Tool wear model and prediction of high-temperature nickel-based alloy cutting

Abstract

During the processing of high-temperature nickel-based alloy components, the cutting tools will face serious wear problems, which seriously affect the processing efficiency and workpiece quality. To address the insufficient accuracy in predicting tool wear during the cutting process of high-temperature nickel-based alloys, an improved limit gradient boosting regression wear prediction model that integrates time series features and attention mechanism is proposed. By designing a single-factor cutting experiment, multiple signals such as cutting force and temperature are collected, and time-domain, frequency-domain, and time-frequency domain features are extracted to build high-dimensional feature vectors. Sliding windows and attention mechanisms are introduced to enhance the ability to characterize dynamic wear processes. The proposed model had a maximum accuracy of 99.2% in predicting tool wear, with an average relative error controlled within 6.67%, which was significantly better than that of traditional methods. Within the parameter range of cutting speed 40-70 m/min, feed rate 0.10-0.25 mm/rev, and cutting depth 0.3-1.3 mm, the model exhibited good generalization ability and real-time performance. The inference time was only 6.5 ms/sample, which could meet the high-precision real-time monitoring and life prediction requirements of tool status in engineering practice.