Research on Failure Modes Converting Method and Identification Method of Autonomous Navigation Equipments Based on Knowledge Graphs

To solve the problem of insufficient real data of autonomous navigation equipment, this paper extracts the failure causes of equipment by constructing a knowledge graph and converting human errors into equipment errors. Firstly, this paper searches relevant websites to obtain the collision accident investigation reports as source data. Web crawlers are applied to get the report that compliance with keywords is efficient. After conversion and filtering operations, the target files are filtered out and they are then transformed into structured data. Text annotation is introduced to process data for entity extraction and relation extraction, and to get an exact data set. A knowledge graph in terms of failure modes of normal vessels is established based on the extracting results, which serves as a knowledge base to support the subsequent work. This step takes failure cases in plain text form and organizes them into a structured, visual form that can be easily retrieved to summarize the main human errors of conventional ships’ collisions. A convert model is created to convert human errors to device errors of autonomous vessels. Typical failure modes identification of the autonomous navigation equipment is performed through FMEA to clarify the critical components and their potential failure modes.