Research On The Multi-source Information Fusion Based Fault Diagnosis System Of Hydro-power Generating Unit And Shaft Orbits Recognition

With the development of the gaint and large capacity hydroupower set, the technology of fault diagnosis on the operation of hydro-unit stands out. Researches on fault diagnosis of hydro-unit have been carried out deeply and many achievements have been obtained. But Limitted by the level of the researches, many kinds of faults with uncertainty could not be diagnosed exactly. It is a challenge for us to research the fault diagnostic method of high performance for hydro-unit. The achievements would help us understand the fault characteristics and information fusion of hydro-unit and improve the diagnostic accuracy.By summarizing the studies published in recent years, the dissertation introduces a Racah moment based method to identify the shaft orbits automatically and gives an evaluation for it. Racah moment based method has the highest classification capability. Compared with some classical methods, the experimental results reveal that the performance of the proposed method is superior to that of other method. It is much suit for practical application.And, in order to resolve the problem of large quantity and high redundancy for the samples, we utilize the rough theory to mine useful information. By using the advantages of processing fussy and uncertain information, we establish the data mining model to extrat the instinct relation between the faults and their calculated characteristics. When simplifying the redundancy in the attributes, the refined information would reduce the operating time and improve the diagnostic accuracy.The mappings between the faults and the characteristics of samples are not monotonous. And the behaviors of the components for hydro-unit have good couple each other. These factors tell us that univocal diagnostic technology cannot obtain the definite operation status of the hydro-unit. Then the D-S Evidence Theory is introduced to fuse the diagnostic results of various fault diagnosis technology.Furthermore, in order to integrate the diagnostic results of various methods to obtain consistent conclusion, the dissertation design a new kind of mlti-agent diagnosis model based on rough set and D-S evidence theory for hydro-unit. The model has four properties to improve its diagnostic efficient. First, in allusion to the classic model of MAS-based fault diagnosis system, it makes status-monitored subsystem to get abnormal signals directly from factory to increase sensitivity of system. At the same time, an attribute value reduction agent is design to mine the instinct and wrought relationship from very large mount of simples. Second, a fusion agent is constructed to integrate the diagnostic results received from the diagnosis sub-agents to chieve the consistent result. Third, it improves information interactive features between mission-controlled subsystems and diagnosis task decomposition subsystem to increase the rate of control signal transmission. Fourth, in fault diagnosis subsystem of real-time fault diagnosis system, an MAS-based interactive parallel working structure is designed to meet requirements of high reliability and good real-time. A java-based software so called as JAFMAS is used to build platform to construct a multi-Agent coordination mechanism.Finally, a multi-agent fault diagnosis system for hydro-power unit based on rough set and D-S evidence theory is established. Experimental results show the effectiveness of the proposed system.