摘要
针对不同工况条件下轴承振动数据分布不一致、源领域与目标领域自适应过程中适配不足或过度适配的难题,提出一种基于子结构最优传输的跨工况轴承故障诊断方法。通过小波变换提取轴承振动数据中的故障特征,构建故障样本集;再对源领域及目标领域轴承故障样本集进行聚类,生成源领域与目标领域故障样本数据的子结构,并自适应的对源领域数据子结构赋予不同权重,目标领域数据子结构赋予相同权重,完成对源领域数据子结构的映射;利用映射的源领域数据子结构及其所对应的标签,训练支持向量机模型并通过训练后的模型实现对目标工况轴承的故障诊断。将所提方法在机械综合故障模拟试验平台及凯斯西储大学轴承数据集上进行验证,并与传统机器学习及其他迁移学习方法进行对比,试验结果表明该方法的有效性与优越性。
Here,aiming at problems of inconsistent distribution of bearing vibration data under different working conditions,and insufficient or excessive adaptation in adaptive process of source domain and target domain,a bearing fault diagnosis method under cross working conditions based on substructure optimal transmission was proposed.Firstly,fault features of bearing vibration data were extracted with wavelet transform to construct fault sample sets.Then,bearing fault sample sets in source domain and target domain were clustered to generate sub-structures of fault sample data in source domain and target domain,adaptively assign different weights to data sub-structures in source domain,and assign the same weight to data sub-structures in target domain for completing mapping of data sub-structures in source domain.Finally,the mapped source domain data sub-structures and their corresponding labels were used to train a support vector machine model,and the trained model was used to realize bearing fault diagnosis under target working condition・The proposed method was verified on a mechanical comprehensive fault simulation test platform and the bearing data set of Case Western Reserve University,USA,and compared with traditional machine learning and other transfer learning methods.The test results showed that the proposed method is effective and superior.
作者
朱良玉
崔倩文
胡超凡
何水龙
ZHU Liangyu;CUI Qianwen;HU Chaofan;HE Shuilong(School of Mechatronic Engineering,Guilin University of Electronic Technology,Guilin 541004,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2023年第7期273-280,332,共9页
Journal of Vibration and Shock
基金
国家自然科学基金(51965013)
广西自然科学基金项目(2020GXNSFAA159081)
桂林电子科技大学研究生教育创新计划项目(2022YCXS017)。
关键词
故障诊断
子结构级匹配
迁移学习
跨工况
fault diagnosis
substructure level matching
transfer learning
cross working conditions
作者简介
第一作者:朱良玉,男,硕士生,1998年生;通信作者:何水龙,男,博士,教授,1983年生。
