par Laurent Simon Dim 13 Jan 2019 - 19:23
Peut-être faudrait il un fil sur l'automatisation, l'IA,
dans les cockpits ?
http://aviationweek.com/commercial-aviation/opinion-redefining-path-forward-pilot-automation-interaction
... Safety above all. As the discussion unfolds there are indications that more automation is not necessarily going to lead to a higher level of safety. But it is clear that a well-informed pilot defined as an integral part of the system, and enabled to make important decisions overriding systems in critical situations, is. Enhanced training will be key.
...
While the industry has been good at learning from its mistakes, a look at the past also illustrates that it is not good at predicting future failures and mistakes. And with so many new technical features and innovations around the corner, that deficiency is important to remember.
As in other sectors, the aviation industry has entered a new phase and is developing at an increased pace. At the same time, there is more pressure than ever on costs. It is therefore not viable to try to avoid change; too many constantly changing factors are in play. The industry needs further improvement and innovation, but the process must be well-controlled.
The aircraft types entering the market will be flown for the next 30-40 years, at least. Their hardware is designed for a two-pilot cockpit and cannot easily be modified. Also, current CRM concepts only work with a two-pilot crew. Nevertheless, we are seeing more automatic features being offered: The autopilot can fly a reliable and accurate resolution advisory following a traffic collision avoidance warning, and now the Airbus A350 can fly an automatic emergency descent down to altitudes at which pilots and passengers can survive without additional oxygen. As these features and improvements are now configured in a more modular technology that can mostly be added to existing aircraft, already certified aircraft types will evolve in steps.
Some recent accident reports might lead to the conclusion that pilots as the last line of defense made mistakes, thereby pinpointing human error as the major reason for the disaster. But caution is advised: With today’s technical reliability and so few accidents, pilot error will almost always be part of the chain of events. Little is reported about the countless cases every year in which pilots were capable of interrupting a potentially disastrous process early. So it is not surprising that there is rising pressure to reduce the influence of pilots or even take them out of the equation. Why pilots, in the end, were not able to interrupt a dynamic development and avoid the accident is a question that is often not asked loudly enough.
What are the strengths of humans in the cockpit, and what is technology able to replace? The autopilot, of course, does not get tired and is able to fly with maximum precision. The electronic support has all numbers available and calculates at superhuman speed. But even with the increasing capacity of computers, humans will, for the foreseeable future, be better at judging complicated situations, developing alternative options for action and sometimes making unconventional decisions.
Computers are not innovative and are unable to use intuition from experience to be creative; they cannot use common sense and plausibility to set the right priorities. It is impossible to preprogram decisions for all conceivable situations. Already, software in civil airliners is so complex that it cannot be tested completely.
...
Rather than designing pilots out of the cockpit, should aircraft manufacturers first design an aircraft that can revert to a “human pilot design” in case of failure? If an accident in a type of autonomous design was attributed to that factor, it could result in passengers not wanting to board the specific type any more, which could mean commercial demise for the airline and manufacturer. If we look at the automotive industry, an accident is readily attributed to the autonomous capabilities of a new design, with worldwide perception, even if the accident would have happened with a conventional car as well.
In the next decades, there will be efforts to reduce the cockpit crew on long-range flights to two pilots, with both at the controls during takeoff, climb, descent and landing, but alternately resting during the long cruise phase. The automatic emergency descent feature might be a hint toward that direction. There are many unanswered questions about such a change to current designs and concepts, as FAA and NASA studies have pointed out. The current level of safety also rests on decision-making processes by two pilots who are not alerted in time to immediately support and question a dynamic and complicated sequence of events.
As pilots increasingly fly highly complicated aircraft, below the surface this progression requires changes. To be able to intervene, pilots need more knowledge, more training and better skills when avionics get conflicting data inputs or fail—and not less. Pilots still train for engine failure at the most critical speed on takeoff, which is good practice, but that is very unlikely today compared to the time when the rule book for every simulator check was written.
More focus on automation and protection systems, their behavior during failures, and flying with reduced automation should be added to the training syllabi. That might require additional hours during type ratings and annual repetitive training. At the same time, pilots should actively use opportunities to manually fly their aircraft when airline procedures allow that.
Pilots will continue to make mistakes, but today’s unseen safety record is an achievement of the interaction in a human-centered design. Therefore, the industry is well-advised to work on improvements of this concept, while adapting to technological refinements and carefully implementing them without adverse effects.
Accueil
Portail
Galerie
S'enregistrer
Connexion
tient son nom d’un principe physique appelé chauffage par effet joule que l’on retrouve dans nos radiateurs électriques. Leur fonctionnement réside dans l’utilisation d’une résistance qui émet de la chaleur en étant traversée par un courant électrique.
bon ...
Liebherr