A CNN-BiGRU-Attention Model for Temperature Forecasting in Motorized Spindles under Unbalanced States

Introduction: In the process of manufacturing and installation of a high-speed motorized spindle, due to the uneven mass distribution of the rotor system, the dynamic balance often fails. In order to study the influence of unbalanced mass on the temperature of motorized spindle during operation, the spindle was changed into an unbalanced state by adjusting the counterweight block on the dynamic balance ring. The temperature of front bearing and rear bearing was tested, and the relationship between unbalanced mass and temperature was analyzed. Then, based on the experimental data, a CNN-BiGRU-Attention temperature prediction model was proposed and compared with the CNN-BiGRU model. The results show that the prediction errors are significantly reduced, which significantly improves the prediction performance and verifies the accuracy of the model in the temperature monitoring of motorized spindle. The proposed CNN-BiGRU-Attention model and its optimization strategy offer potential for patent application due to their novelty and effectiveness in enhancing the thermal monitoring and early warning capabilities of high-speed motorized spindles. Uneven mass distribution in high-speed motorized spindles causes dynamic imbalance, leading to abnormal vibration and a rise in temperature that compromises machining accuracy. This study aims to design a temperature prediction model for motorized spindles under unbalanced states, capable of accurately forecasting temperature trends and providing early warnings to prevent thermal faults. Methods: A CNN-BiGRU-Attention model integrates spatial feature extraction, bidirectional temporal learning, and attention-based feature weighting. It was trained on temperature data from front and rear bearings under four unbalanced mass conditions. Results: The proposed model achieves a superior prediction performance compared to the CNNBiGRU baseline. It effectively captures temperature fluctuations and steady-state behavior under various unbalanced states, demonstrating strong adaptability and prediction accuracy. Discussion: The proposed model effectively captures complex thermal dynamics through its multicomponent architecture, demonstrating strong potential for predictive maintenance in highprecision manufacturing. conclusion: It shows that the established CNN-BiGRU-Attention temperature prediction model can be used to predict the temperature trend of high-speed motorized spindle. Conclusion: The CNN-BiGRU-Attention model provides a reliable and efficient solution for temperature prediction in unbalanced motorized spindles. This method offers a valuable reference and a practical tool for the intelligent monitoring and thermal safety management of high-speed precision equipment, contributing significantly to the advancement of predictive maintenance in smart manufacturing.