WGAN-GP结合CBAM-VGG16轻量化网络滚动轴承故障诊断  

Rolling Bearing Fault Diagnosis with WGAN-GP and Lightweight CBAM-VGG16 Network

作　　者：闫向彤[1] 罗嘉伟 曹现刚[1] YAN Xiangtong;LUO Jiawei;CAO Xiangang(College of Mechanical Engineering,Xi′an University of Science and Technology,Xi′an 710054,China)

机构地区：[1]西安科技大学机械工程学院,西安710054

出　　处：《噪声与振动控制》2024年第5期120-127,共8页Noise and Vibration Control

基　　金：国家自然科学基金重点资助项目(51834006)。

摘　　要：针对滚动轴承故障诊断中存在的故障数据不平衡且诊断效率低的问题,提出一种将改进的生成对抗网络(Wasserstein Generative Adversarial Network with Gradient Penalty,WGAN-GP)和轻量化卷积神经网络相结合的滚动轴承故障诊断方法。首先,利用连续小波变换(Continuous Wavelet Transform,CWT)生成二维时频图,并通过WGAN-GP进行数据增强;其次在视觉几何群网络(Visual Geometry Group Network-16,VGG16)的基础上,引入Ghost模块和全局平均池化(Global Average Pooling,GAP)对其进行轻量化改进;再次,利用卷积注意力模块(Convolutional Block Attention Module,CBAM)和带重启的余弦退火衰减法提高VGG16模型的性能,构建CBAM-VGG16轻量化卷积神经网络模型,将增强后的数据进行预处理后输入到模型中进行训练,建立故障诊断模型;最后采用西储大学轴承数据集进行模型验证和分析。实验结果表明:该方法证实了故障数据不足时进行故障诊断的可行性,缩短了模型的训练时间、诊断时间并缩减了模型的大小和参数量,提高了故障诊断的效率和准确率。Aiming at the issue of imbalanced fault data and low diagnosis efficiency in rolling bearing fault diagnosis,a method combining Wasserstein Generative Adversarial Network with Gradient Penalty(WGAN-GP)and a lightweight con-volutional neural network(CNN)is proposed.Firstly,two-dimensional time-frequency maps are generated by using the con-tinuous wavelet transform(CWT),and the data is enhanced by WGAN-GP.Secondly,the Visual Geometry Group Network-16(VGG16)model is improved by introducing the ghost module and Global Average Pooling(GAP)for lightweight optimi-zation.Thirdly,the Convolutional Block Attention Module(CBAM)and cosine annealing with warm restart are used to im-prove the VGG16 model′s performance,and the lightweight CBAM-VGG16 CNN model is constructed.The augmented da-ta is preprocessed and input into the model for training to establish a fault diagnosis model.Finally,the Case Western Re-serve University bearing dataset is used for model validation and analysis.The experimental results show that this method has the feasibility of fault diagnosis when the fault data is insufficient,reduces the training time,diagnosis time,model size and parameter number,and improves the efficiency and accuracy of fault diagnosis.

关 键 词：故障诊断 滚动轴承 生成对抗网络 Ghost模块 卷积注意力模块 

分 类 号：TH133.3[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]