| Power station fans belong to large-scale rotate machines. At present, under the background of energy conservation and emision reduction, large-scale boiler systems in domestic heat engine plants almost implemented transformation of energy-saving by increasing all kinds of dedust, desulfuration and denitration systems which cause increasing power of induced draft fan in boiler’s exhaust gas system and desulphurization booster fan in desulfuration system. Sometimes above two kinds of fans become one in large-capacity thermal power units which increase its drive power close to 10000kW, and make itself become super-large-scale rotate machine next to steam turbine generator unit. Once occurred faulty in rotate machine with this power level is a significant cause of economic loss, so it’s necessary to perform continuous on-line monitoring on its operation state.There are monitoring points for operation parameters(pressure, flow, vibration etc) installed on all kinds of fan machines in domestic large-capacity thermal power generating units. For big sample interval of the monitoring data, it can only reflect variation trend of equipment operational condition, have an overstep alarm and provide protection for equipment, but can not provide detailed diagnosis information of related machines fault.Fault diagnosis technology and system based on vibration monitoring has been applied on large steam turbine-generator widely, but application on power station fan is not much, so there is a strong practical engineering requirement. But the operating environment of all kinds of fan equipment in heat-engine plant is complex, and field measurement vibration signal contain strong interference noise component, which makes identification of fault information difficult. And because of difference in fan types, structure, layout of wind-net system, operation condition, etc. in different heat-engine plants, maybe occur variou different kinds of abnomal vibration. So it has a practical engineering significance for improving safe reliability of fan equipment operation that researching effective fault feature extraction methods from measured vibration signal in depth and developing vibration monitoring and fault diagnosis system suitting for all kinds of fan in power station.This paper focused on high-power plant fan, aimed at the issues, serious interference existed in measured vibration signal and difficulty in identify fault feature, caused by the complex structure of fan bearing, various operation condition and bad environment. The main research contents and achievement are as follows:(1) Analysis on basic structure and operation feature of power station fans. Aimed at 600MW thermal power generating units, the effect in constant speed and frequency control induced draft fan for quick load change is discussed by comparative analysis on in field historical operation data of all kinds of power plant fans recorded in field SIS system. The analysis showed that for primary air fan, forced draft fan and induced draft fan with constant speed, under changing-load operation, although fan current changes along with load to regulate output, there is difference in vibration feature of fan bearing. For frequency control induced draft fan, reduced speed can reduce fan current to save energy by a large margin, but speed varying in large range causes obvious fan structure resonance problems, in the range of analysed speed, there are resonance peak existed in three different speed, which should be paid attention to.(2) Measured vibration signal analysis and fault feature extraction on power stationfans. On the basis of analysing fan structure vibration feature and fault mechanism, apply vibration signal analysis and fault feature extraction technology to actual power plant fans and summary fan vibration feature by measuring in field. On the one hand, the structure of power plant fan is complex, there are more excitation souces which interfere each other seriously and structure vibration response contains complex superposition, modulation, mixture, etc.,on the other hand, measured fan vibration signal in field effected by measurement noise, which makes fault information submerged in noises, fault feature extracted by traditional signal analysis less prominent and identification difficult.(3) Improved envelopment analysis based on wavelet transform. Aimed at the issue that it is difficult to achieve automative fault feature extraction because envelopment analysis needs artifical identification filter bandwidth identified by artificiality, apply muti-scale envelopment analysis based on complex wavelet transform to bearing fault diagnosis working in complex environment, which showed that the method has a prominent ability of showing the weak fault information and identifying fault impact compenent submerged in strong interference noise effectively. Muti-scale band-pass filtering and amplitude envelopment extraction is achieved at the same time by muti-scale envelopment analysis of complex wavelet transform, disadvantage of traditional envelopment method needing foreseed fault band is overcome, the efficiency of signal analysis is improved. The analysis effect is similar with Hilebert envelopment spectrum in the transverse section of multi-scale complex wavelet envelope spectrum diagram, but vertical section at rotational frequency is most sensitive to the bearing fault, which suits for antifriction bearing weak fault alarm.(4) Wavelet fractal fault feature extraction method based on wavelet and fractal. Vibration signal in complex mechanical system like power plant fans has statistical self-similarity characteristics, so operation status of system can be described using correlation dimension to extract fault feature. Aimed at strong interference noises contained in measured vibration signal, which cause the effect of fault feature extraction unobvious, wavelet analysis and fractal are combined, and according to the different performance characteristics in wavelet transform coefficients of fractal signal and random noise, using the wavelet scale domain filtering to eliminate interference noise, accuracy and stability of computed correlation dimension can be improved. Research shows that computed correlation dimension agrees with nonlinear system feature in actual situation after processing measured vibration signal with the wavelet scale domain filtering refactoring and computed value of correlation dimension is more stable and can’t change along with test time and the length of value.(5) Research and development of vibration monitoring on high power induced draft fan. Isolate multi-channel shaft vibration signal sources for many large steam turbine unit by using frequency domain ICA analysis. The result shows that frequency domain ICA can describe source signal more clearly and the result of isolation agrees with the actual situation of steam turbine unit. The frequency domain ICA can isolate probable weak fault information from measured shaft vibration signal, and the physical significance of isolation result is more clearly. ICA feature extraction method using singel-channel measured vibration signal is researched, and the result shows that the primary function isolated by merely single-channel ICA is not enough to reflect the changes in the operation status of unit. So each primary function should be processed to validate the effectiveness of feature information. |
PDF Full Text Request