| 引用本文: | 唐伯宇,邵星,王翠香,皋军.基于双通道CNN与SSA-SVM的滚动轴承故障诊断[J].控制理论与应用,2024,41(9):1626~1635.[点击复制] |
| TANG Bo-yu,SHAO Xing,WANG Cui-xiang,GAO Jun.Rolling bearing fault diagnosis based on dual channel CNN and SSA-SVM[J].Control Theory and Technology,2024,41(9):1626~1635.[点击复制] |
|
| 基于双通道CNN与SSA-SVM的滚动轴承故障诊断 |
| Rolling bearing fault diagnosis based on dual channel CNN and SSA-SVM |
| 摘要点击 3089 全文点击 48 投稿时间：2022-05-19 修订日期：2024-05-09 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2023.20413 |
| 2024,41(9):1626-1635 |
| 中文关键词 故障诊断 滚动轴承 卷积神经网络 支持向量机 |
| 英文关键词 fault diagnosis rolling bearing convolutional neural network support vector machine |
| 基金项目 国家自然科学基金项目(61502411, 62076215), 中国高校产学研创新基金“新一代信息技术创新项目”(2020ITA02057), 盐城工学院研究生科研与 实践创新计划项目(SJCX22 XZ035, SJCX22 XY061)资助. |
|
| 中文摘要 |
| 为有效提取滚动轴承故障信号特征, 解决分类器对提取特征存在较强依赖的问题, 本文提出了一种双通道特征融合卷积神经网络(CNN)与麻雀搜索算法优化的支持向量机(SSA-SVM)相结合的滚轴承故障诊断方法. 通过建立一维卷积神经网络和二维卷积神经网络并行的双通道结构对轴承数据中的特征进行提取, 并将双通道CNN提取到的故障特征在融合层融合, 将全连接层结果作为SSA-SVM分类层的输入. 通过SSA对SVM的参数进行优化来提高模型的分类精度. 最后, 在凯斯西储大学轴承数据集上将双通道CNN与SSA-SVM模型跟传统一维卷积神经网络和二维卷积神经网络进行对比以验证其有效性. 实验结果表明, 该模型有着更高的故障识别准确率. 本文中所有代码与实验结果均已开源, 可在https://github.com/suisuisuiaa/tbysuisui获取. |
| 英文摘要 |
| In order to effectively extract the features of rolling bearing fault signals and solve the problem that classifiers have strong dependence on the extracted features, this paper proposes a rolling bearing fault diagnosis method combing dual-channel feature fusion convolutional neural network (CNN) and support vector machine optimized by sparrow search algorithm (SSA-SVM). The features of bearing data are extracted by establishing a parallel dual channel structure of onedimensional CNN and two-dimensional CNN. Then the fault features extracted by dual channel structure are fused in the fusion layer, and the results of full connection layer are used as the input of SSA-SVM. In addition, SSA is exploited to optimize the parameters of SVM to improve the classification accuracy. Finally, the proposed method is compared with the traditional one-dimensional CNN and two-dimensional CNN using Case Western Reserve University bearing data set to verify its effectiveness. Experimental results demonstrate that the proposed method has higher fault identification accuracy. All code and experimental results of the method have been open source and available at https://github.com/suisuisuiaa/tbysuisui. |
|
|
|
|
|