Design and Realization of Diesel Fuel System Fault Diagnosis Based on Wavelet Neural Networks
|Course||Control Theory and Control Engineering|
|Keywords||Diesel engine Fuel system Fault Diagnosis Wavelet Analysis Neural Networks|
Currently, the theory and method of diesel engine fuel system fault diagnosis has become a hot research topic. As a typical representative of the reciprocating machinery, fuel system fault showing the complexity and diversity of the characteristics of application of traditional theory and method of fault diagnosis is difficult to achieve the intended purpose and requirements. Neural network theory continues to mature, new ideas, new methods for the diesel engine fuel system fault diagnosis. Feature extraction is an important part of the fault diagnosis is the key to determine the diagnostic success. Wavelet transformation, due to its localized nature of having a space, has become an important means of the feature extraction. Therefore, the combination of wavelet transform and neural network - Wavelet neural network in fault diagnosis of diesel engine fuel system concern. Modern detection techniques, wavelet transform and artificial neural network theory, the paper, the design of the diesel engine fuel system fault diagnosis system, and applied to the project, and achieved good results. The paper work is reflected in the following aspects: 1, the use of a new type of non-disintegration installed the solid signal acquisition method. The clamp-on sensors, indirectly, by the high pressure oil chamber pressure waveform, diesel engine oil pressure waveform without disassembly online collection. 2 based on wavelet transform feature extraction method. On the basis of the analysis and interpretation of the basic theory and method of wavelet transform characteristics for hydraulic waveform, research the following method to extract the oil pressure waveform characteristics: (1) wavelet coefficients modulus maxima method. Diesel engine oil pressure signal from the spray point and the maximum injection pressure points often contain important feature of the fault signal. For this feature, this paper wavelet coefficients modulus maxima multi-scale signal edge detection and analysis of fault feature extraction of oil pressure signal. The results show that the wavelet coefficients modulus maxima method can extract oil pressure signal failure characteristics. (2) wavelet packet frequency bands energy analysis. According to the band energy analysis technologies, energy - Fault Act, hydraulic signal fault feature extraction. The results show that the band energy analysis method is superior to the wavelet coefficients modulus maxima method, it extracts the energy characteristics more suitable as input vector of the neural network fault diagnosis. 3, the study of the neural network diagnosis of diesel engine fuel system failure. Easy to fall into the shortcomings of local minima and slow convergence of BP network following diagnosis of fuel system failure. (1) fuel system fault diagnosis based on RBF network. Make full use of the RBF network has a unique best approximation and an important feature of local minima, fault diagnosis of diesel engine fuel system. The diagnostic results show that the adoption the RBF of this new network structure, the diesel engine fuel system fault diagnosis, can avoid network into local minima and slow convergence problem fundamentally, in order to accurately and quickly diagnose the fault. (2) to explore the fault diagnosis based the SOFM network of fuel system. According to the the SOFM theory build a network model for diagnosis of fuel system failure. The results showed that the the SOFM network diagnostic model input sample vector requirements, but more accurate diagnostic results. 4, wavelet packet neural network is applied to a new method of diesel engine fuel system fault diagnosis. The paper describes the application of wavelet transform extract feature information and the basic structure of the neural network theory, wavelet packet energy and RBF network combination used to fuel system fault diagnosis. Wavelet transform loose type of neural network, greatly improving the accuracy of diagnosis. 5, 165-type diesel engine as an object of study, design of the diesel engine fuel system fault diagnosis system, and conduct on-site verification. The host computer is designed based on VC lower machine 89C52 microcontroller-based design, both through the RS-232 serial communication. Site to verify the feasibility of the system to achieve real-time on-line diagnosis of diesel engine fuel system. Finally, a brief summary of the work done in this article, as well as the results obtained experience, analysis of the inadequacies of this article, as well as problems that need to be solved.