Wavelet processing-based intelligent fault diagnosis of spacial photoelectrical encoders

A fault diagnosis method is put forward to localize the faults of photoelectrical encoders used in space, in which the monitored signals obtained by remote measuring is analyzed using wavelet transform. After the demonizing, the main feature of the signal used as the training sample of neural network is obtained by wavelet decomposition. Then the network was trained as a fault classifier using the recursive localized least square method (LRLS). Practical space optical communication in ground development stage shows that this method can classify fault correctly, and is also sensitive to faults never trained before.