System Architecture for Real-time Condition Monitoring and Anomaly Detection on Ships

Sea-based rescue organizations play a vital role in maritime civil security by resolving offshore emergencies, securing sea routes, and monitoring critical naval infrastructure. However, increasing marine traffic makes rescue operations more difficult, demanding better organization and orchestration of available resources. Therefore, ensuring efficient maintenance, availability and timely deployment of rescue cruisers is of great importance, particularly in rough sea conditions. To achieve this, we design an AI-based system for predictive maintenance and condition monitoring that enables real-time analyses of the ship’s sensory data and the detection of anomalies in the system behavior. In this paper, we detail our software and hardware architecture for these purposes, and we discuss the respective requirements regarding onboard and off-ship data analysis from various data sources (sensor data, camera data, weather data). We underpin our approach by emulating data flows derived from sensory data. Finally, we provide the first results in AI-based anomaly detection, allowing, for instance, early engine malfunctions identification before the actual failure occurs.