Research And Design Of Fault Diagnosis System For Vibration Of Hydropower Unit

Along with the enhancing proportion of hydroelectricity in the whole electricity system in our country, stabilization of hydropower unit is more and more important to the whole system operation, which convert hydraulic energy into electric power, and their running conditions determine whether the hydro-plant can supply electric power safely and economically. Therefore, the effective system of state monitoring and fault diagnose for HGS, which can improve the repair pertinency and enhance the security and reliability of HGS, have important researchful significance.Since vibration is the important factor that influences the security and reliability of hydropower unit, Based on the analysis of characteristics of fault diagnosis for HGS vibration, this dissertation emphasizes on the vibration fault mechanism and extracting ways for the vibration fault symptoms. According to the three kinds of causes, the vibration faults of HGS are divided into mechanical vibration fault, electromagnetic vibration fault and hydraulic vibration fault. This dissertation analyzes the fault mechanism, summarizes the fault symptoms and presents the solutions in order to supply academic base for system development.An overall structure of monitoring and fault diagnosis system for Hydroelectric generating sets is designed. The system hardware design are finished, concluding test parameters, reasonable arrangement of the test spots, choosing of sensors and A/D cards and so on. On the basis of the above parts, a set of vibration monitoring and diagnosis software is developed by using object-oriented programming method.On the base of the studying traditional Expert system, the Neural network Expert system is adopted, it will conquer the limitation of the traditional Expert system, such as can not self-study, self-adaption. the BP algorithms are selected to capture the information. The modes of the neural network are used to express the knowledge, so the superiority of self-study of Neural network can be used sufficiently. In this thesis, A BP neural network expert system is also constructed. The network is trained using the sampled fault data, and the hybrid intelligent diagnosis to some fault is realized sequentially.During the training process of the fault samples, The emulational results show that all of the fault samples can be convergent to the setting error value.