基于扩展空间森林算子和自适应卡尔曼滤波的轴承剩余使用寿命预测  

Prediction of remaining useful life of bearings based on extended space forest operator and adaptive Kalman filter

作　　者：张溧栗 温志鹏 曹菁菁[2] 韩鹏[2] 赵强伟 曹小华[2] ZHANG Lili;WEN Zhipeng;CAO Jingjing;HAN Peng;ZHAO Qiangwei;CAO Xiaohua(Huadian Heavy Industries Co.,Ltd.,Shanghai 200126,China;School of Transportation and Logistics Engineering,Wuhan University of Technology,Wuhan 430063,China)

机构地区：[1]华电重工股份有限公司,上海200126 [2]武汉理工大学交通与物流工程学院,湖北武汉430063

出　　处：《机电工程》2025年第3期420-431,共12页Journal of Mechanical & Electrical Engineering

基　　金：湖北省重点研发项目(2023BEB046)。

摘　　要：针对传统单一轴承退化指标所含信息不足,有限数据样本条件下轴承剩余寿命(RUL)难以预测等问题,提出了一种基于扩展空间森林算子和自适应卡尔曼滤波轴承剩余使用寿命(ESF-AKF)的预测方法。首先,提取了原始轴承振动数据的均方根值和整流平均值两个退化指标;然后,根据扩展空间森林算子对两个退化指标进行了特征扩展,提出了新的动态退化评估准则,选择了两个新的退化指标融合构建综合轴承退化指标;接着,设计了自适应卡尔曼滤波预测模型以估计维纳过程的未知参数,即引入自适应因子代入先验误差协方差矩阵,实时调整了滤波的卡尔曼增益;最后,采用IEEE PHM2012公开数据集进行了轴承剩余寿命预测验证。研究结果表明:与两个传统退化指标相比,基于提出的综合轴承退化指标的预测结果平均误差分别降低了5.69%和21.10%;与卡尔曼滤波和粒子滤波相比,基于自适应卡尔曼滤波的平均误差分别降低了45.41%和10.92%;与其他模型相比,平均均方根误差、平均绝对误差分别降低了48.56%、29.11%。该研究结果验证了该轴承使用寿命预测方法的准确性和有效性。To address the issues of insufficient information in traditional single-bearing degradation indicators and the challenge of predicting remaining useful life(RUL)under limited data sample conditions,a prediction method for the remaining service life of bearings based on extended space forest operator and adaptive Kalman filter(ESF-AKF)was proposed.Firstly,two degradation indicators,the root mean square value and absolute value of the rectified signal average,were extracted from the original bearing vibration data.Secondly,feature extension was performed on these two indicators using the extended space forest operator,and a new dynamic degradation evaluation criterion was introduced.The two new degradation indicators were fused to construct a comprehensive bearing degradation indicator.Then,an adaptive Kalman filter prediction model was designed to estimate the unknown parameters of the Wiener process,by introducing adaptive factors into the prior error covariance matrix and adjusting the Kalman gain of the filter in real time.Finally,the IEEE PHM2012 public dataset was used to verify the RUL prediction model for bearings.The experimental results show that,comparing with the two traditional degradation indicators,the average error of the RUL prediction based on the proposed comprehensive bearing degradation indicator respectively reduces by 5.69%and 21.10%.Comparing to Kalman filtering and particle filtering,the average error of the adaptive Kalman filtering model respectively reduces by 45.41%and 10.92%.Comparing to other models,the average root mean square error and mean absolute error respectively reduces by 48.56%and 29.11%.The experimental results verify the accuracy and effectiveness of the method in predicting the RUL of bearings.

关 键 词：滚动轴承 剩余使用寿命 扩展空间森林 维纳过程 卡尔曼滤波 预测与健康管理 

分 类 号：TH133.3[机械工程—机械制造及自动化]