Towards accident prevention on liquid hydrogen: A data-driven approach for releases prediction

Abstract

Hydrogen is a clean substitute for hydrocarbon fuels in the marine sector. Liquid hydrogen (LH2) can be used to move and store large amounts of hydrogen. This novel application needs further study to assess the potential risk and safety operation. A recent study of LH2 large-scale release tests was conducted to replicate spills of LH2 inside the ship's tank connection space and during bunkering operations. The tests were performed in a closed and outdoor facility. The LH2 spills can lead to detonation, representing a safety concern. This study analyzed the aforementioned LH2 experiments and proposed a novel application of the random forests algorithm to predict the oxygen phase change and to estimate whether the hydrogen concentration is above the lower flammability limit (LFL). The models show accurate predictions in different experimental conditions. The findings can be used to select reliable safety barriers and effective risk reduction measures in LH2 spills.