应用Hilbert时频谱和神经网络对发电机转子故障进行分类的方法研究

Classification for Fault of Generator Rotor Based on Hilbert Time-frequency Spectrum and Neural Network

提出了一种新的发电机转子故障检测的信号分类方法H ilbert时频谱,它是一种新的分析非平稳、非线性的时频分析方法。这种方法用经验模式分解法将一维信号分解成内蕴模式函数,进而计算有意义的多分量信号的瞬时频率。将其应用于故障信号的分析,可以提供新的时频属性;然后计算这种时频谱的矩和边缘以及时频熵,并将其作为特征向量。应用RBF概率神经网络作为分类器,可以实现不同故障模式的自动分类。对发电机的不同转子故障模式的信号研究表明了该方法的精确性和稳定性。

A new fault detection method for signal classification in generator was presented. It is a new method for processing non - stationary, signal. This method decomposes the one- dimensional signals into intrinsic mode functions （IMFs） using empirical mode decomposition method and then calculates the meaningful multi -component instantaneous frequency. Applied to a fault signals analysis, it can provide more new time - frequency attributes, Then the moments, margins and entropy of the time - frequency spectrum can be calculated as the feature vectors. The probabilistic neural network can be used to classify different fault modes. The accuracy and robustness of the proposed methods were investigated on signals of different fault condition in generator rotor.