Militaries spend billions of dollars on satellites to relay sensitive information and provide real-time battlefield data, so it’s imperative to avoid breakdowns and expensive rescues of wayward orbiting objects. That’s why a growing field of engineers, robotics experts and mathematicians are developing AI algorithms to perform what is known as predictive maintenance.
Using tested models that have proven successful with aircraft, experts are using AI to predict a satellite’s performance, temperature spikes and future power failures so that one day maintenance can occur before breakdowns, keeping battlefield awareness and communications intact. The goal is to usher in an era of preventive maintenance that extends satellite life and reduces mission failure.
One expert in predictive analytics believes AI will be key to conquering the highest military domain. “We developed a generic, proven [predictive maintenance] model using historical data for aircraft to try to predict anomalies,” Abdel-Moez E. Bayoumi, an associate dean and mechanical engineering professor at the University of South Carolina’s Molinaroli College of Engineering and Computing, told Apogee. “We found that the model we used for aircraft would also work for satellites as well.”
The implications of Bayoumi’s work could be profound. Space agencies estimate that there are more than 11,000 actives satellites are on Earth’s orbit. They perform civil and defense functions that include weather forecasting, telecommunications and military reconnaissance. To achieve optimal efficiency for satellite fleets, Bayoumi and research scientist Rhea Matthews and their team completed a two-year project in late 2024 designed to predict satellite failures before they occur. The $400,000 project was funded by the South Carolina Department of Commerce in partnership with the Los Alamos National Laboratory and the South Carolina Fraunhofer USA Alliance.
CPL. MICAH THOMPSON/U.S. MARINE CORPS
The team analyzed historical satellite data and real-time environmental factors, such as solar flares, which provided insights into satellite performance and improved maintenance decision-making. “With predictive analytics, you can extract the data, characterize the performance and determine a component’s lifespan,” Bayoumi said on the university’s website. “Then a correction can be made since the user can move from monitoring the data to having control for solving the problem.”
The team first had to find sources for identifying space weather anomalies. It decided on data from the National Oceanic and Atmospheric Administration and relied on a web-based platform called Seradata for historical information on satellite performance.
“We had hundreds of parameters,” so the team performed an analysis on which ones provided the most valuable data, Matthews said at the university’s website. “From there, we figured out how these features connected to describe the relationships that we wanted.”
The team developed two models. One predicted whether an anomaly would occur on a given day, while another estimated its potential severity. Bayoumi’s team then built machine learning models to predict anomalies. The model also estimated remaining useful life and assessed satellite health.
“It sensed the type of weather and parameters that could lead to certain anomalies. These were related to the health and performance of the satellite,” Matthews said.
The team learned it could develop forecasting models with real-time data if it designed a user-based dashboard. The user interface displayed space weather parameters, current readings and forecasts. Seradata lets users search for specific satellites.
Bayoumi’s team provided an overview of their work to NASA and other stakeholders who could benefit from the analysis. A collaboration developed with a space weather expert at Los Alamos. The team’s work will include refining predictive algorithms for accuracy, incorporating additional data, and exploring new applications of machine learning and predictive analytics in satellite operations. “Phase 0 produced a solid framework and prototype that laid a strong foundation for future expansions,” Bayoumi said.
on the subway system. AFP/GETTY IMAGES
Origins of predictive maintenance
Bayoumi has been working on predictive analytics with the U.S. military for years to prolong the lives of aircraft. As director of his university’s Center for Predictive Maintenance, he partnered with Clemson University in South Carolina, the Fraunhofer-Gesellschaft research organization and the South Carolina Department of Commerce to produce a virtual reality (VR) demonstration of how the gearbox in the tail rotor drivetrain of an Apache helicopter degrades over time.
The demonstration revealed wear patterns caused by stress on gear teeth, and showed the stress eventually could lead to the component’s failure. Bayoumi said this VR tool is a steppingstone to provide manufacturers with a methodology to detect faults and fix them on the fly or even produce zero-defect products. “We discovered a problem with the gearbox in the tail rotor of an Apache helicopter,” Bayoumi told Apogee. “If that problem was discovered before [it failed], we could save $53 million a year for that fleet by addressing a single component.”
The experiment showed that grease freely moved from the main gear into the static mast. Leaks in the seal depleted lubricant on the gear mesh surfaces, which led to catastrophic failures. These revelations led to changes in maintenance practices, and a new seal removal tool is now used in the field.
This type of AI-powered analysis could do more than improve safety and save millions of dollars, Bayoumi said. It also could prevent the grounding of a fleet and the tactical disadvantage that would bring. “When you have a gearbox that is essential to that aircraft, you might have to ground that aircraft,” he said. “Mission availability could be extremely important because you don’t want to lose one of your aircraft.”
Bayoumi said the Apache AH-64 helicopter, which first was tested in 1975, is evidence that AI already is extending the lives of military equipment. “The Apache was designed and built in the late 1970s,” he said. “They are still flying now because of the repairs. You can extend not only a component, like a gearbox, but the overall structure of the aircraft. You can improve the skin and the structure of the aircraft.”
Bayoumi has conducted similar research on the V-22 Osprey, a tilt-rotor military transport and cargo aircraft. The progression of AI research continues with the Marine Corps variant of the Osprey, the MV-22. The Marine Corps in November 2025 announced they had hired California-based firm Virtualitics for a contract to deploy the company’s Integrated Readiness Optimization platform to enhance the “sustainment and mission readiness of the MV-22 Osprey fleet,” according to a company news release.
The platform uses AI to provide commanders with analytics to forecast maintenance, Virtualitics said in the release. “The MV-22 is one of the most versatile and critical utility assets in modern warfare, and helping its crews sustain operational readiness with trusted AI is exactly what our mission is about,” Virtualitics chief revenue officer Rob Bocek said in the release. “By delivering AI to the warfighter, at the edge, where decisions matter most, we’re advancing how the Department of War prepares, responds, and maintains superiority in every domain.”
A growing field
The implementation of AI and machine learning into predictive analytics will be evident in many facets of daily life, Bayoumi said. From supply chain logistics to cancer diagnoses and public transit, intelligent machines are increasingly playing larger roles.
Bayoumi points to the COVID-19 pandemic as an example of how AI-enabled processes could prevent costly shortages of medical and personal protective equipment. “What action could be taken ahead of time to be sure you have what you need?” he asked.
Public transportation agencies already use AI to keep trains running. New York City’s Metropolitan Transportation Authority (MTA) in February 2025 announced a predictive maintenance program for its subway tracks. Working with Google, the MTA’s project aimed to detect potential track defects before they escalated into disruptions for customers.
The TrackInspect prototype integrates sensor hardware with advanced cloud and AI capabilities to detect track problems. Google Pixel smartphones were placed on subway cars to capture vibrations and sound patterns through built-in sensors equipped with a microphone, according to an MTA news release.
The sound and vibration data was sent in real time to cloud-based systems, where artificial intelligence and machine learning algorithms generated predictions on trouble spots. “By being able to detect early defects in the rails, it saves not just money but also time — for both crew members and riders,” New York City Transit President Demetrius Crichlow said in the MTA release. “This innovative program — which is the first of its kind — uses AI technology to not only make the ride smoother for customers but also make track inspectors’ jobs safer by equipping them with more advanced tools.”
Predictive analytics also are shaping habits in the medical field, specifically in the fight against cancer. According to a November 2025 report from the American Journal of Managed Care, AI is on pace to “become the most rapidly adopted technology in health care, whether clinicians are ready for it or not. From ambient scribes transcribing doctor-patient conversations, to AI-powered digital twins replacing placebo arms in clinical trials, AI is being inserted throughout oncology,” the report said.
It’s a pattern of man embracing machine that shows no signs of slowing in many diverse fields. Virtualitics CEO and co-founder Michael Amori is banking on that in the military sector, saying his new contract with the 2nd Marine Aircraft Wing shows “how explainable AI solutions are transforming aviation sustainment, helping the Marine Corps maintain fleet readiness, reduce downtime, and strengthen mission success at speed and scale.”