基于鲁棒局部线性嵌入投票的轴承故障诊断

Bearing Fault Diagnosis Based on Robust Locally Linear Embedded Vote

针对基于局部线性嵌入的特征选择方法对噪声和K近邻点敏感等问题,提出一种基于鲁棒局部线性嵌入投算法(Robust locally lineally embedding vote,RLLE vote)。将矩阵的1范数和2范数(L1和L2正则化技术)引入到局部线性嵌入的高维重构模型中,使其能够自适应选择样本的K-近邻点,进而提高了图形保持框架的鲁棒性;通过采用相对保守的最小角回归弹性网络(Least angle regression-elastic net,LARS-EN)迭代算法计算得到一个稀疏的重构权重矩阵以实现对噪声的抑制;利用度量准则衡量重构特征与原始特征之间的差异选择出原始高维数据中最具有代表性的特征。在标准轴承故障数据集(CWRU)和从实际操作平台采集的轴承数据集上的实验表明,与传统的方法相比,实验结果证明了所提方法的优越性和有效性。

Aiming at the problem that the feature selection method based on local linear embedding is sensitive to noise and K-nearest neighbors,a Robust locally lineally embedding vote(RLLE vote)is proposed。The L1 and L2 regularization techniques are introduced into the locally linearly embedded high-dimensional reconstruction model,which enables it to adaptively select the K-nearest neighbors of the sample,and improves the robustness of the graph-preserving framework;A sparse reconstruction weight matrix is calculated by adopting the relatively conservative Least angle regression-elastic net(LARS-EN)iterative algorithm to achieve noise suppression。Using measurement criteria to measure the difference between the reconstructed features and the original features to select the most representative features in the original high-dimensional data。The experiments on the standard bearing fault data set(CWRU)and the bearing data set collected from actual bearing test platform show that compared with the traditional method,the experimental results prove the superiority and effectiveness of the proposed method.