基于多源信息融合的滚动轴承故障监测方法

Fault Monitoring Method for Rolling Bearings Based on Multi-Source Information Fusion

提出了一种基于多源信息融合的滚动轴承运行状态监测方法。首先,获取2种不同类型的传感数据信息(轴承振动和电动机电流),对单一类型的数据进行时域、频域及时频域的特征提取,得到24维的信息特征矩阵,基于主成分分析法(PCA)对矩阵降维,将第一主成分归一化后的向量作为初始证据源;其次,利用熵权法对每个类型的数据进行权重评估,并对初始证据源进行修正;然后,通过孙全组合规则改进D-S证据理论的融合决策,重新建立新的决策规则并引入离散度对融合结果进行评价;最后,借助帕德博恩大学公开的轴承数据信息作为数据源进行方法验证。试验结果表明:本方法可有效融合不同类型的状态特征信息,在同等级故障下,改进D-S融合效果是改进前的8倍以上;在相同加载条件下,改进D-S融合效果是改进前的4倍以上。

A monitoring method for running condition of rolling bearings is proposed based on multi-source information fusion.Firstly,two different types of sensing data information(bearing vibration and motor current)are obtained,and the single type of data is subjected to feature extraction in time domain,frequency domain and time-frequency domain to obtain a 24-dimensional information feature matrix,and the dimensionality reduction is carried out for matrix based on principal component analysis(PCA),and the vector normalized by the first principal component is used as initial evidence source;secondly,the entropy weight method is used to evaluate the weight of each type of data,and the initial evidence source is revised;then,the fusion decision of D-S evidence theory is improved by SUN Quan's combination rule,a new decision rule is re-established,and the dispersion is introduced to evaluate the fusion results;finally,the method is validated with the help of publicly available bearing data information from Paderborn University as data source.The experimental results show that:the method can effectively fuse the different types of state feature information,and the improved D-S fusion effect is more than eight times of that before improvement under same level of fault;the improved D-S fusion effect is more than four times of that before improvement under same loading conditions.