基于融合模型的齿轮传动故障诊断研究

Study fault diagnosis of gear based on the fusion model

针对齿轮故障诊断系统的特点,提出基于多尺度小波能量谱、扩展邻域粗糙集和多分类支持向量机融合的故障诊断方法.首先使用多尺度小波变换有效提取变工况条件下的故障特征;然后以广义欧式距离为基础构建扩展邻域粗糙集模型,在不需先验知识和理论公式的前提下,完成噪声剔除、各能量带重要度权重分配;最后利用支持向量机在小样本、非线性空间中良好的分类能力,提高故障诊断的决策准确性.将该融合模型应用到齿轮副的故障诊断中,验证其有效性.

For fault diagnosis of gear,this paper presented the fusion method based on multi-scale wavelet energy spectrum,expanded neighborhood rough set and multi-classifier SVM.Firstly,multi-scale wavelet transform was used to effectively extracted failure feature in variable work conditions.Secondly,without prior knowledge and theoretical formula,extended neighborhood rough set based on generalized euclidean distance was used to remove noise and distribute weight of every energy band.Then with better classification ability in small sample and non-linear space,SVM was used to improve the decision accuracy of fault diagnosis.Finally,the fusion model was applied to fault diagnosis of gear pair to verify its effectiveness.