期刊文献+

Kalman融合模型在无人装备关键部件寿命预测中的应用 被引量:1

Application of Kalman Fusion Model in Life Prediction of Unmanned Equipment's Key Components
下载PDF
导出
摘要 无人装备一般数量众多、执行任务时间长、环境恶劣,因此剩余使用寿命(RUL)预测尤为重要。综合性能指标序列使用单一模型的预测精度较低,为解决此问题,提出基于Kalman融合模型的RUL预测方法。首先,采用面积最大值法提取无人装备关键部件综合性能指标的退化阶段;其次,利用具有指数特征的GM(1,1)模型、线性支持向量机SVR模型、非线性极端学习机(ELM)模型对综合性能指标进行预测,每种模型可以捕捉综合性能指标的不同特征;最后,通过Kalman框架将3种模型的预测结果以迭代最小二乘的原则进行融合。实验结果显示,Kalman融合模型的预测方法可显著提高对综合性能指标的预测精度,与ELM,SVR和GM(1,1)单一模型相比,拟合精度分别提高了16.96%,1.61%和39.84%,预测精度分别提高了45.06%,38.35%和74.12%。 The number of unmanned equipment is generally large.Remaining Useful Life(RUL)prediction is especially important due to long mission time and harsh environments.The prediction accuracy of comprehensive performance indicators sequence using a single model is low.In order to solve this problem,an RUL prediction method based on Kalman fusion model is proposed.Firstly,the area maximum method is used to extract the degradation phase of the comprehensive performance indicators of key components of unmanned equipment.Secondly,the GM(1,1)model with exponential characteristics,the linear support vector machine SVR model,and the nonlinear Extreme Learning Machine(ELM)model are used to predict the comprehensive performance indicators.Each model can capture dfferent characteristics of the comprehensive performance indicators.Finally,the Kalman framework is used to fuse the prediction results of the three models according to the principle of iterative least squares.The experimental results show that the prediction method of Kalman fusion model can significantly improve the prediction accuracy of comprehensive performance indicators.Compared with that of single models of ELM,SVR and GM(1,1),the fitting accuracy is increased by 16.96%,1.61%and 39.84%respectively,and the prediction accuracy is increased by 45.06%,38.35%and 74.12%respectively.
作者 孙兴奇 赵爱罡 葛春 钟建强 许倍榜 刘茜萱 寇峰 SUN Xingqi;ZHAO Aigang;GE Chun;ZHONG Jianqiang;XU Beibang;LIU Xixuan;KOU Feng(Rocket Army Sergeant School,Qingzhou 262000,China)
机构地区 火箭军士官学校
出处 《电光与控制》 CSCD 北大核心 2023年第6期107-113,共7页 Electronics Optics & Control
基金 国家自然科学基金(61773389)。
关键词 剩余寿命预测 GM(1 1)模型 极端学习机(ELM) SVR支持向量机 Kalman融合模型 RUL prediction GM(1,1)model Extreme Learning Machine(ELM) SVR support vector machine Kalman fusion model
  • 相关文献

参考文献14

二级参考文献127

共引文献195

同被引文献6

引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部