摘要
针对用传统的故障诊断方法难以对非线性非平稳的柴油机故障信号进行准确高效诊断的问题,提出基于小波时频图与Swin Transformer的柴油机故障诊断方法。该方法可以有效结合小波时频分析在处理非线性非平稳信号方面的优势和Swin Transformer强大的图像分类能力,通过连续小波变换将原始信号表示为小波时频图,将小波时频图作为特征图输入到Swin Transformer进行训练,实现柴油机故障状态识别。实验结果表明,与对比方法相比,所提方法具有较好的故障识别精度及稳定性,在公开数据集和实验室实测数据中的整体故障诊断准确率分别达到100%和98.88%,为柴油机故障诊断提供了一种新的思路。
In view of the problem that it is difficult to accurately and efficiently diagnosis the nonlinear and non-smooth diesel engine fault signals by traditional fault diagnosis methods,a diesel engine fault diagnosis method based on wavelet time-frequency diagram and Swin Transformer is proposed.The method can effectively combine the advantages of wavelet time-frequency analysis in processing nonlinear non-stationary signals and the powerful image classification capability of Swin Transformer,represent the original signals as wavelet time-frequency diagrams through continuous wavelet transform,and input the wavelet time-frequency diagrams as feature maps to Swin Transformer for training to achieve diesel engine fault status identification.The experimental results show that the proposed method has better fault identification accuracy and stability compared with the comparison methods,and the overall fault diagnosis accuracy reaches 100%and 98.88%in the public data set and the laboratory measured data respectively,which provides a new idea for diesel engine fault diagnosis.
作者
刘子昌
白永生
李思雨
贾希胜
LIU Zichang;BAI Yongsheng;LI Siyu;JIA Xisheng(Shijiazhuang Campus,Army Engineering University of PLA,Shijiazhuang 050003,China;Hebei Provincial Key Lab of Condition Monitoring and Assessment of Mechanical Equipment,Shijiazhuang 050003,China)
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2023年第9期2986-2998,共13页
Systems Engineering and Electronics
基金
国家自然科学基金面上项目(71871220)资助课题。
