Data-driven and Knowledge-based predictive maintenance method for industrial robots for the production stability of intelligent manufacturing

The service stability of industrial robots (IRs) is considered the basis for ensuring intelligent manufacturing operations. Knowledge-based work plays a central role in the practical application of intelligent manufacturing because the staff have professional knowledge of production and manufacturing after learning, undergoing training, and thinking. They can use their knowledge to analyze complex states, assess them accurately, and make innovative decisions. Therefore, expressing and constructing IR data and knowledge is a key issue in the application of knowledge to the predictive maintenance (PdM) of IRs. Considering the intelligent management of IRs as the research objective of this study, a PdM method based on data and knowledge was developed. A running-state feature-recognition model based on a long short-term memory network was first established to recognize future running states using the history and real-time running data of IRs. Furthermore, the k-nearest neighbor algorithm was used to analyze the correlation between the running-state feature data and faults to predict possible faults. The prediction results were then input into the knowledge graphs (KGs) of IRs for reasoning. PdM strategies were automatically formulated based on the KGs. Finally, a PdM system for IRs was designed and developed. The effectiveness of the proposed method and system were verified by applying them to the welding robots in a new energy automotive welding workshop. The findings of this study provide new concepts and tools for the PdM of IRs, as well as theoretical and methodological support for the production stability of intelligent manufacturing.1

Automated summary

The service stability of industrial robots is crucial for intelligent manufacturing operations. Knowledge-based work plays a central role in intelligent manufacturing, and the expression and construction of robot data and knowledge are important for predictive maintenance (PdM). This study proposes a PdM method based on data and knowledge, which automatically formulates PdM strategies using a running-state feature-recognition model and fault prediction. The effectiveness of the method is verified through application to welding robots.