Intelligent condition prognostics for helicopter main gearboxusing discrete wavelet and Kalman filter
LIU Li-Sheng , YANG Yu-hang
1. College of automation, University of Science and Technology, Nanjing 210094, China;2. Department of Reliability, Army Aviation Research Institute of General Staff, Beijing 101121, China
Abstract:Main gearbox (MGB) is a key component of helicopter transmission system, it often runs in the atrocious condition of high revolving speed and high burthen, it is very important to achieve condition prognostics for the safe of helicopter. This paper puts forward a intelligent condition prognostic system for helicopter MGB using discrete wavelet transform (DWT), Kalman filter and Elman neural network. The mother wavelet of Daubechies 44 (db44) is selected to extract feature vectors in the process of DWT. Kalman filter is used for feature vectors prognostics, and Elman neural network is taken for fault distinguish and classification. In the algorithm of Kalman filter, a new prognostic method is proposed, and it is verified in the system by experiment. The results indicate that the prognostic outcome of the Kalman filter with this method is better, which is more applicable for the prognostics of feature vectors, and the intelligent condition prognostic system composed of DWT, Kalman filter and Elman neural network is feasible, which can predict the future condition of MGB accurately.