Digital twin system of thermal error control for a large-size gear profile grinder enabled by gated recurrent unit

Large-size gear profile grinder is extremely important for high-accuracy machining of large-scale and high-performance gears. A lot of heat is generated during the service process, leading to thermal errors of large-size gear profile grinder. In recent years, the data-based modeling and prediction methods have been widely used in manufacturing systems to control thermal errors. However, a limited bandwidth and latency characteristic of industrial Internet has brought serious challenges to the real-time and efficient processing of a large-volume data. To solve the above problems, a digital twin system of thermal error control is proposed for large-size gear profile grinders. The theoretical-based modeling method is used to prove the memory behavior, and then the feasibility of applying gated recurrent unit (GRU) to thermal error control is demonstrated due to its strong long-term memorizing capability. Then the hyper-parameters of GRU model is optimized by an improved bat algorithm (IBA), and the self-learning IBA-GRU error model is proposed. Finally, a digital twin system of thermal error control is established based on a new haze-cloud computing architecture to improve the executing efficiency, and the self-learning IBA-GRU error control model is embedded into the digital twin system. With the implementation of the digital twin system, the reduced percentages for the maximum tooth profile tilt deviations of $$f_{H\alpha l}$$ and $$f_{H\alpha r}$$ for the left and right tooth flanks are 58.43% and 64.16%, respectively, and the reduced percentages for the maximum total tooth profile deviations of $$F_{\alpha l}$$ and $$F_{\alpha r}$$ for the left and right tooth flanks are 28.78% and 34.53%, respectively. The saved ratio of the transferred data is up to 76.30% in 5 months. The executing efficiency of the digital twin system with IBA-GRU as control model is higher than that with other models and architectures.