將旋轉機械振動信號頻譜特徵，以隸屬度函數(membership function)模糊化後，訓練模糊類神經網路(fuzzy neural network)，用來評量故障的嚴重程度。當故障與信號關係可以解耦時，模糊類神經網路據此分解成獨立子網路，分立使用時，故障程度評量失統一的標準，因此，本文提出一種子網路共用基準的故障嚴重程度的評量方法，並以三個故障馬達的實例，使用類神經網路(neural network)、模糊類神經網路，及以本文所提方法進行診斷，以相互比較。

The signals of frequency symptom fuzzed by membership function and diagnosed the order of severity of faults by suing fuzzy neural network. When faults and signal symptoms can decouple, the structure of fuzzy neural network decomposed to several independence sub-net-works and the degree of faults evaluation lose uniform standard by computing separately. This study proposed a method that using independence sub-networks could evaluate the order of severity of faults in uniform standard in three case situations. Furthermore, in these cases analyses, this method is compared with neural network and fuzzy neural network.