslip). Mark E. Dreier’s book goes into this in some detail, and is worth checking out. It very similar to the existing method of airfoil lookups, but instead of 2 forces and one moment as a function of alpha, it’s 3 forces and 3 moments as a function of alpha and beta.

  11. This is excellent news Austin. It is the continuous focus on physics that sets XP apart from all other offerings in the market place. I would love to see an even larger focus of the team on advancing the physical aspects of the simulation. I would love to see LR hiring engineers, scientists and physicists that can help with the various physical sub systems – aerodynamics, ground physics, water physics, engine simulations, athmospheric modelling, hydraulics, electrics, pneumatics, etc.

    I’m looking forward to the versioned flight model and new jet engine simulations!

    1. I second this.
      By the way, why not trying to improve the stall behavior?
      Flying near the stallspeed and using ailerons extremly should have an impact (airplane rolling upside down)
      With the current model stalling is not dangerous at all.

      1. I haven’t tested stalling recently, but the easiest way to force a stall is this:
        Reduce speed with a slight “nose up” pitch. Then, when you are close to stall speed, rapidly “pull up”. While the Cessna is very good-willing (having the center of gravity below the wings), those with the “wings down” (low wing, like Pipers) are a real challenge to get out of stall.
        (Sorry I’m native German, so I don’t know all the exact phrases)

  12. That’s very empirical, but sounds like it will be a good improvement to the flight model. Thanks for that!

  13. Wow. I have the same color Model S as AM. *Nerd Starstruck*
    Great write up. Does this study impact modeling airflow from the propeller across the fuselage in crosswind situations to “better” model the wind vane effect at high and low power, low speed ground operations (with a moving, round rubber pivot point on the pavement)?

  15. A great approximation for every Boeing/Airbus product ever….but maybe not a Cessna or Lancair fuselage or that new Diamond twin. It would be very interesting to see how the center-of-pressure locations for those bluff bodies compare to your model. These xplane acf models are so close to a CFD mesh, it’s droolingly tantalizing. Just think…a RANS model in PlaneMaker. Run for a few days…apply some corrections…

  16. Dear Austin, you can’t imagine my thrill seeing X-Plane R&D working hard on raw engineering, the gorgeous old engineering! This makes me really looking forward for the next releases.
    You went bold on this, isolating just the tube of an airplane and studying its aero behavior its as finesse in spades (bridge joke sry) since fuselages are done to have the little impact possible on aircraft aerodynamics. I really hope you can find the desirable response of your laboratory but I guess it will be marginal changes.
    One think really bugs me: Since your model and fuselages are not mirrored in side view, the front and aft part of it are different shaped I would not expect to balance point being exactly at 50% as you find out. This is amazing and tells me that the (longer) conical part at the rear somehow balance with the ogival front and builds more drag per centimeter. Do you confirm that? Isn’t is awesome?
    Do you happen to use some aerodynamic software to compare results?

  17. Very cool Austin, love how dedicated you are and I do share you passion.
    Watched Michaels video of this test and found it very int.

  18. Why made real home-made wind tunnel? I don’t think airspeed is too much important, so you can use some fan, and scale to measure exact forces.

    2. Problem you have with a fan is that it makes very choppy air, and it’s harder to get good measurements. By using a car, you can get smooth air flows and reliable data. A wind tunnel would be the ultimate choice to perform experiments, but they’re very expensive to run. There’s a huge closed-loop wind tunnel just 250m away from where I live. That would be a great place to get data.

  19. Austin, perhaps the fuselage IS just a multi faced airfoil and blade element theory can be applied in segments around 360 degrees circumference of the nose?

Comments are closed.