A Holistic Anomaly Detection System: Improving Safety in Maritime Engine Rooms by Utilizing Methods Mimicking Human Perception

Abstract

Mimicking human senses and perception with digital solutions is surprisingly complex. Doing it safely in a marine environment with modern machinery is even more difficult. To support the development trend towards increased autonomy in ships, the author described the most used sensors, data processing and algorithm training techniques used for anomaly detection within the industry. Insights and learning were benchmarked from the aviation industry in how to build human-centric, ethical, and trustworthy AI solutions for both the development and operation of assets. The work identified typical challenges and solutions for how to counteract them.

To better understand how humans react in situations where machinery fails, the author took a deep dive into how human intelligence, perception, and cognition are similar to or differ from the artificial equivalents, when reaching situational awareness. The author especially investigated what a maritime crew does and how they act in both normal and abnormal situations, to find out how to support their goals using holistic anomaly detection systems, both from a technological and psychological point of view. Finally, the work presented suggestions on developing, implementing, and validating a holistic AI-powered anomaly detection system in a maritime engine room environment.