Maneuvering Speed and Broken Airplanes

It’s been more than nine years since the vertical fin of an Airbus A300 operated by American Airlines broke, causing the wide body jet to crash shortly after takeoff from JFK Airport. All aboard were killed.

The NTSB determined that rapid and repeated movement of the rudder pedals in both directions by the pilot overstressed the vertical fin, causing it to snap off. What alarmed most pilots is that the Airbus was flying slower than maneuvering speed (Va) when the failure occurred.

As far back as I can remember, we were told that our control inputs could not overload and break the airplane if it were flying slower than Va. The exact words out of the FAA-approved limitations sections of the pilot’s operating handbook for my Baron says, “Do not make full or abrupt control movements above this speed.”

The book doesn’t exactly say that “full and abrupt” control movements at slower than Va airspeed are harmless, but it implies that they are. In fact, generations of flight instructors have told pilots that flying slower than Va protects the airplane from overload either by the pilot or from the gust loads of turbulence. And dozens, even hundreds of magazine articles – yes, some written by me – have repeated the same.

And, in a clinical way, there is an element of truth in the belief that you can’t break the airframe when flying slower than Va. That kernel of fact applies only to the wing and probably the horizontal tail. The reason you probably can’t break the airplane by applying full elevator travel at an airspeed slower than Va is that the wing will pitch to the stalling angle of attack before structural design loads are exceeded. The stall would then unload the wing, protecting it from failure.

I say Va “probably” protects the wing structure from an abrupt elevator input because in a real-life situation, turbulence or unintended bank angle change could add unexpected loads.

The fact that flying slower than Va could protect an airframe from failure was never completely true, whether the loading came from the pilot moving the controls or from severe turbulence. But the concept made us as pilots feel good. If things got really rough we could slow down to Va and take comfort in the legend that turbulence couldn’t break the airplane because the wing would stall before failing. Or, if we wanted to throw the airplane around and show off a little, doing so at speeds slower than Va would make it okay even if the airplane were only certified in the normal category.


What the Airbus crash taught us—or at least should have taught us—is that Va certification standards, and certification flight test results, protect the airplane from only a single control input in only one direction at a time. Any combination of control inputs that rotate the airplane around more than a single axis creates loads for which Va does not necessarily consider or test.

Flying slower than Va also only protects the airframe from moving a flight control – elevator, ailerons, or rudder – to its full travel in a single direction, not from stop to stop. So certification calculations and flight testing show that moving the ailerons fully and abruptly full left at a speed slower than Va, for example, will not break the airplane. But if the ailerons are suddenly moved fully back to the right without the airplane stabilizing in a steady attitude, Va offers no guarantee.

These airframe structural loadings have been understood by airplane designers and certification authorities for many decades. That’s why an airplane can be certified in the aerobatic category but be restricted to certain maneuvers in which the loads are understood and have been tested. Only in unlimited aerobatics have all possible loads been considered and sufficient structural strength designed into the airframe to handle them.

Building an airframe with enough structural strength for normal flying – with an additional 50-percent strength margin between limit load and ultimate load – is a useful tradeoff between a low enough empty weight to provide a high useful load, and enough strength to handle powerful turbulence and normal maneuvering. Modern unlimited aerobatic airplanes, and the almost unbelievable combinations of maneuvers pilots put them through, are proof that it is possible to build an unbreakable airplane. But that works only if you want to carry one or, at most, two people, and not much fuel or baggage. Taking your family to Disney World in your Extra isn’t going to work.

The military did a good job of teaching its pilots to avoid combinations of rolling and pitching maneuvers. But the general aviation pilot was told Va is his insurance against breaking the airplane.

Finally, the FAA is convinced that many, even most, pilots do not understand the true significance of Va and its limits on control inputs and has just issued one of its Special Airworthiness Information Bulletins (SAIB) to try to explain what Va really means.

The SAIB is written pretty much in FAA speak, so allow me to translate: Even when flying at or slower than the correct Va airspeed adjusted for the actual weight of the airplane, there is no guarantee that the airframe won’t break if you make full or abrupt control movements in combination, or if you move a control abruptly and fully from stop to stop.

There aren’t many guarantees in flying, but the protection of flying slower than Va seemed to be one. It’s not.

This entry was posted in Airmanship, Mac Clellan's Left Seat Blog. Bookmark the permalink.

28 Responses to Maneuvering Speed and Broken Airplanes

  1. As far as I am concerned, its much simpler: Airbus produced JUNK.
    The design of the fin-attach for the vertical Stab was a chain reaction: 1.Design Failure, 2.Production Quality Control Failure, 3.layout of the additional fasteners failure when they tried to improve it after the initial production (Basically Minimum-hole-distance-rule-violation in a composite structure. -This is how I remember it. Its 9 years now, but one of my friends was on the investigation team, He told and showed it THAT way and said outcome of investigation was political, not real. Another friend of mine is design engineer with Airbus, he talked about Tailheavyness and weight saving worries and attempts… I am personnally to the point that “if its aint Boing- I’m not going!!!

    • larry m parrish says:

      Thank you sir, I could not have said it better. I refuse to get into a scarebus let alone leave the ground in one.

    • Gordon Arnaut says:

      I think it is overdoing it a bit to say Airbuses are inferior. Their overall safety record is just as good as Boeing’s.

      As an engineer, what I do see is that the move to composite airframe parts is more troublesome than some people thought it would be. The 787 program has required a wing redesign, for example. And it remains to be seen if the weight savings will actually materialize.

      What I did notice on the 787 is that its (composite) fin is attached in a different way than those composite tails on the Airbuses. The Boeing fin has long spars that pass through into the fuselage tail cone and attach at several points to the bulkhead, while the Airbus fin simply bolts to three attach brackets on the fuselage surface.

      A surface attach point is fine for aluminum structures, but composites have some peculiarities (they are plastic, which means they can deform over time, especially under pressure, such as a bolt), and one must be especially careful at the attach points. It is usual practice to “harden” the attach points with metal inserts embedded in the composite matrix, but it appears to me that this may not always be the case.

      Now supposedly the Airbus tail was able to withstand twice the limit load it was designed for. I looked at the pictures of the retrieved attach points in the NTSB report and what is interesting is that the metal (titanium I think) parts remained intact, but the composite parts failed.

      Bottom line is there is a long learning curve ahead before the switch to composite airframe parts for transport category airplanes is a mature technology. The same learning curve happened with aluminum–for example the fatigue factor, which only became thoroughly understood after years of aluminum airframe experience.

      So looking at the 787 fin, it seems that Boeing has learned a lesson from the Airbus fin separation. Unfortunately there may be more lessons to learn in the future.

      As for the NTSB report, I have to wonder if the copilot flying the plane at the time actually did make full alternating inputs, as is clamed. According to colleagues he was not known as heavy-handed (or footed) on controls and lots of airmen have expressed the opinion that he is being unfairly scapegoated.

      Even full, alternating rudder input should not break the tail off at maneuvering speed, and this is indeed part of the flight test routine during certification. The large planes have automatic airspeed limits for the rudder too, so it does not make sense to me that the pilot’s inputs broke the tail off.

      Unfortunately for the Airbus flight, the wake turbulence could have contributed enough additional load to overload that tail.

      Also worth noting that a vertical tail can be stalled, just like any flying surface. When stall occurs there is a reversal of the rudder hinge moment which means you have to press really hard on opposite rudder to bring it back to neutral.

      On a large plane with hydraulic assist this is not a problem. But the airplane can start into a bank, due to dihedral effect. In the days before hydraulic controls large-plane pilots would have to use asymmetric throttle input to counter this because it would be impossible to get enough leg force to unstick the rudder.

      Bottom line is that I find the official line on this mishap somewhat unsatisfactory. No the tail should not break, even with full alternating inputs, and that’s what airline pilots were trained in the past. Of course there can always be a gust that exceeds the gust standards used in certification testing.

      I think after this accident the airlines modified their training to reflect the “new” line of thinking that full and alternating control inputs should not be used. Rather than come out and say that we are still learning how to build plastic airplanes, the authorities are finding it more convenient to advise caution about pilot control inputs.

  2. This is new? When I went to Bonanza specific pilot training some 15 years ago (give or take a couple) the Air Safety Foundation was teaching this back then. They even (gasp) taught that flying at or under Va in level flight was no guarantee that a stronger than “non standard” vertical gust wouldn’t break your airplane. It’s even in the manual they provided.

    We do need to remember than all level turns involve *three* control inputs, (aileron, rudder, AND elevator), but they are not abrupt, nor are they putting lots of stress on the airplane. At least they aren’t supposed to. Sooo the question arises, “how do you not pitch and roll at the same time” every time you turn?

  3. Paco says:

    At the risk of sounding aeronautically ignorant, doesn’t making a normal turn involve two control inputs at once: roll and yaw? So, what’s the rule? Are we weakening the structural integrity of our plane, however slightly, each time we turn?

    • Ken Woodard says:

      Kenneth S.Wodard

      Incorrect. After you roll and apply rudder you must increase elevator to account for the loss of lift. Another way to think is that in a coordated turn
      you are actually climbing. Of course these control inputs should all be done
      smoothly and coordinated.

  4. Mac says:

    A normal maneuver often involves the airplane rotating around more than one axis. But, if you want to make an abrubt or full control input the certification rules account for one at a time. So, if you are going to make a very steep level turn and stay in the design standards you would bank to a steep angle and halt the rolling moment. At that point you would pull up on the elevator to hold altitude. Rudder would not be needed. The primary concern is to halt the momentum around one axis before initiating movement around another.
    Of course, none of this matters if the control inputs are not full or abrupt. Normal maneuvering about a combination of axis is just fine.

    Mac Mc

  5. Planes I’ve flown that did not have interconnected ailerons and rudder do need rudder input. Otherwise as you roll left the nose goes right (Adverse Yaw). I was reminded of this the first time I flew a 172 after flying a Cherokee and Bo for nearly 20 years. Even with the interconnected ailerons and rudder in an F33 Bo rolling into a turn aggressively does take substantial rudder if you don’t want to skid.

    They taught at the Bo specific training to coordinate the turn. As you roll into the turn, keep the nose tracking with the rudder and start adding elevator gradually, particularly if doing steep turns the way we used to have to do them (60 degrees and 2G). The key word is gradually (or smooth). There should be no need for abrupt control inputs of a magnitude that would cause problems even when recovering from unusual attitudes in normal and utility category aircraft. That’s one of the reasons for continued practice so the pilot doesn’t get surprised and do something foolish.

  6. Bill McClure says:

    The same take on the A-300 JFK gets repeated over and over. The NTSB once again would not blame Airbus for the true cause of this accident, The control inputs made that caused the vertical stabilizer were roughly 3/4″ one way, the other, then back. And the vertical stab ripped off. How about we rig any light plane to do the same? The problem was with the control laws programmed to limit rudder throw as the aircraft accelerated. Most transport jets are designed to do the same, using different methods. This accident primarily occurred because the rudder responded to the small inputs by going full throw one way, the other, and back. Another example of blaming deceased pilots rather than place the blame squarely where it belongs.

  7. Paul Nuss says:

    I would think the pilot operating the controls of the Airbus 300 wouldn’t have been using any more control input (to include alternating rudder inputs) than he thought necessary for keeping the aircraft pointed where it was supposed to be. Were his inputs violent, and/or stop to stop? Airline pilots normally have safety and the comfortable ride of the passengers as a priority. I can’t help but think the aircraft should have been able to withstand what was needed within the parameters of the control corrections needed for gusts, wakes, or turbulence, without breaking. Would a similar-sized Boeing, at similar speeds, and similar circumstances, have broken off the vertical fin and crashed, killing all aboard? I suspect the engineering and/or the construction integrity of the particular aircraft.

  8. Bob Simmons says:

    As I understand it, the Va of an airplane that is less than gross weight will be LOWER than it is at gross weight. I wonder if the accident Airbus was at gross weight, or well below.

  9. Frank D. Szachta says:

    As a retired USAF Pilot I flew the B-47 from 1957 to 1962. My first assigned Aircraft Commander at MacDill AFB, Tampa, FL. was killed in his local transition flight in Florida. This was a local area transition flight, with local Thunderstorms in the area. Boeing did not give life spans for this plane since the 13 1/2ft wing tip movement on take-off was extreme. The crew (3-pilots & Nav upfront to seek paths through weather). The wing peeled off 10 miles after take-off as they hit a T-Storm. Boeing did the milk-bottle repair shortly after many crashes, since we were the nuclear shield at the time. No wings peeled off after the “morale plates” were bolted on. This seems unrelated to this story, but it tells a tale many military crews faced.
    This story follows many stories of Airbus crashes over the years, since there seemed many software problems in the early days of their fleet. Are these accidents related to fly by wire??

  10. A couple of points addressing the “fly by wire” concept. In today’s planes they control inputs and control surface movements are not directly connected. You have a computer, or more correctly computers between the two that sense the input. Then based on a number of parameters they each decide (and vote) on what the outcome should be. In the case of the A300 airspeed was one with lower airspeed giving a greater deflection for a given input. What other parameters are taken into consideration? I don’t know. As computers are my field I do know they only know what and how to do what they have been told. IF there is a mistake in the program, a parameter left out such as vertical acceleration in the case of the B47, then it *could* be possible for a computer to make a mistake and “break the airplane”. But remember the “hard over” problem with the 737s. Under certain conditions the rudder actuator could give a complete deflection, or even a control reversal. What does a high AOA do to the airspeed indication in the A300?

    I know what it does in my Debonair. Any accurate reading is purely coincidental but at those speeds and AOA I’m not really concerned with the accuracy. I’m looking out at the runway.

    The point is: Full and abrupt control deflections (regardless of the reason)*under* Va can break your airplane.

    • Brad says:

      The A-300 is not fly-by-wire. The only rudder deflection limiting would be provided by cleaning up (raising flaps) and/or airspeed parameters fed to the plane’s air data computer. Still, feet on the pedals, moving cables on pulleys to move hydraulic actuators in a flight control can easily overdo it.

  11. Pingback: Maneuvering Speed and Broken Airplanes | Aviation Blogs

  12. Paul Burrows says:

    Interesting article and comments following. I do remember reading about some early findings before the official one(s) and there was mention of another airliner, heavy type, that had flown the same course about three miles I believe ahead of the mishap a/c and that wake turbulence was possibly a factor. If such turbulence had been encountered, resulting in vortex-induced roll, some aggressive control inputs would no doubt would have been used, and I don’t know about anyone else, but I was taught from day one that at slow speeds rudder is the most effective control and that ailerons can lose their effectiveness early and might even induce enough adverse yaw to cause roll in the opposite direction from that intended. At no time have I ever received inputs, military or civilian, to the effect that in such cases rudder would have to be “babied.” It was to be used aggressively if needed to maintain aircraft control. This, of course, only applies at slow speeds. I ‘ll bet lots of pilots came up the same way I did. If that’s the case it just may be that our training and airframe building no longer mesh as well as they used to.

  13. I’m not sure if this would apply but to address the ailerons losing their effectiveness:
    Control effectiveness (and harmony) varies noticeably from aircraft to aircraft. Raising a wing with the aileron increases the angle of attack on that side. If near a stall (slow flight) it can induce a stall on that wing resulting in what appears to be a control reversal. Hence the reason for care with aileron input at slow speeds. Full aileron input during slow flight could give the unwary pilot more excitement than they bargained for. Some aircraft are much more prone to this than others. Some become “rudder only” on the edge of, or into a stall while others behave normally.

    Take for instance the old venerable Cherokee 180 with the Hershey Bar wing. I was able to still makes turns using the ailerons gently while in a full stall. The 180 behaved pretty much as it did in slow flight except for the relatively high rate of descent and the vibration of the stabilator. It never did drop the nose.

    OTOH The Debonair is a “rudder only” airplane on the edge of or into the approach stall. Ailerons appear to be effective all the way to the stall. Its stall is quite abrupt with the warning “bump” often mistaken for the stall. It *wants* to drop the port wing and spin. Adding aileron opposite to the roll will only accelerate the roll rate as it increases the AOA and deepens the stall on the descending wing. Immediate opposite rudder input when the wing starts to drop will stop it. At that speed there is very little resistance to rudder deflection so it’s quite easy to reach full deflection. Too much rudder and the nose will bounce up before dropping the right wing. This happens fairly fast. It’s been my experience that tip tanks (empty) and gap seals are likely to accentuate this tendency. 15 or 20 gallons out on the wing tips would likely turn this into a dangerous maneuver.

  14. Charlie Holleman says:

    My understanding of the Airbus accident is that the pilot applied the second application of rudder as the first yaw reached its limit and the tail was beginning to move back toward the stable or normal flight position.

    There is a maneuver in flight test called a rudder doublet. In this maneuver full rudder is applied, and when the yaw reaches the maximum point, then opposite rudder is applied as the nose starts back. If the opposite rudder is applied before maximum yaw is achieved, the input dampens the maneuver. If the input occurs at the point of maximum yaw, the yaw rate is highly accelerated. This is like pushing a person in a swing. Each push at, or just after, the reversing point accelerates the swing even more. This can generate yaw angles and loads well above those from a one cycle input. This maneuver is sometimes performed to see how many cycles are required for the yaw to stop and how many degrees the nose swings on each dampening cycle after the inputs are removed.

    Another point, which has nothing to do with the Airbus accident, but one that has been brought up about using two controls at the same time. A “rolling pull out” occurs when an airplane is simultaneously pulling positive Gs and rolling. The upward moving wing is now under a much greater strain than in a straight, wings level pull out. This is due to the fact that the aileron is also producing lift on this wing, and providing a greater strain on the wing than would be showing on an accelerometer. Many people are not aware that the limit load factor in rolling pull outs should only be 2/3 of the normal limit load factor. In other words a 6G airplane should only pull 4Gs in a rolling pull out. So the solution is if you want to do a 6G pull out, you roll first, then pull.

    • Brad says:

      Great explanation Charlie, and also to the point of rolling G. That whole area of producing an excitation mode in an aircraft’s axis via a doublet is quite cosmic at full, “square wave” input.

      One person has lamented the FAA’s “blaming a dead” pilot in this case. They’re right on. The amount and rates of rudder application by that First Officer were inexcusable. Encountering a wake behind even a heavy 747 should not be so overwhelming to a pilot of an aircraft the size of an A-300, leading him to make such excessive inputs.

      The Air Force trained for years and years in MITO (Minimum Interval Takeoffs) with dissimilar aircraft (heavy bombers, tankers, E-4/747′s, etc) during the Cold War….12 seconds behind preceding aircraft in the launch stream. Crews were trained to handle these wakes effectively and safely. Massive rudder inputs were never part of that training.

  15. John Clark says:

    The Airbus tail was designed to a limit factor of 1, the required ultimate factor was 1.5. It was designed to a factor of 2 to be conservative and it broke at a factor of 2. All said, it would have met the design requirments if it broke at a factor of 1.

    The alternating rudder inputs were excessive for the upset and accumlative. Charlie H described it well.

    The rudder limiter was a factor. For some airplanes, with the rudder travel limited to 1/2 travel for example, you still have to push full pedal at full force to get the 1/2 travel. For the Airbus, if the rudder travel is limited to 1/2 travel, you have to only push 1/2 pedal at 1/2 force to get 1/2 rudder travel. That makes the design more sensative to pedal input, both force and amount of pedal travel. Force and travel make it harder to overcontrol.

    There appears to be some confusion about simultaneous control inputs. Many manuevers do require simultaneous inputs, such as slips or rolling pullouts. You just don’t want to use one full input and an input in another axis while near maneuver speed. Just note that a full input near Va is not normally encountered in our day to day flying, big airplanes and little. Again, Charlie H gives a nice discussion of that issue.

  16. Mike Young says:

    What impact did the wake turbulence from the preceding B747 have on the Airbus aerodynamics? Was the bus pitching and rolling?

    • John Clark says:

      There were two vortex encounters, the second about the same as the first. The rudder was used on the second encounter and that fed into the roll oscillatons to keep the event going. Wheel input was adequate for the first encounter. The two or three excursions that followed were from the flight control inputs. From the report, the first roll of the second encounter was 23-25 degrees, vertical G dropped from 1 to .6.

      http://www.ntsb.gov/publictn/2004/AAR0404.pdf

  17. Hank Eilts says:

    Does anyone have a reference to the FAA SAIB that was supposedly issued to explain VA? J Mac says it was “just issued”, but the FAA regulatory and guidance library does not seem to show anything recent.

    Thanks.

    Hank

  18. Hank Eilts says:

    Mac sent me an email with the answer to the above question. The SAIB is on the FAA web search page, but hard to find. It has SAIB number CE-11-17 and it’s title is Instruments.

    Here’s a link to the SAIB itself:
    http://rgl.faa.gov/Regulatory_and_Guidance_Library/rgSAIB.nsf/0/3c00e5aa64a2827e8625781c00744393/$FILE/CE-11-17.pdf

    Thanks, Mac…

    Hank

  19. Shahryar Saigol says:

    When one reads Gann’s “Fate is the hunter” or Buck’s “North star over my shoulder” one is struck by the sheer resilience of the aircraft these pilots flew into thunderstorms and the wildest weather one can imagine. Those DC2, DC3, B24 and others were “overbuilt” ie beyond a significant margin for failure. In engineering terms it is called load factor or service factor. If the max g-load for a plane is expected to be 3 g then the plane was built to withstand 9g. This is the reason the first Boeing 707 was rolled by during its demo flight. To roll an airliner with significant dihedral requires crossing the controls. Its pilot knew he could do it with a huge margin of safety. Today, to get maximum seat/mile efficiency these margins are being sliced thinner. So we have an Airbus coming apart over the mid-Atlantic where Gannn and Buck would have happily flown a 707 or a DC4 without any concern.

  20. jdm says:

    I totally agree with Shahryar. As an engineer, we make trade-offs, and in this case, clearly the Airbus trade-off was for something other than aircraft integrity, which, IMHO, is criminal. Unless one does something obviously stupid, an airplane should not break apart, killing all aboard, period. The pilots put in the controls they thought they needed, just like any of us would do, without thinking, “Ooh, is it OK to put in 1/2 deflection now? No, wait….” That’s unrealistic BS. If the margins are that thin, they’re too thin. Saying Va is only good for one control input at a time is bureaucratic dung – if that’s really what Va is, then give us a real V speed where we can fly the airplane, e.g., full slips, side-to-side, w/o killing everyone on board.

    Seriously, how much would it have cost Airbus to have made the vertical stab more robust? How much is one life worth, let alone hundreds?

  21. Borneo Pilot says:

    I think any of us with half a brain all agree that the NTSB unfairly puts the blame on the FO’s rudder inputs instead of the Airbus design flaw. What I’m wondering, is how did Airbus get away with this? I’ve always thought of the NTSB as being neutral and unbiased, but were they paid off by Airbus or something? Or is it the fact that since the plane passed certification, they had to blame someone else? I agree with JDM above…no way should those rudder inputs have broken off the vertical stab.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>