In a soundproof chamber at the Naval Surface Warfare Center, Philadelphia Division (NSWCPD), engineers are testing an experimental machine learning model designed to detect early signs of failure in high-pressure air compressors through vibration analysis. This initiative is part of NSWCPD’s expanding portfolio of artificial intelligence (AI) and machine learning (ML) projects, which show promising potential for future Navy applications.
High-pressure air compressors are critical in submarine operations, supporting various essential ship functions and ensuring reliable performance to maintain maneuverability and mission effectiveness.
The Navy’s Condition-Based Maintenance Plus (CBM+) initiative seeks to merge traditional maintenance practices with AI-driven prognostics, aiming not only to identify issues but also to estimate their progression and the time remaining until equipment failure. The compressor project at NSWCPD represents an exploratory step in this direction, assessing how AI and ML can analyze carefully collected vibration data before considering operational deployment.
“As a warfare center, we are performing applied research into how AI and machine learning can sharpen the tools we provide our Sailors,” said NSWCPD Technical Director Nigel C. Thijs, SES. “Projects like this help us understand where AI adds value, where it still falls short, and how we can align digital innovation with our core mission of delivering warfighting capability in both acquisition and sustainment to the fleet.”
NSWCPD has been cultivating AI/ML expertise on multiple fronts. Engineer Dr. Kaitlyn Sitch recently received a SMART SEED Innovation Award for developing algorithms to predict power-system health, aimed at reducing downtime and enhancing energy resilience across Navy platforms. The command is also advancing the concept of “digital twins” for shipboard systems, such as Enterprise Remote Monitoring (eRM), utilizing Python-based machine learning models to forecast anomalies in hull, mechanical, and electrical equipment. These initiatives frame the compressor work as part of a broader strategy to apply data science to real-world Navy challenges.
The compressor model is intentionally constrained due to the rarity of real-world failures in operational platforms. To address this, NSWCPD engineers constructed a dedicated test loop to induce faults, including simulated air leaks and inlet restrictions, under controlled conditions. Using arrays of accelerometers, they captured every change in vibration, providing a reliable foundation to evaluate how different machine learning techniques differentiate between normal and faulty behavior.
“Our lab tests to date show real promise: on sample data, our machine learning models distill thousands of vibration features into just 10 key indicators that reliably flag common faults, such as leaks and restrictions,” said NSWCPD Machine Learning Engineer Colin Dingley. “The next step is to scale AI with more diverse data and edge hardware to see whether this holds up in real-world conditions — it’s challenging, but the early results are encouraging.”
Submarines face a unique constraint: limited underwater bandwidth prevents the transmission of full-fidelity vibration streams for analysis. This reality necessitates the use of “edge” processing, where compact, energy-efficient hardware is mounted near the machinery to run AI models locally, transmitting only essential health information to operators and shore systems. Luna Labs designed the embedded eCBM node that NSWCPD is testing in the Anechoic Chamber, a controlled environment allowing engineers to evaluate algorithm performance without interference from shipboard noise.
“AI will complement our Sailors,” Dingley stated. “Machine learning and artificial intelligence will become part of a Sailor’s tool belt — another layer of protection. It’s been said that a Sailor has 26 hours of maintenance in a 24-hour day. We’re trying to make that more manageable by using AI to highlight which components truly need their attention.”
NSWCPD CBM+/Prognostic Health Monitoring Lead Sherwood “Woody” Polter traces this research back to 2012, prior to AI gaining widespread recognition and before today’s processors became capable of running extensive models at the edge.
“As a Navy laboratory, we routinely partner with industry, academia, and other government organizations to advance technology,” Polter remarked. “Thanks to computers capable of processing these large models, we can experiment with AI-enabled health monitoring in ways that simply were not feasible a decade ago.”
Polter emphasized that no single machine learning model can serve every shipboard system, as each piece of equipment has unique dynamics, operating ranges, and failure modes. However, the high-pressure air compressor utilized in this project is prevalent across both surface and undersea fleets, making it an ideal testbed.
“This type of compressor is found throughout the fleet,” he noted. “It is critical on submarines, but that is just the tip of the iceberg.”
The long-term vision extends beyond manned platforms. “Besides surface ships and submarines, we are actively pursuing machine learning with eCBM technology for unmanned undersea platforms,” Polter added. “On those systems, where no crew is standing nearby with a wrench and a clipboard, trusted autonomy will depend as much on reliable self-monitoring as on navigation and communications.”
At its core, Polter describes the work as “condition-based monitoring plus prognostics.” He likens it to a car’s owner manual that recommends changing a fan belt at a specific mileage, suggesting an AI-enabled vehicle might predict a belt’s failure based on its unique usage history.
“Apply that to being on a ship,” he explained. “Before something breaks, it bends. Our goal is to find that bend in the data.”
In practice, “finding the bend” involves focusing on vibration data as a rich source of information. “We primarily use vibration sensor data and can detect outlier frequencies of interest, which we compare to baseline measurements,” Polter said. “That enables us to apply data analytics and our algorithms to build machine learning models for anomaly detection.”
Looking ahead, the estimation of remaining useful life (RUL) is regarded as a longer-term goal rather than an immediate deliverable. “Remaining useful life analyzes a lot of data,” Polter noted. “There will be requirements for the system to notify the user or, if it’s an unmanned system, to send remote data to the platform operator. But all of this is only possible if the models are well developed, using big data, to provide an RUL solution.”
For now, NSWCPD leaders view the compressor project as one of several “learning labs” for AI/ML within the command. “This work will directly impact the warfighter,” Thijs concluded. “When we transition AI-enabled health monitoring from the lab to the fleet, as we are already doing with condition-based monitoring and enterprise Remote Monitoring, we can help Sailors better prepare for or avoid casualties, increase operational time, and derive more value from every maintenance dollar.”
As the compressor wound down in the testing facility, the sound faded into silence, but the data it generated continues to contribute to a growing digital ecosystem of models, simulations, and lessons learned, shaping how NSWCPD will integrate AI and machine learning into future machinery health projects. “Every time we improve the Navy’s ability to see a problem before it becomes dangerous, we’re protecting our Sailors,” Dingley remarked. “That’s what this is really about.”
NSWCPD employs approximately 2,700 civilian engineers, scientists, technicians, and support personnel, conducting research and development, test and evaluation, acquisition support, and logistics engineering for non-nuclear machinery and associated systems for Navy surface ships and submarines while also leading cybersecurity efforts for all ship systems.
See also
AI Study Reveals Generated Faces Indistinguishable from Real Photos, Erodes Trust in Visual Media
Gen AI Revolutionizes Market Research, Transforming $140B Industry Dynamics
Researchers Unlock Light-Based AI Operations for Significant Energy Efficiency Gains
Tempus AI Reports $334M Earnings Surge, Unveils Lymphoma Research Partnership
Iaroslav Argunov Reveals Big Data Methodology Boosting Construction Profits by Billions
