基于模平方阈值—FastICA的滚动轴承降噪方法研究

Research on Noise Reduction Method of Rolling Bearing Based on Modulo Square Threshold—FsatICA

针对小波阈值降噪时硬阈值和软阈值存在的连续性差和小波系数高频损失的问题,以及经典独立分量分析(independent component analysis,简称ICA)方法降噪时的欠定问题,提出了模平方阈值与Fast ICA结合的联合降噪方法。首先将采集的单通道信号进行模平方阈值降噪处理,然后将降噪后信号与原振动信号组成输入矩阵进行Fast ICA降噪,最后对降噪后信号进行包络谱分析,实现滚动轴承特征提取和故障诊断。仿真和实验结果分析表明:该方法能使振动信号的峭度值增大,有效地滤除噪声,凸显故障特征信息,证明了该方法的有效性。

Aiming at the problem of the poor continuity and the loss of high frequency when using hard threshold or soft threshold of wavelet to make the noise reduction of vibration signal,and the problem of undetermined parameters when independent component analysis（ ICA） method is used to make the noise reduction of vibration signal,a newcompound fault diagnosis of rolling bearing method was proposed based on modulo square threshold combined with Fast ICA. Firstly,collected single—channel vibration signals was processed by Modulo square threshold denoising. Then the Fast ICA algorithm which input matrix conclude the processing result and the original signal was used for noise reduction. Finally,the characteristic frequency of rolling bearing can be clearly observed through envelope spectrum analysis. The simulation and experimental results showthat this method improves the kurtosis of vibration signal,filters out noise effectively and highlights the characteristic frequency of rolling bearing,proves that the newdenoising method proposed herein is valid.