考虑自愈现象的轴承多阶段退化剩余寿命预测  

Remaining Useful Life Prediction of Bearing with Multi-stage Degradation Considering Healing Phenomenon

作　　者：卢锦枫 吴太欢 罗华耿 LU Jinfeng;WU Taihuan;LUO Huageng(School of Aerospace Engineering,Xiamen University Xiamen,361102,China)

机构地区：[1]厦门大学航空航天学院,厦门361102

出　　处：《振动.测试与诊断》2025年第2期323-330,413,414,共10页Journal of Vibration,Measurement & Diagnosis

摘　　要：轴承服役过程存在“自愈”等非线性退化现象且缺乏训练寿命标签,限制了智能轴承寿命预测方法在实际工程中的应用。针对此问题,提出一种多阶段退化标签构建(multi-stage degradation label construction,简称MDLC)方法。首先,运用深度自编码网络与高斯分布的自适应3σ法则,根据振动信号统计特征识别轴承的初始退化点;其次,利用自下而上分割算法,基于均方根特征值曲线划分轴承退化阶段并分段拟合,构建多阶段退化剩余寿命标签;然后,搭建长短时记忆人工神经网络的寿命预测模型,以有监督的方式训练并优化该模型;最后,利用XJTU-SY滚动轴承加速寿命试验数据集测试所提出的方法,并与经典方法进行了对比。结果表明,该方法不仅能够准确识别轴承初始退化点,且剩余寿命预测误差更小,验证了其有效性与准确性。The existence of nonlinear degradation phenomena such as "healing" and the lack of training life labels in the bearing service process limit the application of intelligent bearing life prediction methods in practical engineering.In this paper,we propose a multi-stage degradation label construction(MDLC) method.Firstly,the first degradation point of the bearing is identified based on the statistical features of vibration signals by using deep autoencoder network and adaptive 3σ law of Gaussian distribution.Secondly,a bottom-up segmentation algorithm is used for dividing the bearing degradation stages based on the root-mean-square eigenvalue curve and fitting them in segments to construct a multi-stage degradation remaining life label.Finally,the life prediction model with long short-term memory artificial neural network is built,and the model is trained and optimized in a supervised manner.The new method is tested with XJTU-SY rolling bearing accelerated life experimental data set and compared with the classical method.The results show that the new method can not only accurately identify the initial degradation point of the bearing,but also decrease the remaining life prediction error,which verifies the effectiveness and accuracy of the new method.

关 键 词：滚动轴承 剩余寿命预测 自愈现象 多阶段退化 

分 类 号：TH17[机械工程—机械制造及自动化] TH165.3[自动化与计算机技术—控制理论与控制工程] TP183[自动化与计算机技术—控制科学与工程]