基於關鍵特徵之微銑削加工精度預測

• Main Authors: Lu-Wen Kung, 龔呂文
• Other Authors: Haw-Ching Yang
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/88445864329892096870

碩士 === 國立高雄第一科技大學 === 電機工程研究所碩士班 === 104 === Accuracy of a machined workpiece is affected by variations of factors including workpiece material, tool characteristic, cutting parameter, and tool status. Although the variations could be evaluated by sensing signals, detecting indistinct variations of the signals with low material removal rate is a challenge in micro-milling process. Therefore, a key for predicting accuracy is how to efficiently extract machining sensing features in machining. This research improves the developed ISU (Intelligent Sensing Unit) to monitor a micro-milling process. The improved ISU not only can timely collect current signals from motors and vibration signals from a spindle but also can enhance automatic segmentation of signals to increase de-nosing capability. Also, the EKI (Expert Knowledge Improved) feature set for micro-milling process is proposed, such as 1/2 harmonic power, skewness, and wavelet packet power, for serving as inputs of the AVM (Automatic Virtual Metrology) system to predict finish accuracy. In addition, an APP is developed to integrate prediction results of the AVM system to monitor on-line machining quality. The experimental results of machining thin cylinder shells (SUS303) with diameter 4.5mm and depth 2mm indicate that the signal to noise ratio can be improved from 11dB (original) to 15.63 dB (de-noised by using mother wavelet db3 with 5 levels decomposition). When adopting the EKI features set and applying one from five sampling policy, the mean absolute percentage errors of predicting inner roundness, inner cylinder, and outer roundness are 8.65 %, 13.2 %, and 6.5 %, respectively. Hence, the improved ISU is a promising tool for predicting accuracy of a micro-milling process.