A digital twin-driven dynamic prediction method for robotized production line operational performance

With the widespread use of robotized production lines (RPL) in manufacturing plants, timely and accurate prediction of production line operation performance is very important, which guarantees reasonable adjustment of production planning and resource allocation. The current prediction methods of production line operation performance are mostly off-line prediction, which cannot adjust the prediction model timely according to the dynamic changing characteristics of the manufacturing process, and have the problems of functional lag and low accuracy. In order to address the above issues, a digital twin-driven dynamic prediction method for the operational performance of RPL is proposed. Firstly, the manufacturing process description model and operational performance evaluation indicator are constructed based on multi-source manufacturing data from products, industrial robot equipment, and system operation dimensions. Exploring the time-varying mechanism of the manufacturing system, the digital twin model of RPL is constructed from multi-dimension. Then the correlation between manufacturing data and operational performance is explored using grey relational analysis (GRA). The online extreme learning machine algorithm is proposed to predict the dynamic performance of RPL, which is self-correcting and self-optimizing to realize the dynamic prediction based on the digital twin. Finally, a case study is implemented to verify the feasibility and effectiveness of the proposed method.