关键词: 水电机组振动信号, 集合经验模态分解, 近似熵, 特征提取, 概率神经网络

Abstract: In recent years, research on the feature extraction method of non-stationary nonlinear vibration signals has been a hot spot in the fault diagnosis of hydropower units, and its effectiveness is a key factor of fault diagnosis accuracy. This paper describes a new feature extraction method based on ensemble empirical mode decomposition (EEMD) and approximate entropy for vibration signals from hydropower units. It uses EEMD to obtain approximate entropy eigenvalues of the intrinsic mode function (IMF), and then inputs them into a probability neural network (PNN) for pattern identification. Results show this method can effectively distinguish the difference in the operation modes of a unit, laying a basis for fault diagnosis in engineering applications.

Key words: vibration signal of hydropower unit, ensemble empirical mode decomposition, approximate entropy, feature extraction, probability neural network