首页|基于状态划分和集成学习的轴承剩余使用寿命预测模型

基于状态划分和集成学习的轴承剩余使用寿命预测模型

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
针对滚动轴承剩余使用寿命(RUL)预测退化起始时间(DST)难以确定,以及单一寿命预测模型精度比较低的问题,提出了一种基于状态划分和集成学习模型的滚动轴承 RUL预测方法.首先,提取了轴承振动信号的特征,利用滑动窗口不断更新 3σ 准则预警范围,结合连续触发机制自适应确定 DST;然后,采用具有自适应噪声的完全集成经验模态分解(CEEMDAN)对退化阶段信号序列进行了自适应分解;最后,构建了集成学习模型,考虑分量的不同特性进行了多步滚动预测,融合预测结果得到了轴承 RUL,采用滚动轴承 XJTU-SY 公开数据集进行了试验验证.研究结果表明:与基于长短时记忆神经网络(LSTM)、反向传播神经网络(BPNN)的预测方法相比,该方法预测结果的平均绝对误差分别降低了 11.7％以及 5.6％,相对均方根误差分别降低了 12.2％以及10.7％,验证了该方法在轴承 RUL预测中的有效性和优越性.
Remaining useful life prediction model of bearing based on state division and ensemble learning
To solve the problems of the difficulty in determining the remaining useful life(RUL)of bearings and the low accuracy of a single life prediction model to predict degradation start time(DST),a RUL prediction method based on state division and ensemble learning models were proposed.First,the DST was determined adaptively by extracting the bearing vibration signal characteristics,constantly updating the 3σ criterion warning range using a sliding window and combining a continuous triggering mechanism.Then,complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)was used to adaptively decompose the signal sequence in the degradation phase.Finally,an ensemble learning model was built to perform multi-stage rolling prediction considering different component characteristics,the prediction results were merged to obtain the bearing RUL,and the public bearing dataset XJTU-SY was used for experimental verification.The research results show that the mean absolute error of the prediction results of the proposed method is respectively reduced by11.7％and 5.6％,and the relative mean square error is respectively decreased by 12.2％and 10.7％,comparing with the prediction methods based on long short-term memory neural network(LSTM)and back-propagation neural network(BPNN).The validity and superiority of the proposed method in the application of bearing RUL prediction is verified.

rolling bearing remaining useful life(RUL)degradation start time(DST)adaptive DST state divisionensemble learning modeldegenerate feature extractioncomplete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)long short-term memory neur

胡志辉、王绪光、王贡献、张腾、李帅琦

展开 >

武汉理工大学 交通与物流工程学院,湖北 武汉 430063

滚动轴承剩余使用寿命 退化起始时间 自适应DST状态划分 集成学习模型 退化特征提取 具有自适应噪声的完全集成经验模态分解 长短时记忆神经网络

国家科技重大专项

2022ZD0119304

2024

10.3969/j.issn.1001-4551.2024.08.011

机电工程
浙江大学　浙江省机电集团有限公司

机电工程

CSTPCD北大核心
影响因子：0.785
ISSN：1001-4551
年,卷(期)：2024.41(8)