Machine Learning for Data Processing in Vessel Telemetry System: Initial Study

In 2016, there are more than 20.000 commercial vessel (ship/boat) in operation inside Indonesian Maritime Zone. Global competition has pushed Vessel owner and Fleet Management to continuously improve their efficiency and effectiveness of their Vessel Operations. One common way of improvement is by implementation of Vessel Monitoring System which help Fleet Management to locates and monitors their vessel out in the open sea in real time. Vessel Monitoring System usually monitor basic GPS data only, like location, heading, and speed. The recent trend is the implementation Vessel Telemetry System (VTeS), which delivers more data from vessel, such as engine status, propeller status, and fuel consumption. While more parameters are delivered to ground station, a deeper analysis is required to understand the current condition of vessel, hence it requires more time and effort. The lack of person or expert who understand how to analyze those data only make the VMS system underutilized and become useless, therefore it's very important to provide a decision support system to resolve this problem. This research is an initial study on how to build a Decision Support System using machine learning to interpret VTeS data. The selected case is how to detect fuel consumption anomaly in vessel operations. Early study indicates that this system is possible, and in the end of paper we propose a machine learning pipeline model for this system.