The continuing shift to data-based predictive maintenance brings rewards and challenges. Richard Brown, a principal at ICF’s aerospace practice, estimates $3 billion in potential maintenance savings alone. One challenge will be exploiting all the data from the new e-enabled aircraft, which Brown thinks will rise from 3% of current fleets to 45% by 2025.
Another challenge is ensuring all that lucrative data gets to companies that can extract the maximum value from it for the industry as a whole. That is a happy challenge, but not without some controversy.
- OEMs have used aggregated operational data to improve maintenance programs and extended intervals
- Predictive maintenance could save $3 billion
- The right-to-use issue is a hot topic among OEMs and maintainers
Absent agreements to the contrary, data is owned by the company that generates it. OEMs own the design and test data that is gold before a new aircraft or engine launches. Airlines and MRO shops own operating and repair data that becomes increasingly valuable as aircraft fly.
Data ownership has been a contentious issue, especially in an aircraft purchase, says David Marcontell, vice president of Cavok, a division of Oliver Wyman. But the consensus is that operating and maintenance data belong to airline operators.
With tens of thousands of data points culled monthly, Pratt & Whitney’s “big data” analytics capability provides tailored support to customers.
Credit: Pratt & Whitney
OEMs might prefer to own this data, but many operators agree to share raw data with OEMs so they can improve their products. And once OEMs aggregate data from multiple sources, OEMs own this valuable aggregation and can do what they like with it. However, airlines have an interest in preventing disclosure of airline-specific data that could harm their competitive position.
Operators often share data from flights, ground operations, in-house maintenance and outsource shops with OEMs, even if they are not on flight-hour support by the OEMs, simply so products can be improved. Data-based improvements by OEMs have been significant such as improved maintenance programs and extended intervals based on nonroutine cards from heavy checks. This kind of sharing is generally freely given; airlines are not in the business of selling data.
Operational and maintenance data is usually the most valuable for predictive maintenance, especially as flying experience is gained. However, maintainers also need the fault logic of a system or component, derived from OEM fault isolation manuals. After that, it is experience in failures and fixing them that airlines accumulate so valuably.
But when a new aircraft is launched, there is no operating data. Early monitoring and predictive systems are populated with data from the OEMs’ design and test programs. This initial data is later refined or superseded by airline operating data.
That is one reason flight-hour support in the first third of aircraft life is almost always provided by the OEMs, which can best estimate frequency and cost of repairs. Usually, flight-hour programs from major integrators become important afterward, although major airline MROs may begin offering valuable predictive maintenance algorithms well before this.
The OEMs that own that initial design and test data can be a combination of airframe prime OEMs, like Airbus and Boeing, and Tier 1 suppliers, which are increasingly given responsibility for design and testing of their systems and may be allowed to handle aftermarkets independently.
Marcontell expects continued debates on data ownership and sharing. But he also expects that data will be shared, one way or another, with the companies that can best exploit it for predictive maintenance.
Brown notes that global airline MROs are entering the predictive market, such as Lufthansa Technik with Condition Analytics and Air France-KLM E&M with Prognos. But their approach differs from OEMs’ who use “big data” to analyze massive data sets; integrators focus on specific and frequent reliability or cost issues caused by certain components.
The OEM Approach
Boeing works with airlines on data analytics to improve fuel efficiency, flight planning, crew assignment and maintenance, including predictive maintenance. Predictive efforts require data from many sources, including OEM design and test data, operational and performance data from airlines, and maintenance and component data from shops.
This major U.S. OEM says operators regularly share data with it and other providers to generate data-driven predictions. The company believes this trend will increase as providers show they can improve predictions. Boeing will continue to develop its predictive tools, including machine learning, text analytics, modeling, simulation and decision-support software.
Engine-makers have been at the predictive game the longest. Lynn Fraga, business analytics manager at Pratt & Whitney, cites engineering design data; configurations of engines delivered; maintenance and part-repair records; operational data such as utilization, one-offs and disruptions; ERP data and health-monitoring and environmental data as key to her needs.
Honeywell’s MyMaintainer app and web-based system is designed to simplify fault analysis.
Fraga says data ownership varies by type and agreements, and partnerships and joint ventures also matter. “Right to use should be a part of data ownership discussions.” She advocates facilitating efficient access to data for safety, reliability and other purposes.
Pratt is most interested in data that improves safety and reliability and supports contract obligations. As companies expand their digital strategies, this may mean obtaining data from nontraditional sources. “Data sharing is about change management, new-use cases and insights, not just technology and data,” Fraga says.
Pratt has formed a cross-industry team focused on best practices in data sharing. The team is identifying the minimum data required by each business segment and examining data-sharing challenges, including which data need retention and where data gaps still exist.
Honeywell provides predictive maintenance on auxiliary power units (APU) for airlines, on engines for smaller aircraft, and is developing predictive algorithms for environmental controls, wheels and brakes. Bharathan Aravamudhan, senior product manager, says several types of data are needed: fault data, operating data such as exhaust gas temperatures, log data on repair actions and Honeywell’s own design and test data.
Typically, Honeywell offers a valuable service to an aircraft operator, which then tends to share data on faults, shop actions and operations. Airlines get the shop data from in-house facilities or MROs.
Honeywell runs the predictive system on APUs, while a service company uses Honeywell algorithms and operator data to advise clients on engines. Aravamudhan says the biggest challenge is making data-sharing work. That’s why: “We create pipes in the air or on the ground to pull the data.”
Airline MROs Stress Experience
For predictive maintenance, Lufthansa Technik (LHT) needs all data generated by aircraft and components each day, including flight, maintenance and shop data. Holger Appel, program manager for monitoring, diagnosis and prognosis, says this enables LHT to monitor aircraft systems with precision and merge information with failures and component behavior on the ground. “This combination of data sources enables us to make precise predictions on a component level,” he notes. Going one step further, data on weather and airports may find links between component failures and these variables, such as harsh climates or runway lengths.
LHT uses shop data to validate removals based on predictions. The challenge here is optimizing predictions so “No Fault Founds” are minimized and “Mean Time Between Removals” maximized. Optimization is only possible with shop data on the condition of removed components.
Lufthansa Group generates and owns all these data sets. It does not need to find workarounds for missing data or negotiate with other firms. Even if it does not get all the data from its predictive maintenance clients, the group has enough to do predictions.
Appel says OEM design and test data are not used by LHT for predictive maintenance: “This data is not relevant for actual problems; here, actual operations data are more valuable.”
In any case, Appel says there are now no general agreements for sharing this data today, except when it affects flight safety, in which case all relevant data is shared. Sharing nonsafety data would require decisions on the value of data and what the data owner gains by sharing it. “Do I strengthen a possible competitor by sharing? What is the aim of the company I give my data to? Will I benefit?” he asks.
Nevertheless, “the more data you gather, the more analysis you can run,” stresses Rodolphe Parisot, vice president for digital and innovation at AFI-KLM E&M. The practical effect of this general principle depends on the complexity of system to be predicted.
For example, Prognos predictions for the Airbus A380 are far more accurate when AFI-KLM E&M combines maintenance data with aircraft condition and usage data. Parisot says this is an advantage of having both airline and maintenance experience. The MRO can improve predictions with maintenance data from component and engine shops.
In free markets, data, like other goods, will flow to those who can extract the maximum value. The only real question is on what terms. c