基于数据融合驱动和DLSTM网络的轴承RUL预测  

BEARING RUL PREDICTION BASED ON DATA FUSION DRIVE AND DLSTM NETWORK

作　　者：段桂英[1,2] 姜洪开 Duan Guiying;Jiang Hongkai(Ministry of Education of Public Courses,Shandong Art College,Jinan 250014,Shandong,China;School of Aeronautics,Northwestern Polytechnical University,Xi’an 710072,Shaanxi,China)

机构地区：[1]山东艺术学院公共课教育部,山东济南250014 [2]西北工业大学航空学院,陕西西安710072

出　　处：《计算机应用与软件》2021年第12期22-29,共8页Computer Applications and Software

基　　金：国家自然科学基金项目(51475368)。

摘　　要：针对滚动轴承的剩余寿命预测问题,提出一种基于多传感器信号融合的深度长短期记忆网络预测模型。利用深度学习和长短期记忆网络组合来构造深度长短期记忆网络;将多个传感器信号数据进行融合处理,从而通过深度学习结构能够发现传感器时序信号中隐藏的长期依赖关系;通过网格搜索策略、自适应矩估计算法(Adaptive Moment Estimation Algorithm,AMEA)优化深度长短期记忆网络的网络结构和参数,并且引入一种主动丢弃法以缓解过度拟合问题。实验结果表明该方法具有更高的预测准确性和稳定性。Aiming at the problem of residual life prediction of rolling bearing,a deep long short term memory(DLSTM)network prediction model based on multiple sensor signal fusion is proposed.The combination of deep learning(DL)and DLSTM networks were used to construct deep long short-term memory networks;the multiple sensor signal data were fused,so that the hidden long-term dependence could be found through deep learning structure;the network structure and parameters were optimized by grid search strategy and adaptive moment estimation algorithm(AMEA),and an active loss method was introduced to alleviate the over fitting problem.The experimental results show that the proposed method has higher prediction accuracy and stability.

关 键 词：滚动轴承 深度学习 长短期记忆网络 寿命预测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] V263.6[航空宇航科学与技术—航空宇航制造工程]