摘要
针对传统时域、频域统计指标构建的特征矩阵在处理复杂非线性数据时的局限性,提出一种基于数据层特征融合和卷积神经网络(convolutional neural network,简称CNN)相结合的轴承故障诊断方法。首先,将采集到的信号进行变分模态分解(variational mode decomposition,简称VMD),选择峭度值大于阈值的分量进行重构;其次,从时域、频域、能量和稳定性等方面计算构建重构信号的多维度复合特征矩阵,将数据特征进行融合;然后,对特征矩阵进行核主成分分析(kernel principal component analysis,简称KPCA)降维处理,去除冗余信息;最后,将得到的低维矩阵输入到批量归一化(batch normalization,简称BN)层优化的CNN模型中进行故障识别与分类,并通过辛辛那提大学智能维护系统(intelligent maintenance systems,简称IMS)、旋转机械振动分析与故障诊断实验平台(QPZZ-Ⅱ型)2组实验数据进行数据验证。结果表明:所提方法对轴承故障分类具有较好的处理效果和稳定性。
To address the limitations of feature matrices constructed using traditional time domain and frequency domain statistical indicators in handling complex nonlinear data,a bearing fault diagnosis method combining data layer feature fusion and convolutional neural network(CNN)is proposed.First,the collected signals are processed using variational mode decomposition,and components with kurtosis values exceeding a predefined threshold are selected for reconstruction.Second,a multi-dimensional composite feature matrix of the reconstructed signals is calculated and constructed from the aspects of time domain,frequency domain,energy,and stability,integrating data features.Then,the feature matrix undergoes kernel principal component analysis for dimensionality reduction to remove redundant information.Finally,the obtained low-dimensional matrix is input into a CNN model optimized by batch normalization layers for fault identification and classification.Data validation is conducted using experimental data from the intelligent maintenance systems of the University of Cincinnati and the rotating machinery vibration analysis and fault diagnosis experimental platform.The results indicate that the proposed method achieves high classification and stability in bearing fault classification.
作者
张雄
李嘉禄
武文博
董帆
ZHANG Xiong;LI Jialu;WU Wenbo;DONG Fan(Hebei Key Laboratory of Electric Machinery Health Maintenance&Failure Prevention Baoding,071003,China;Department of Mechanical Engineering,North China Electric Power University Baoding,071003,China;Tianjin Metro Electronic Technology Co.,Ltd.Tianjin,300380,China)
出处
《振动.测试与诊断》
北大核心
2025年第6期1120-1127,1272,共9页
Journal of Vibration,Measurement & Diagnosis
基金
国家自然科学基金资助项目(52105098)
河北省自然科学基金资助项目(E2024502052,E2021502038)
中央高校基本科研业务费专项资金资助项目(2025MS137)。
关键词
滚动轴承
特征矩阵
降维
卷积神经网络
rolling bearing
characteristic matrix
dimensionality reduction
convolutional neural network
