基于关联维数和线段聚类的滚动轴承故障诊断

Fault Diagnosis of Rolling Bearing Based on Correlation Dimension and Segment Clustering

提出一种基于关联维数的滚动轴承故障特征提取方法。线性标度区域的识别是影响关联维数准确度的重要因素,针对关联维数线性标度区对应的二阶导数在零上下波动这一特征,将二阶导数数据点转化为线段,再利用线段的聚类方法进行两次聚类分析,并应用统计学准则排除粗大误差,最后对数据拟合得到特征值。对经典的Lorenz混沌系统进行仿真分析,具有良好的效果,并对滚动轴承4种状态信号进行特征分析,实验表明该方法能更加准确地识别出轴承故障信号。

A new fault feature extraction method for rolling bearing based on correlation dimension is proposed. Linear object recognition region is an important factor for the accuracy of correlation dimension,this method is characterized by the fluctuation of the two order derivative corresponding to the linear scaling region of the correlation dimension,the two order derivative data points are transformed into line segments,then the two clustering analysis is done by line segment clustering method,and the gross error is eliminated by statistical rule. Finally,the eigenvalues are fitted to data. This method is used to simulate and analyze the classical Lorenz chaotic system,and the effect is good. Four kinds of signal of rolling bearing are identified,and the experiment shows that the new method can identify the bearing fault signal more accurately.