THE VALUE OF BIG DATA TO MROS

It’s clear. Big data is one of the important movements in aerospace with the potential to radically improve efficiencies. To end any confusion and explain why big data truly matters, especially to MROs, Aerospace Tech Review sought the wisdom of GE Aviation, MTU Maintenance and Pratt & Whitney to see where the use of big data stands today and where it is heading.

What is Big Data?

The definition of big data is “an accumulation of data that is too large and complex for processing by traditional database management tools.” Big data information will be used to analyze systems and organizations can systematically extract information from, or otherwise deal with data sets that are massive and complex, much more than can be dealt with by traditional data-processing application software.

“Big data is a type of data that is diverse in nature (structured and unstructured) and growing at an exceptional rate,” agreed Dinakar Deshmukh, GE Aviation’s vice president of Data and Analytics. “Conventional software and tools are not capable of handling this volume, velocity and variety.”

Where Aviation Big Data Comes From and Why It Matters

In the aviation world, big data encompasses all of the onboard performance and diagnostic information generated by aircraft systems in real-time – and more.

“A significant portion comes from sensor data from aircraft,” Deshmukh said. “We also have data sharing agreements with airline customers to capture other data from operations, such as borescope images and video. To accompany this sensor and operational information, GE also tracks data from our overhaul and repair shops of the maintenance work performed to return an engine to service.”

All this information is stored on GE Aviation’s cloud-based infrastructure, which allows appropriate employees with authorized access to perform various in-depth analyses, he added. “Analysis of big data available to GE Aviation helps us provide more proactive maintenance recommendations so that unscheduled customer disruptions can be minimized.”

This is the benefit of big data to aviation. When you are able to sift through all this information to detect equipment performance trends and hints of possible future problems, the potential for improved customer service and fewer Aircraft on Ground (AOG) situations is obvious and impressive.

“This is why big data is expected to transform the aviation industry, as the large volume of data will enable proactive, predictive analysis,” said Dr. Friedhelm Kappei, head of Industrial Engineering at MTU Maintenance. “For instance, it can be used to improve machining processes, or for earlier and faster failure detection in aircraft operation. The aviation industry is on its way to discovering potential use cases for big data.”

Like GE Aviation, MTU gathers its big data from various sources. “The data we work with comes from our customers’ aircraft, our worldwide facilities, engine trend monitoring tool as well as airlines on-wing maintenance tools,” Dr. Kappei said.

“It is stored on MTU servers. The data is then analyzed by our tools and reviewed by our experts to provide accurate recommendations for customers.”

Data Collection is Nothing New

Before we go any further down the big data rabbit hole, we should acknowledge that data collection is not a new activity for aircraft operators. What has changed, thanks to modern software tools, is their ability to mine it for information in full.

“Airlines have been capturing and storing full flight data from all their aircraft for many years in order to support their FOQA and MOQA programs,” said Arun Srinivasan, Pratt & Whitney’s associate director of Strategy and Engine Health Management.

“While full flight data was being captured and stored, the airlines were also sending us snapshots of data during the flight via the ACARS system. We have been using this data for 15-20 years to provide airlines with Engine Health Management services.”

In the same vein, “engines have been generating data for many years, going all the way back to the JT8D engine and even earlier,” he said. However, what has changed is the amount of data being generated by succeeding generations of P&W engines. “With the sophistication of the engine controls improving, the amount of data generated by the engines has been increasing. Our latest GTF series of engines incorporate about 40% more sensors than the V2500 family of engines. An average two-hour flight for a GTF engine generates about 4 million data points.”

Today, thanks to the capture of full flight data, Pratt & Whitney has access to the performance of its newer engines over the entire flight and access to a much larger data set. When these numbers are crunched using big data analytical methods, this data allows P&W to trend and monitor the gas path of the engine, plus the various other elements/pieces that make it such a complex machine.

While big data analytical tools and techniques are improving, so too are the quality and breadth of data sources being mined. A case in point: “Companies such as Teledyne Controls, a data delivery solutions provider, have devices on board aircraft that capture this data during the flight and provide the airline with capabilities to download the data on a regular basis,” said Srinivasan. “Working with Teledyne Controls, Pratt & Whitney will expand its capabilities for collecting full-flight data to a larger population of aircraft by using a service that many airlines already subscribe to.”

Extracting Useful Information

Back to our big data: Like the 40 million-plus books housed in the Library of Congress, there’s a lot of useful information waiting to be mined in big data when you have the right software tools to collect, analyze, and draw conclusions from it — assuming that you do.

“When it comes to big data, analysis is certainly one of the largest challenges we see in the industry,” explained Dr. Kappei. The reason: “Processing big data (both storing and analyzing) usually requires new IT infrastructure, new process designs and often a new way of thinking as well as creativity to generate value from the data,” he said. “And since big data needs big storage solutions and a lot of processing power, business cases and commercial aspects need to be considered as well.”

A further way to address the challenge of running big data analyses is by digitizing legacy data (including paper manuals and maintenance files) and ensuring that all of it is accessible to an MRO’s big data software tools.

Accomplishing this massive task — plus ‘digitalizing’ the entire company’s work environment so that it entirely exists in digital form — offers substantial benefits but takes time. For instance, at MTU Maintenance “we expect the digital change process to be a continual process over the next five to ten years,” Dr. Kappei said. “The change process has been initiated and we are currently in the process of setting up the tooling that we are going to use in order to store and analyze big data.”

AI Central to Analyzing Big Data

The reason big data has become a buzzword is because software algorithmic tools have been developed to tackle its immensity; allowing MROs (and others) to mine it successfully for useful information. The most sophisticated of these tools are driven by artificial intelligence (AI) and machine learning (ML).

Together, AI/ML tools make it possible for big data software to ingest extremely large data sets, and to analyze them for trends and other indicators upon which conclusions may be drawn. Add the fact that AI/ML systems can ‘learn’ as they do this – that their past analytical conclusions can be drawn upon to speed up the current analytical process and to infer new insights previously unthought of – and one can see why AI/ML systems are so vital to effective big data analysis.

“Artificial intelligence is a great tool (among other algorithms) for analyzing large data sets and produces the best results when both volume and data similarity are given,” said Dr. Kappei. “For instance, artificial neural networks are trained based on example data, allowing them to then process new data sets easily.”

“The volume of data that can be processed this way is so huge, that it would take years for a human to analyze,” he noted. “In cases where the work might be monotonous (as when images are all of the same item shot from the same perspective over time) and require a high level of concentration, some of this work can be taken on by the computers and then checked by employees. AI also gives the opportunity to create self-learning or supervised learning processes for big data analytical systems.”

It is worth noting that dig data has been instrumental in the advancement and improvement of AI/ML systems. “Without big data, AI solutions would be suboptimal,” said Deshmukh. “This is because AI solutions are self-learning. They become better working with data over time, but to do this they need to be fed with more and more data that has been validated.”

“In other words, dig data makes AI perform better,” he said. It’s a mutually beneficial relationship.

How GE Aviation, MTU and Pratt & Whitney Are Using Big Data

To assess the true usefulness of big data in MRO operations, one needs to know how these companies are employing it in their businesses. Here is what GE Aviation, MTU Maintenance, and Pratt & Whitney told us.

GE Aviation uses big data to help monitor its on-wing aircraft engines, so that this MRO can predict operational disruptions before they happen. “We use the same information to improve Service operations when an engine is inducted in the shop,” said Deshmukh. “Big data also gives us a better idea of what engines will arrive at our shops at what time, what kind of workscopes will need to be performed when the engines do arrive at the shop, and what kind of material should be available on site to carry out the shop visits.”

“From an MRO perspective, big data supports engine removal forecasting and maintenance planning for an optimized shop visit,” he added. “Big data also supports GE’s monitoring of engines on-wing for preventative maintenance, focused on identifying engine issues early to keep engines flying.”

These services are backed by GE Aviation’s analytics-based maintenance technology, which is an AI-based solution. “Analytics-based maintenance uses digital copies or digital twins of engines and engine parts developed using aircraft sensor data and operational data, such as geographical locations and borescope images,” said Deshmukh. “The digital versions of the engines are aggregated to help predict in advance the likelihood of engine removal.”

MTU Maintenance has a slightly different but generally similar approach to using big data.

“At MTU, we work with smaller data sets; for instance, from our proprietary engine trend monitoring system which gathers data from operations (e.g. region, derate and engine performance, and shop information) and helps us to forecast remaining on-wing time and optimal engine removal points,” Dr. Kappei said. “Furthermore, data from operations (on-wing) is combined with actual results found in the shop to calculate scrap rate probabilities. Engine trend monitoring data is used to assess engine deterioration. Through this analysis, workscopes can be adjusted and improved. This information can be particularly helpful to our purchasing and sourcing departments and has a positive effect on prices for our customers.”

Due to the power of big data, MTU has been able to integrate its engine trend monitoring system into the company’s engine fleet management software. Thanks to this analytical enhancement, MTU’s engine fleet management software can now generate scenarios, engine MRO strategies and workscopes to maximize the use of existing client assets while avoiding unnecessary service expenditures — “all at the touch of a button,” said Dr. Kappei. “We can also provide accurate and timely recommendations as to which engine to reactivate in a restart scenario, for example. Our system can generate workscopes down to the module level and include engine leasing and greentime variables into its assessments as well.”

MTU is also applying AI-enabled data analysis to borescope inspections. To this end, it has been working with Aiir Innovations on that company’s AI-enabled automated boroscoping tool, which collects/analyzes data and images from within an engine.

“We’re currently exploring this technology further and working towards maturing the technology for more reliable results,” Dr. Kappei said. “It is worth mentioning that such inspections are not intended to replace an actual engine check or boroscope inspection (BSI). They are not certified by aviation authorities and are merely an additional support to the engineer at this time — both in terms of gaining insight when one is not physically present on location, but also to gain a comprehensive view of what is going on in an engine and aiding in removing the margin for human error in such checks.”

Pratt & Whitney’s Srinivasan agreed with Deshmukh and Dr. Kappei. He added that, “The ultimate goal is to contact a customer ahead of a potential issue; not have a customer contact you with a problem. The ability to access full flight data has vastly improved our ability to detect and proactively prevent maintenance events. We’re now able to provide critical services in a matter of hours and days, rather than months and years.”

Pratt & Whitney has been using access to engine data to address specific situations with its customers as well as to manage fleet trending through aggregation of data benefiting all of its customers. Its engine health management service ADEM (Advanced Diagnostics and Engine Monitoring), employs a suite of web-enabled software tools to provide expert analysis of engine health data for more than 8,000 engines in service.

“While some analytics applications can be generalized, we recognize that each of our customers is unique and we want to be able to offer them the personalized customer integration they expect from an OEM leader such as Pratt & Whitney,” said Srinivasan. “Based on our experience, machine learning and automated decisions have to be used in conjunction with our OEM expertise and the applications to a particular mission, fleet, region or customer needs to optimize our engine availability for their operations. Collecting fleet data enables us to maximize a customer’s specific engine performance and engine time on-wing, while maintaining predictable MRO spend.”

A Qualified Success

There is no doubt that big data analysis can provide substantial benefits to aircraft owners and operators. Thanks to the power of AI-enabled analysis tools, problems can be detected and prevented before they become disruptive. Money can be saved by minimizing unplanned maintenance servicing and Aircraft on Ground (AOG) situations, to everyone’s benefit.

“Our customers benefit from better and more accurate planning and cost information, fast response times, and verified options and recommendations for fleet planning/decision-making,” said MTU’s Dr. Kappei. “Ultimately, we believe the aviation industry will benefit most from big data when the jump is made from prescriptive maintenance to becoming truly and accurately predictive. We are excited to see the developments that come next!”

At the same time, GE Aviation’s Deshmukh cautions against MROs and their clients expecting too much from big data analysis at the present time.

“The use of big data to improve aircraft engine services is a journey and we still have room to grow in the journey,” he said. “As well, the foundation of AI-based analysis requires access to robust data streams. This is why GE Aviation has efforts focused on capturing data appropriately and validating that data to scale solutions across the business and for our customers.”

These concerns explain why Pratt &Whitney stresses the importance of having skilled human operators assess the conclusions drawn by big data analyses and using their expertise to apply these conclusions in the best and most appropriate ways possible.

“The engineers who are monitoring the engines for our customers are able to tap into the vast knowledge within the engineering community here at Pratt & Whitney and provide key insights to our customers to keep their fleets flying,” said Srinivasan. “The advantage of the human-in-the-middle solution that we offer is that we can minimize the burden on the airlines by filtering out potential false positives.”

The Moral of the Big Data Story

Qualifications notwithstanding, the application of AI-analyzed big data in aviation maintenance is already delivering big benefits to MROs and their clients. As the data pool generated by aircraft grows going forward, and the tools to analyze it gain additional power and sophistication, the result should be enhanced preventative maintenance for all aircraft in service, leading to fewer unexpected failures and more predictable MRO costs for everyone in the aviation business.