System Design Of Swashplate Rolling Bearing Fault Diagnosis

Helicopters have a very important role in the defense and civilian fields, swashplate rolling bearing is one of the important parts of the helicopter, and its health status impact on the of the helicopter’s safety badly. The Fault Diagnosis of Rolling for the helicopter Swashplate carried great significance. This task was entrusted by some aviation units, Swashplate rolling bearing fault diagnosis system was developed which based on vibration signal analysis, and fault diagnosis method was studied. The main work and contents for the research are as follows:(1) The Swashplate Bearing Fault Diagnosis System was developed. Firstly, the user needs was analyzed according to the client’s needs which contained system requirements, functional requirements and technical indicators. Secondly, the general design of the system was achieved which includes structural design and interface design which were on the basis of the demand analysis; Finally, The features of the software modules implemented which based on the detailed design of software.The system includes preprocessing, feature extraction and fault diagnosis modules. The software preprocessing module includes time averaging, wavelet packet filtering and morphological filtering de-noising method; feature extraction module includes time domain feature, et.al, frequency domain feature, LMD feature and EMD feature, et.al; fault diagnosis module includes BP RBF neural network and support vector machine classification method. The software system has the features of flexible, open, and meets the needs of the principal of the proposed software system.(2) Swashplate rolling element bearing of fault diagnosis method was studied. For the problem of Local Mean Decomposition(LMD) was easily affected by noise interference when in the extraction of fault features, a rolling bearing fault diagnosis method which based on LMD and Independent Component Analysis(ICA) was proposed. Firstly, original signal was decomposed into a series of production functions(PF) by the LMD method. Secondly, the estimate of PF was obtained after the PF components had been separated by ICA method,and the noise was effectively removed. Then, mutual information, correlation coefficient and approximate entropy which were extracted from the estimate of PF components were grouped together as a feature vector. Finally, the fault feature vectors were classified by SVM. The results of the feature extraction and fault diagnosis experiments show that the fault recognition rate of LMD-ICA method is significantly better than the traditional LMD method.