- 學位論文
小波轉換於轉子故障與切削顫振偵測之應用
The application of wavelet transform on diagnosis of rotor failure and milling chatter
摘要
轉子故障與切削顫振發生時，轉子與刀具振動特性皆會改變，本研究透過小波轉換與傅利葉轉換兩種訊號處理方法來擷取振動訊號之特徵值，根據待判別訊號與轉子故障或切削顫振之特徵值的接近程度，即可決定故障種類與發生機率。 電腦輔助轉子故障診斷系統包含訊號收集、訊號處理、故障診斷。訊號收集乃以位移計和加速規配合Hp3566，蒐集轉子的運轉振動訊號。本研究收集3種故障訊號一外部衝擊力，異常摩擦，和不平衡轉子等引起之故障。訊號處理是將資料經小波轉換萃取特徵值：平均值，最大值，均方根，峰值比(最大值與均方根值之比)，傅利葉轉換後，萃取特徵值：PSD(Power Spectral Density) 之最大峰值，特徵頻率(最大峰值所在之頻率)作為轉子故障線上診斷之依據。 小波轉換本質上是時域分析工具且具有區域性，可察覺微小的動態異常訊號；傅利葉轉換本質上是頻域分析工具且具有全域性，可觀察整體的穩態頻率變化。由於摩擦訊號屬於穩態訊號，其在頻域響應上和正常訊號有明顯不同；衝擊訊號與不平衡轉子所引起之訊號是屬於動態訊號，適合以小波轉換作訊號處理。因此，判斷上將結合小波轉換與傅利葉轉換作訊號處理之工具。 故障診斷方法主要是把多次實驗之訊號處理所得特徵值之平均值作為參考點，依大小排列參考點分佈圖後，依未知訊號特徵值在參考點分佈圖中的位置，計算未知訊號特徵值與已知訊號特徵值參考點的接近度，依接近度大小判斷訊號發生故障種類的機率。 顫振之偵測方法大致與轉子故障之偵測方法相同，本研究收集CNC工具機作端銑時不同加工條件的切削振動訊號來分析。
並列摘要
In order to avoid the loss or damage caused by the failure of the rotor system and milling chatter, it is necessary to have a efficient and reliable diagnosis system that can on-line detect the occurrence of failures and chatter. The methodology of diagnosis system development proposed in this study is to process and analysis the vibration signals of the rotating rotor and machine center measured by capacitance probes and accelerometers. The diagnosis scheme is developed for detecting three major failure causes: abnormal friction, imbalance of rotor, and external impact force. The diagnosis method is first collecting the normal signal and failure signals, and converting the signal through wavelet transform and Fast Fourier Transform (FFT). Because the converted failure signals exhibit different characteristic value （ratio of peak value to root mean square value of signal）, a distribution of these characteristic values is used to form a diagnosis map that can reflects the possibility of occurrence of failure. When a diagnosis is performing, the characteristic value of detected and converted failure signal is contrasted with the diagnosis map. Finally, the possibility of occurrence for failure mode can be determined by approach degree of function, which is induced based on membership function of fuzzy control theory. Similar algorithm is applied to the diagnosis scheme for milling chatter. Since milling chatter exhibits different dynamic characteristics from failure signals of rotor system, real vibration signals were collected and analyzed to determine proper characteristic value diagnosis. Experiments were conduced to verify the proposed diagnosis theory.