基于特征融合和改进PSO的轴承剩余使用寿命预测  

Bearing remaining useful life prediction based on feature fusion and improved PSO

作　　者：杨洵 文传博(指导)[1] YANG Xun;WEN Chuanbo(School of Electrical Engineering,Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电机学院电气学院,上海201306

出　　处：《上海电机学院学报》2025年第1期27-32,共6页Journal of Shanghai Dianji University

摘　　要：针对深度学习方法进行轴承剩余使用寿命(RUL)预测时存在精度不够的问题,提出了一种基于改进GRU的特征融合以及改进PSO算法的轴承RUL预测方法。首先,利用CEEMDAN提取信号中的退化信息,筛选出有价值的特征;然后,结合GRU以及点积注意力机制完成特征融合的任务;最后,利用Logistic映射来自适应调整PSO的权重,来完成RUL预测。通过在PHM2012数据集上进行实验,验证本文所提方法具有良好的RUL预测精度。There are challenges in achieving sufficient accuracy when predicting the remaining useful life(RUL)of bearings using deep learning methods.To address this issue,a bearing RUL prediction method is proposed,which combines an improved gated recurrent unit(GRU)for feature fusion and an enhanced particle swarm optimization(PSO)algorithm.Firstly,the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)method is employed to extract degradation information from the signal and select valuable features.Then,the GRU network,in conjunction with a dot-product attention mechanism,is used to perform feature fusion.Finally,logistic mapping is applied to adaptively adjust the weights of the PSO algorithm to improve the RUL prediction.Experimental results on the PHM2012 dataset demonstrate that the proposed method achieves high RUL prediction accuracy.

关 键 词：滚动轴承 寿命预测 粒子群算法 门控循环单元 

分 类 号：TH133.3[机械工程—机械制造及自动化]