Systematic knowledge modeling and extraction methods for manufacturing process planning based on knowledge graph

Process knowledge is an important part of intelligent manufacturing, which has become the trend of global manufacturing industry. However, process planning in manufacturing heavily relies on experience and historical knowledge that cannot be effectively reused. Also, process knowledge is complicatedly structured, widely distributed, and weakly associated, which makes it more difficult to be shared and reused. Moreover, existing methods of organizing process knowledge are labor-intensive, inefficient and inevitable to extensive manual annotation. To overcome the above problems, a systematic method of knowledge graph construction for process knowledge is presented. First, through key concepts and relationships analysis, a domain ontology for process knowledge is proposed. Second, a pattern-based bootstrapping framework with a 2-level masking technique is established to perform process knowledge extraction, which does not require manual annotation and avoids overfitting issues. Third, a meta path-based question answering over knowledge graph approach is presented to support process planning, which can effectively capture keywords of the input question, present the intention and match the designed path. Finally, taking the production process of radiator as a research object, a corresponding manufacturing process knowledge graph is constructed and applied to process planning scenario, and experiments validate the feasibility of the proposed methods.