A knowledge graph-based structured representation of assembly process planning combined with deep learning

The assembly process is a significant foundation for reference in the procedure of product assembly. Aiming at the problem that assembly process design overly depends on expertise and plenty of existing assembly process information is not exchanged and reused, a knowledge graph-based structured representation of assembly process planning that combines expert experience and a large amount of existing process information is proposed. In the process of assembly process knowledge graph (APKG) implementation, the upper conceptual schema layer of APKG is constructed in a top-down manner. Named entity recognition is a critical component of constructing APKG and generates the instance data required for the APKG. In this study, a deep learning approach is applied to identify named entities of assembly process text based on existing assembly process data. With the use of BERT-CNN-BiLSTM-CRF deep neural network, the accuracy acc of the recognized entities of assembly process text is 96.48%, which is 1.39% better than the accuracy acc of the BERT-BiLSTM-CRF model. The results demonstrate that BERT-CNN-BiLSTM-CRF has excellent performance in named entity recognition of assembly process text, which can effectively identify entity boundaries and improve the accuracy of the named entity recognition task. Finally, the knowledge graph APKG is constructed to verify the effectiveness of the construction method by taking the assembly process records of a company's scanning device equipment as an example.