基于滑动峰态算法的信号弱冲击特征提取及应用

机械故障振动信号中往往含有故障引起的弱冲击成分，冲击信号具有显著的非高斯特性，而零时滞四阶累积量即峰态能够描述信号偏离高斯分布的程度；基于峰态这一特性，本文提取一种基于滑动峰态算法的弱冲击特征提取方法，首先对原信号进行滑动峰态计算，获得一个新的峰态时间序列，然后对该峰态时间序列进行傅立叶变换，提取出信号中冲击成分的频率特征。通过强背景信号及噪声环境下弱冲击特征提取的仿真研究，证明了该方法具有很好的冲击特征提取能力。以实测齿轮断齿信号分析结果证明了该方法的有效性。

A new method for extracting weak impulse characteristic based on a sliding kurtosis algorithm

The weak impulse component is always contained in the mechanical fault vibration signal, and the impulse signal possesses the remarkable non-Gaussian characters. The kurtosis could depict the departure degree away from the Gaussian distribution, so a new method was proposed for the weak impulse characteristic based on the sliding kurtosis algorithm. Firstly, a new time series, kurtosis series, is obtained by sliding kurtosis computation of the original signal; then FFT of the kurtosis series is done to extract the frequency of the weak impulse signal. The simulation was done by the new method to the strong signal and noise signal with weak impulse components, and the results illustrated that the method has strong ability for extraction of the weak impulse component. In practice, a broken gear vibration signal was analyzed by the method and the weak impulse component was extracted effectively.