An industry-oriented digital twin model for predicting posture-dependent FRFs of industrial robots

The advantages of industrial robots over machine tools for machining large parts, such as their cost-effectiveness and expansive workspace, are well-recognized. However, robotic machining is extremely prone to vibration instability due to low stiffness characteristics. The complexity of industrial robots is their posture-dependent dynamic response. In this paper, an industry-oriented digital twin model is proposed to accurately predict the posture-dependent FRFs (Frequency response functions) of industrial robots. First, a multibody model with flexible joints is built for an articulated industrial robot. Then, the measured FRFs and the fitting algorithm are used to identify the elastic parameters of the joints. Next, the simulated FRFs obtained from the multibody model in the virtual space and the FRFs measured in physical space with a 70% reduction in sample size are used for training to obtain a mapping between the simulated and actual frequency bin functions (The set of real and imaginary parts of FRFs with different postures at a specified frequency interval.). Simulations and experiments validate the effectiveness of the proposed model. The results show that the proposed industry-oriented digital twin model is superior to the data-driven model in terms of both accuracy and interpretability, and the implementation process is consistent with the data-driven model.