基于多重分形理论与神经网络的齿轮故障诊断

Gear fault diagnosis based on multifractal theory and neural network

针对齿轮故障振动信号具有多重分形特征,提出多重分形与神经网络相结合的机械故障诊断方法。采用多重分形理论计算出振动时间序列的多分形谱f(α)和广义分形维数D(q),并将多分形谱能和广义分形维数谱能作为特征量,构成二维特征向量。将该特征向量作为概率神经网络的输入参量,对采自齿轮故障台的振动信号进行故障分类。作为对比,将关联维数作为特征量输入同样参数的概率神经网络并进行故障识别,结果表明,所提出的方法具有更高的识别率。

The mechanical fault diagnosis method based on the multifractal theory combined with the neural network was proposed according to multifractal characteristics of gear fault vibration signals.The multifractal theory was used to calculate multifractal spectra f（α）and generalized fractal dimensions D（q）of vibration time series.Taking the multi-fractal dimension spectral energy and the generalized fractal dimension spectral energy as characteristics,a two-dimensional characteristic vector was formed,it was taken as input parameters of a probabilistic neural network,and with it the fault classification for vibration signals collected from a gear fault platform was done.Experimental results showed that the proposed method can effectively be used in gear fault diagnosis,and it has a higher recognition rate.