Application of Support Vector Machine and Vibration Feature Extraction for Identifying Milling Status under Different Rotating Speed

本研究主要針對工具機在不同加工參數條件下，銑削加工時透過主軸振動訊號來辨識加工狀態。透過放置於主軸上的加速規收集振動訊號並進行分析，運轉狀態分成空轉、進刀和穩定加工三類。訊號處理分析分別於時域和頻域進行，前者將原始訊號經由經驗模態分解法拆解後去除趨勢項，再透過多尺度熵和方均根等方法進行分析來提取時域特徵;後者將原始訊號經由經驗模態分解法拆解，將訊號分解成若干個固有模態函數，選取含物理現象的特徵頻帶進行傅立葉轉換結合轉速正規化並去除非轉速相關之頻率，來提取頻域特徵;再經由費雪分數找出時域訊號和頻域訊號較具區別性的訊號特徵。最後，將篩選的訊號特徵以支持向量機進行加工狀態辨識，結果顯示有相當高的準確率。This study focuses on the identification of cutting status of machine tool under different preset cutting parameters through using the vibration signal analysis. The vibration signals were collected by the accelerometer on the spindle. The identified tooling status includes the idle cutting, initial feeding and stable cutting. The signal analysis is performed in the time domain and frequency domain. In the time domain analysis, the original signal is separated by using the empirical mode decomposition (EMD) method to abandon the trend term, and then the multi-scale entropy (MSE) and root mean square (RMS) of the signal are determined as the time domain features. For the feature extraction in frequency domain, the intrinsic mode functions (IMFs) that contain the physical features in certain frequency band are analyzed through the fast Fourier transform (FFT) process. The rotation speed normalization approach is employed to remove the influence of spindle rotating speed. The factors of natural frequencies of the structural vibration in machine tool are also neglected. The features of higher priorities are selected by evaluating their fisher scores. The results show that the support vector machine (SVM) can identify the different cutting status with high accuracy.摘要 i Abstract ii 目錄 iii 圖目錄 vi 表目錄 xi 第一章 緒論 1 1-1 前言 1 1-2 文獻回顧 3 1-3 問題描述與研究動機 7 1-4 論文大綱 8 第二章 理論 9 2-1 希爾伯特-黃轉換(Hilbert-Huang Transform) 9 2-1-1 固有模態函數(Intrinsic Mode Function) 10 2-1-2 經驗模態分解法(Empirical Mode Decomposition) 10 2-2 多尺度熵 (Multi-scale Entropy) 12 2-2-1 能量熵(Entropy) 12 2-2-2 取樣熵(Sample Entropy) 13 2-2-3 多尺度熵(Multi-scale Entropy) 16 2-3 費雪分數(Fisher Score) 18 2-4 支持向量機(Support Vector Machine) 19 2-4-1 超平面(Hyper-plan) 19 2-4-2 支持向量(Support Vector) 23 2-4-3 多分類支持向量機 23 第三章 實驗架構及實驗方法 25 3-1 實驗架構與儀器 25 3-1-1 實驗加工之工具機 25 3-1-2 實驗加工之刀具 26 3-1-3 實驗加工之工件 27 3-1-4 實驗加速規之規格 28 3-1-5 DAQ NI 9234擷取卡 29 3-2 實驗規劃 31 3-2-1 加工參數配置 31 3-2-2 訊號擷取流程 32 3-3 實驗之訊號特徵提取流程 32 3-3-1 時域訊號特徵提取方法 33 3-3-2 頻域訊號特徵提取方法 36 3-3-3 非轉速相關之頻率 39 第四章 實驗結果與討論 41 4-1 時域訊號分析 41 4-2 頻域訊號分析 45 4-3 訊號特徵篩選 46 4-4 支持向量機分類 64 第五章 結論與未來展望 69 5-1 結論 69 5-2 未來展望 70 參考文獻 7