Wavelet Application In Fault Diagnosis For Air Defense Missile

Throughout fault diagnosis development we could see that it is essential to have signal preprocessing and fault signal analysis in order to obtain accurate result during fault diagnosis while methods to implement are just as important. This dissertation examines the wavelet theory and its application on signal preprocessing and data feature extraction during fault diagnosis based on its recent ten years development.First we discuss the selected topic and its background about this paper, and then go into the basic theory of wavelet transform. For real world implementation of wavelet theory, we mainly focus on the selection of suitable wavelet bases and related algorithm applied in the following different aspects:In this paper, the selection of wavelet basis function is presented in detail Through the investigation of traditional selection of wavelet basis and the theory of lifting scheme, an adaptive wavelet transform is put forward , and the adaptation is come from adaptive choosing between a class of linear predictors within the lifting framework according to the local gradient of the signal. We investigate the central issues such as the structure of adaptive frame and the calculation of corresponding wavelet basis function.Then we investigate the application in signal de-noising with wavelet transform. Due to the specialties of fault signals in missile fault diagnosis, we proposed a novel de-noising method applying wavelet analysis scale space filter by using the properties of the signal and noise modulus maxima across scales. And modified it by using information on neighboring coefficients into the decision making. This de-noising method can reduce noise to a high degree while preserving most of the important features of the signal such as edges and other singularities. This algorithm applied to real system produces a good result.Third we investigate the method for feature extraction, and put forward two methods base on wavelet analysis: coefficients character obtained by wavelet multi-resolution analysis and energy spectrum analysis using wavelet pack technique,from which a feasible feature vector is created. Also ground is laid out for further study in fault diagnosis based on integrated neural networks. In study of the air defense missile system, since the range is variant, so we processed the remain signal and extracted comparable data features out of it. Parameters for diagnosis are also selected based on two different criteria: the cluster divergence of sample datas and the diagnosis reliability of parameter candidates.Finally we applied our investigation on real missile fault diagnosis system, and achieve a success. Our research is proved to be important for improving efficiency on fault diagnosis and speeding up the progress of missile trial-produce.