基于组稀疏学习与AVOA-XGBoost的轴承故障分级诊断

Bearing fault classification diagnosis based on group-sparsity learning and AVOA-XGBoost

针对工业设备中轴承振动信号在噪声环境下故障分级诊断准确率低的问题,提出一种基于组稀疏学习与非洲秃鹫优化算法优化极端梯度提升树(African vultures optimization algorithm-extreme gradient boosting,AVOA-XGBoost)的轴承故障分级诊断方法。首先,利用组稀疏学习对轴承振动信号进行重构,以降低噪声水平并更有效地表征故障脉冲。然后,对重构后的信号提取时域、频域和熵值特征并构建特征集。最后,利用AVOA自适应优化XGBoost超参数以建立稳健的XGBoost模型,进而高效实现轴承故障分级诊断。试验结果表明,经过组稀疏学习重构的信号具备更强故障特征表示能力,相较于传统机器学习模型,采用AVOA-XGBoost模型进行分类能够取得更高准确率,所提方法能够有效诊断轴承故障类型及故障程度。

In response to the challenge of low accuracy in bearing fault classification under strong background noises in industrial equipments,a bearing fault classification method based on group-sparsity learning and African vultures optimization algorithm-extreme gradient boosting(AVOA-XGBoost)was proposed.First,the bearing vibration signals were reconstructed using group-sparsity representation,which reduces the noise level and characterizes fault impulses more effectively.Then,time-domain,frequency-domain,and entropy features were extracted from the reconstructed signals and a feature set was constructed.Finally,the super parameters of XGBoost were adaptively adjusted by AVOA,which establishes a robust XGBoost for efficient bearing fault classification diagnosis.The experimental results demonstrate that the signals reconstructed by group-sparsity learning exhibit stronger fault characteristic representation,and AVOA-XGBoost achieves higher classification accuracy compared with traditional machine learning models.The proposed method can effectively diagnose the types and degrees of bearing faults.