基于特征选择与Transformer-LSTM的滚动轴承寿命预测  

Rolling Bearing Life Prediction Based on Feature Selection and Transformer-LSTM

作　　者：李沁远 雷文平[1] 闫灏 娄永威 陈阳[1] LI Qinyuan;LEI Wenping;YAN Hao;LOU Yongwei;CHEN Yang(School of Mechanical and Power Engineering,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学机械与动力工程学院,郑州450001

出　　处：《组合机床与自动化加工技术》2025年第2期200-206,211,共8页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(51775515)。

摘　　要：滚动轴承作为旋转机械设备中的关键部件,影响着设备的可靠性运行。针对以往剩余使用寿命(RUL)预测方法对轴承退化信息挖掘不充分、忽视不同特征贡献度和不同特征组合对预测模型精度的影响,提出一种基于特征选择与Transformer-LSTM的剩余使用寿命预测模型。首先基于单调性、趋势性以及最大相关最小冗余特征选择算法对振动信号的时域、频域、时频域特征进行重要性排序和筛选,从而捕获特征与剩余寿命以及特征之间的相互的关系。然后将筛选后的特征输入Transformer-LSTM预测模型中,深度挖掘输入特征与RUL之间的复杂映射关系,从而更准确地进行预测。通过公开的轴承数据集进行实验验证,与其他RUL预测方法相比,所提方法的预测性能更优越。Rolling bearings,as key components in rotating machinery,have a significant impact on the reliability of equipment operation.In view of the insufficient exploration of bearing degradation information and the neglect of the influence of different feature contributions and combinations on the accuracy of previous methods for residual life prediction(RUL),this paper proposes a residual life prediction model based on feature selection and Transformer-LSTM.The monotonicity,trend,and maximum correlation minimum redundancy feature selection algorithm are used to rank and select the importance of time-domain,frequency-domain,and time-frequency domain features of vibration signals,this enables capturing the mutual relationship between features and residual life,as well as the relationship among features.The selected features are then input into the Transformer-LSTM prediction model to deeply explore the complex mapping relationship between input features and RUL,thus achieving more accurate predictions.Experimental verification is conducted using publicly available bearing datasets,and the proposed method demonstrates superior prediction performance compared to other RUL prediction methods.

关 键 词：剩余使用寿命 特征选择 最大相关最小冗余 Transformer-LSTM模型 

分 类 号：TH133.3[机械工程—机械制造及自动化] TG66[金属学及工艺—金属切削加工及机床]