Fault Diagnosis Method Of Helicopter Gearbox Based On Vibration Signal

The gearbox is the key component of the helicopter’s transmission system.If the gearbox fails,the transmission system will fail,even causing the helicopter to deviate from the ideal operating state.At the same time,the helicopter gearbox is difficult to disassemble in the routine maintenance,if the status is unknown,it may lead to artificial breakdowns.With the growth of gearbox working time and the change of external environment,breakdowns of the equipment will inevitably occur.Therefore,in order to ensure the helicopter to fly stably and reliably.It is of great practical significance to carry out efficient and precise fault diagnosis and status monitoring for gearboxes.Based on the above situation,in order to extract the fault features and complete the diagnosis work accurately from the complex and changeable signals of the gearbox.In this paper,the gear in the gearbox is taken as the research object.Based on the acceleration vibration signal collected by the test bench,we make an intensive study of the signal processing,feature extraction and fault pattern recognition.Specific work list as follows:Firstly,The operating principles of helicopter gearboxes are analyzed,and the types of failures are examined.Based on these modes,introducing the vibration mechanism of the gearbox,including the generation of vibration,vibration transmission paths.At the same time,making a detailed analysis of the characteristics of the gear failure information.Then,establishing a connection between the fault characteristics and vibration mechanism.It transfer the key problem of helicopter gearbox fault diagnosis into a vibration signal-based fault extraction problem,which provide a direction for subsequent fault diagnosis.Secondly,the technical means of vibration signal processing is explored.The traditional wavelet transform has the disadvantages of frequency aliasing and low accuracy in signal processing.A double-tree complex wavelet transform(DTCWT)is used to process the vibration signal with a known state,obtaining the components containing fault characteristics.Based on kurtosis index and correlation coefficient,determine the key components that contain the main fault characteristics.and performing maximum entropy spectral estimation(MESE)on the components to stabilize signal characteristics and reduce noise interference.After feature extraction,the energy entropy of the decomposed signal is used to form the fault feature matrix.The matrix is used as training and testing samples of neural network model.Thirdly,a fault diagnosis model of Helicopter Gearbox Based on improved probabilistic neural network is proposed.The smoothing factor of the probabilistic neural network is the key parameter to determine the diagnosis rate of the model.it is necessary to optimize the parameters of probabilistic neural network for improving the diagnosis rate and generalization performance of the model.In order to improve the inertia weight selection problem in traditional particle swarm optimization and search for the best smoothing factor of probabilistic neural network,the linear decreasing weight particle swarm optimization algorithm is adopted to avoid the search falling into local optimum by introducing the dynamic characteristic of linear weight decline.The probabilistic neural network with optimized parameters is used to train and test the fault samples,then,the accurate diagnosis and recognition of different operating states of gearbox are realized.Finally,on this basis,a system has been developed for helicopter gearbox fault diagnosis,which is convenient to collect and manage data,process information,this system has good engineering application value.