Knowledge Graph Construction for Railway Electrical Accident Analysis

Rail system's electrical accidents happen frequently; yet currently existing accident data are presented/stored as unrelated information, which makes it difficult for us to achieve data correlation. In order to discover hidden connections between different types of railway electrical accidents and integrate these seemingly independent data into a structured body of knowledge, a semi-automated construction process is explored to build a knowledge graph for all railway electrical accidents for the past 8 years in China. The experiment results show that CNN classifier can obtian perfect classification performance, and needn't diagnose the faulty equipment of accidents artificially, which can greatly save time and effort. The knowledge graph we constructed in this paper is not only used to analyze and diagnose the faulty equipment of railway electrical accidents, but also can help us discover trends and changes of these accidents. In addition, the knowledge graph also lays a solid data foundation for the progressively intelligent railway electrical systems.