The virus crisis has been very tough on digital innovators throughout commercial aviation. Indeed, airlines slashed IT spending from nearly $50 billion in 2019 to a little over $20 billion in 2020, according to SITA’s 2020 Air Transport IT Insights. The cuts in airport IT spending were also drastic, from almost $9 billion in 2019 to about $3.5 billion last year. It’s safe to assume that cuts in commercial MRO IT spending were also very sharp. 

Yet several MROs still attempted to push IT and innovation forward, despite cash restraints, travel restrictions, managers’ focus on survival and other virus-imposed limits. MRO innovation is not happening as fast as it would have, but it is happening.  

In the last six months Lufthansa Technik’s Aviatar team developed new machine learning models trained on sensor measurements coming from full flight data, says lead data scientist Dimitri Reiswich. This was a non-trivial task. The data had to be parsed from its raw format, and the team had to deal with missing, noisy and faulty sensor measurements. Further, the sheer volume of data and required computational resources posed challenges. 

Reiswich says Aviatar customers, which now include United Airlines for its Boeing 777s and Airbus A320s, should benefit from these more advanced predictors powered by Artificial Intelligence models. “The predictors show a superior performance [compared with] simpler models as they are able to extract failure signatures after accounting for normal system behavior under various flight conditions.”

The Aviatar team will continue to develop new predictors based on machine learning algorithms. In addition, Lufthansa Technik is moving toward running the improved algorithms in a production environment. “This is a common challenge to all industries and is known as machine learning operations, or MLOps,” Reiswich explains.

MLOps moves machine learning from an analytical tool that can be used only by specialists to create custom models to help production line manager make decisions, to a constant collaboration between data scientists and line managers in real time on a wide variety of practical problems.
 
“Despite the difficult market environment, AVIATAR continued its strategic investment into predictive maintenance,” Reiswich stresses. He says LHT has built a dedicated predictive analytics infrastructure that is a highly scalable state-of-art machine learning platform. “This infrastructure allows us to train and deploy machine learning models with ease at reduced costs and also adds new research capabilities.”

The virus forced AAR to abort one potential application of augmented reality, but provided a perfect opportunity to use AR in some other applications, according to director of digital product management Matthew Kammerait.

When the crisis hit, AAR had just piloted using AR to eliminate wait time and improve visibility on a cross-functional authorization to proceed during a heavy check. “Just as we were in the process of optimizing the use case and developing a plan to scale up the implementation, COVID travel restrictions made further progress incredibly difficult, and we were facing the tough decision to put the project on hold,” explains Kammerait.
 
But these same travel restrictions meant onsite audits by FAA and onsite collaboration with customers could not proceed as normal. “We realized AR could be an enabling technology under these new constraints,” Kammerait says. AR and wearable technology enables someone who can't be onsite to see what staff onsite are seeing and thus approximates an onsite visit. Several AAR sites demonstrated AR capability to local FAA authorities and some of these authorities used AR to do site visits and audits.

In a similar case, AAR’s warehouse team proposed using AR to enable socially-distanced communication and collaboration within AAR facilities. The MRO, which also supports defense units, had to find safe and effective ways to continue operating through the pandemic. “We've moved into an active pilot phase with this use case as well and are collecting data about how AR and wearable technology can integrate with our existing systems and processes so we can determine how best to scale up hardware and software implementations once we establish best practices,” Kammerait says.

So AAR plans to us AR for remote support of production by engineering and is now doing remote quality inspections. It has discussed using AR for remote training of technicians. “Eventually, we want to weave both AR and virtual reality into our training processes as we see them as a bridge between digital educational content and real world practice,” Kammerait says.
 
But there are challenges to exploiting AR because AR technologies are still highly fragmented. “Many aspects of solution design or architecture are left to the end user or implementer to figure out for themselves,” Kammerait says. 

Further, each current hardware or software solution involves a number of tradeoffs involving factors such as battery life, wearability and connectivity. The AAR exec predicts progress will be made in the next few years toward a single AR platform with much better performance for all applications and in all environments. And early adopters or experimenters like AAR will have a lot to say about the technology’s development.

Apart from AR, AAR is still working on drone-based inspections, digital maintenance workflows, and artificial intelligence capabilities embedded in its core business applications.

Across the Pacific, GAMECO has also been active in AR, notes CEO Norbert Marx. “We have researched and compared different AR glasses and invested in about a dozen sets,” Marx says. Most of GAMECO’s AR glasses are deployed in line maintenance, especially in domestic outstations like Shenzhen, but also in international stations in Australia and New Zealand. “The main purpose is to use AR for remote support of production by engineering.” So far this has happened not routinely, but only in selected cases. 

GAMECO is also working with some of its customers that want to use AR to oversee maintenance of their aircraft during the virus crisis.

For training mechanics, Marx says his MRO has found other technologies to be more efficient than AR. These tools include online training, for example on aircraft types and aviation regulations, two-dimensional simulators for aircraft type training and three-dimensional simulators for engine run-up training. “We are experimenting with remote quality inspections,” Marx says. “But for regulatory reasons, they are not in daily use.”

Even smaller shops are pursuing digital innovation. Over the last six months, AJW Technique has audited its processes to find the hurdles distracting technicians from touch time on components, explains strategy and business development manager Monica Badra. “We have set up a digital incubator team to act as change agents focused on process reengineering and adoption of new technologies,” Badra explains.
 
During the last quarter of 2020, the MRO focused on personal interviews to highlight recurring problems and design a digital roadmap to counter key pain points and improve efficiency. “Digital advancement has gone from nice-to-have to essential for survival,” Badra stresses.

Short-term, AJW Technique has been deploying tablets on the shop floor to reduce paper printing and streamline entries into enterprise resource planning system. “The technology needs to work for us, not the other way around,” Badra says.

The MRO has used a gaming system to give mechanics visibility of contracted turnaround times, prioritize work flows across all teams and manage performance management. It is now looking into asset tracking to ensure real-time visibility of units both on the shop floor and en-route to customers.

Medium- to long-term, AJW Technique will integrate with customers through portals that show work order status in its facility. This sounds a bit similar to the way major airframe shops show check status to their customers.

Badra says software will also enable accurate forecasts of work on both scheduled and unscheduled removals. And the MRO will use a flexible resource allocation model to obtain insights into the training and qualifications required for mechanics to fulfill work plans. In addition, the model will improve material forecasting, reducing holds on piece parts.