Historically, some of the worst aircraft disasters have been attributed to faulty or overlooked maintenance—and as recently as last year, maintenance was still factoring into the top five causes of domestic aircraft delays. It’s prompted aerospace companies to take a hard look at implementing analytics and artificial intelligence (AI) in order to predict potential maintenance failures on aircraft before the failures happen.

As an aerospace manufacturer, Airbus is taking proactive steps to improve performance and reliability in the area of aircraft maintenance. It is doing this by migrating historical maintenance information from aircraft and fleets to a cloud-based data repository known as Skywise. Airbus is also installing systems on each aircraft in an airline’s fleet to collate and record thousands of data parameters in real-time. After each flight, this data is uploaded to Skywise to be analyzed and to enable maintenance predictions for the future.

Airline P2

“As an example, this data can record how the pressure of a certain type of hydraulic pump is gradually reducing over time,” said Norman Baker, senior vice president, Digital Solutions, at Airbus. “The Skywise analytics engine could then interpolate and ‘flag’ to the airline that the pump will likely be fine for the next five flights, but that failure is very likely to occur within ten more flights.”

Baker said that what he was describing was a “one aircraft scenario”—however, the power of the system really comes into play when this recorded data is correlated with similar data that is continuously recorded across a whole fleet consisting of hundreds of aircraft – across all operators globally.

“There are two types of maintenance performed on aircraft: planned and unplanned,”explained Baker. “With planned maintenance, all the necessary resources (manpower, materials, hangar slots, etc.) are prepared and allocated in advance, and are put in place at just the right time to receive the aircraft. In addition, for the aircraft, if it needs to spend some time out of service while the maintenance is undertaken, and if this maintenance is planned well in advance, then another aircraft can be already allocated in its place so that the airline continues to fly its revenue-paying passengers. In short, planned maintenance is good, because it avoids an aircraft suddenly being withdrawn from service.”

Airline P3 Photo courtesy: Airbus

In contrast, unplanned maintenance is the result of something going wrong unexpectedly. “Unplanned maintenance has two consequences,” said Baker. “The aircraft or engine, etc., goes into the hangar for work, and resources need to be urgently allocated to fix it, or a spare part has to be ordered. All unplanned maintenance puts a major strain on airline operations and scheduling, not to mention disruptions of passengers’ journeys. It also strains the maintenance department and its supply/logistics chain, which urgently must put resources in place.”

The Skywise analytics and AI system used by Airbus alerts aerospace operators of predictive maintenance needs and timelines so they can take proactive steps that enable them to sidestep maintenance issues before they appear. “In a nutshell, you need to have a history of all the parts that, if they fail, have previously caused unplanned maintenance,” said Baker. “This history needs to be accessible and usable via a powerful analytics engine.”

In order to feed this analytics and AI engine, airline maintenance departments must keep records of all maintenance actions, the parts affected, and the causes of the failure. This is easier said than done, because much of the information is buried either in thousands of paper forms or in Excel spreadsheets.

“Now that we have big data, analytics and AI, what we have been working on with technology vendors like Palantir, and also with our customer airlines, is a way to move all of this paper-based and spreadsheet data into the Skywise cloud-based data lake,” said Baker. “We structure the data it so that maintenance trends can be meaningfully analyzed by using special software.”

There is a major push in aerospace to use predictive maintenance analytics and AI as a means of reducing the occurrence of unplanned maintenance. This can be done by systematically predicting approaching ends of component life and also the probability of component failure long before a failure happens.

“What we want to achieve is conversion of unplanned maintenance into more planned maintenance, because operators now have a system that enables them to replace parts at a convenient time before those parts ever actually fail. They can do this because predictive analytics give them notice in advance,” said Baker. “Likewise, if you can predict weeks in advance that an aircraft needs to spend some time in a hanger, then the airline can already schedule another aircraft to take its place so the passengers’ booked journeys are not disrupted.”

Airline participants in Skywise have already seen improvements. “Delta Airlines and easyJet were two of the early adopters that ran pilot projects with us,” said Baker. “We worked with them during the pre-launch Skywise development phase, and they were very confident of the operational savings and efficiencies which Skywise Predictive Maintenance would confer. In fact, Skywise Predictive Maintenance already helps these airlines to: mitigate unplanned maintenance downtime; anticipate maintenance tasks, move from 'major' to 'minor' repair, and increase the efficiency of spares inventory management.”

In current project phases, Airbus is migrating historical maintenance data into the Skywise cloud. It is also installing systems on each aircraft in an airline’s fleet so each craft’s maintenance data can be tracked and given a maintenance score after each flight. “This helps us make future maintenance predictions,” said Baker. “The Skywise prediction algorithms let the operator know when a part will need to be repaired, and it can also automatically call-up the exact digital repair manual and repair procedure. The repair manuals are now authored into a digital format, so this is quite straightforward to accomplish in the Skywise platform.”

Airline P4 Skywise Reliability complements internal data with worldwide anonymised fleet data so that airlines can proactively assess reliability performance against industry benchmarks

Skywise works in conjunction with onboard diagnostic systems which can also generate an ‘alert’ while an aircraft is in flight, transmitting details of the problem to the airline’s technical ground staff before the aircraft lands. The transmission system, which uses VHF radio and/or satellite communications, is called Aircraft Communication Addressing and Reporting System (ACARS).

“Upon receiving the report, hopefully there will be enough time for a technical drew and a spare part to be ready for when the aircraft lands to fix the problem during the aircraft turnaround,” said Baker. “However, while this approach can reduce the disruption following an unexpected problem, it cannot guarantee to totally eliminate disruption. This is where the Predictive Maintenance system comes in, because it lets an airline know days or weeks in advance when a part needs to be repaired or replaced.”

A major challenge for airlines is migrating all of their historical data to the Skywise cloud, since the information has to be categorized, structured and formatted before it can be part of the Skywise data lake, We work with our customers to make this transition as smooth as possible,” said Baker. “We’ve also gained significant experience by working with our early adopters in pilot projects that enabled them to be up and running before the official launch of our full platform. These approaches have worked well for us, and we've now set up a dedicated Skywise Campus, where airlines can learn all there is to know about Skywise, its implementation and operation.”

For others considering an AI system for aircraft maintenance, what "words of wisdom" and recommendations does Baker have?

“Embarking on a cloud-based predictive maintenance and analytics capability is no something that any airline should try to accomplish on its own,” said Baker. “Aside from the many resources that they would have to invest in the associated infrastructure from scratch, as well as structuring and migrating their entire fleet’s data into it, an isolated system developed in-house by an airline would never offer anything like the power of a truly global platform, where each operator member can benefit from the vast amount of operational experience aggregated from the many fleets of all the participating airline members.”