Impact of hull cleaning and crew performance on bunker consumption: Classification and optimization of underwater hull cleaning intervals under data uncertainty

Abstract

This thesis investigates the impact of underwater hull cleaning and crew performance on bunker consumption using noon reports. Biofouling imposes increased resistance on oceangoing vessels over time, and hull cleanings are subsequently performed to remove marine growth and reduce resistance. With uncertain data, a classification model is proposed to identify hull cleaning dates. The hull cleaning dates classified by the proposed model outperform the company-reported dates in terms of fitting expectations. A model to economically optimize hull cleaning intervals is further defined, achieving savings of 0.3 – 1.4 % over a three-year period, by applying optimal intervals to vessels with two hull cleanings. Adding an additional hull cleaning resulted in fuel savings of 2.1 – 3.2 %. Although results are found under strict assumptions, they are similar to savings made by advanced continuous monitoring systems. Individual crew members are analyzed to find whether certain crew over- or underperforms in terms of fuel expenditure. Findings suggest that several masters and chief engineers have significant deviations in mean consumption even after controlling for all known covariates, although the causality of deviations remains unexplained. Using fixed effects regression models, the impact of hull cleaning on fuel consumption is estimated to be approximately 1 % for Panamax and Medium Range vessels, and 9 % for Suezmax vessels. Crew members are estimated to explain between 3 – 4 % of variation in fuel consumption. Several machine learning models were tested to measure effects on prediction accuracy. Linear models achieved prediction accuracies of 63.5 – 67.5 %, increasing by 3 – 8 %, while advanced non-linear models achieved prediction accuracies of 77.1 – 78.2 %, increasing by 2.5 – 3 %. The thesis’ findings contributes to existing literature by quantifying the impact of underwater hull cleaning and crew performance on bunker consumption under data uncertainty, by providing a model to identify hull cleanings and to observe potential savings of optimizing intervals.