Aircraft sensors and onboard equipment gather large amounts of data. In addition to data storeable in a database, ‘big data’ includes volumes of unstructured data like engineering drawings, simulation files, engine telemetry, and old maintenance write-ups.
Big data technology is used to broadcast, aggregate, index and process a very large amount of data from many sources and with heterogeneous formats, explained Rodolpe Parisot, AFI KLM E&M’s chief digital officer.
The volume and diversity challenges traditional computing. Yet modern algorithms, aided by digitization and the Internet of Things (IoT), can uncover patterns in what would otherwise be a meaningless jumble.
“Sufficiently sized data sets enable the generation of a statistical baseline for the detection of trends or exceedances,” explained Uwe Zachau, head of industrial engineering with MTU Maintenance. Nevertheless big data is just getting off the ground: “We believe the majority of monitoring programs are in the diagnostic phase though the aim of all providers is, of course, to enter the predictive sphere as soon as possible.”
Big data’s largest effect on the aftermarket will be “tak[ing] MRO management and condition analytics into the predictive sphere,” Zachau said. Thus MRO providers will not only be able to detect and rectify negative trends early on — as is the case today — but they also “be able to analyze larger patterns, better predict performance in the field, and, as a result, better plan shop visits, parts logistics, and fleet management.”
The most valuable engine MRO market data is the data that is collected “directly from operations” – combined with maintenance records – “as this provides a context for any anomalies or patterns [that are] discovered,” added Zachau. “The real ‘art’ is in the data analysis and producing the predictive models from this large pool of inputs.”
Infrastructure investment is necessary, yet the larger issue at the moment is amassing the data to feed the databases and create enough data points to be able to establish patterns across engines, regions, and operations.
Data analytics helps to create customized, intelligent workscopes, said Lynn Fraga, senior director, aftermarket programs, for Pratt & Whitney (P&W). While these techniques improve visibility into the overall health of the fleet, however, “no two operators are the same [and] operators have different aircraft and engines, geographic routes, operational needs, and environmental conditions. Collecting fleet data helps maximize the customer’s specific engine performance and time-on-wing while maintaining predictable MRO spend, Fraga said.
According to Parisot, AFI KLM E&M considers digital transformation and big data as a strategic move in its portfolio of services. The Franco-Dutch MRO sees these market trends as so powerful that it is transforming itself into “a fully data-driven organization,” able to deliver the best digital services to its customers.
Aircraft sensors and onboard equipment gather up to a terabyte (1 TB) of data during a flight, revealled Ryan Chapin, chief product portfolio manager for GE Aviation’s Digital organization. This data characterizes key measurements that are important to operations, maintenance, training, and equipment optimization. GE engines on the latest-generation aircraft produce on the order of 25 megabytes (MB) per flight hour per engine. All of the ‘snapshot data’ currently available is captured.
Traditionally, an aircraft transmitted single-snapshot engine reports in different flight modes; about 10 kilobytes (KB) per snapshot report, said Zachau. “Newer developments use continuous data from the whole flight, including snapshots each second. Although this data is still below one gigabyte (GB), it is already exceeding current in-flight data transmission capabilities and is therefore downloaded via wireless quick access recorders after arrival, he said.
Airplanes like the 787 and A350 “collect 10,000 times more data than 1990s or early 2000s-era aircraft,” explained Joel Reuter, vice president of public affairs for Rolls-Royce North America. “That is because more parameters are being measured at higher frequencies, using broader transmission pipelines.” He predicts that data from the next generation of airplanes entering service will increase by another order of magnitude.
P&W’s Geared Turbofan (GTF) family incorporates 40 percent more health and performance monitoring sensors than the V2500, Fraga explained. Data flows from the airplane to the airline and then to Pratt & Whitney. The data then is normalized for more accurate trending and anomaly detection. But in both GTF and non-GTF applications data also can be sent to P&W in flight. With the GTF’s increase in sensors and expanded data capture through the full-flight profile, P&W continues to develop new reporting, trending, and monitoring capabilities, Fraga affirmed.
The GTF also expands field data capture for engine health management to nearly four million pieces of data per engine per revenue service flight, Fraga says. The new powerplant family also provides “deeper integration of sensor data within the FADEC [full-authority digital engine control] for an increase in fidelity.”
There is a growing list of sensors or components that send data over the available bus infrastructures such as ARINC 429 and ARINC 717, said Jan Stoevesand, head of analytics and data intelligence, information management, for Lufthansa Technik (LHT). Modern equipment can record 4,000 or more data points in parallel. But the sheer amount of collected data often is used as a synonym for the potential to generate insights, he noted. “Our experiences in the last two years of data analytics lead us to the conclusion that ‘it is not the amount of data, it is the right amount of right data’” that makes a difference.
The same is true with aircraft sensors, Stoevesand said. Their rising numbers do not automatically improve the quality of the analytics results. The key factor, rather, will be to pick the right set of sensors and to collect the data when it is needed at the frequency that it is needed.
“Airplanes like the 787 and A350 collect 10,000 times more data than 1990s or early 2000s-era aircraft. That is because more parameters are being measured at higher frequencies, using broader transmission pipelines.”
– Joel Reuter, Vice President of Public Affairs, Rolls-Royce North America
Boeing’s Airplane Health Management (AHM) uses analytics to evaluate two million conditions each day to determine when alerts should be generated across 4,000 airplanes, explained John Maggiore, managing director for maintenance and leasing solutions for the Digital Aviation unit.
AHM is “the remote monitoring of airplane data to understand its current or future serviceabilities,” Maggiore said. Data can include real-time and post-flight collections, mechanic write-ups, and shop findings. AHM enables pro-active maintenance management and maintenance scheduling to avoid schedule disruptions. “It’s all about ‘no surprises,’” he stated.
AHM predictive alerts, for example, indicated that a 777integrated drive generator would fail in the near future, Maggiore recalled. A scheduled inspection found and corrected the issue before it created an in-service delay or cancellation and saved up to $300,000 in repair costs.
Qantas, an early adapter of AHM, used prognostic alerts on its 747-400 fleet. In 2010 the airline also adopted AHM for its 737NG fleet. The carrier used AHM’s predictive alerts on the narrow-body fleet to address and fix issues during regularly scheduled maintenance – before they turned into costly delays and cancellations, Maggiore said. Qantas, for example, was able to reduce pneumatics systems-related pilot reports and delay events significantly on its 737NG fleet, with logbook activity dropping over a four-year period by more than 80 percent, thanks to AHM.
Boeing also provides an Optimized Maintenance Program (OMP). These customized programs reduce scheduled maintenance labor and material cost by more than 20 percent and associated ground time by more than 30 percent on average, while maintaining or improving fleet on-time performance and reducing in-service maintenance activities, Maggiore stated. OMP uses techniques such as text analytics, parametric modeling, and diagnostic analysis to optimize maintenance program scope, contents, and intervals.
One large European 737 operator used OMP to reduce scheduled C-checks from six to four, slashing the carrier’s scheduled maintenance cost by 47 percent, according to Boeing.
A 777 operator reaped OMP benefits, as well, after program implementation and for three years post-OMP. Its delay rate decreased from 0.69 before OMP to 0.27 after OMP. And the carrier reduced scheduled maintenance-related delays by 55 percent and scheduled maintenance non-routine findings by 14 percent, Boeing revealed.
According to GE’s Chapin, the “key ingredients to success in the industrial Internet” – central to modern aircraft data analysis – include focus on data, domain-specific analytics, and a cloud-based platform designed for scale and reuse.”
Domain-specific analytics combines physics and data science expertise to build “digital twins” – the digital model of a jet engine, for example. The virtual image of an engine might enable its overhaul interval to be increased. GE Aviation uses its 3-year-old Predix, cloud-based platform and operating system to create new twin models and applications.
GE Aviation also is driving toward a concept of wide-ranging data exchange. It recently launched the Configuration Data Exchange initiative for the aviation industry “to drive asset productivity and maintenance optimization across the aviation ecosystem,” Chapin said. The data pipeline will enable two-way digital exchange of aircraft configuration data and the exchange of operations, maintenance, and configuration data between participants, such as airlines, MROs, lessors, OEMs, and parts brokers. The pipeline will be agnostic to the various information technology systems of record, the company says.
Rolls-Royce also cites digital twin techniques, “whereby in-service data (as flown) is being combined with simulation data (as designed), and even coordinate measuring machine data (as manufactured), to gain more accurate understanding of the actual condition and performance of engines,” Reuter said. The company is also interested in the prospects of automated analytics presented by machine learning and other forms of artificial intelligence, he added.
Rolls-Royce also stresses the increasing width and depth of the data that it collects and analyzes. “Today we are able to connect the end-to-end system,” said Reuter. The company combines design, simulation, test, and manufacturing data with operational, and maintenance data, along with inputs such as fuel consumption, pilot behaviors, and context data, such as weather and traffic controls, in order to increase customers’ control of and insight into their operations.
P&W cites its ADEM (Advanced Diagnostics and Engine Management) and eFAST (enhanced flight data acquisition storage and transmission) programs. These help the OEM reduce the operator maintenance burden, such as borescoping, and connect trend analytics to remote on-wing-/near-wing maintenance, Fraga says.
ADEM uses a suite of Web-enabled tools and captures flight conditions, temperatures, pressures, low/high rotor speeds, fuel flow, and vibration parameters at takeoff and cruise. For GTF engines ADEM will be expanded to capture these parameters at engine start, climb, and during thrust reverser usage. eFAST, P&W’s next-gen engine health offering, includes a highly secured acquisition, storage, and transmission infrastructure that can record aircraft/engine full-flight data, generating reports and offloading data and reports to a remote ground station.
But MROs like LHT and AFI KLM E&M have by no means yielded the field to the OEMs. LHT has decided to build its technology stack around the Hadoop ecosystem, Stoevesand confirmed. “This gave us the flexibility to choose from a large variety of tools and … to quickly scale to our rising demands.”
But a technology stack is not a market differentiator anymore, he said. Lufthansa Technik’s key advantage is having the data scientists and engineers who “jointly develop use cases on top of that infrastructure,” he added. “They developed algorithms that are able to predict the wear of expendables or … the decrease of precision of certain components long before a component failure.”
LHT also builds ‘digital images’ of the aircraft it works on, collected via the combination of data from internal and external systems, customer and public data, and structured and unstructured data. This enables the creation of “models that can predict the condition of the aircraft or even specific components within the aircraft,” Stoevesand confirmed. Among the data sources are aircraft sensors, maintenance data, airline ops data, logbook data, weather, and environmental information, he says.
AFI KLM E&M has developed Prognos, a suite aimed at predictive maintenance and engine health monitoring, which is applicable to Airbus, Boeing, GE, CFMi, and Rolls-Royce products, among others.
Prognos is proving its value. “Every single component removed following a Prognos recommendation has been confirmed faulty at the test bench,” Parisot cocluded, addint that many no-go potential faults have been anticipated and avoided, thanks to this predictive maintenance tool.