奇异值分解与迁移学习在电机故障诊断中的应用  

Application of SVD and transfer learning strategy on motorfault diagnosis

作　　者：沈飞[1] 陈超[1] 严如强[1] 

机构地区：[1]东南大学仪器科学与工程学院,江苏南京210096

出　　处：《振动工程学报》2017年第1期118-126,共9页Journal of Vibration Engineering

基　　金：国家自然科学基金资助项目(51575102)

摘　　要：针对变转速、变负载条件下的电机故障诊断问题,提出了一种基于自相关矩阵奇异值分解(Singular Value Decomposition,SVD)的特征提取和迁移学习分类器相结合的诊断方法。对于Hankel矩阵提取的奇异值向量,设计了平均曲率区分度指标来描述特征差异性,迁移学习TrAdaBoost算法通过迭代过程中调节辅助振动数据的权重来帮助目标数据学习,提升了分类正确率,同时利用向量夹角余弦进行可迁移度检测从而避免负迁移。试验结果表明,SVD无需利用故障先验知识,具有通用性,且迁移学习相比传统机器学习在目标振动数据较少条件下性能得到显著提升。A novel approach utilizing singular value decomposition（SVD）for feature extraction and transfer learning for classification is presented in this paper aiming to improve the motor fault diagnostic performance under various operating conditions.A discrimination index is designed to describe the difference between the vibration features which are extracted by the Hankel matrix in SVD.The main idea of transfer learning is to utilize selective auxiliary data to assist target data learning,where a weight adjustment is involved in the TrAdaBoost algorithm for enhanced diagnostic capability.In addition,negative transfer is avoided through the similarity judgment between the target and auxiliary data.Experimental study indicates that transfer learning improves the diagnosis accuracy in complex conditions as compared to the traditional machine learning strategy.

关 键 词：故障诊断 奇异值分解 迁移学习 可迁移度 特征提取 

分 类 号：TH165.3[机械工程—机械制造及自动化] TP181[自动化与计算机技术—控制理论与控制工程]