The research of digital twin models for CNC spindle systems based on deep residual shrinkage network

Abstract This paper studies the algorithm implemented for digital twin model training for CNC spindle systems. Firstly, the framework of digital twin models for reliability analysis of CNC machine tool spindle is built, which includes a physical perception layer, an information transmission layer, and an information application layer. The built procedures of digital twins consist of five steps: reliability analysis, parameters collection, data transmission and processing, digital twin model training and evaluation, and model optimization. Then, the training algorithm based on a deep residual shrinkage network is studied, which provides an average accuracy of 98.17% and nice performance under different signal-to-noise ratios. The feasibility and effectiveness of the training method have been verified, and the rationality of the digital twin model with a 3-layer structure has been proved. This paper provides an important research basis for further research of digital twin models for complex electromechanical equipment, and optimization directions for training algorithms.