摘要
基于数据驱动的滚动轴承智能故障诊断得到广泛研究,但多数研究中均假设训练数据与测试数据同分布,考虑到旋转机械实际运转中复杂多变的工况往往导致数据分布产生偏差,使得识别方法的通用性差、实际识别效果不佳。将域适应引入轴承故障诊断过程中,基于迁移学习提出了一种特征空间域和标签概率分布同步适应的迁移学习网络。该网络将一维稠密卷积网络及注意力机制融合实现复杂故障特征的自动提取;域适应处理通过联合最小化特征概率分布差异和标签概率分布差异来约束网络学习域不变特征;最终对变工况滚动轴承故障实现了高准确度的识别。实验结果表明了该方法的可行性及良好的性能。
Intelligent fault diagnosis of rolling bearing based on data-driven has been widely studied,but most studies assume that training data and test data are distributed in the same way.Complex and changeable working conditions in actual operation of rotating machinery often cause data distribution having deviations to make the universality of the identification method poor and the actual identification effect not good.Here,domain adaptation was introduced into bearing fault diagnosis process.Based on migration learning,a migration learning network with synchronous adaptation of feature space domain and label probability distribution was proposed.The network could integrate one-dimensional dense convolution network and attention mechanism to realize automatic extraction of complex fault features.Domain adaptation constrains invariant features of network learning domain by jointly minimizing feature probability distribution difference and label probability distribution difference.Finally,high accuracy identification was realized for faults of rolling bearing under variable working conditions.Test results showed that the proposed method has the feasibility and good performance.
作者
孙洁娣
刘保
温江涛
时培明
闫盛楠
肖启阳
SUN Jiedi;LIU Bao;WEN Jiangtao;SHI Peiming;YAN Shengnan;XIAO Qiyang(School of Information Science and Engineering,Yanshan University,Qinhuangdao 066004,China;Hebei Key Laboratory of Information Transmission and Signal Processing,Yanshan University,Qinhuangdao 066004,China;Key Laboratory of Measurement Technology and Instrumentation of Hebei Province,Yanshan University,Qinhuangdao 066004,China;School of Artificial Intelligence,Henan University,Zhengzhou 475000,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2022年第17期204-212,共9页
Journal of Vibration and Shock
基金
国家自然科学基金(61973262)
河北省自然科学基金(E2020203061)
河北省高等学校科学技术研究项目(QN2019133)
河北省重点实验室项目(202250701010046)
河南省青年人才托举计划(2021HYTP014)。
关键词
轴承故障诊断
变工况
稠密卷积网络
注意力机制
类别标签辅助
bearing fault diagnosis
variable working condition
dense convolution network
attention mechanism
class labels aided
作者简介
第一作者:孙洁娣,女,博士,教授,1975年生。
