De-noising of signal processing is crucial for fault diagnosis in order to successfully conduct feature extraction and is an efficient method for accurate determination of cause. In this paper, the empirical mode decomposition (EMD) thresholding-based de-noising method and probabilistic neural network (PNN) are respectively used in the de-noising of the vibration signal and rotor fault diagnosis and compared with wavelet thresholding-based de-noising technology and back propagation neural network (BPNN). The results show that the clear iterative EMD interval thresholding performs better than wavelet thresholding in the de-noising of the vibration signal, and avoids the determination of wavelet basis and decomposition level. In addition, the PNN created by feature samples does not require training and has a higher accuracy than BPNN. on the fault diagnosis of hydropower units, application of artificial intelligence in electric power system, and modeling, analysis, design and control of hydropower equipment. Peng Lihong received B.E degree in thermal and power engineering from School of Energy and Electrical Engineering, Hohai University, Nanjing, P.R. China, in 2015. She is currently a Master candidate in hydraulic hydroelectric engineering from School of Power and Mechanical Engineering, Wuhan University. Her research interests include modeling simulation and optimization control of the hydroelectric generating set, condition monitoring and fault diagnosis technology.