基于SC-DCGAN的不平衡数据轴承故障诊断  

Unbalanced Data Bearing Fault Diagnosis Based on SC-DCGAN

作　　者：廖珂 荆晓远 李双远 刘雨晖 刘飞 LIAO Ke;JING Xiaoyuan;LI Shuangyuan;LIU Yuhui;LIU Fei(College of Information and Control Engineering,Jilin Institute of Chemical Technology,Jilin Jilin 132000,China;College of Computer,Guangdong University of Petrochemical Technology,Maoming Guangdong 525000,China)

机构地区：[1]吉林化工学院信息与控制工程学院,吉林吉林132000 [2]广东石油化工学院计算机学院,广东茂名525000

出　　处：《机床与液压》2024年第24期208-213,共6页Machine Tool & Hydraulics

基　　金：国家自然科学基金面上项目(62176069)。

摘　　要：针对滚动轴承故障诊断过程中因数据不平衡而导致的少数类样本诊断精度低的问题,提出一种基于统计特征条件深度卷积生成对抗网络的不平衡数据故障诊断方法。该方法在条件生成对抗网络中引入振动信号统计特征,得到新的融合条件模型,引导生成器更稳定地生成符合真实样本分布的数据以平衡数据集;再采用卷积网络模型在平衡后的数据集上进行分类识别。选择多个不平衡比例,在某装备故障诊断重点实验室数据集上进行实验。结果表明:相对于其他经典模型,文中所提方法能够有效地处理不平衡故障分类问题,并提高对少数类样本的识别能力。In order to address the problem of low diagnostic accuracy of minority class samples due to data unbalance in the rolling bearing fault diagnosis process,a diagnosing method for imbalance data was proposed based on statistical feature condition generative adversarial network.The statistical characteristics of vibration signals were introduced into the condition generative adversarial network to obtain a new fusion condition model,which could guide the generator to generate more data matching the real sample distribution to balance the data set,and then the convolutional network model was used to classify and identify the balanced dataset.Experiments were conducted on the bearing dataset from a equipment fault diagnosis key laboratory,considering various unbalanced ratios.The results show that compared with other models,the proposed method can effectively handle the unbalanced fault classification problem and the identification ability of the minority class samples is improved.

关 键 词：类不平衡 故障诊断 条件生成对抗网络 统计特征 

分 类 号：TH165.3[机械工程—机械制造及自动化]