Analysis of ship voyage data based on chow-liu tree augmented naïve bayes-method to support biofouling management

Abstract

The Baltic Sea is a sensitive marine environment and a unique brackish water basin. In spite of environmental protection policies of its coastal states, the sea is still under unsustainable load. Environmental impact of shipping is one of the concerns, as the area is a major transport corridor for both passengers and cargo. Many restrictions and environmental regulations direct the shipowners actions. However, there is still issues yet too complicated to be regulated by laws. One such an issue is the spreading of harmful invasive species. In order to gather more data that the future regulations can be based on, the EU Interreg Baltic Sea Region programme funded a multinational research project called “COMPLETE”. Ballast waters are well known pathways for invasive species, but migration by immersed hull structures is a lesser known vector. The role of the South-Eastern Finland University of Applied Sciences in the project is to conduct onboard measurements and data collection on ships’ performance. The data is collected 2018…2019 during normal operation of the ships, and consists of information extracted from voyage data; shaft power, fuel consumption, propeller pitches and rotation speeds, trim, draught, DWT, speed over ground, AIS data and weather conditions provided by coastal meteorological stations. The aim is to demonstrate on individual ship level the impact of biofouling on resistances and fuel consumption, and, in consequence, increase the shipowners’ motivation to keep immersed hulls clean and thus save fuel, reduce gaseous emissions and prevent the spreading of invasive species. However, causation network of collected data includes powerful mixers such as weather conditions. The objective of this paper is to present experiences of utilizing Chow-Liu tree augmented Naïve Bayes formulation for analysing the voyage data. The method is computationally efficient and usually gives quite reliable results even if the collected data contains a lot of noise or inadequacies.