![AI graphic](/sites/default/files/styles/crop_freeform/public/2024-08/ai-graphic-ifs-article.jpg?itok=e8zs3wMJ)
Last month, enterprise software company IFS acquired artificial intelligence-powered MRO software provider EmpowerMX as part of its plans to accelerate its AI strategy. Rob Mather, IFS’s vice president of aerospace and defense industries, sees several viable ways in which AI could transform the MRO industry—now and into the future.
“When we talk about AI in MRO, we have to talk about different time thresholds,” Mather tells Aviation Week Network, noting that he thinks of the technology’s potential in three waves. “The first wave is what we can do now with existing regulatory frameworks, largely in a supporting role. The second is what we can do in the relatively near future, but this still involves humans being the ultimate decision-maker. The third wave is much further in the future and involves some significant regulatory changes in terms of how we do things in aviation maintenance, which ultimately results in much more AI-driven automation.”
Mather notes that AI-driven optimization tends to be overlooked in today’s AI discussions because it has been around for longer and does not utilize generative AI, but he asserts that this use case for AI has the potential to positively impact supply chains and maintenance scheduling. For instance, AI can not only enable when and where maintenance tasks and visits are scheduled, but it can also be used to optimize the sequence in which tasks are performed and to whom tasks are assigned. Mather says this can drive down the cost of maintenance by minimizing travel distance and wasted time while maximizing efficiency and utilization.
Mather says the industry is currently able to use AI for error detection and reclassification. “This can be done after the fact using specialized large language models or deep learning to review records,” he says.
“An easy example is around ATA classification of faults. This is often a tricky prospect for technicians—particularly when the symptom is showing on a different system than the root cause,” says Mather. “By improving the ATA classification accuracy, you can improve your reliability data and provide a better foundation for predictive maintenance models. If we shift it to a proactive or real-time mode with a co-pilot prompting the technician with suggestions for the ATA chapter as the record is being entered, the benefits are even greater.”
One company pursuing this type of use case is American Airlines, which has been exploring the use of natural language processing to more accurately identify which ATA codes should be used to classify faults during maintenance.
Another example Mather cites as a potential use case is automated failure, troubleshooting and repair identification. “This is one of the areas where technicians spend a lot of their time. AI can be used to identify failures and suggest the appropriate troubleshooting tasks for particular faults, even including additional useful information like success rates,” he says. “It can suggest repairs and proactively provide the instructions or references necessary to the technician, avoiding the need to search through reference manuals. In fact, utilizing a co-pilot to help navigate through reference documents is another valuable case by itself. Anything that speeds up the technician is incredibly valuable because they are a scarce and valuable resource.
Perhaps one of the most widely accepted use cases is predictive maintenance and anomaly detection. “We could do simple, backward-looking predictive maintenance without AI. OEMs have been able to provide health monitoring for years. However, applying machine learning to predictive maintenance makes it more cost-effective and more accurate,” says Mather.
“Couple that with sensor data hooked up to anomaly detection and suddenly you can combine the understanding of the historical data with interpretation of the real-time data in a way never before available to operators,” he adds. “Predictive maintenance today has the power to reduce AOGs by replacing a component before it fails, at a location where you have the time, materials and skills to replace it without delaying flights, improving your on-time departure rates. This generates both direct and indirect cost savings.”
To unlock the most value of AI for predictive maintenance, Mather asserts that the industry will need to move from a model based on scheduled inspections at fixed intervals to a model in which an aircraft’s condition is constantly being monitored so that maintenance only needs to be performed as required.
“This requires the buy-in of the regulators—something that organizations like IATA are working to lay the groundwork for with the regulators,” he says. “There are also cases today that are paving the way, such as Rolls-Royce’s IntelligentEngine program that can extend the life on in-service engines and reduce the amount of maintenance needed based on how they are performing and how they have actually been flown.”
Rolls-Royce has been using IFS Maintenix software since 2019 to exchange data with airlines that operate its engines. Earlier this year, the engine OEM also signed a five-year contract with AI systems developer Aerogility to utilize its enterprise digital twin system.
Editor’s Note: Stay tuned for this week’s Fast 5 interview with Rob Mather, where he shares more insights about IFS’s AI plans and the technology’s potential benefits—and challenges—for the MRO industry. The September issue of Inside MRO magazine will also feature insights from Mather and other AI experts about data strategies for implementing the technology successfully.