针对滚动轴承振动信号具有较强的非线性，且包含较多冗余和无关特征，导致提取本质特征和故障识别困难，提出一种基于联合局部线性嵌入和稀疏自表示(joint locally linear embedding and sparse self-rep-resentation，JLLESSR)与参数优化支持向量机的滚动轴承故障诊断方法。该方法构造了一个统一的特征提取框架，依靠局部线性嵌入(locally linear embedding，LLE)挖掘高维数据的局部几何结构，同时通过稀疏自表示(self-representation)在低维空间挖掘高维数据的全局几何结构，得到表征滚动轴承运行状态的嵌入特征。然后，将得到的特征输入至交叉优化支持向量机(cross-validation support vector machine，CV-SVM)中进行故障识别。最后，在常见滚动轴承故障数据集上对所提出的方法进行测试，实验结果表明提出的方法能有效识别出滚动轴承不同类型的故障，并且故障诊断精度可达98。5％。
Fault diagnosis method of rolling bearing based on joint LLE and SSR
A rolling bearing fault diagnosis method based on joint locally linear embedding and sparse self-representation(JLLESSR)and parameter-optimized support vector machine is proposed for rolling bearing vibration signals with strong nonlinearity and containing more redundant and irrelevant features,which leads to difficulties in extracting essential features and fault identification.The method constructs a unified feature extraction framework,relying on local linear embedding(LLE)to mine the local geometric structure of high dimensional data,and self-representation to mine the global geometric structure of high dimensional data in low dimensional space,to obtain the embedding features characterizing the operating state of rolling bearings.Then,the obtained features are fed into a cross-validation support vector machine(CV-SVM)for fault identification.Finally,the proposed method is tested on a rolling bearing fault data set,and the experimental results show that the proposed method can effectively identify different types of rolling bearing faults,and the fault diagnosis accuracy can reach 98.5％.
rolling bearingvibration signallocally linear embeddingjoint locally linear embedding and sparse self-representationfault diagnosis
NETL
NSTL
万方数据
