TE-PF及其在轴承寿命预测中的应用
Trajectory Enhanced Particle Filter and Its Application in Bearing Remaining Useful Life Prediction
罗鹏 1胡茑庆 1沈国际 1张伦1
作者信息
- 1. 国防科技大学智能科学学院 长沙,410072;国防科技大学装备综合保障技术重点实验室 长沙,410072
- 折叠
摘要
针对构建科学的预测模型以及估计合适的模型参数是极大限制粒子滤波(particle filter,简称PF)方法计算效率与稳定性的瓶颈问题,提出了一种基于轨迹强化粒子滤波(trajectory enhanced particle filter,简称TE-PF)的滚动轴承剩余使用寿命(remaining useful life,简称RUL)预测方法.从退化速率跟踪和退化轨迹强化的角度出发,构建了一种面向PF方法的通用预测模型,利用历史样本以及粒子生成样本的退化趋势信息,有效指导通用预测模型的参数估计,最终获取多信息融合的轨迹增强预测模型.实验结果表明,相较于已有方法,TE-PF方法具有更高的计算效率与更强的趋势预测稳定性,观测样本累积情形下能够获取置信区间内较高的预测精度.
Abstract
The particle filter(PF)method is widely used in prediction of the remaining useful life(RUL)of rolling bearings.However,constructing a scientific prediction model and estimate appropriate model parameters is still a bottleneck.This limitation greatly affects the computational efficiency and stability of the PF method.To solve this problem,an RUL prediction method for rolling bearings is proposed based on trajectory enhanced particle filter(TE-PF).From the perspective of degradation rate tracking and degradation trajectory strengthen-ing,a universal prediction model based on the PF method is constructed,and the degradation trend information of historical samples and particle-generated samples are effectively used,which effectively guide the parameter estimation of the universal prediction model.Finally,a multi-information fusion trajectory enhancement predic-tion model is obtained.Experimental verification results show that the TE-PF method has stronger computa-tional efficiency and trend prediction stability than the traditional methods.In the case of observation sample ac-cumulation,the prediction accuracy in the confidence interval is higher.
关键词
滚动轴承/剩余使用寿命预测/退化速率跟踪/轨迹强化粒子滤波Key words
rolling bearings/remaining useful life prediction/degradation rate tracking/trajectory enhanced particle filter引用本文复制引用
基金项目
国家自然科学基金(51975576)
国防基础科研项目(WDZC20205500301)
出版年
2024
NETL
NSTL
万方数据