一种多阶段机械装备剩余寿命预测新方法及应用  

A Novel Multi-stage Approach and Application of Remaining Life Prediction for Mechanical Equipment

作　　者：吕明珠[1,2] LüMingzhu(School of Automatic Control,Liaoning Equipment Manufacturing Vocational and Technology College,Shenyang 110161,China;Liaoning Open University,Shenyang 110034,China)

机构地区：[1]辽宁装备制造职业技术学院自控学院,沈阳110161 [2]辽宁开放大学,沈阳110034

出　　处：《机电工程技术》2024年第7期23-28,共6页Mechanical & Electrical Engineering Technology

基　　金：2021年度辽宁省教育厅高等学校基本科研项目(重点项目)(LJKZ1286);2021年度辽宁省教育厅高等职业教育开放办学合作项目(2021360-191)。

摘　　要：针对机械装备剩余寿命预测精度不高的问题,提出了一种基于深度迁移学习的多阶段剩余寿命预测新方法。该方法的核心思想是先确定机械装备的健康状态,对已进入退化状态的装备启动预警和剩余寿命预测机制,以大幅提高剩余寿命预测的准确性。在第一阶段,采用卷积自编码器和皮尔逊相关系数相结合的方法建立健康指标,通过快速搜索和发现密度峰聚类方法进行在线健康识别;在第二阶段,将故障数据输入一个多通道可迁移的双向长短时记忆网络预测模型,通过添加领域适配模块逐步减小特征分布差异,以便获得最优的训练模型,得到更具有泛化能力的回归结果。以IEEE2012PHM轴承全寿命数据集为例,与其他相关方法相比,获得了最小的预测误差。实验结果验证了所提方法的有效性和准确性,且无需进行手动阈值设置,具有很高的推广及应用价值。A new multi-stage remaining life prediction method based on deep transfer learning is proposed to address the issue of low accuracy in predicting the remaining life of mechanical equipment.The core idea of this method is to first determine the health status of mechanical equipment,and then activate the warning and remaining life prediction mechanism for equipment that has entered degradation status.This can greatly improve the prediction accuracy of remaining life.In the first stage,a combination of convolutional autoencoder and Pearson correlation coefficient is used to establish health indicators,and online health recognition is carried out through fast search and discovery of density peak clustering methods;In the second stage,the fault data is input into a multi-channel transferable bidirectional long short-term memory network prediction model,and the feature distribution differences are gradually reduced by adding domain adaptation modules to obtain the optimal training model and obtain regression results with more generalization ability.Taking the IEEE2012PHM bearing life dataset as an example,the minimum prediction error is achieved compared with other related methods.The experimental results verify the effectiveness and accuracy of the proposed approach,without the need for manual threshold setting,which has high promotion and application value.

关 键 词：机械装备 深度迁移学习 多阶段 剩余寿命预测 

分 类 号：TH17[机械工程—机械制造及自动化]