摘要
针对复杂工况下滚动轴承振动信号特征提取困难及诊断精度易受噪声干扰等问题,提出一种白鲸(BWO)优化变分模态分解(VMD)算法和CNN-Transformer相结合的故障诊断方法。采用BWO对VMD中的模态个数K及惩罚因子α进行优化,并对原信号进行分解和提取最佳模态分量(IMFs);构建CNN-Transformer双通道特征模型,通过卷积神经网络(CNN)挖掘信号的局部时频特征,并结合Transformer捕捉全局时序依赖关系,实现对多尺度故障特征的有效表征;通过全连接层与Softmax分类器实现精确的故障识别。基于CWRU轴承数据集,将该模型与其他模型相比,其故障识别率达到99.1%以上,为滚动轴承故障诊断提供了具有创新性和可靠性的技术支持。
Addressing challenges in extracting vibration signal features and maintaining diagnostic accuracy under complex operating conditions for rolling bearings,this study proposes a fault diagnosis method combining the Beluga Whale Optimization(BWO)-optimized Variational Modal Decomposition(VMD)algorithm with a CNN-Transformer dual-channel feature model.First,BWO is employed to optimize the number of modes K and penalty factor α in VMD,decomposing the original signal and extracting optimal mode components(IMFs).Second,a dual-channel feature model combining CNN and Transformer is constructed:CNN extracts local spatiotemporal features,while Transformer captures global temporal dependencies,enabling effective representation of multiscale fault characteristics.Finally,a fully connected layer and Softmax classifier enable precise fault recognition.Evaluated on the CWRU bearing dataset,the proposed model achieves a fault recognition rate exceeding 99.1% compared to other models,providing innovative and reliable technical support for rolling bearing fault diagnosis.
作者
李朝阳
李晓勤
孙宇乔
李昊
刘书梅
Li Zhaoyang;Li Xiaoqin;Sun Yuqiao;Li Hao;Liu Shumei(School of Mechanical and Electrical Engineering,Tarim University,Alar Xinjiang 843300,China;Alar Wanda Agricultural Machinery Co.,Ltd.,Alar Xinjiang 843300,China)
出处
《机械管理开发》
2025年第12期79-82,86,共5页
Mechanical Management and Development
基金
国家自然科学基金项目(NO.51565052)
校长基金技术开发项目(TDZKJS2022003)。
