Fault diagnosis for gear pump based on feature fusion of vibration signal

Information fusion arises in a surprising number of fault diagnosis applications. In this paper, common faults are designed in the experiment according to the gear pump vibration mechanism. Fault signal is collected from vibration sensors of different positions, and wavelet packet energy percentage and RMS are extracted as features of the signal. RBF neural network is adopted to fuse thiese features which are used to learn and train the network. The testing results prove that this approach possesses higher diagnostic precision and better diagnostic effect than single signal fault diagnosis.