摘要
基于学习模型的轴承剩余使用寿命(RUL)预测方法通常假设训练和测试数据具有相同的数据分布,为解决现有方法在不同工况或不同轴承RUL预测精度大幅下降的问题,提出一种基于对抗域自适应的轴承RUL预测方法。首先进行轴承健康阶段划分,使用等渗回归对振动数据进行预处理,平滑退化信号中的随机波动,再通过测量滑动窗口内的退化梯度进行健康阶段识别,表征退化趋势并识别跳跃点,从而划分健康阶段;在此基础上,选择源域和目标域的轴承退化阶段的振动数据作为模型输入,使用源域数据预训练特征提取器和寿命预测模块;然后设计域判别器网络对抗性地训练特征提取器,以最小化源域特征与目标域特征之间的分布差异;最后使用更新参数的目标特征提取器提取目标域的特征并进行RUL预测。使用IEEE PHM Challenge 2012轴承数据集验证了本文方法的有效性,与现有模型的对比试验表明本文方法在实现不同工况下轴承RUL预测迁移问题上表现更好。
The prediction method for remaining useful life(RUL) of bearings based on learning model is usually assumed that training and test data have same data distribution, In order to solve the problem that the RUL prediction accuracy of existing methods decreases greatly under different working conditions or different bearings, an prediction method for RUL of bearings based on adversarial domain adaptation is proposed. Firstly, the health stage of bearings is divided, and the vibration data are preprocessed by isotonic regression to smooth the random fluctuations in degraded signals. The health stage is identified by measuring the degradation gradient in sliding window to characterize the degradation trend and identify the jump points, so as to recognize the health stages. On this basis, the vibration data of bearing degradation stage in source domain and target domain are selected as model input, and the feature extractor and life prediction module are pre-trained with source domain data. Then, the domain discriminator network is designed to train the feature extractor adversarially to minimize the distribution difference between source domain feature and target domain feature. Finally, the target feature extractor with updated parameters is used to extract the feature of target domainand predict the RUL. The IEEE PHM Challenge 2012 bearing data set is used to verify the effectiveness of proposed method. The comparison test with existing model shows that the proposed method has better performance in predicting the migration of RUL of bearings under different working conditions.
作者
徐娟
蒋瑞
陈为伟
王东峰
郑昊天
XU Juan;JIANG Rui;CHEN Weiwei;WANG Dongfeng;ZHENG Haotian(School of Computer Science and Information Engineering,Hefei University of Technology,Hefei 230009,China;School of Mechanical Engineering,Hefei University of Technology,Hefei 230009,China;Shanghai Aerospace Control Technology Institute,Shanghai 201109,China;Luoyang Bearing Science and Technology Co.,Ltd.,Luoyang 471003,China;Luoyang Bearing Research Institute Co.,Ltd.,Luoyang 471039,China)
出处
《轴承》
北大核心
2023年第2期113-120,共8页
Bearing
基金
国家重点研发计划资助项目(2018YFB2000505)
安徽省重点研发计划资助项目(202104a04020003)。
关键词
滚动轴承
自适应
剩余使用寿命
退化
寿命预测
振动
rolling bearing
adaptation
remaining useful life
degradation
life prediction
vibration
