Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (13): 139-148.DOI: 10.3778/j.issn.1002-8331.2203-0398

• Pattern Recognition and Artificial Intelligence • Previous Articles     Next Articles

Research on Construction of Event Logic Knowledge Graph of Robot Fault Diagnosis

DENG Jianfeng, WANG Tao, CHENG Lianglun   

  1. 1.School of Automation, Guangdong University of Technology, Guangzhou 510006, China
    2.School of Computer Science and Technology, Guangdong University of Technology, Guangzhou 510006, China
  • Online:2023-07-01 Published:2023-07-01

机器人故障诊断事理逻辑知识图谱构建研究

邓健峰,王涛,程良伦   

  1. 1.广东工业大学 自动化学院,广州 510006
    2.广东工业大学 计算机学院,广州 510006

Abstract: Knowledge graph technology has important guiding significance for efficient and orderly fault diagnosis of robot system. In view of the coarse knowledge conceptual granularity of fault diagnosis ontology, entity recognition model has the problem of inaccurate feature extraction. A top-down construction method of fault diagnosis event logic knowledge graph is proposed. Firstly, the fault diagnosis event knowledge is modeled, and the fine-grained event logic knowledge ontology model is constructed. Secondly, an attention-based stacked BiLSTM with improved capsule network for event argument entity recognition is proposed. This model generates character features through BERT pre-trained model, uses stacked BiLSTM to obtain deep context features. Combined with the attention mechanism of the key features of the event argument entity, the key information of the entity is highlighted in the forward and backward context. An improved capsule network is proposed to encode the character position features, which increases the model’s attention to the character position features. Experimental results show that the proposed event argument entity recognition model can improve the result of entity recognition. Further, according to the sentence pattern matching, the argument entity matching relation is completed, and an event logic knowledge graph of robot system fault diagnosis is constructed to provide knowledge support for autonomous fault diagnosis.

Key words: event logic knowledge graph, fault diagnosis ontology, event argument knowledge extraction, stacked BiLSTM, attention mechanism, improved capsule network

摘要: 知识图谱技术对机器人系统高效有序的故障诊断具有重要的指导意义。针对故障诊断本体知识概念粒度较粗,实体识别模型存在特征提取不够准确的问题。提出了一种自顶向下的故障诊断事理逻辑知识图谱构建方法。对故障诊断事件知识建模,构建细粒度事理逻辑知识本体模型。提出基于注意力机制的堆叠BiLSTM和改进胶囊网络的事件论元实体识别模型。通过BERT预训练模型生成字符特征,利用堆叠BiLSTM获取深层上下文特征;结合事件论元实体关键特征注意力机制,对前向和后向上下文突出实体关键信息;提出改进胶囊网络对字符位置特征进行编码,提高模型对字符位置特征的关注。实验结果表明,提出的事件论元实体识别模型能提高实体识别效果。进一步,根据句式匹配完成论元实体匹配关系,构建机器人系统故障诊断事理逻辑知识图谱,为自主故障诊断提供知识支持。

关键词: 事理逻辑知识图谱, 故障诊断本体, 事件论元知识提取, 堆叠BiLSTM, 注意力机制, 改进胶囊网络