基于GAF与CNN的齿轮箱故障诊断  

Fault Diagnosis of Gearbox Based on Gramian Angular Filed and Convolutional Neural Network

作　　者：项岱军 王煜伟 凌峰[2] 许猛 邓艾东[2] Xiang Daijun;Wang Yiwei;Ling Feng;Xu Meng;Deng Aidong(CHN Energy Jiangsu Power Co.,Ltd.,Nanjing 215433,China;National Engineering Research Center of Power Generation Control and Safety,Southeast University,Nanjing 210096,China)

机构地区：[1]国家能源集团江苏电力有限公司,南京215433 [2]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

出　　处：《信息化研究》2022年第4期51-57,共7页INFORMATIZATION RESEARCH

摘　　要：文章针对传统的旋转机械故障诊断方法过于依赖专家知识,未能充分挖掘运行数据中的隐含特征,且在强噪声下无法保证故障诊断效果的问题,提出了一种基于特征图像与卷积神经网络(CNN)的齿轮箱故障诊断模型。首先对齿轮箱信号进行Hilbert变换获取信号的包络谱,然后采用格拉姆角场图像编码将齿轮箱信号的包络谱序列转化为特征图,最后通过CNN进行故障识别。实验结果表明,齿轮箱信号经过格拉姆角场转化后更能显示出其隐含特征,具有良好的诊断效果,能够满足强噪声下的故障诊断需求。Since the traditional intelligent fault diagnosis methods for rotating machinery rely too much on expert knowledge, they fail to fully exploit the hidden features of signals and do not achieve satisfactory fault diagnosis under strong noise. In order to solve the above problems, this paper proposes a gearbox fault diagnosis model based on feature images and convolutional neural networks(CNN for short). Firstly, the Hilbert transform is applied to the gearbox signal to obtain the envelope spectrum of the signal;then the envelope spectrum sequence of the gearbox signal is transformed into the feature map by using Gram’s angular field image coding;finally, the fault identification is performed by CNN. The experimental results show that the gearbox signal can show its hidden features more after the Gram’s corner field transformation, which can meet the fault diagnosis demand under strong noise.

关 键 词：齿轮箱 故障诊断 格拉姆角场 卷积神经网络 

分 类 号：TH133.33[机械工程—机械制造及自动化]