A data-driven method combining knowledge graph with deep learning for constructing machining process knowledge base
Liping Ma,Xitian Tian,Lijiang Huang,Fan Yang,Xiaolin Shi
DOI: https://doi.org/10.1177/09544054241272920
2024-10-05
Proceedings of the Institution of Mechanical Engineers Part B Journal of Engineering Manufacture
Abstract:Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, Ahead of Print. Under the paradigm of intelligent manufacturing, web-based intelligent CAPP will become the trend of machining process planning. Machining process knowledge base is the foundation for implementing web-based intelligent CAPP. Massive historical machining process planning data accumulated in different sources become the main carrier of machining process knowledge. Hence, web-based intelligent CAPP puts forward the demand for data-driven construction of machining process knowledge base. Heterogeneous knowledge and unstructured data pose challenges for data-driven construction of machining process knowledge base. This article proposes a data-driven method combining knowledge graph with deep learning for constructing machining process knowledge base to provide more effective support for web-based intelligent CAPP. Firstly, a unified ontology model of machining process knowledge which integrates multi-domain machining process knowledge at the fine-grained semantic level is proposed, providing a unified semantic structure for the acquisition and organization of machining process knowledge based on distributed and unstructured process data. Secondly, a deep learning model consisting of Bert, Bi-LSTM, and CRF models deployed within the federated learning framework is proposed to automatically extract structured machining process knowledge from unstructured process texts with improved performance. Eventually, the prototype system for machining process knowledge graph construction can verify the effectiveness of the proposed method.
engineering, mechanical, manufacturing