基于MSAM-CGAN的轴承碰摩故障识别方法研究  

Research on the Method of Rub-Impact Fault Recognition Based on the Conditional Generative Adversarial Nets

作　　者：张秋生 吴志刚 张文亮 邓艾东[3,4] 许猛 Zhang Qiusheng;Wu Zhigang;Zhang Wenliang;Deng Aidong;Xu Meng(CHN Energy New Energy Technology Research Institute Co.,Ltd,Beijing 102209,China;CHN Energy Dingzhou Power Generation Co.,Ltd,Dingzhou 073000,China;National Engineering Research Center of Power Generation Control and Safety,Nanjing 210096,China;School of Energy and Environment,Southeast University,Nanjing 210096,China)

机构地区：[1]国家能源集团新能源技术研究院有限公司,北京102209 [2]国家能源集团定州发电有限责任公司,定州073000 [3]东南大学大型发电装备安全运行与智能测控国家工程研究中心,南京210096 [4]东南大学能源与环境学院,南京210096

出　　处：《信息化研究》2025年第1期48-56,共9页INFORMATIZATION RESEARCH

基　　金：江苏省碳达峰碳中和科技创新专项资金(No.BE2023854);中央高校基本科研业务费专项资金资助(No.2242024k30046,No.2242024k30047)。

摘　　要：针对机械转子碰摩AE故障样本少,采样困难,并难以有效构建数据驱动的设备故障诊断模型的问题,本文提出一种基于MSAM-CGAN的轴承碰摩故障识别方法。通过条件生成对抗网络(CGAN)实现了转子故障特征样本的有效扩充,并构建了多尺度注意力机制(MSAM)故障识别网络,使得特征样本生成与故障识别在同一网络完成,实现了对转子碰摩故障的自动端到端检测。实验结果表明,本文所提方法能改善现有数据不足的瓶颈问题,鲁棒性较强,能有效进行轴承碰摩状态的识别。Aiming at the problems of few samples of mechanical rotor touch-friction AE faults,sampling difficulties,and difficulties in effectively constructing data-driven equipment fault diagnosis models,a bearing touch-friction fault identification method based on MSAM-CGAN is proposed.Through the Conditional Genera-tive Adversarial Network(CGAN)to achieve the effective expansion of the rotor fault feature samples,and constructing a multi-scale attention mechanism(MSAM)fault identification network,the feature sample gener-ation and fault identification is completed in the same network,to achieve the automatic end-to-end detection of rotor touching faults.The experimental results show that the method proposed in this paper can improve the bottleneck problem of insufficient data,has strong robustness,and can effectively identify the bearing grinding state.

关 键 词：声发射 条件生成对抗网络 碰摩故障 多尺度注意力机制 

分 类 号：TH133.33[机械工程—机械制造及自动化]