This paper proposes a sequential diagnosis method for rotating machinery using the fuzzy neural network called Partially-Linearized Neural Network (PLNN) and relative ratio symptom parameters (RRSPs) in frequency domain, by which the fault types of rotating machinery can be precisely diagnosed on the basis of probability distribution of the RRSPs. The RRSPs in frequency domain are defined for reflecting the features of time signals measured in each state of rotating machinery. Sensitive evaluation method for selecting good symptom parameters using principal component analysis (PCA) is also proposed for detecting and distinguishing faults in rotating machinery. The practical examples of fault diagnosis for misalignment state (M),　unbalance state (UN) and looseness state (L) of a rotating shaft verify the effectiveness of the method proposed in this paper.