AI: Friend or Foe in the Aerospace Industry

Artificial intelligence(AI) is an entity which by means of computer algorithms, computer hardware and miscellaneous technologies enable a machine or a computer system to reason, learn, solve problems, comprehend, interpret language and eventually to make decisions. The use of AI in the recent years has seen a radical rise in many walks of life.  Aerospace Industry is no different.  It can be described as a system of systems. The aircraft itself consists of many systems and subsystems that make the flight safe and efficient.  Many other systems are needed for aircrafts operational success. From ticketing, check-in, flight planning, baggage handling, fuel distribution, ground transportation to name a few.  Most of these systems are reaping the benefits of AI in increased efficiency and better customer experiences. 

However software used on the aircraft has to meet design assurance levels (DAL). These levels define the severity of outcome should the software malfunctions.  These are categorised as levels A to E.

Level A – Catastrophic – example of a system classified as DAL A is the Engine Control System. A failure in this system could lead to catastrophic outcomes, including loss of engine power and potential fatalities.

Level B – Hazardous – A software malfunction has negative impact on safety or performance, or reduces the ability of the crew to operate the aircraft due to physical distress or a higher workload, or causes serious or fatal injuries among the passengers. An example here would be the Flight Control System.

Level C – Major – The Landing Gear Control System can be classified under DAL C. While a malfunction might not endanger lives directly, it could cause significant discomfort or minor injuries during landing if the landing gear does not deploy correctly.

Level D – Minor – An example of a DAL D system is an Environmental Control System (ECS) that regulates cabin temperature and air quality. If this system fails, it may cause passenger inconvenience but does not pose a direct threat to safety.

Level E- No Effect – A typical example for DAL E would be an In-Flight Entertainment (IFE) System. The failure of this system would only result in passenger dissatisfaction without affecting safety or operational capabilities of the aircraft.

The certification requirements for aircraft software DAL A-C would require the output from any computational element to be deterministic. Since the output of AI driven software is not deterministic, it is not permitted in software that falls under DAL A-C.  For design insurance levels D-E use of AI driven software would be possible. AI could be used very efficiently in certain aspects of improving the customer experience and also optimising the flight, for example recently, British Airways was able to perform route planning using large datasets that optimised how the route need to be flown, resulting in 1% fuel savings which saved British Airways an estimated USD10 million.

Another brilliant application of AI technology is Searidge Technologies’ DATMS (Digital Airport Traffic Management System) it uses AI to automate traffic control at airports, reducing the risk of human error, helping reduce tarmac incidents, radar and video cameras monitor the traffic, which is then regulated to increase the safety and avoid congestion. AI software solutions within the aerospace sector are finding uses in flight planning to optimize flight routes, predict maintenance needs, and to manage air traffic.

There is however limitations of what AI can do in terms of aircraft design.  AI is not sufficiently evolved to produce a workable designs. It may make suggestions based on the data set it has,  the output generated by AI requires real intelligence to interpret and correct.   It is possible for AI to assist in certain aspects, but this does require experience and fairly, well managed keywords and constraints to guide the AI output.   AI is like a team of eager and skilled assistants. But they need focus, direction and plain common sense to persuade them into creating a useful and accurate result. That is what is provided by Real Intelligence.

Aircraft systems containing many sensors generate a lot of data, which when analysed can yield valuable insights for improving performance, fuel efficiency, and decision-making.  As we embrace data-driven approaches and the versatility of machine learning and the LLM with predictive capabilities opens up opportunities in the areas of predictive maintenance by identifying likely component failures before they actually occur, the rapid processing of vast data aids in better aircraft performance, safety enhancement, and operational efficiency. 

So is AI a “friend” or “foe”?  In the hands of an experienced and a trained person, it is a friend.  In the hands of an untrained and inexperienced user, it is a foe.  As real intelligence is needed to coerce and correct the output from AI.