Helicopter Main Gearbox Fault Diagnosis And Failure Prediction Technology Research

Main gearbox (MGB) is a key component of helicopter, its condition effects the reliability and safety of helicopter. The technology of prognostics and diagnostics can diagnose and predict the fault of MGB, which make its maintenance from afterwards and prevent (or hard time) to on-condition and condition monitoring.According to the research condition of mechanical failure at home and abroad, this paper investigates the fault technologies of diagnostics and prognostics for MGB, puts forward a diagnostics system that compose of discrete wavelet transform(DWT), neural network and prognostics system that is a combination of DWT, Kalman filter and neural network. Based on these two systems, an integrated system of fault diagnostics and failure prognostics for MGB with DWT, Kalman filter and neural network is proposed. The systems are verified by eight signals that collected by sensors, and the follow conclusions are got.(1) Based on Shannon theory and Parseval's theorem, the vibration signal here decomposed into nine layers, the minimum entropy is got when wavelet bases is "db45";(2) In the four neural networks (BP, Elman, RBF and GRNN), the training time of GRNN is shorter and its diagnostics accuracy is highest. The training time of improved wavelet neural network (WNN) whose hide layer is Gaussian wavelet is shorter than that of Morlet, but their diagnostics accuracy is almost same.(3) The improved prognostics algorithm of Kalman filter can better forecast eigenvectors of MGB, and the approach accuracies of every sensor prediction values are high.(4) To the system that real-time is high, we suggest to use DWT, Kalman filter and GRNN to construct fault diagnostics and failure prognostics system. However, to the one that requirement of real-time is not strict, but high diagnostics accuracy, the neural network can choose the one whose hidden layer is Gaussian function.The results of the present work imply that the fault diagnostics and prognostics system for helicopter MGB using DWT, Kalman filter and neural network is feasible and effective, it can diagnose and predict the fault of MGB accurately, which will provide new technical references for the development of healthy and usage monitoring systems (HUMS), prognostics and health management (PHM), condition-based maintenance (CBM) and CBM+.