Fault knowledge graph construction for coal mine fully mechanized mining equipment

The existing fault diagnosis methods for coal mine fully mechanized mining equipment lack systematic management and application of historical fault data of fully mechanized mining equipment. In response to this problem, knowledge graph technology is introduced to systematically manage the fault data of fully mechanized mining equipment. The top-down approach is used to construct the ontology of fully mechanized mining equipment fault knowledge. The knowledge of fully mechanized mining equipment fault is classified into four categories: fault location, fault phenomenon, fault cause, and treatment method. And the naming of the knowledge is standardized. The universal naming entity annotation method BIOES is used to manually annotate the fault knowledge of fully mechanized mining equipment. By combining bi-directional long short-term memory (BiLSTM) and conditional random field (CRF), the BiLSTM-CRF model is constructed. The marked fault knowledge of fully mechanized mining equipment is identified by the named entity, and the fault knowledge extraction is realized by manually extracting entity relationships. Combining the entity recognition results of the BiLSTM-CRF model with the manually extracted entity relationships, a Neo4j graph database is used to store the fault knowledge of fully mechanized mining equipment. A fault knowledge graph of fully mechanized mining equipment is constructed. The experimental results show that compared to the BiLSTM model and BiLSTM-Attention model, the acurracy of the BiLSTM-CRF model is significantly improved, reaching 87%. The F₁ value also has a certain increase, reaching 69%. The construction of fully mechanized mining equipment fault knowledge graph can provide support for the effective analysis, management, and application of large-scale and multi-domain fully mechanized mining equipment fault data.