HMMs Based Faults Diagnosis for Rotor-Gear-Bearing Transmission System

This paper proposes a new fault diagnosis scheme based on continuous density Hidden Markov Model (HMM) for vibration signals. Features extracted from vibration signals of rotor-gear-bearing transmission system are used to train HMMs to represent various running conditions. The feature vectors based on the node energies of wavelet packet decomposition are extracted from the vibration signals. Faults can be identified by selecting the HMM with the highest probability. The proposed method was tested by measuring the data of rotor-gear-bearing transmission system and has been demonstrated to be accurate and feasible.