Integrated Signal Processing, Intelligent Fault Diagnosis

Intelligent fault diagnosis is a new rising discipline and has wide applications in areas of aviation, aerospace, nuclear power plant and automatic control systems Intelligent fault diagnosis systems combine the measurements from many sensors, and process them to attain the information of the running states of the diagnosed systems. Based on the information, intelligent fault diagnosis systems produce the diagnosis and prognosis for the diagnosed systems. This dissertation focuses the key points of the researches on the signal processing techniques for intelligent fault diagnosis systems, especially on the application of wavelet analysis (WA) and neural network (NN) technology in the field of the intelligent fault diagnosis. The main research work includes:1. The fault diagnosis based on wavelet transform (WT) is studied.According to the theory of the signal singularity detection based on WT, a fault diagnosis method based the B-spIine wavelet is given, and the good result is obtained in the simulation experiment.Making use of the performance differences between the WT of the signal and the WT of the white noise, a kind of denoising method is proposed. The simulation experiment shows that the method could filter the white noise from the signal effectively. It provides a solid foundation for raising the correct rate of the fault diagnosis.By the using of the wavelet packet transform, the energy frequency-band of signal is analyzed. At the same time, corresponding diagnosis methods and results are given.2. The fault diagnosis based on NN is studied. It is studied deeply that BP algorithm of feed forward neural network that is used in fault diagnosis. An improved BP algorithm is proposed, and applied to the fault diagnosis of the rotating machinery.3. This paper introduces the structure models and learning algorithms of a wavelet neural network (WNN), and discusses its applications in the fault diagnosis systems.