轴承故障的排列熵特征提取与GK模糊识别方法

为了提高轴承故障诊断准确率,提出了参数优化多尺度排列熵的特征参数提取方法和加权GK模糊聚类的识别方法。在特征提取方面,以多尺度排列熵序列偏度最小为优化目标,使用多作用力微粒群算法优化多尺度排列熵参数,实现了排列熵特征参数在轴承不同故障状态下的完全分离;在故障识别方面,提出了加权GK模糊聚类的识别方法,使用ReliefF算法计算特征参数权重,为高敏感度特征参数赋予更大的权值,从而提高GK模糊聚类的聚集度。经轴承故障实验验证,文章提出的排列熵特征参数提取和GK模糊聚类识别方法在此次实验中能够精准识别轴承故障类型,说明文中提出的特征提取和模式识别方法具有一定借鉴意义。

Permutation Entropy Feature Extraction and GK Fuzzy Recognition Method of Bearing Fault

To improve accuracy rate of bearing fault diagnosis,feature extraction based on parameters optimized multiple scales entropy and recognition method based on weighted GK fuzzy clustering are proposed.In aspect of feature extraction,minimizing multiple scales entropy series skewness is as optimizing object.Using multiple interactions particle swarm algorithm to optimize multiple scales entropy parameters,complete separation of entropy feature parameters under different fault state overcomes.In aspect of fault recognition,weighted GK fuzzy clustering is put forward.Feature parameters weight is calculated by ReliefF algorithm,and enhanced sensitive feature is given big weight,so that degree of aggregation of GK fuzzy clustering is improved.Clarified by bearing fault trial,the method proposed by the essay can recognize bearing fault accurately in this trial,which means feature extraction and model diagnosis method possesses certain reference significance.