首页|预测轴承寿命的gate递归单元特征融合域自适应模型

预测轴承寿命的gate递归单元特征融合域自适应模型

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
采用现有的数据驱动模型对不同工况下的轴承剩余使用寿命(RUL)进行预测时,精度会大幅下降.针对这一问题,提出了一种基于门控递归单元特征融合领域自适应(GFFDA)模型的轴承RUL预测方法.首先,采用信号分析方法对轴承振动信号进行了特征提取,并采用特征评价的方法选择出了5 个最优特征,在最优特征的基础上,采用粒子群算法优化后的支持向量机的方法对轴承的健康阶段进行了划分;然后,选择目标域和源域退化阶段的最优特征子集作为GFFDA模型的输入,采用源域数据对特征提取器和寿命预测模块进行了预训练;最后,更新了目标特征提取器和寿命预测模块,对目标域的RUL进行了预测;并使用西安交通大学的轴承数据集对该GFFDA模型的有效性进行了验证.研究结果表明:相比于现有的数据驱动模型,GFFDA模型具有更好的跨工况分析能力和更出色的信息提取能力;同时,在对变工况的轴承寿命进行预测时,采用GFFDA模型具有更好的性能.
Adaptive model of gate recursive element feature fusion domain for predicting bearing life
Aiming at the problem of the decrease accuracy of existing data-driven models in predicting the remaining using life(RUL)of bearings under different operating conditions,a bearing prediction method based on gated recursive unit feature fusion domain adaptive(GFFDA)model was proposed.Firstly,the signal analysis method was used to extract features from the bearing vibration signal,and the feature evaluation method was used to select 5 optimal features.Based on the optimal features,the support vector machine optimized by particle swarm optimization was used to partition the health stages of the bearing.Then,the optimal feature subsets of the degradation stage in the target domain and source domain were selected as inputs for the GFFDA model,and the feature extractor and lifespan prediction module were pre-trained using source domain data.Finally,the target feature extractor and lifespan prediction module were updated to predict the RUL of the target domain.The proposed method was validated using the bearing dataset from Xi'an Jiaotong University.The research results indicate that the GFFDA model has better cross condition analysis ability and excellent information extraction ability compared to existing data-driven models,and has better performance in bearing life prediction tasks under different working conditions.

rolling bearingsremaining using life(RUL)feature evaluationadversarial adaptationgated recursive unit feature fusion domain adaptive(GFFDA)modeldata-driven model

曾玉海、程峰、魏春虎、杨世飞

展开 >

江南大学 机械工程学院,江苏 无锡 214122

南京凯奥斯数据技术有限公司,江苏 南京 210012

滚动轴承 剩余使用寿命(RUL) 特征评价 对抗自适应 门控递归单元特征融合领域自适应(GFFDA)模型 数据驱动模型

国家重点研发计划政府间国际科技创新合作项目山东省重大科技创新工程自主项目

2022YFE0143002019JZZY020111

2024

10.3969/j.issn.1001-4551.2024.04.006

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(4)
  • 27