Avionics Digital Edition
Found inFeature

Airlines are Increasingly Connecting Artificial Intelligence to Their MRO Strategies

Predictive maintenance is still in its infancy for commercial airlines, but in the future will evolve into intelligent maintenance for large-fleet commercial operators.

Predictive maintenance is still in its infancy for commercial airlines, but in the future, predictive will evolve into intelligent maintenance for large-fleet commercial operators.

The use of artificial intelligence (AI) is expanding as a decision-making tool for airline maintenance teams at large fleet commercial airlines.

Airlines based in the U.S., Europe and Asia have been quietly adopting AI tools in the form of intelligent agents for data modeling and simulation to the use of cognitive computing. The use of AI within airline maintenance strategies is evolving into an advanced and expanded use of predictive data analytics.

A challenge exists for airline maintenance teams dealing with the large amount of data being produced by newer generation aircraft: the need for an intelligent application, bot or computer program capable of generating a specific work order task for maintenance technicians, rather than large volumes of data that they have to aggregate and analyze to produce an actionable result. In some cases, an action isn’t even taken, and a technician or engineer simply discovers a no-fault found situation.

Right now, Delta Air Lines is working on adopting artificial intelligence and machine learning into its aircraft maintenance strategy. James Jackson, Delta Air Lines manager of predictive technology engineering, provided a look at how airlines are approaching the use of artificial intelligence within their maintenance strategies during an “intelligent maintenance” themed presentation at the 2019 AEEC/AMC and MMC general session.

“We want to integrate some of the more advanced technologies such as machine learning, artificial intelligence, natural language processing and deep learning into our predictive maintenance process. With the increased digitalization of data, we want to have our technical airplane specialists focusing more on validation rather than the aggregation and analysis of maintenance data,” said Jackson.

While Delta is not the only airline thinking about the use of artificial intelligence for maintenance, Jackson’s approach to the use of AI shows how it can be an effective tool for airline mechanics well into the future. Jackson’s presentation focused on the use of artificial intelligence primarily to replace today’s human tasks of ingesting, aggregating and analyzing raw data transmitted by aircraft.

Instead, Jackson wants to use artificial intelligence to generate an accurate work order straight from the analysis of the data.

“If I have an alert that is a single failure mode, then why can’t I automate that and have the alert trigger out prescriptive instruction in our [maintenance information system], to send those out to maintenance to include the parts, tooling, the routing of the aircraft. That way, we have our experts focused on responding to alerts that include instances where their specialized skills are needed, rather than a single failure alert,” said Jackson.

Jackson also explained how one of the primary reasons why Delta wants to adopt an intelligent maintenance strategy is a result of not only the amount of aircraft that the airline has within its fleet, but also the variety of their aircraft models as well. This is also a reflection of how and why the broader commercial airline industry is adopting AI as a decision tool for aircraft maintenance.

Delta’s global fleet, according to its annual report filed February 15, 2019, stands at well over 1,000 aircraft between their mainline and regional brands. According to Jackson, at peak operations the airline operates 3,500 flights per day. Their maintenance team is responsible for 23 individual aircraft types and 25 different engine models.

Jackson said that historically, Delta’s started to transition toward a digital and predictive-focused maintenance strategy when aircraft engines first began shipping with full authority digital engine controls and smarter sensors capable of capturing and transmitting larger amounts of data. Now, as onboard quick access recorder and data acquisition technologies make access to the rest of the airframe’s data more ubiquitous and easier, their focus is on reducing the number of human maintenance technician and engineering labor hours being used for aggregation and analysis of data, replacing those hours with automated decision making.

Delta has a five-year plan for officially adopting artificial intelligence into its predictive maintenance strategy, according to Jackson.

Other presentations during the 2019 AEEC AMC and MMC conference included a predictive maintenance strategy outlined by Air France KLM Head of Strategy Vincent Metz.

While Metz did not explicitly discuss the use of artificial intelligence and machine learning within Air France KLM’s maintenance strategy, these are among the concepts being researched at the airline’s MRO Innovation Lab. Air France KLM uses its MRO Lab to partner with universities, manufacturers and software developers within and outside of the aviation industry to explore how they can make new ideas and concepts a reality within aviation MRO.

But the predictive maintenance strategy discussed by Metz already has the digital infrastructure — as do modern airplanes — to enable the future utilization of artificial intelligence for data mining by the Air France KLM. Developed in Air France KLM’s MRO Lab, Prognos is a predictive maintenance software designed to capture data from aircraft in-flight and on the ground across available connectivity links. That data is then stored and analyzed using algorithms that trigger alerts for components according to a pre-defined set of a parameters. Those results are then uploaded in real time to an airline’s maintenance control center, leading to a maintenance work order.

“One of the things we really focus on in our models is we need to have more length in predicting. Because if we find out 30 cycles before it fails, then suddenly what happens is maintenance that was unplanned, we can turn it into a planned one and then we save a lot of costs,” said Metz.

EasyJet is also adopting AI tools for predicting maintenance, using London-based startup Aerogility’s decision support tool set that features intelligent software agents capable of representing every aircraft in the low cost carrier’s fleet. Every aircraft, including its individual software parts and upgrades, modifications and operating profiles are represented Aerogility’s web-based application and SQL database capable of configuration and simulation output data, including analytics, schedules, and model configuration parameters.

The tool is used by EasyJet to automate daily maintenance planning for its fleet, including the forecasting of heavy maintenance, while simultaneously factoring in existing plans with third-party suppliers and incorporating individual fleet modification and upgrade schedules. EasyJet first started using the new tool in December 2017 and has continuously upgraded its capabilities, which now include forecasting of engine shop visits and landing gear overhauls.

On the aircraft manufacturing side of commercial aviation, Airbus has also emerged as an industry leader actively looking to introduce the increased use of artificial intelligence into airline maintenance operations. Airbus has already established Skywise as its official predictive maintenance and advanced data analytics platform. It serves as a singular access point to data analytics that combine multiple sources into one secure cloud-based platform, including work orders, spares consumption, components data, aircraft/fleet configuration, onboard sensor data and flight schedules.

Las Vegas-based Allegiant became the most recent airline to adopt a new form of this platform, Skywise Health Monitoring. which itself has AI-like capabilities. Airbus has confirmed in testing of Skywise Health Monitoring that it can analyze up to 600,000 data occurrences within 0.1 seconds. That’s an exponential improvement over Airbus’ predecessor health monitoring technology, Airman, which was capable of handling just 7,000 events with a 30-second response time.

Christian Toro, vice president of maintenance and engineering at Allegiant Air, said the airline is using the platform to provide its maintenance information system with predictive capabilities.

“Allegiant uses Skywise to predict component failures and product maintenance actions within our aircraft pneumatics/bleed systems, anti ice system, hydraulics, gears and brakes and auxiliary power unit,” said Toro.

But Airbus is now taking its approach even further, establishing a new online platform, the Airbus AI Gym, seeking to identify new and unexpected changes in the behavior of monitored systems, as well as analyze suspicious behavior for potential faults and failures more efficiently. During an October 2018 speech at the 2018 Venture Beat Summit, Adam Bonnifield, VP of artificial intelligence at Airbus, summarized the company’s future artificial intelligence goals — and they look very similar to the type of challenges Delta’s predictive maintenance manager Jackson is looking to resolve.

“We need help understanding how to [parse] … technical diagrams that have a lot of captions and annotations,” Bonnifield. “A key lesson we learned was that bringing … data together is only solving the first part of the problem. The second part of the problem is understanding how that data interoperates.”