The most important next step in predictive maintenance is simply its wider adoption and use, according to Jim Fitzgerald, aerospace sales director at Spark Cognition. “Innovative commercial and military aircraft operators are already using machine learning-based predictive analytics, so the next steps are simply increased adoption through increased familiarity with the capabilities of this technology,” Fitzgerald says. “We believe this has already begun, given the profound advantages provided by Artificial Intelligence-based technologies in the very competitive aerospace market.”
Other Artificial Intelligence tools could also play a larger role in the future. Fitzgerald says digital twinning will play a more important role in predictive maintenance, and the infrastructure required for digital twinning is basically similar to what is required for machine learning-based predictive analytics. “Digital twinning can tie together the entire lifecycle of a given aircraft platform or subsystem, from design to MRO to the sunset of the platform.
Conceptually, traditional use of sensor data for component monitoring through a decades-old approach like thresholding or other physics-based models could be an input to digital twinning. If insights from sensor data were enhanced by Machine Learning, there would most likely be new relationships uncovered between individual sensors and subsystems across the entire platform. This would help predict anomalous behaviors earlier and increase readiness levels and dispatch rates.”
Another Artificial Intelligence tool, Natural Language Processing, is already boosting efficiency and reducing costs through what Fitzgerald calls prescriptive, rather than predictive, maintenance. “It's empowering maintainers and technicians to diagnose aircraft problems faster, codify tribal knowledge in places where the publications may not have been recently updated, and fix aircraft issues, correctly, the first time.”