|Most pilots have heard the acronym “PIO” (pilot-induced-oscillation) at some point in their careers, and most assume it is related to poor piloting skills. Recognition and recovery from PIO is not even taught in the airline world. This is coupled with a lack of understanding of the topic by many accident investigators. In fact, outside of the aircraft certification community, there is a general lack of appreciation for these issues.Recognizing this problem, the National Research Council established a committee to study the topic. The committee included participants from industry and government. Members included all the major manufacturers of transport aircraft, flight test pilots from the military, FAA and manufacturers, engineers and ALPA representatives. The final product was a publication titled “Aviation Safety and Pilot Control – Understanding and Preventing Unfavorable Pilot-Vehicle Interactions”. The lack of attention to the topic was noted by the National Research Council (NRC):The committee was disturbed by the lack of awareness of severe APC (PIO) events among pilots, engineers, regulatory authorities and accident investigators.i
Traditionally, the letters P.I.O. have been taken to mean “Pilot-Induced-Oscillation”. Today there has been some movement towards redefining PIO to move away from the traditional “blame the pilot” mindset. In that vein, PIO came to be defined as Pilot-Involved-Oscillation. The USAF Fight Test Center prefers the term as “pilot-in-the-loop oscillation”, and many in the field have replaced the letters PIO completely by a new acronym, “APC”, or Aircraft-Pilot-Coupling, although APC can refer to either an open or closed loop event.1 The NRC states: “Aircraft-pilot-coupling (APC) events” are inadvertent, unwanted aircraft attitude and flight path motions that originate in an anomalous interaction between the aircraft and the pilot.ii
Pilots need to understand what APC is, what factors can lead to it, and how to recover from it. The importance of reporting APC events cannot be overstated. Current data recorders lack the sampling rates to adequately ascertain that a APC has occurred and accident investigators are often not familiar with the dynamics. As a result, some recent APC events were erroneously blamed on the pilot.
On July 31, 1997, a FedEx MD-11 crashed on landing at the Newark International Airport. The NTSB stated that the probable cause was the “captain’s overcontrol of the airplane during the landing and his failure to execute a go-around from a destabilized flare.” In the analysis section of the accident report was the following statement:
The captain’s large and rapid elevator control reversals, which resulted in an increasing divergence above and below the target pitch attitude, were consistent with a “classic” pilot-induced oscillation (PIO). Essentially, the captain made each increasingly larger elevator input in an attempt to compensate for the input he had made in the opposite direction about 1 second earlier. PIO in the pitch axis can occur when pilots make large, rapid control inputs in an attempt to quickly achieve desired pitch attitude changes. The airplane reacts to each large pitch control input, but by the time the pilot recognizes this and removes the input, it is too late to avoid an overshoot of the pilot’s pitch target. This, in turn, signals the pilot to reverse and enlarge the control input, and a PIO with increasing divergence may result.iii
Similarly, the Japan Accident Investigation Board (AIB) faulted the Captain of JAL 706 for another MD-11 accident in which PIO occurred – this one descending through approximately 17,000 feet. The AIB faulted the Captain despite its own analysis that clearly demonstrated problems in the aircraft handling qualities. These two events highlight the need to educate accident investigators about APC.
While the NTSB and AIB analysis did cite PIO as a factor in the above accidents, they both concluded with a probable cause which was directly contrary to studies on the topic by NRC researchers:
It is often possible, after the fact, to carefully analyze an event and identify a sequence of actions that the pilot could have taken to overcome the aircraft design deficiencies. However, it is typically not feasible for the pilot to identify and execute the required actions in real time.iv
This sentiment is also echoed in a paper written by USAF Test Pilot School flight test engineers:
Pilots must be in the loop for a PIO to occur, but pilots do not induce these unwanted oscillations. If anything, it is the airplane that induces them. This is easily shown by noting that the same pilot, flying two different airplanes in the same manner may experience many PIOs in one but never experience PIO in the other.v
Factors related to APC
Aircraft are designed to be inherently stable, however, there are limits to how stable (or unstable) an aircraft can be and still be controllable. If an aircraft were too stable it would not be controllable due to lack of response (a problem in some pre-Wright brother’s designs), and if too unstable, it would be difficult to control due to the extremely high workloads involved in maintaining the desired flight path. Between those two extremes, aircraft are fairly stable yet do have several dynamic modes, some of which are oscillatory.
Oscillations in pitch can be simplified for analysis to two separate and independent motions, a short period and a long period (also known as a phugoid).vi The short period essentially assumes no change in the path2 of the aircraft, and is the pitch change due to the aircraft’s natural stability. The longer phugoid period involves change to the actual aircraft path, and, being a longer and relatively slower oscillation, is not a problem for the pilot to control. Not necessarily so with the short period.
The short period is normally a fairly fast cycle, where the pilot will see the aircraft pitch changing to correct itself as soon as what ever caused the deviation is removed. If the cycle is a bit slow, the pilot is apt to intervene to correct the pitch to the desired attitude. A problem could occur if the pilot does this just as the natural stability is restoring the aircraft attitude on its own. The combination of the aircraft pitching on its own coupled with the pilots input can lead to a very fast “correction” and a resulting “overshoot” of the desired attitude, with a repeat of the process in the reverse direction. The result is an oscillation, with the pilot’s attempts to arrest the pitch changes actually contributing to them. A PIO occurs when a pilots inputs combine with the natural frequency of the aircraft’s motion such that the pilot inputs sustain and/or perhaps even increase the amplitude of the motion. A PIO can occur in any axis, although pilots are most familiar with oscillation in pitch and roll.
A faster natural cycle is preferable as the pilot can predict the changes more easily and is thus less likely to become part of the oscillatory dynamic.
The oscillation frequency itself is influenced by several factors, including (but not limited to): • The basic stability of the aircraft (more stable, faster rate). • The amount of damping in the design, (the larger the horizontal stabilizer and/or the more moment arm, the more damping, (think of a paddle at the end of a broomstick, the longer the handle or the larger the paddle the more it wants to align with the wind). • Moment of Inertia, (the amount of mass that is forward or aft of the wings (or, in the case of roll, way from the longitudinal axis), which means more inertia for the natural stability to counter once the oscillation starts – more moment of inertia nets a slower rate). • The altitude (high altitude thinner air damps less so slower rate). • Advanced flight control inputs, (Fly-By-Wire (FBW), Stability Augmentation Systems (SAS).
Some of the above are fairly intuitive while others are not. Stretching an airplane has a well known (and accounted for) effect on the moment of inertia (even if the CG is the same). The same physics hold true for cargo load distribution on a freighter aircraft.
Unfortunately, current regulatory and flight test requirements for cargo consider only the CG and structural limit weights of the cargo. Although the moment of inertia is far greater with a full cargo load distributed as it would be for line operations, flight testing is not required for this condition. The same CG does not necessarily mean that the short period oscillation rate is the same. More research should be conducted on the handing qualities of freighter aircraft and certification standards should be modified accordingly.
FBW and SAS has allowed for aircraft designs where the natural stability of the aircraft is less of a factor. The systems can, and do, mask the natural dynamics of the aircraft such that in some designs the control of the aircraft can require a different thought process than in conventional aircraft.
Finally, it is important to remember that high altitude tends exacerbate any unfavorable handling characteristics due to less natural aerodynamic damping.
There are many ways to manipulate the controls and still perform with the certification standards required to pass checkrides. Some pilots fly by making small inputs, predicting what is needed (known as “low-gain”), while at the other end of the spectrum we have pilots who tend to use relatively large control inputs to accomplish the same task (high-gain). While this is, to some extent, attributable to pilot experience and technique, there are events that can drive a “low gain pilot” towards the “high gain” side, and high-gain can precipitate APC.
Pilots are incredibly adaptable to the way an aircraft handles. Pilot ability to compensate for unfavorable handling qualities is a large part of what keeps flying as safe as it is. Part of the key to that is pilot experience with the feel of a wide variety of aircraft. As pilots gain experience in a particular aircraft they are also able to adapt and “work around” problem areas that might be peculiar to a particular aircraft. In terms of handling qualities, this obviously requires the pilot to spend some time hand-flying the actual aircraft. While this is does not excuse the manufacturers from designing aircraft that handle well, it could make the difference should a pilot find his or herself in an unexpected situation requiring superior flying skills.
Being proficient in handling the aircraft makes it less likely that a pilot will experience PIO to begin with. However, not all airline pilots spend a lot of time hand-flying. In fact, hand flying has been discouraged by many in the industry. Autopilots can fly the aircraft more precisely, leading to fuel saving and greater passenger comfort, plus the design of modern cockpits is such that hand-flying can lead to unacceptably high workloads. Airlines train pilots to fly using the autopilot with less emphasis on hand-flying. Consequently, the first time some line pilots spend more than a few moments hand flying their aircraft at altitude could be after a malfunction disengages the autopilot. The surprise of being presented with an aircraft that is not flying the way one expects also contributes to the pilot response. A “startled” pilot is more likely to fly with “high gain”. That adrenalin surge is not always helpful.
The “startle factor” is actually a term used in the test flight community. It can be due to a system failure, ice accumulation, TCAS RA or an unexpected mode change in the flight control system. For example, a pilot is flying in cruise when the autopilot disconnects due to an out of trim condition or similar anomaly. The pilot now tries to bring the aircraft back to the assigned cruise altitude with a large control input. The sudden necessity for the pilot to intervene often results in a much bigger correction than needed, and if conditions are right, a PIO may result.
Another factor is the task. Higher precision tasks, such as landing, tend to require higher gain. A sudden shift from a low precision task to a high precision task can lead to the pilot gain increasing suddenly. TCAS RA is an example of this as the TCAS may direct the pilot to fly within a relatively narrow vertical speed window. Response to a TCAS RA could force a pilot to make rapid pitch changes to hold the aircraft within a very narrow vertical speed. A TCAS RA therefore includes facets of the “startle-factor” as well as a shift in task precision. This in turn can lead to a sudden onset of PIO in aircraft that are susceptible. Similarly, turbulence at high altitude has also been linked to several severe PIO events in transport aircraft.vii
Mixed-mode control, where an aircraft is flown using partial automation (such as hand-flying with autothrottles engaged) has been identified as a PIO trigger.viii Unlike older systems, where the vertical path is controlled entirely with pitch, and airspeed controlled separately by power, advanced flight control systems are designed such that the automatic system has control of the entire flight regime – which includes interrelating the pitch and power inputs for the vertical mode much similar to what a pilot does when hand-flying. When flying the aircraft entirely manually, the pilot will naturally add power immediately as pitch is increased and vice versa. When both are engaged, the autopilot and autothrottles on the newer aircraft work together in a similar manner, as the flight control system “knows” what each component is doing.
Clearly this is not possible if using only part of the system. When hand-flying with authothrottles on, the autothrottles will lag behind manual changes in pitch, attempting to hold the selected airspeed. In order to maintain the desired vertical path the pilot needs to augment the pitch inputs to make up for the lag of the throttle response. The net result is the pilot is essentially flying with higher gain, with associated higher risk of APC.
The more complex flight control systems in modern transports can create a mismatch between what the pilot is doing with the controls and how they are actually responding, particularly when the stability augmentation or flight control systems inputs combine with pilot inputs to the same control surfaces. The NRC stated:
The pilot, unaware that the systems are operating at their limits, may call for a greater response from the control surface than is allowed by the system’s rate or position limits for that (control surface).ix
Accident investigators should be aware that flight control surfaces are not always moving the way the pilot is commanding, and in many modern aircraft the pilot has no feedback to that effect.
If there is a bit of a delay in the control response (rate limiting), this may not be apparent when the pilot is making small, smooth (low gain) inputs, but may become very evident during abrupt (high gain) inputs. This could result in a handling qualities “cliff”, where the pilot suddenly encounters PIO. In certain cases it is very difficult to reduce the control inputs (gain) once the PIO is encountered. For example, if PIO is encountered in landing, the pilots natural reaction is to work harder to get the situation under control, possibly exacerbating the situation.
As aircraft flight control design increases in complexity we could see more PIOs during line operations. Flying qualities requirements for transport aircraft are not well defined in the FARs or JARs. More comprehensive military specifications are available, but they are more fighter oriented and the problem has been further complicated by newer flight control systems (FBW, SAS) which have made existing handling qualities criteria obsolete.x It is interesting to note that every FBW aircraft has, at one time or another, experienced PIO.
System failures are another trigger for PIO. On November 25, 2000, a FedEx MD-11 experienced a PIO event while climbing out of Newark, New Jersey. The first officer was hand flying and at FL260 the aircraft began to pitch up and down at a fast rate. The PIO continued until the autopilot was engaged. Post event analysis found a fault in one of the electronically controlled hydraulic actuators that essentially gave the stability augmentation system (SAS) more authority than it was designed to have, leading to the oscillation. The autopilot had no problem flying the aircraft as the flight control computer had full control of all the inputs.
The discussion of failures listed as PIO triggers by the National Research Council provide further insight:
System failures can alter the effective aircraft dynamics either by changing the aircraft’s response to control inputs or by changing the feedback with the pilot. Control system failures, such as failures in the hydraulic system, actuator failures, or uncontrolled changes in aircraft trim, may significantly compromise the controllability of the aircraft. Intermittent control system failures can result in highly nonlinear or discontinuous control responses that act as potential triggering events.
Sensor and display failures that alter the feedback dynamics to the pilot or control system are also potential triggers. Even a simple mechanical failures, such as a loose pilot’s seat, can alter the acceleration feedback the pilot receives and has been observed to trigger an APC event.xi
Complexity in the automation is another area of concern. If the automation is extremely complex a pilot may not be able to fully understand the nuances of the system, and may develop an ad hoc understanding based on normal operations. This can lead to incorrect response in an unusual or emergency situation.xii
The design engineers do attempt to eliminate any unfavorable handling qualities during the design phase without sacrificing other design requirements, such as maneuverability. After the prototype is built, much of the task of rooting out any handling problems is transferred to the flight test pilots and flight test engineers. Often times, problems in aircraft handling qualities do not manifest themselves until very late in the certification process, as noted by one researcher:
On most cases of PIO experienced in the past, the problems were discovered in a relatively late stage of development, or even, during routine operation.xiii
There have been several instances in flight test where the first several test pilots did not experience any problems, but a subsequent test pilot did experience PIO. This illustrates the need to have as many pilots as possible involved in the test process.
The situation is somewhat analogous to new software on a home PC (and, indeed, in newer aircraft the handling qualities are literally new software!). The test program could be considered the “beta test” phase, where a relatively small group of users “test” a new program to work out any “bugs”. Even with a large pool of beta testers (a much larger group than test fly new airplanes), new software problems are virtually always found once the software is released to the public. The larger numbers of users leads to more variation thus exposing “bugs”. Similarly, when an aircraft is flown by a large number of line pilots, each with different handling techniques and skill levels, any residual problems in handling are likely to manifest themselves.
Another potential problem area occurs when flight computer software code is modified to address a specific issue. In many cases, the aircraft is only tested for that flight specific task or flight regime without regard to the possibility that the particular modified computer code might affect other flight regimes as well. Ideally, full flight regime testing should be used after any change to control law, and operators should ensure that the line pilots receive adequate training in the changes to the control laws.
Finally, sometimes problems have been dismissed with statements such as “pilots wouldn’t fly like that”. This is a convenient “out” for schedule-driven engineering managers, but one that clearly not conducive to ensuring good handling qualities.xiv
First, it should be recognized that by definition, APC/PIO cannot happen unless the pilot is making inputs that are sustaining the maneuver, i.e., the pilot is in the loop. Consequently, the first step is to get out of the loop. This presents three primary possibilities:
• The pilot freezes the controls. • The pilot releases the controls. • The pilot significantly reduces the aggressiveness of control.xv
Because a pilot may be highly focused on a task when PIO develops, it is important that the pilot-not-flying assist the pilot-flying in recognizing the situation. It may take forceful intervention to get the pilot to reduce his gain, freeze the controls or, in particular, release the controls altogether. If near the ground, this means a go-around is likely the only option, as was used by a NWA A-320 crew when they experienced a lateral PIO in April, 1995.3
Reducing aggressiveness of control is a lot easier said than done. Reducing the size of the inputs is tough, even for experienced test pilots. A few can increase their gain when asked, but it is rare that a pilot, once in a “high-gain” situation can choose to reduce it.xvi It also may be possible to exit the PIO by engaging the autopilot, but this may not be a realistic option in many circumstances and would depend on the design of the particular flight control system.
There are many that argue that an APC/PIO represents a loss of control, as the situation is out of control until the pilot takes action that is probably not conducive to completing the task when the PIO began. It would then be logical to conclude that if a pilot has to abandon the task (if only temporarily) to maintain control, that the aircraft was out of control previous to that abandonment. While in most cases this type of loss of control may not lead to a crash, incidents of APC should be categorized as loss of control events by safety and regulatory organizations.
Reporting of Events
An order of magnitude increase in the data rates utilized for accident investigations is required to establish that APC is, or is not, a primary causal factor.xvii
Recognition of APC by accident investigators has been problematic not only due to lack of training but also a lack of data. Unfortunately, many current FDRs lack the parameters to spot a PIO, such as control column position, and even the new generation FDRs that do include those parameters do not have adequate sampling rates to ascertain whether a PIO has occurred or not. Many parameters are currently only recorded once per second, or 1 hertz. Current data recording ability would easily accommodate at least 20 hertz, which is adequate to identify PIO.
The current lack of data recording capability adequate to identify APC makes it all the more important that pilots report APC should it be encountered. Pilots should be trained to identify APC. The NRC notes:
Operational line pilots have little or no exposure to APC potential and are not trained to recognize the initial symptoms or to understand that APC does not imply poor airmanship. This may limit reporting of APC events.xviii
If a PIO is ever experienced it is vital that it be reported and not dismissed as the fault of a new pilot. Pilots tend to blame themselves rather than realize that the PIO may be indicative of a fundamental problem in the aircraft itself. It is a recognized problem that APC events are under-reported, and without such reporting it is difficult for manufacturers and regulators to fix real problems. Systems such as ASRS and FOQA should be utilized to identify and report APC events. All too often, pilots only submit an ASRS if they are worried about a violation, however, pilots should utilize ASRS any time that they experience something that does not seem quite right, even if they do not feel that the flight was in jeopardy.
1 “Closed Loop” in this context refers to an event where the pilot’s actions act to continue the unwanted behavior, while “Open Loop” refers to an event that does not continue in a repetitive cycle. PIO is, therefore, a subset of APC that is a closed-loop, where pilot feedback into the system is an essential component. 2 The Center of Gravity continues along a straight path. 3 An interesting footnote to that incident was that the manufacturer had issued a temporary flight bulletin to the operating manual, but it was not implemented – nor was it required to be.
i National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997) Page 3.
ii National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997) Page 14.
iii NTSB DCA97MA055
iv National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997) p.2.
v Twisdale, Thomas R. and Nelson, Michael K. (1999). A method for the Flight Test Evaluation of PIO Susceptibility, (NASA Dryden PIO Workshop, 1999) p.2.
vi Carpenter, Chris. Flightwise – Stability and Control. (Shrewsbury, England: Airlife Publishing Ltd. 1997) p.113.
vii National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997), Page 50.
viii National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997), p.53.
ix National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997)Page 15.
x Rossitto, Ken F. and Field, Edmund J., Boeing Phantom Works, Criteria for Category I PIOs of Transports Based on Equivalent Systems and Bandwidth. (NASA Dryden PIO Workshop, 1999).
xi National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997), p.52.
xii National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997), p.55.
xiii Weerd, Rogier van der, Delft University of Technology/Aerospace Engineering, The Prediction and Suppression of PIO Susceptibility of Large Transport Aircraft. (NASA Dryden PIO Workshop, 1999). P.189.
xiv National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997),, p.111.
xv Twisdale, Thomas R. and Nelson, Michael K. (1999). Pilot Opinion Ratings and PIO, (NASA Dryden PIO Workshop, 1999) p.2.
xvi Lee, B.P. Airplane Handling Qualities, Boeing Commercial Airplane Group. PIO Flight Test Experience at Boeing (Puget Sound) – and the need for more research. (NASA Dryden PIO Workshop, 1999). P.152.
xvii A’Harrah, Ralph A. and Kaseote, George, A Case for Higher Data Rates, (1999, NASA hq, FAA hq).
xviii National Research Council. Aviation Safety and Pilot Control (Washington, D.C.: National Academy Press, 1997), p.163.