基于N-BEATS神经网络的滚动轴承剩余寿命预测  

Prediction of Residual Life of Rolling Bearing Based on N-BEATS Neural Network

作　　者：时培明[1] 苏世敏 马慧中 许学方 韩东颖[2] SHI Pei-ming;SU Shi-min;MA Hui-zhong;XU Xue-fang;HAN Dong-ying(Key Laboratory of Measurement Technology and Instrument of Hebei Province,Yanshan University,Qinhuangdao,Hebei 066004,China;School of Vehicles and Energy,Yanshan University,Qinhuangdao,Hebei 066004,China)

机构地区：[1]燕山大学河北省测试计量技术及仪器重点实验室,河北秦皇岛066004 [2]燕山大学车辆与能源学院,河北秦皇岛066004

出　　处：《计量学报》2023年第8期1240-1247,共8页Acta Metrologica Sinica

基　　金：国家自然科学基金(61973262);中央引导地方科技发展资金(216Z4301G,216Z2102G);河北省自然科学基金(E2020203147)。

摘　　要：针对轴承剩余使用寿命预测时采集到的信号序列复杂且不平稳,导致预测精度和性能较低的问题,采用经验模式分解对信号序列进行自适应分解,利用动态时间规整算法筛选主要退化特征,提取信号序列趋势特征,使用具有残差原理的深层神经网络N-BEATS进行预测。对于预测历史数据较少的问题,采用递归与直接相结合的预测结构对剩余寿命进行多步预测。将N-BEATS与长短时记忆神经网络、灰色预测模型进行比较,结果表明在不同工况下,所提出的方法预测结果的平均绝对误差相较于LSTM、灰色预测模型分别提升了3.2%以及3.3%,相对均方根误差分别提升了3.5%以及3.1%。In order to effectively predict the remaining service life of the bearing,and to solve the problem of complex and unstable signal sequences collected during the prediction,resulting in low prediction accuracy and performance,the empirical mode decomposition is used to adaptively decompose the signal sequence,the dynamic time warping algorithm is used to screen the main degradation features,and the trend features of the signal sequence are extracted.The deep neural network N-BEATS with residual principle is used for prediction.For the problem of less prediction history data,the prediction structure combining recursion and direct is used to predict the remaining life in multiple steps.Comparing N-BEATS with long-term and short-term memory neural network and grey prediction model,the results show that the average absolute error of the prediction results of the proposed method is increased by 3.2% and 3.3% respectively compared with LSTM and grey prediction model under different working conditions,the relative root mean square error increased by 3.5% and 3.1% respectively.

关 键 词：计量学 滚动轴承 剩余寿命预测 动态时间规整(DTW) N-BEAT神经网络 

分 类 号：TB936[一般工业技术—计量学] TB973[机械工程—测试计量技术及仪器]