The time has arrived and the technology is available – airlines need only embrace it. That’s the message delivered by experts of predictive maintenance, a subject long discussed by the civil aviation industry, but one that for years seemed one step removed – a goal for the future, not a reality.
The idea: that increasingly complex aircraft generate enough data to fundamentally advance how airlines maintain fleets, allowing carriers to schedule formerly unplanned maintenance, cutting delays and cancellations.
But those benefits are already possible, say manufacturers, which continue improving predictive models and algorithms.
Still, airlines face barriers to fully benefiting, insiders say. They advise that airlines let go of maintenance ideas of the past, adopt proactive philosophies and be more open to data sharing – the ultimate predictive maintenance fuel.
“We are certainly evolving at a much faster rate and pace than we have ever in the past,” says Pratt & Whitney’s vice-president of engine services, Eva Azoulay. “In the last two years it really has taken… a significant step forward in our knowledge base.”
John Maggiore, maintenance and leasing solutions director in Boeing’s commercial aviation services business, says: “The industry has already turned a corner and recognised the value of predictive maintenance. We can avoid disruptions by proactively changing components. That is something that is happening with a great amount of frequency.”
But airlines need to change with the times.
“The end users – the airlines and MROs – they are lagging behind. They haven’t really thought this through,” argues Espen Olsen, business development and sales director at Swedish aviation software company IFS.
Rick Wysong, director of the transportation and logistics practice at consultancy PwC, says: “We think folks are just barely beginning to move to a predictive environment. It’s a pretty big culture shift… You have to believe enough in the forecast.”
MORE AND MORE DATA
Any discussion of predictive analytics includes mention of the jargon terms “big data” and “the Internet of Things”. Big data encompasses real-time information recorded by aircraft sensors, plus operating and historic maintenance information, says Jeff Cass, vice-president of strategy and chief technology officer at Canadian aviation maintenance company Mxi Technologies, a division of IFS.
“Big data is all three of those, and the various players are coming at them from different angles,” he says. “At the end of the day… they all have to come together.”
All that data supports the Internet of Things – a concept meaning the connectivity between devices, sensors, software and the broader internet-connected world. “The Internet of Things remains just a buzzword,” says Olsen at IFS. “The Internet of Things is the ability to capture the data and use the data.”
Aircraft and engines have long generated data, but new types such as Boeing 787s, 737 Max and Airbus A350s have catapulted forward the data evolution, experts say. A 787’s sensors, for instance, can detect 45,000 faults, five times more than a 1980s-era 767, according to a 2017 paper written by IFS vice-president of aerospace and defence Kevin Deal.
A 787 generates about 28MB of data per flight, says Boeing’s Maggiore, while General Electric’s turbofans can monitor 1,000 parameters and generate between 50MB and 200MB per flight, says Ryan Chapin, chief product and portfolio manager in GE’s digital division.
Azoulay says P&W’s new geared turbofans have 40% more sensors than earlier-generation V2500s made by International Aero Engines, which is owned partly by P&W. “There is a significant evolution with regard to the quantity of sensors… and how we integrate those sensors with the FADEC,” he says, referencing the engine's full-authority digital engine control unit. “That simple fact means we have access to a much greater level of data knowledge.”
Meanwhile, large-scale data transmission has become cheap and practical. Airlines once transmitted most data through aircraft communications addressing and reporting systems (ACARS), but increasingly use satellites or, on the ground, wireless internet, says PwC’s Wysong.
Cass at Mxi says: “Ten years ago, people were talking about the cost of storage. Today, I don’t ever hear [that] discussion.”
MANUFACTURERS TAKE THE LEAD
The proliferation of aircraft data caught the early attention of OEMs, which began collecting data and creating processing software.
Meanwhile, many OEMs – particularly engine makers – expanded their maintenance presence with popular by-the-hour maintenance plans, through which they gained more access to their machines. Airbus offers the AIRMAN aircraft maintenance analysis product, and Boeing in 2004 launched Airplane Health Management (AHM).
That product now performs two million analytical calculations daily – information Boeing uses to help minimise operational “surprises”, says Maggiore. For instance, Boeing’s analytics have identified integrated drive generator problems on 777s before those problems became evident, allowing one airline to avoid a $300,000 repair bill, says Maggiore. The technology has also helped carriers address problems with 737 environmental control systems and 777 high-pressure shutoff values, he adds.
More than 4,000 aircraft, including nearly all 787s and more than 90% of 777s, are enrolled in AHM, and the number of aircraft enrolled across all Boeing’s predictive maintenance products increases by double-digit percentages annually, Maggiore says.
Almost all Dreamliners in service benefit from Boeing's AHM health management system
AirTeamImages
GE has tackled analytics with Predix, a software platform launched three years ago to advance GE’s role in what executives call the “industrial internet” – the complex interplay between heavy machines, data and software. A company-wide effort, Predix can process data generated by virtually any asset that creates it – from aircraft to wind turbines, power grids to oil pumps on the ocean floor, says Chapin.
The system already processes data from some 36,000 aircraft engines, including “snapshot” parameters captured during specific flight phases and “continuous engine operation data”, which can encompass 1,000 different parameters measured throughout flights, says Jayesh Shanbhag, executive director of digital strategy in GE’s aviation services business.
Predix incorporates satellite, maintenance, operating, environmental and other information, and engineers can use it to create virtual representations of machines called “digital twins”, executives say.
Data lets GE model how turbofans operate across fleets and in specific conditions, says Chapin. For instance, if an aircraft frequently flies to airports near salt water, Predix could anticipate if components might face early corrosion. GE has developed related technologies such as Flight Phase Analyzer, which enables airlines to identify anomalies in their fleets and compare their operations with other fleets. He adds: “Many airlines are waking up to this opportunity.”
P&W recently rolled out its “enhanced flight data acquisition storage and transmission” (eFAST) system, a product that collects and transmits engine data from aircraft to the ground, says Azoulay. The company installs eFAST on Bombardier CSeries aircraft, which are powered by PW1500Gs. Although P&W does not yet offer eFAST on other GTF-powered aircraft, engine data can be collected in other ways. P&W can merge engine data with “severity” information – which includes geographic, environmental and operational conditions, she adds.
“Engine performance differs when it’s in China, India, North America, Latin America,” Azoulay says. “It’s not just if you operate in the Middle East. It’s if you operate in the Middle East, and where did you fly?” P&W also sees opportunity to equip thousands of still-in-service V2500s with eFAST.
The V2500’s FADEC can capture a wide range of data, but due to earlier limitations P&W typically receives just moment-in-time snapshot data from V2500s, says Azoulay. The company expects to begin installing eFAST this year on Airbus A320-family aircraft powered by V2500s, enabling those engines to transmit “full-flight data”.
Kevin Deal at IFS envisions predictive maintenance evolving into “prescriptive maintenance” – using software and data to suggest to airlines the best maintenance approach. “Predictive analytics answer, ‘What will happen, when and why?’” Deal writes in his 2017 MRO trends paper. “Prescriptive goes one step further, [offering] ‘what if’ scenarios to show how each possible event will impact operations.”
AIRLINES LAGGING?
Insiders say airlines – particularly smaller carriers – have tended to lag OEMs, either by not effectively collecting or using data or by ceding data to manufacturers. Slow uptake partly results from software development costs, but some carriers also failed to grasp the increasing importance of data, some say.
“The airlines, they don’t really see that this is something they should get their hands on. This could only be in their favour to analyse this data,” says Olsen at IFS. “They need to treat the data as a high-value asset, like the aircraft.”
Although carriers receive data coming off aircraft, many have struggled to shift from a reactive to a predictive maintenance mindset, says Wysong. “We think folks are just barely beginning to move to a predictive environment,” he says. “It’s a pretty big culture shift… You have to believe enough in the forecast.”
Many carriers do not adequately use even relatively simple fault messages generated by earlier-generation aircraft, and others use only a fraction of available data, he adds. “We’ve even heard carriers say, ‘We are just dumbing it down to the same parameters we used to get’.”
DIFFERENT CONCERNS
Airlines and OEMs approach data from different directions, says Mxi’s Cass. OEMs have manufacturing experience but lack experience running airlines, and airlines lack product expertise but hold a “massive repository of operational history” – data that OEMs value, he says. “Someone independent will be the aggregator of both sides."
Few US airlines responded to requests from FlightGlobal for comment. “Outside of engine data, we are not yet doing predictions by mining big data,” Hawaiian Airlines tells FlightGlobal. “We are listening to vendors at this stage.”
PwC’s Wysong thinks predictive technology will flower once airlines believe analytics can actually prevent schedule disruptions. “The real money is in failures that will necessarily cause a cancellation or delay,” he says. “Finding something that is going to actually interrupt the flight operations… Now you will get people’s attention.”
Therefore, PwC developed algorithms to forecast when parts will fail and, more specifically, when failures might actually cause delays or cancellations.
The algorithms predict “true positives”, which are forecasted failures that actually occur and that cause schedule disruptions, and “false positives”, which are forecasted failures that do not occur, Wysong says. At first, PwC predicted five false positives for every one true positive. But the ratio has improved to 1:1. “Even at 5:1 it made economic sense to take action on our alerts, but at the improved 1:1 it really is compelling,” he says.
WHO OWNS WHAT?
The rush towards predictive analytics continues amid what some people describe as broad uncertainly about data ownership. “The question is, who owns the data? Is it the OEM or engine manufacturer, or the end users?” says Olsen. “We believe that this is going to be more important in the coming years.”
The question has legal implications, but consultants such as those at PwC suspect data ownership lies with airlines, and engine makers agree. “There’s… a big debate in the industry around who exactly owns the data," says GE’s Chapin. “GE has taken the stance that it’s really the customer’s data. They bought the asset.”
Azoulay says: “My view is that the operator owns their specific fleet data.” She adds that P&W combines data with institutional knowledge and processes it using sophisticated algorithms, and “that’s Pratt’s”.
Boeing’s Maggiore declines to specifically say who owns data coming off Boeing aircraft, but insists the issue “is not as contentious as it may seem”.
“To us, it’s not a big issue,” he says. In his view, customers understand that Boeing possesses the institutional knowledge needed to make data more valuable. “Customers absolutely are willing to pay for analytic solutions which add real value in their operations and save them money. They are glad to provide the data to do this.”
Maggiore touches on a point made by several OEMs: that accurate analytics rely on institutional knowledge as much as raw data. “A key element is domain knowledge. That is, understanding the problem you are solving and having the underlying knowledge about the asset,” he says.
Shanbhag at GE says: “Data has enabled us to marry physics and analytics. We are able to move away from average assumptions to more individualistic engine-related assumptions. Earlier that was not possible.”
In other words, GE’s analytics unite data with engineers’ understanding of basic engine components and how operating factors influence those components. “It is the system interaction that makes it very difficult, from a purely data-science approach to… make sense of all that data,” Chapin says.
LOOKING AHEAD
Predictive capabilities continue a rapid advance, and insiders predict that the technology will increasingly guide airline operations.
Boeing’s Maggiore sees a time when aircraft data will be meshed with data related to other facets of an airline’s operation, such as crew scheduling, aircraft assignment and flight planning. “I think the next horizon is to… start to link these things together so that you can have a full predictive operation,” he says. “In 10 years, my personal prediction is that we will be there.”
GE’s Shanbhag thinks predictive maintenance can help airlines “predict workscopes” – speeding engine maintenance by, for instance, ordering parts before an engine comes off the wing.
P&W’s Azoulay says customers will increasingly have “full visibility” into the state of engines and other components, enabling faster maintenance, more efficient engine removals and engine time-on-wing intervals customised to operators. “That’s what I see as one of the big evolutions in the future – a more planned environment,” she says.
Likewise, P&W expects to increasingly use information gleaned from one airline’s fleet to troubleshoot issues at other airlines, improving the broader fleet’s operation.
Fred Cleveland, PwC’s managing director of the transportation and logistics practice, suspects that more MROs and a broader swath of component manufacturers will jump on predictive maintenance. Makers of radios and navigation equipment, for instance, could use sensor data to better predict failures, allowing airlines to replace such equipment proactively. “It’s already happening in the engine world. I see a day when it happens to the rest of the aircraft,” he says.
But data must flow more freely before predictive maintenance will fully advance, some say. “We aren’t going to get any better until we share back and forth among the airlines,” says Cleveland.
Therein lies another barrier. Data flow raises competitive concerns, and some airlines resist giving OEMs more maintenance control.
“When it comes to sharing unique, sensitive, proprietary performance and usage data, some of this is competitive advantage stuff,” Cleveland says. “Airlines should be at least storing and owning data coming off the aircraft, so they have options to avoid future monopoly pricing by the OEMs.”
But such concerns must not hold back progress. “It should not restrict airlines from collaborating,” Cleveland adds. “You shouldn’t be afraid to pay the OEMs, but you should protect yourself.”
Big data is about people, not just metal
Underlying today’s aerospace industry is a virtual world of 3D digitised design, enabling engineers to know pretty much what is going to happen through the lifetime of an aircraft, component or system before any metal is cut. For this industry – and many others – a driver of that capability is Dassault Systèmes, the software house behind the popular CATIA and Solidworks packages.
But Dassault’s influence goes way beyond design, to the extent that it likes to talk about the whole-product-life “3D Experience”; here, “big data” meets real life. One example the company likes is coffee; for growers, a kilo of beans returns a few pennies’ profit. For retailers like Starbucks, armed with vast amounts of data about their stores, customers and suppliers, that same kilo returns many dollars; they are selling a high-value coffee shop experience, not just coffee.
For aerospace, that example is not trivial. One clear industry trend is that aircraft OEMs want to expand their revenue from maintenance – which gives a much higher profit margin than aircraft sales – so are interested in business models that would see aircraft supplied and paid for by use, much like the “power by the hour” models being used by engine makers.
Michel Tellier, vice-president aerospace and defence at Dassault Systèmes, sees a clear assumption that at some point in the next decade or so a percentage of the fleet will be not sold but, in essence, rented. The same applies to major systems, though to a lesser extent.
There is an ongoing shift in mindset, from being a supplier of a product to being a supplier of a service. Airbus and Boeing, for example, are principally engineering businesses but are increasingly focused on the very profitable supply of spare parts, to the extent that there is a trend toward preventing suppliers from selling directly to the aftermarket.
As Tellier sees it, the “holy grail” is predictive analysis – to know that a failure is coming and to be able to tell the customer in advance that the best time to take preventative action will be during a stop at airport X, and that spare parts will be waiting when the aircraft arrives. The key to making such a model work is data: actual service life of specific parts, operating conditions, weather, part stores location, aircraft location, etc.
And, he adds, more data is better because it validates predictions, thereby improving forecasting. Dassault Systèmes’ expertise is extensive in the “highly-sophisticated” information technology that can make such a system work. Semantic search engines, for example, can scour otherwise incompatible databases to gather and compile useful information; machine-learning techniques can crawl through that data to find cause-effect correlations. “Post-processing” technology turns it into actionable information.
But should airlines resist this trend? On one level, says Tellier, losing control of aircraft data opens the door to collusion between service providers, leaving airlines at risk of paying too much. But, he adds, a trend to service-oriented aircraft and systems supply is already happening, so is less a disruption of the industry than a “realignment” of business cases.
Airlines are, at root, about service. To focus on the “end-to-end customer experience” makes business sense from two perspectives. One, it may well be most efficient to leave the owning and maintaining of aircraft to specialist suppliers. Two, says Tellier, airlines should not lose sight of the other aspect of Big Data, which is very much in their control: passengers and their preferences.
On 6-7 December 2017, FlightGlobal is hosting the Aerospace Big Data Conference, focused on the business benefits of digital transformation. Find out more here
Source: Cirium Dashboard
