The invention relates to a method for diagnosing a bearing fault, which comprises the following steps of: (1) acquiring a
vibration acceleration signal of a bearing by using an acceleration
transducer; (2) acquiring a form spectrum of the
vibration acceleration signal by adopting a morphologic closing computation mode; (3) determining parameters of a morphologic
structural unit according to the form spectrum in the step (2), and extracting a fault
impact sequence; and (4) observing whether an obvious peak is present at a fault characteristic frequency on the form spectrum in the step (2), or observing whether the extracted
impact sequence in the step (3) has obvious periodicity and corresponds to the corresponding fault characteristic frequency to judge whether a rotating
machine has a fault. By the method, signals are not required to be filtered, and a window effect is avoided. The method is simple and has better adaptability, and the parameters are not needed to be determined in advance.
Impact pulses generated due to the bearing fault can be directly extracted, other unrelated information is eliminated, and the method has obvious physical significance. The structure is simple, the computing speed is high, and real-time monitoring can be realized.