Technology Of Short Signal Analysis And Its Application In Fault Diagnosis

The thesis devote to DESAMS (Double-channel Electro-hydraulic Servo Adaptive Motor Simulator-rig) condition monitoring and accessing in theory and application. Based on short signal analysis theory, the thesis explores the representative vibration faults of complex and non-linear equipment by both theoretical and experimental methods systematically.The thesis gives a very comprehensive discussion of the technology for short signal analysis. The analysis of the FFT arithmetic for short signal serials posts the essential reasons of spectrum leakage for traditional methods of spectral analysis. The thesis, applies the auto-adaptive theory to the arithmetic for MESE (Maximum Entropy Spectral Estimation), and presents new LSLL arithmetic for MESE, which is based on method of the recursive least square with forgetting factor.The thesis researches deeply the order selection for the Maximum Entropy Spectral Model. Corresponding with the particularity of the short signal serials, the paper concludes the rules of the order selection for MESE and presents the experiential formula. A great lot of analysis and experimental results indicate that the experiential formula can give more exact order than traditional methods, such as A1C, B1C, CAT and SVD.Based on Texa Instrument Company's TMS320LF2407DSP microprocessor, data acquisition module, MESE arithmetic module and serial communication module are embedded in the single-chip microcomputer. The single-chip microcomputer can carry out condition monitoring and accessing in real time.In the field of application of fault diagnosis with MESE, using short signal analysis technology can get characteristics vector of faults. Then fuzzy reasoning and neural networks are used to construct fault diagnosis system. The fault diagnosis system analyses the characteristic vector by self-learning. Also the thesis improves the diagnosis model of fuzzy neural networks, and establishes the composite sub-nets based on RBF (Radial Basis Function). Thus the ability of the system to classify the faults is enhanced obviously. The multi-layers fuzzy neural networks fault diagnosis system is set up by using the hierarchy decomposing strategy to carry out condition monitoring and diagnosing. Finally, combinedwith the technology of short signal analysis, the application software of fuzzy neural networks fault detecting and diagnosis system (FDDS) for DESAMS is developed independently with the Microsoft Visual C++6.0.The simulative tests by computer and experiments by test-bed all can indicate that the established modularization and multi-layers fault diagnosis system for DESAMS is not only simple but also specific and real-time. The embedded single-chip microcomputer can process data in real time, and the precision of the analysis can meet demand of fault diagnosis. The PC industrial controlling computer can identify accurately fault and possess ability of self-learning for new fault. The FDDS is quite flexible and applicable to the actual system. All these properties make it available to do further research on the fault diagnosis technology of DESAMS.