Abstract: In recent years,knowledge-based reasoning technology has been widely used in many fields,but the research in the field of CNC machine tool equipment maintenance is relatively scarce.From the perspective of knowledge reasoning,combined with the scattered and incomplete CNC equipment maintenance data,a new knowledge graph completion method based on attention mechanism was proposed,which coupled with CNN and BiGRU that named ConvBiGRU modules.ConvBiGRU module mainly encoded multiple inference paths between entities as low-dimensional embedding,and used the attention mechanism to capture the semantic correlation between candidate relationship and each path between two entities.The Embedding method was adopted to realize the multi-step knowledge question and answer in the field of CNC machine tool equipment maintenance.ComplEx was used to embed the knowledge graph of CNC machine tool equipment maintenance,while RoBERTa model was used to embed the user problems.The superiority of the proposed method in the field of CNC machine tool had been verified from many ways.

Key words: CNC machine tools, knowledge graph completion, attention mechanism, knowledge embedding, knowledge question answering

摘要： 近年来,基于知识推理技术在很多领域得到了广泛应用,但是面向数控机床设备维护领域的研究比较匮乏。研究从知识推理的角度出发,结合数控设备维护数据零散、不完整的特点,提出了新的知识图谱补全方法。该方法基于Attention机制并与ConvBiGRU模块进行耦合。其中ConvBiGRU模块主要是将实体之间的多个推理路径编码为低维嵌入,而采用Attention机制来捕获候选关系与两个实体之间的每个路径之间的语义相关性。研究采用Embedding的方法实现数控机床设备维护领域的多跳知识问答,针对数控机床设备维护知识图谱采用ComplEx进行嵌入,而对于用户问题则采用RoBERTa模型进行嵌入处理。经多角度证明了该方法在数控机床领域具有优越性。

关键词: 数控机床, 知识图谱补全, 注意力机制, 知识嵌入, 知识问答