A knowledge graph-based inspection items recommendation method for port state control inspection of LNG carriers

The safe navigation and operation of Liquified natural gas (LNG) carriers is crucial in maritime transportation. The Port State Control (PSC) inspection is the primary method for identifying deficiencies in vessels. Different from other vessels, the PSC inspection of LNG carriers requires knowledge of specialized transport equipment. Therefore, effective management of inspection knowledge and precise targeting of inspection items are particularly crucial. Knowledge graph is an efficient way to manage knowledge in the context of artificial intelligence. In this study, an LNG carrier PSC inspection knowledge graph is constructed for fusing multisource knowledge data from PSC inspection. Additionally, a knowledge graph-based recommendation model, namely the improved knowledge graph convolutional network combined with pretrained translation embedding (PT-KGCN), is developed to recommend inspection items and knowledge. The PT-KGCN model first carries out knowledge graph embedding. It then predicts possible deficiencies on the basis of historical PSC inspection data. Finally, it recommends inspection items by correlating the predicted deficiencies with knowledge from the knowledge graph. The results show that more than 87% of defective items in historical data are correctly predicted. The research findings can provide ideas for the practical application of knowledge data in the inspection fields.