本篇的主要目的在於使用後處理總體經驗模態分解法來改善之前總體經驗模態分解法的缺點，即模態混雜與端點效應。首先將旋轉機械之振動訊號透過後處理總體經驗模態分解法分解成一組數個無模態混雜且符合內稟模態函數基本條件的內稟模態函數，並選擇其中含有故障特徵的內稟模態函數進行分析。 軸承故障特徵的提取則是將內稟模態函數的包絡線經過希爾伯特轉換後繪出時頻譜，最後計算出希爾伯特邊際譜，觀察是否含有軸承故障特徵頻率以達到故障檢測目的，本篇除了檢測軸承故障的種類外，並將進行軸承故障程度上的辨別。 In order to improve the drawbacks of Ensemble Empirical Mode Decomposition (EEMD), such as mode mixing and end effect problem, we present an improved HHT approach based on post-processing of EEMD to solve the problem in this paper. Once the Intrinsic Mode Functions (IMFs) are obtained from the decomposition process, the crucial step is to extract the fault features from the information-contained IMFs. The amplitude modulation (AM) phenomenon can be discovered in the IMFs with fault information. In this paper, we not only classify the types of bearing fault but also identify the level of the fault.
