Chen-Yao WangPo-Han ChenShang-Yu LinPEI-ZEN CHANGYuh-Chung Hu2025-01-212025-01-21202424058971https://www.scopus.com/record/display.uri?eid=2-s2.0-85213005252&origin=resultslisthttps://scholars.lib.ntu.edu.tw/handle/123456789/724985This work introduces a non-destructive, wireless sensor system for real-time monitoring of cutting temperatures to predict thermal errors in tools during dry cutting, reducing environmental pollution compared to coolant-based methods. Specifically, the system employs a thermocouple sensor positioned between the disposable insert and the tool shank, secured by the clamping force of the screw, and transmits the temperature data through Bluetooth to achieve real-time measurement. The volume of the cutting slot per tooth is the main factor influencing cutting temperature, and the cutting temperature directly affects the magnitude of tool center point (TCP) thermal error. Therefore, this sensor system offers a solution to the thermal error issue by collecting temperature and thermal deformation data to train a model capable of predicting thermal errors in the cutting tool. This model comprises temperature and thermal error models, both utilizing four predictive models-Random Forest (RF), Gradient Boosting (GB), Extra Trees (ET), and eXtreme Gradient Boosting (XGB)-to identify the most accurate thermal error model. The results show that the ET model performs best for temperature prediction, with a root mean square error (RMSE) of only 1.05°C, while the RF model achieves the lowest error for thermal error prediction, with an RMSE of 1.07 µm. Consequently, this sensor system provides a method for manufacturers to predict thermal errors in real-time during the machining process.truecutting tool thermal errorembedded thermocouplemachine learningmilling temperaturewireless temperature measurementjournal article10.1016/j.procir.2024.10.3302-s2.0-85213005252