Combining deep learning with knowledge graph for macro process planning
Yajun Zhang,Shusheng Zhang,Rui Huang,Bo Huang,Jiachen Liang,Hang Zhang,Zheng Wang
DOI: https://doi.org/10.1016/j.compind.2022.103668
IF: 10
2022-09-01
Computers in Industry
Abstract:The macro process, as the core process content of the whole machining process for a design part, serves as a guide for the complete machining process planning. It determines the final machining quality and machining cost of the part to a large extent. However, due to the various bottlenecks in machining process knowledge representation, matching, and inference, macro process planning still depends heavily on the knowledge and experience of process designers. In this paper, a macro process decision-making approach combining knowledge graph and deep learning technology is proposed. Firstly, based on deep learning, the macro process reasoning function is learned to model the mapping relationship between the process elements of the design part and macro process analysis rules. Then the process elements of the design part are fed into the reasoning function to activate the applicable process analysis rules of the process knowledge graph. Next, according to the association relationship between the activated process analysis rules and the predefined machining methods of the macro process graph, a series of working steps nodes and the directed edges are activated, which constitutes the feasible solution space for the macro process. Finally, the swarm intelligence algorithm is applied to search for an effective and low-cost macro process scheme from the feasible macro process solution space. In experimental studies, the slot cavity parts are taken as examples to verify the feasibility and effectiveness of the proposed approach.
computer science, interdisciplinary applications