摘 要
在本研究中，提出振動模態之平均頻譜峰度演算法，可有效應用於軸承損壞診斷分析。首先以小波包絡函數擷取軸承振動模態，再藉由平均頻譜峰度分析演算法，最後可獲致頻譜峰度視窗長度對頻譜峰度值之二維圖。由理論分析可發現，對正常軸承之頻譜峰度值小於3，近似隨機訊號之峰度值；然而，對損壞軸承之頻譜峰度值則大於3，其二維圖呈現具山峰形狀之尖峰現象，其山峰形狀之中心值，可轉換對應其軸承損壞型式之特徵頻率，藉此可用以確認損壞軸承元件。
關鍵詞：軸承損壞、小波包絡函數、頻譜峰度、模態振動
Abstract
This paper proposes that a mean spectral kurtosis analysis algorithm can be effectively applied to mode vibration for bearing defect diagnosis. First, the wavelet enveloping function is used to derive a mode vibration of bearing. Then, the mean spectral kurtosis analysis algorithm is applied to the mode vibration. Finally, the two dimensional diagram of window length vs. mean spectral kurtosis could be derived. The theoretical study shows that for normal bearing the mean spectral kurtosis is lower than 3 in the diagram, which is similar to the kurtosis of a random signal. On the other hand, for defect bearing the mean spectral kurtosis is higher than 3 in the diagram, which takes the shape of a pointed mountain peak and corresponds to the characteristics of bearing defect. Accordingly, the mean spectral kurtosis analysis algorithm could also identify the defect type of a bearing system.
Keywords: Bearing Defect, Wavelet Enveloping Function, Spectral Kurtosis, Mode Vibration