• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2021, Vol. 57 ›› Issue (2): 21-29.doi: 10.3901/JME.2021.02.021

• 仪器科学与技术 • 上一篇    下一篇

扫码分享

子空间嵌入特征分布对齐的不同工况下旋转机械复合故障诊断

陈仁祥1,2, 吴昊年1, 张霞1, 汤宝平2, 胡小林3, 蔡东吟1   

  1. 1. 重庆交通大学交通工程应用机器人重庆市工程实验室 重庆 400074;
    2. 重庆大学机械传动国家重点实验室 重庆 400030;
    3. 重庆工业大数据创新中心有限公司 重庆 400056
  • 收稿日期:2019-12-12 修回日期:2020-11-03 出版日期:2021-01-20 发布日期:2021-03-15
  • 通讯作者: 汤宝平(通信作者),男,1971年出生,博士,教授,博士研究生导师。主要研究方向为无线传感器网络、故障预测与健康管理。E-mail:bptang@cqu.edu.cn
  • 作者简介:陈仁祥,男,1983年出生,博士,教授,博士研究生导师。主要研究方向为智能测试技术与信号处理,机电装备安全服役、寿命预测与可靠性分析。E-mail:manlou.yue@126.com;吴昊年,男,1993年出生。主要研究方向机电设备故障诊断与健康评估、迁移学习。E-mail:296018167@qq.com
  • 基金资助:
    国家自然科学基金(51975079)、重庆市技术创新与应用示范(cstc2018jscx-msybX0012)、重庆市教育委员会科学技术研究(KJQN201900721)和交通工程应用机器人重庆市工程实验室开放基金(CELTEAR-KFKT-202002)资助项目。

Compound Fault Diagnosis of Rotating Machinery under Different Conditions Based on Subspace Embedded Feature Distribution Alignment

CHEN Renxiang1,2, WU Haonian1, ZHANG Xia1, TANG Baoping2, HU Xiaolin3, CAI Dongyin1   

  1. 1. Chongqing Engineering Laboratory for Transportation Engineering Application Robot, Chongqing Jiaotong University, Chongqing 400074;
    2. The State Key Laboratory of Mechanical Transmission, Chongqing University, Chongqing 400030;
    3. Chongqing Innovation Center of Industrial Big-Data Co., Ltd., Chongqing 400056
  • Received:2019-12-12 Revised:2020-11-03 Online:2021-01-20 Published:2021-03-15

摘要: 针对不同工况下复合故障诊断时跨域特征对齐与分布差异自适应调节的问题,提出子空间嵌入特征分布对齐的不同工况下旋转机械复合故障诊断方法。利用相关对齐方法在目标域子空间对齐源域与目标域对应特征,有效抑制域偏移;在该空间训练基分类器为目标域预测伪标签,用于定量估计两域边缘分布与条件分布各自权值,以适配两域特征分布差异;通过结构风险最小化框架构造核函数,建立分类器以传递上述两步学习规则,并通过迭代更新获得最优系数矩阵完成复合故障诊断任务。在两组多类别复合故障诊断实验证明了所提方法的可行性和有效性。

关键词: 不同工况, 复合故障, 子空间嵌入, 特征对齐, 分布差异自适应调节

Abstract: Aiming at the problem of cross domain feature alignment and distribution difference self-adaptive adjustment in different working conditions of composite fault diagnosis, a method of composite fault diagnosis for rotating machinery under different working conditions of subspace embedded feature distribution alignment(SEDA) is proposed. Using correlation alignment(CORAL) effectively suppressed the domain shift by aligning the corresponding features of the source domain and the target domain in the target domain subspace; In this space, the training base classifier is used to predict the pseudo label of the target domain, which has been used to quantitatively estimate the weights of the edge distribution and the conditional distribution of the two domains, so as to adapt to the differences of the feature distribution of the two domains; Through the structural risk minimization framework(SRM) constructed a kernel function, establish a classifier to transfer the above two-step learning rules, and obtain the optimal coefficient matrix through iterative updating to complete the composite fault diagnosis task. The feasibility and effectiveness of the proposed method are proved by two groups of multi-class composite fault diagnosis experiments.

Key words: different working conditions, compound fault, subspace embedded, feature alignment, adaptive adjustment of distribution difference

中图分类号: