As airliners become more automated and a cadre of military-trained airline pilots hands over to commercially sourced flyers, training needs to adapt

Aircraft manufacturers’ assumptions about how pilots will react to the unexpected, particularly technical faults or anomalies, are in the safety spotlight in the aftermath of a series of incidents including two fatal Boeing 737 Max crashes. This reflects a growing recognition of just how much flightdeck design, aircraft systems and flight management tools have changed over the past three decades. But despite these changes, aircraft manufacturers have continued to assume that pilots would react to anomalies in exactly the same way as they did in the 1970s to 1980s – and just as quickly.

737 Max c Shutterstock

Source: Shutterstock

Two fatal crashes of 737 Max jets have forced Boeing to rethink its design and manufacturing processes

Any study intending to reassess the performance of today’s pilots in today’s flightdecks must inevitably examine changes not just to aircraft design, but also to changes in pilot education, sourcing, selection and training over the same period (see below).

The European Cockpit Association says that commercial training schools are not producing the required standards: “Pilot training traditionally concentrates on training for a licence. What has become clear in recent years, though, is that pilots need an additional set of competencies for a successful career. The ‘training for a licence’ philosophy has led to a situation where the market was flooded with licence-holders who were not up to the airline standards. With this as a starting point, we looked at how a new toolbox for training can help to address this issue: training which encompasses everything from the initial basic flying skills to the management competencies needed in an airline environment.”

It is not only Europe that is worried. US airlines suffered a fatal accident last year in which the pilots did not react as expected. On 23 February 2019, an Atlas Air Boeing 767 freighter, operating for Amazon Prime, went out of control and crashed in a steep dive. The National Transportation Safety Board (NTSB) is still investigating this event, but it also raises questions about what US carriers should be able to assume about their crews’ competencies today.

A US congressional committee reported in March that Boeing had made over-optimistic assumptions about pilot responses when designing the 737 Max series. The committee was studying the failure of pilots in two Max accidents to control the aircraft’s stabiliser trim behaviour when a faulty angle of attack (AoA) sensor wrongly triggered the machine’s Maneuvering Characteristics Augmentation System. The committee stated that Boeing had made over-optimistic assumptions about the speed and accuracy of pilot fault recognition and reaction in a case like this, because the latest generation of aircraft are so much more technically complex than their predecessors.

This conclusion by the congressional committee mirrors findings and recommendations in the Indonesian final report into the Lion Air 737 Max crash of October 2018 concerning pilot reaction assumptions. Boeing’s new chief executive, David Calhoun, has acknowledged this concern. He says Boeing is working to update its corporate culture, explaining: “It will be built around the level of light we shed on safety processes. It will be built on the engineering disciplines and what we do for pilots around the world, not just pilots in the USA.”

Right now, in the ongoing Atlas Air investigation, the NTSB is testing evidence that suggests pilot disorientation by somatogravic illusion might be pivotal in what happened. During a normal descent towards its destination airport the aircraft’s power suddenly increased to maximum, then it dived steeply and at high speed into the surface.

A common example of somatogravic illusion – which is induced by acceleration – is the feeling that airline passengers get when their aircraft begins to accelerate along the runway for take-off; they perceive the cabin to be tilted nose-upward, but a glance out the side window shows the aircraft is level, the nosewheel still on the ground.

External visual input, if available, is the dominant human sensory input, and it will correct the illusions caused by the reaction of the body’s inner-ear balance organs to a linear acceleration.

On the day of the accident, the Atlas Air 767 freighter was inbound to Houston Intercontinental airport from Miami, and the flight phase in which things began to go wrong was a routine descent. The crew were receiving vectors to avoid weather while positioning for an approach to runway 26L. As the aircraft was descending, in cloud, through about 10,000ft, cleared to 3,000ft, the crew were flying a vector heading of 270˚, and were told to expect a turn north onto a base leg to intercept the westbound final approach.

A pilot voice called for “flaps 1”, the aircraft levelled briefly at 6,200ft, climbed very slightly, and its airspeed stabilised at 230kt (426km/h). But shortly after that, the engine power increased to maximum and the aircraft pitched about 4˚ nose up.

It is at this point that somatogravic illusion – if that is indeed what occurred – appears to have kicked in powerfully with the pilots. They had no external visual horizon because the aircraft was in cloud.

According to the NTSB, almost immediately the aircraft began a dramatic pitch down to -49˚, driven by elevator deflection. The airspeed ultimately increased to 430kt, and although the pitch-down angle was eventually reduced to -20˚, impact was inevitable.

The factor the NTSB is examining now is what triggered the sudden – apparently unwarranted – massive power increase. The cockpit voice recorder has captured a sound that may indicate the activation of the go-around button on the power levers. But neither of the pilots mentioned a need for go-around power nor an intention to select it.

About 10s after the power increase, caution alarms began to sound. The inquiry says the control column remained forward for 10s. According to “The aircraft transitioned from a shallow climb to a steep descent. Five seconds after the alarm commenced, one of the pilots exclaimed: ‘Whoa,’ and shortly afterwards, in an elevated voice: ‘Where’s my speed, my speed?’ Three seconds later, a voice loudly declared: ‘We’re stalling.’”

There appears to have been a degree of crew confusion here that matches the total lack of situational awareness on the flightdeck of Air France’s notorious Flight 447 in 2009 when, after a momentary interruption to the accuracy of the indicated airspeed, a perfectly serviceable Airbus A330 descended, at night, into the Atlantic ocean in a high-AoA stalled condition, with the crew voicing the belief that it was over-speeding in a dive.

The 767 flight data recorder (FDR) gives the lie to the pilot’s stalling perception, because the AoA at that moment was safely below the stalling level, according to the FDR.

During these rather panicky pilot remarks the thrust levers were brought to idle for about 2s, then were advanced again to their high power setting. During the transition from nose slightly up to nose steeply down, there were negative g-forces for nearly 11s.

The reason somatogravic illusion is on the NTSB investigation menu in this event is that, every time a big engine-power increase takes place in flight, forward acceleration is the result, possibly combined with a pitch-up moment from the underslung engines. If this happens at night or in cloud with no external visual horizon, there is a risk that somatogravic illusion will affect the pilots. “For this reason,” observes the NTSB in a remarkable understatement, “it is important that pilots develop an effective instrument scan.”

Airlines in most nations are now required to carry out at least a modicum of upset recovery training (UPRT) in flight simulators. Unfortunately simulators are not good at reproducing sustained acceleration forces, so realism is lacking. The European Union Aviation Safety Agency (EASA), aware of both the essential part simulators play in pilot training and of their limitations, is working with simulator manufacturers to upgrade their performance fidelity at the edges of the flight envelope to improve UPRT. The manufacturers have been struggling with EASA deadlines for meeting the performance targets and the effort is ongoing.

What seems to be needed is a way of inculcating in pilots, despite the limitations of UPRT in simulators, an acute awareness that acceleration in flight brings with it the risk of disorientation, and therefore of emphasising the need to ignore all other sensory inputs except the visual picture provided by the flight instruments. This “effective instrument scan” would enable the pilot to control the aircraft attitude and power accordingly.

Atlas Air Amazon Prime 767

Source: Shutterstock

Atlas Air suffered fatal loss of a 767 freighter operating for Amazon Prime in February 2019

Recent accidents demonstrate that awareness of this risk – too often – is missing among African, Asian and now – it appears – some US pilots.

Loss of control in flight (LOC-I) has, since the late 1990s, been the biggest killer accident category for airlines. LOC-I linked to somatogravic illusion has frequently occurred, two of the most dramatic recent examples being the March 2016 FlyDubai Boeing 737-800 crash at Rostov-on-Don, Russia, and the August 2000 Gulf Air Airbus A320 crash at Bahrain International airport. Both occurred at night; both involved a go-around.

The FlyDubai pilot reaction to somatogravic illusion during a night go-around was a dramatic push-forward into a dive, like Atlas Air, and the aircraft smashed steeply into the runway.

The Gulf Air manoeuvre was an abandoned night visual approach from which the captain elected to climb and turn left into a 3,000ft downwind leg to make a second approach. In the latter case, the changes in attitude and power were less dramatic, but as the captain advanced the power and began the climbing turn to the left over the night sea, he would have lost sight of the airfield and city lights, and should have transitioned fully to flight instruments. It seems he did not. The aircraft described a shallow spiral into the dark water.

Somatogravic illusion makes highly disciplined instrument flying essential, but more difficult because of the need to reject the balance organs’ misleading input. A clear natural horizon in daylight completely overcomes those misleading feelings, and although the flight instrument panel – especially in modern flightdecks – provides an intuitive visual display, it is still not as compelling as the real thing. And there is a long list of LOC-I accidents in the past two decades that involved more subtle sensory inputs resulting in pilot disorientation leading to disaster.

In the past two months, FlightGlobal has conducted an informal poll of professional airline pilots with one of the US majors to find out what issues they believe are thrown into relief by recent fatal accidents, including the Atlas Air one and the two Max crashes. There was a fair weight of opinion that the Atlas Air crew must have been poorly paired in terms of experience, because the pilot monitoring – the captain – should have had sufficient time to pick up on what the pilot flying – the co-pilot – was doing, and to correct it. Apart from that, there is a general uneasiness with the perception that recurrent training does not attempt to compensate for the insidious effect on pilot cognitive and manual skills of operating with high levels of automation almost all the time.

In the USA the Air Carrier Safety and Pilot Training Aviation Rulemaking Committee recommends that airlines, in their flight manuals, indicate when and where, on revenue flights, manual flying may be carried out by pilots to help maintain their handling skills. In Europe, however, with its denser airspace and stricter rules on where visual flight rules flying may be carried out, manual “flying” practice is basically only available in simulators or base training.

There is no shortage of agonising – in the training industry and the airlines – about how to ensure pilots react well to unexpected demands. The only certainty is that they are not prepared to throw money at the problem. It looks, therefore, as if aircraft manufacturers are going to have to downgrade their expectations of pilot reactions in the meantime, and ultimately make the task of managing their aircraft – faults and all – more intuitive.

FlyDubai A6-FDN

Source: AirTeamImages

March 2016 loss at Rostov-on-Don of this FlyDubai 737-800 was linked to somatogravic illusion at night during a go-around attempt


Up until the beginning of the 1980s, airlines recruited a high proportion of their pilots pre-trained from the military. A relatively small proportion were trained commercially.

Today, the vast majority of airline pilots are graduates of commercial training organisations.

The effects of this radical change in pilot training provision have not been studied as such. The industry needs to analyse what it is losing as the military retire, and what the differences are of operating with an entirely commercially trained pilot workforce, because they are not identical.

Until about 20 years ago, the commercially trained and military trained groups had some basic training experiences in common; they both did their early flying in simple aircraft with primitive flight and navigation instruments. Since then, however, the airline sector has grown massively and the military has shrunk, both in relative and in absolute terms. Although there are still military trained pilots in the airlines, their numbers and influence are reducing rapidly.

This change matters, because civil airliners have been designed by their manufacturers according to assumptions about pilot performance that were set in the 1950s and 1960s. In fact, an industry-wide debate about the issue of pilot performance expectations has been reignited recently by the final report on the Lion Air Boeing 737 Max crash near Jakarta, Indonesia.

To understand the differences, consider that military pilots, almost from the beginning of their training, are exposed to far more flying at the edges of – and even outside – the flight envelope. They are taught aerobatic flying to a competent level during basic training. Even before being selected for pilot training, they had been chosen for having the robustness of personality required of a military officer in the theatre of war.

In advanced flying training, the military pilots practise missions. That is, they are given an operational task to perform in addition to “aviating and navigating”, while also conducting each flight so as to survive hostile action. Thus, successful military pilots have to demonstrate they have mental and physical capacity to spare above and beyond meeting the demands of navigating an aircraft safely from A to B.


Change, in the meantime, has not stopped. In anticipation of a confidently forecast long-term pilot supply shortage, substantial new Approved Training Organisations (ATO), like Gloucester airport, UK-based Skyborne Air Academy, are equipped and ready to go They are hoping that the disastrous effect of the coronavirus epidemic on airline operations will not stop wannabe pilots training to be ready when business gets airborne again.

The most senior commercially trained pilots in the airlines today will have begun their training in simple aircraft equipped with vacuum-driven artificial horizons and directional gyros. They learned about gyroscopic precession, and developed a scepticism about what they saw on their instrument panel. That developed in them an instinct for checking one set of readings against an alternative source to see if it made sense – like checking the artificial horizon against the turn-and-slip indicator and the compass. Those were also the days of beacon-to-beacon navigation using raw data. No navigation display, no magenta lines, but also limited exposure to the demands of area navigation.

The more junior arrivals on the line today are deft with the flight management system, adept at handling the control and display unit. They are at home with computers of all kinds, and have been drilled in accurate area navigation for hours in fixed-base procedure trainers and flight simulation training devices during their training.

An increasing proportion of the total have been trained specifically as airline co-pilots, almost from their first solo onward. And most will indeed be good co-pilots from the start, providing that nothing unexpected happens. And usually, nothing does.

The problem is that if they are forced by system failure, to go back to basics, they have had so little time with raw navigation and raw flying that they might be overloaded by the demands because their self-confidence in that arena will be low.

For that reason, some airlines do far more than the regulatory minima in their recurrent training to keep their pilots in touch with their aircraft. And some approved training organisations take the same approach. But not all of them do.

Therefore, airlines are increasingly flying into a future in which, on the rare occasions when something unexpected happens in flight, they cannot be certain how their crews will cope.

For this reason, manufacturers are under pressure now to review very carefully the pilot performance presumptions they have relied on for about 60 years when they design their aircraft and systems. As a result of lessons learned in the two fatal Boeing 737 Max accidents, the company’s chief executive, David Calhoun, has said Boeing is committed to doing this