This paper proposed a new method called a “sequential fuzzy neural network” to diagnose gear equipment failures automatically and precisely. The symptom parameters in time domain, by which each gear equipment failure can be detected sequentially, were selected according to values calculated from the signals measured in each gear condition. To express the relationship between the gear condition and the symptom parameters, the probability density functions were translated to possibility distribution functions by possibility theory. The diagnostic process can be carried out automatically by a neural network combined with sequential fuzzy inference. Examples of practical diagnosis are shown to verify the efficiency of this method.