February 17, 2016      

In late 2015, the Boeing Co. announced the establishment of the Boeing/Carnegie Mellon Aerospace Data Analytics Lab in Pittsburgh. Researchers at the lab, which is expected to cost Boeing $7.5 million over the next three years, will use analytics and artificial intelligence to improve aircraft maintenance and safety.

The researchers hope that smart use of AI and big data will enable them to analyze the information generated during the design, construction, and operation of Boeing’s aircraft.

Such capabilities would enable aircraft owners to develop maintenance schedules that are determined by each airplane’s history and component performance, for example, rather than relying on historical norms. The founders said the lab, which is expected to employ 20 people, could also help improve aircraft performance and passenger comfort.

Led by Jaime Carbonell, head of Carnegie Mellon University‘s Language Technologies Institute, the lab hopes to combine data gathered during design and construction with data collected by onboard sensors and embedded computers.

“The mass of data generated daily by the aerospace industry overwhelms human understanding,” Carbonell said. “But recent advances in language technologies and machine learning give us every reason to expect that we can gain useful insights from that data.”

Carbonnel’s expertise in language technologies includes developing AI systems that can automatically analyze written reports for evidence-based predictive maintenance. Experts from CMU’s Machine Learning Department are also involved; these researchers bring expertise in finding patterns in large databases.

“A Boeing aircraft such as the 787 Dreamliner combines thousands of on-board sensors, text from pilots and mechanics, structured engineering data bases, across the entire fleet collected from each of the client airlines,” Carbonell told The Washington Post. “This provides a golden opportunity, merging CMU’s capabilities and Boeing data to address problems such as predictive analysis for preventive maintenance — rather than after-the-plane-is-grounded maintenance.”

Carbonnel also revealed that the long-term goal of the lab is to develop self-healing airplanes.

“The audacious plan is closer to self-healing airplanes: evidence-based predictions of what may not be working right tomorrow, to enable preventive inspection or replacement before a failure, and hence to lower costs of coping with real unscheduled failures and to increase safety,” he said.

Block diagram of intelligent fault-tolerant flight-control system

One researcher is working on a fault-tolerant
flight-control system.

The Boeing-CMU collaboration is all part of a move towards increasing use of AI across what is known as “the Internet of Aircraft Things” (IoAT), and it’s poised to transform aviation.

Can AI lead to ‘uncrashable’ planes?

In Japan, for example, Shinji Suzuki, an aeronautics professor at the University of Tokyo’s School of Engineering, is investigating the use of artificial intelligence to keep even damaged planes in the air.

Suzuki analyzed differences in the flight-control techniques of experienced pilots and novices and found that experienced pilots do not rely solely on flight instruments to assess attitude, altitude, and speed. Instead, they have a knack for immediately grasping the picture.

“Say part of a plane’s main wing breaks off during a flight,” Suzuki said. “The pilot will try to keep the aircraft flying by adjusting its position, engine power and other factors.”

Suzuki is trying to create a system that can test such variables much faster and find the best solution to prevent a disaster.

Suzuki’s AI system — which he claims could lead to “uncrashable” planes — has been tested in partnership with Fuji Heavy Industries Ltd. and the Japan Aerospace Exploration Agency.

Aviation AI is being further assisted by a growing number of sensors in jet engines. At last year’s Paris Air Show, Bombardier Inc. showcased a jetliner that uses a Geared Turbofan (GTF) engine fitted with 5,000 sensors.

The GTF engine generates up to 10 GB of data per second. This data is used to predict engine demands in order to control thrust. GTF engines have reportedly enabled a reduction in fuel consumption by 10 to 15 percent, as well as bringing “impressive performance improvements” in engine noise and emissions.

Sensing the right vibes

Meanwhile, an algorithm developed by Rodrigo Teixeira, a researcher at the University of Alabama in Huntsville, could greatly increase accuracy in diagnosing the health of complex mechanical systems, including those used in aviation.

Announced in early February of this year, the algorithm is designed to extract “dependable and actionable information from the vibration of machines,” said Teixeira.

Vibration analysis seeks out anomalies in the vibrations of engines and gearboxes to detect wear and wear and trigger maintenance responses. One of the major challenges for vibration analysis is dealing with the large amount of noise generated by complex machinery during normal operations.

Whereas traditional modeling tools assume that vibrations are static, Teixeira’s algorithm has been trained to incorporate basic principles of physics that govern faults in a vibrating environment. The algorithm has already been trialed in U.S. Army helicopters.

The combination of artificial intelligence, big data, and sensor-filled equipment that’s emerging today promises increased safety, reduced costs, and greater efficiency across the aviation industry.

And it’s no small beans we’re talking about in terms of cost-savings: Worldwide aviation maintenance costs alone amount to around $40 billion each year, according to Predictive Aviation Analytics Inc., a company that develops software designed to predict aircraft component failure.

Add in reduced fuel costs and greater efficiency, and those cost savings grow rapidly.

Now, if only artificial intelligence could do something about airplane food, we’d have even more reasons to celebrate.