| Drive system is one of the important components in helicopter and it is also easy toget abnormal, whose work status will directly determine the performance and flightsafety of the helicopter. Thus the researching of the fault diagnosis approach for thedriving system is practical significant to protect the flight safety of the helicopter. In thispaper, the in-depth research is conducted on the Fault Diagnosis Method for HelicopterTransmission, which is sponsored by the Aviation Science Fund Project and theHorizontal Subject of a Helicopter Research and Development Institute. The main workand achievements are as follows:(1) A fault diagnosis approach for the driving system based on wavelet packetde-noising and local mean decomposition (LMD) is proposed. LMD method is a newadaptive time-frequency analysis method. However, LMD method is sensitive to noise.In order to eliminate noise influence the result of diagnosis, firstly, wavelet packet isused to remove noise from the signal. Then, that result is decomposed by LMD, and thecorrelation coefficient between the PF and the signal is used as the standard of judgment,so that the redundant low-frequency PF can be rejected. Finally, the effective PF isselected to analyze the power spectrum and extract the fault feature. The experiment ofthe simulation data and the actual roller bearing fault diagnosis data show that thismethod is effective.(2) In order to realize the association between the fault features and the fault typethe diagnosis method established on the basis of the combined with wavelet packet,LMD decomposition and Radial Basis Function Neural Network （RBF）is proposed.Firstly, wavelet packet is used to remove noise from the signal. Then, the energy ratio ofthe PF as the feature vector and is input to the RBF neural network. For comparison,select the BP neural network as the fault mode classifier. The experimental analysisshow that the energy ratio of the PF the wavelet de-noising is better to reflect thecharacteristics of different fault than that has not been the wavelet packet de-noising.The RBF neural network has better classification ability than the BP neural network.The experiment results show that this method can effectively classify the roller bearingfault. |
