Page 135 - 《振动工程学报》2025年第8期
P. 135

第 38 卷第 8 期                       振  动  工  程  学  报                                  Vol. 38 No. 8
               2025 年 8 月                      Journal of Vibration Engineering                       Aug. 2025


                          不均衡样本下轴承故障的 LSGAN⁃Swin


                                          Transformer 诊断方法



                                          刘     杰, 谭玉涛, 谷艳玲, 杨                 娜

                                           (沈阳工业大学机械工程学院,辽宁 沈阳 110870)


              摘要: 针对轴承在复杂环境下工作时故障数据难以大量获取,正常数据与故障数据比例严重失衡造成的深度模型训练不充分、
              诊断精度低等问题,提出一种基于 LSGAN‑Swin Transformer 的轴承故障诊断方法,利用最小二乘生成对抗网络(LSGAN)扩
              充不均衡或缺少的轴承数据集,引入窗口自注意力网络进行轴承故障状态识别,使用两种数据集验证所提方法的有效性,并分
              别与 SGAN、WGAN 进行对比,证明 LSGAN 生成的数据训练模型具有更高的准确率。在小样本条件下训练 LSGAN,将所提
              Swin  Transformer(Swin‑T)模 型 与 CNN、AlexNe 和 SqueezeNet 进 行 对 比 ,诊 断 准 确 率 分 别 提 升 了 34.85%、13.45% 和
              12.95%。通过 t‑SNE 可视化分析对模型分类效果进行评估,结果表明,LSGAN‑Swin‑T 模型在训练样本数量较少时仍能较好
              地满足故障诊断中的需求,为不均衡数据下的轴承故障诊断研究提供思路。

              关键词: 故障诊断; 滚动轴承; 不均衡样本; 最小二乘生成对抗网络; Swin Transformer
                             +
              中图分类号: TH165 .3; TH133.33    文献标志码: A    DOI:10.16385/j.cnki.issn.1004-4523.202308023

                         LSGAN-Swin Transformer diagnosis method of bearing fault

                                              under unbalanced samples


                                         LIU Jie, TAN Yutao, GU Yanling, YANG Na
                        (School of Mechanical Engineering, Shenyang University of Technology, Shenyang 110870, China)

              Abstract: Aiming at the problems of bearings working in complex environments, where fault data are difficult to obtain in large
              quantities and the serious imbalance between the ratio of normal data and fault data resulting in insufficient in-depth model training
              and low diagnostic accuracy, a bearing fault diagnosis method based on LSGAN-Swin Transformer is proposed. The least-squares
              generative adversarial network is utilized to expand the imbalanced or lack of bearing dataset, and the windowed self-attentive net‑
              work is introduced for bearing fault state identification.  The proposed method is validated by using two date sets, and compared
              with SGAN and WGAN respectively. It is demonstrated that LSGAN generates data training models with higher accuracy. The
              proposed Swin Transformer (Swin-T) model is compared with CNN, AlexNet and SqueezeNet under small sample conditions,
              and the accuracy is improved by 34.85%, 13.45%, and 12.95%, respectively. The classification effect of the model is evaluated
              by t-SNE visualization, and the results show that the LSGAN-Swin-T model can still meet the requirements in fault diagnosis bet‑
              ter when the number of training samples is small, which provides a new idea for the research of bearing fault diagnosis under unbal‑
              anced data.
              Keywords: fault  diagnosis;  rolling  bearings;  unbalanced  sample;  least  square  generative  adversarial  network;  shifted  windows
                       transformer (Swin Transformer)


                  滚动轴承在各类机械设备中具有重要作用,在实                         外,轴承特殊的工作位置容易造成其多种故障类型数
              际工程应用中,由于设备运行数据监测过程的工作                            据采集不均匀,进而造成模型训练不充分,严重影响
              环境复杂,信号一般呈现非平稳及非线性的特点,且                           其诊断性能,难以达到故障诊断模型的训练要求。
              部分状态数据难以大量采集,正常数据样本通常远                                 为解决轴承故障诊断任务中标记样本有限的
              远大于故障数据样本,数据之间比例严重失衡。另                            问 题 ,通 常 应 用 数 据 增 强 方 法(data augmentation


                  收稿日期: 2023-08-11; 修订日期: 2023-11-06
                  基金项目: 辽宁省教育厅面上项目(LQGD2020016)
   130   131   132   133   134   135   136   137   138   139   140