基于改进变分模态分解与流形学习的滚动轴承故障诊断

Fault Diagnosis of Rolling Bearing based on Improved Variational Mode Decomposition and Manifold Learning

滚动轴承早期损伤信号特征量缺失且易被环境噪声掩盖,根据分形理论,结合灰狼优化算法(GWO)提出改进变分模态分解方法(Improved Variational Mode Decomposition,IVMD),求解各模态多种非线性特征量,并采用随机近邻嵌入理论(t-distributed Stochastic Neighbor Embedding,t-SNE)进行降维分类,以实现无监督故障诊断。基于轴承损伤实验数据,验证所提方法的可靠性。结果表明:采用IVMD所获模态与多种非线性值构建的特征矩阵更具代表性,可诊断轴承微弱故障;与现有方法相比,所提方法聚类表现更清晰,分类准确率更高,且具有良好的鲁棒性。

Early damage of rolling bearing lacks characteristic signals and such signals are drown out easily by environmental noise.Based on the fractal theory and the grey wolf optimization algorithm(GWO),the improved variational mode decomposition(IVMD)is proposed to solve nonlinear characteristics in different modes.The theory of t-distributed stochastic neighbor embedding(t-SNE)is applied to carry out dimension reduction and classification,in order to achieve fault diagnosis without supervision.Based on the experimental data of bearing damage,the reliability of the proposed method is verified.The results show that the eigenmatrix constructed by the modes obtained from IVMD and a variety of nonlinear values are more representative,and the proposed method can be used to diagnose weak faults of bearings.Compared with the existing methods,the proposed method has clearer clustering performance and higher classification accuracy,as well as good robustness.