Underwater data centers are increasingly considered for deploying computing infrastructure in constrained environments; however, maintaining such systems after deployment remains challenging due to limited physical access and harsh operating conditions. In many cases, faults are detected only after noticeable degradation or service disruption has already occurred. Most existing monitoring approaches remain reactive in nature or focus mainly on cooling and protection, which provides limited support for identifying early-stage degradation. In this work, an edge-intelligent predictive maintenance system for underwater data centers is presented based on multi-domain sensor fusion. The system integrates pressure, temperature, conductivity, moisture, vibration, acoustic, turbidity, and electrical health sensors, with data processing performed directly on an embedded edge platform. A Raspberry Pi is used to acquire sensor data and carry out preprocessing, feature extraction, and time-series analysis locally. Machine learning methods are employed to learn normal operating behavior and to identify abnormal patterns across correlated sensor signals. Instead of relying on fixed threshold limits, the proposed approach evaluates combined sensor trends to support condition-based maintenance decisions. Experimental evaluation under simulated underwater conditions indicates that multi-sensor analysis can provide earlier and more reliable fault indication when compared with single-sensor monitoring approaches. Overall, the proposed system demonstrates a practical and scalable predictive maintenance framework for underwater data center environments and supports improved long-term reliability through embedded intelligence and data-driven condition monitoring.
Keywords: Predictive maintenance, underwater data centers, sensor fusion, edge intelligence, Raspberry Pi, anomaly detection, condition monitoring.
How to cite this article: Suryawanshi A, Borde S, Rangdale S, Kale N. An edge-intelligent predictive maintenance system for underwater data centers using multi-domain sensor fusion. Int J Drug Deliv Technol. 2026;16(7s): 143-155; DOI: 10.25258/ijddt.16.7s.18
Source of support: Nil.
Conflict of interest: None