摘要
提出了一种基于SL(S-L2UWE)变换和SA-ConvNeXt模型相结合的故障诊断方法,以解决永磁同步直线电机导轨磨损、滚珠磨损和滚珠缺失等多类型故障精准识别诊断的问题。首先对采集的振动信号进行S变换处理,将一维时域信号转化成二维时频图像;利用L2UWE算法对S变换后的图像进行增强,实现故障时频图像颜色和形状的细节特征增强显示;将自注意力模块(self-attention,SA)引入到ConvNeXt模型中,从特征通道层面上统计图像的全局信息,使网络关注重点特征,提高模型分类精度。最终,基于SL变换和SA-ConvNeXt深度学习模型对于直线导轨故障的识别准确率高达93.8%,能准确识别导轨和滚珠磨损、缺失等故障。对比试验和鲁棒性试验,验证了本文所提方法的有效性和优越性。
Here,a fault diagnosis method based on combing S-L 2UWE(SL)transform and SA-ConvNeXt model was proposed to solve problems of accurate identification and diagnosis for multi-type faults of guide rail wear,ball wear and ball missing in permanent magnet synchronous linear motor.Firstly,SL transform was performed for the collected vibration signals to convert 1-D time-domain signals into 2-D time-frequency images.L 2UWE algorithm was used to enhance the S-transformed images and realize detailed feature enhancement display of color and shape for fault time-frequency images.The self-attention(SA)module was introduced into ConvNeXt model to statistically analyze the global information of images at the feature channel level,and make the network focus on key features and improve classification accuracy of the model.Finally,the effectiveness and superiority of the proposed method were verified through comparative experiments and robustness experiments.It was shown that based on SL transform and SA-ConvNeXt deep learning model,the recognition accuracy of linear guide rail faults can reach as high as 93.8%,so the proposed method can accurately identify faults of guide rail wear,ball wear and ball missing.
作者
吴艳萍
陆思良
宋俊材
王骁贤
吴先红
丁伟
WU Yanping;LU Siliang;SONG Juncai;WANG Xiaoxian;WU Xianhong;DING Wei(College of Electrical Engineering and Automation,Anhui University,Hefei 230601,China;College of Internet,Anhui University,Hefei 230039,China;College of Electronic and Information Engineering,Anhui University,Hefei 230601,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2024年第19期28-36,共9页
Journal of Vibration and Shock
基金
国家自然科学基金(52207036)
安徽省自然科学基金青年项目(2208085QE167)
安徽省教育厅自然科学重点项目(KJ2021A0018)。
