Turbine Rotor Vibration Faults Diagnosis Based On Artificial Immunity And Information Fusion Technology

The failure rate of steam-turbine set is very high and the harmfulness is very great, about 30 to 50 percent of halting problem in heat-engine plant is due to the rotary such as steam-turbine generator, fan and pumping machine. So the fault forecasting and diagnosis are regarded as the key factor by the related research institutions, enterprises and administration departments, it is also important in the application of present fault diagnosis technology. The vibration wave pattern provides sufficient fault sign information. How to fully and exactly extract these signals will has significant meaning for the fault type determination, fault trend forecasting and predictive maintenance of steam-turbine unit.The subject is the experimental researches which use artifical immunity and D-S theory combined with wavelet packet to diagnose and forecast the steam-turbine vibration fault systematically.According to the common fault of steam-turbine vibration, rubbing, loosing, uncountershaft and mass unbalance can be simulated. With the Fast Fourier Transform about the collected vibration data, the wave pattern and the spectrum analysis can be realized. It is showed in the results that the disorder level of fault wave pattern will vary with different type of fault and the concerned elements and power are also different, therefore wave pattern and spectrum analysis can be regarded as the basic reference in fault diagnosis.Every signal (stationary or non-stationary) can be decomposed as a basic function race expanded and contracted by the wavelet through wavelet packet analysis. The information content is complete. Through the decomposition and recomposition for the signal in different criterion, the detailed information for the initial signal in different frequency band can be obtained, as well as the sudden changing time point. Based on the characteristic, decomposing and reconstructing the vibration signal can extract the mutation signal and non-stationary signal, filter the noise and other unimportant frequency component, and reserve the important signalImmune system is essentially the organism's own fault diagnosis system. Antigen and antibody immune system of mutual recognition process has a very strong pattern recognition ability of antibodies against the antigen, like keys and locks, a class of antibody is only able to identify a specific and similar antigens.In the process of practical fault diagnosis, due to the complex operating conditions, the parameter that is sensitive to all faults does not exist, and the same phenomenon of fault is often the result of several faults. Therefore, the diversity and uncertainty of the failure and the complexity of the linkages between the various failures together constitute the difficulties of fault diagnosis. The information fusion technology, based on the D-S evidence theory, integrates multi-sensor data as well as other channels of information, can recognize and describe diagnostic objects more accurate and comprehensive, and thus makes the right judgments and decision-making of the complex faults.The result indicates: Wavelet packet analysis method has good ability to extract non-stationary signal; the application of wavelet packet and artifical immunity provides a new diagnostic method of turbine rotor vibration faults; Combining wavelet packet analysis and D-S evidence theory for the fault diagnosis can enhancing the probability of the fault diagnosis and reduce uncertainty.