由於瞬間轉速變化大的機械設備所產生的信號模式大部分為非穩態信號，其 特徵信號會隨分析時間長度而平均化，無法突顯信號特徵，導致在故障診斷或辨 識上之困難，本文提出小波時頻階次譜方法，透過結合小波轉換（Wavelet Transform，WT)與轉速頻率階次化，其階次特徵不因轉速變化而改變，可有效 作為機械設備在非穩態狀態下之故障辨識。此外，經由主成分分析法（Principal Components Analysis，PCA)，提取小波時頻階次譜之主要特徵，進行資料量降維； 並結合自組織映射圖（Self Organizing Maps，SOM)，以作為非穩態狀態下之智慧 故障診斷之方法。本文以齒輪-轉子實驗平台’驗證小波時頻階次譜方法之可行性。 嚴遂涔：非穩態信號，小波轉換，主成分分析，自組織映射圖，小波時頻階次譜 方法

The signal patterns generated in instantaneously fluctuating speed by the mechanical equipment are largely non-stationary. The features of the signal patterns are averaged in correspondence with the length of analysis time, thus making it impossible to highlight the signal characteristics, which results in raising the difficulties from identifying or diagnosing malfunctions. In this paper, a wavelet order spectrum method using combination of wavelet transform (WT) and speed frequency ordering is proposed to make sure that the feature of order does not change with speed variations. In this way, identification of non-stationary malfunction in mechanical equipment can be effectively improved. In addition, Principal Components Analysis (PCA) is used to extract the main features of the wavelet order spectrum and reduce the volume of data. This is combined with Self Organizing Maps (SOM) to devise an artificial intelligence method for malfunction diagnosis in non-stationary states. Lastly, the wavelet order spectrum method is verified by using a gear-rotor test platform that proofs the feasibility of the theory.