Predictive AI Systems Could Revolutionize Aviation Maintenance

Nov. 5, 2024

As business aircraft become increasingly complex, aerospace software engineers are designing revolutionary new systems aimed at maintaining them, thanks to advances in artificial intelligence (AI).

Even now, AI-assisted computer algorithms are being developed to predict when
aircraft parts will need to be replaced before they fail. Armed with this data, business aircraft maintenance technicians could benefit from more effective and efficient scheduling that could ensure maintenance is performed at the right time, reducing costs and minimizing aircraft downtime.

Machine learning technology has the potential to add an even more powerful layer to predictive aviation maintenance, allowing computers to use historical maintenance data to actually learn how to detect patterns, anomalies and trends.

One of the aviation industry’s most knowledgeable authorities on predictive maintenance, AeroTechna Solutions Managing Director Len Beauchemin, said there are three key things to keep in mind about the pending integration of AI and predictive maintenance.

First, it’s all about the data. “Data extracted from certified components and sources is certified data,” said Beauchemin, who’s based in Kennesaw, GA. “That may sound redundant, but it is crucial to maintaining continued airworthiness.”
Second, the first step for any operator looking to learn more about predictive maintenance is FAA Advisory Circular AC 43-218, which provides guidance for developing an integrated aircraft health management system.

Third, Beauchemin said, AI should complement an effective scheduled maintenance program. Currently, he said, too many “unuseful” tasks continue to be accomplished for no other reason than they’ve “always been done.” He recommends maximizing the use of the data gleaned from the aircraft to drive action.

Collecting and cataloging data only works if it is acted upon, said Beauchemin.

Strides Toward the Future

Like Beauchemin, Fern Campos, aviation director of maintenance at Disney Aviation Group, believes AI will be a vital part of aircraft maintenance. “The technology is there and the industry is ready,” said Campos. “There have already been great strides within the industry to move toward the future.”

Campos and his team use a third-party predictive maintenance program to monitor Disney’s fleet, including specific data sets on certain flight factors, such as flap over-extension, or aircraft landing characteristics. “These practices are just on the cusp of what is coming,” Campos said. “Now, one can download flight data, speed, temps and operating parameters.”

But if third-party vendors analyze an operator’s maintenance data, who owns that data? According to Beauchemin, the operator owns the data when it applies to continued airworthiness.

Vast Potential

The possibilities for using AI for aircraft predictive maintenance appears to be vast and no one is quite sure yet how far the technology can be applied in the industry. Joe Domaleski, founder of Country Fried Creative, is currently at Georgia Institute of Technology working with AI, machine learning systems and tools such as Bayesian networks.

Bayesian networks use probabilistic reasoning to model the relationships between aircraft systems and their components. Based on observed data, these models can calculate the likelihood of component failures and suggest potential causes. Bayesian models are particularly effective for decision-making under uncertainty.

Domaleski said deep learning models could be used in aviation predictive maintenance. They’re increasingly used to process vast amounts of sensor data and detect more complex failure patterns. These models can automatically extract essential features from raw data, making them suitable for complex aircraft systems with intricate interactions between components.

According to Domaleski, key concepts in predictive maintenance include using AI for anomaly detection by identifying patterns that deviate from normal behavior and helping to predict failures in components or systems. Also, an important concept: asset optimization. Domalesky says using predictive analytics to ensure aircraft components are maintained and replaced at the right time will improve overall operational efficiency.

Safety: Priority One

Overall, experts agree: When it comes to ability of AI-assisted predictive aircraft maintenance to improve dependability and efficiency in aviation, the bottom line must always be safety.

“The development of AI under the concept of establishing airworthiness must never negatively impact the baseline of airworthiness we have today,” said Beauchemin. AI should never increase the level of risk, he warned. And it’s important to keep in mind that perfecting the “decision-making tree” will take time Beauchemin said, pointing out that the International Civil Aviation Organization, FAA and other governing bodies must approve any alterations to continued airworthiness.

All of this head-spinning technology has the potential to bring industry-changing improvements across multiple areas of aircraft maintenance. It certainly could be a critical tool in the aviation maintenance toolbox, but when it comes to making final return-to-service decisions, experts emphasize that humans should – and will – always be in the loop.