电子学报 ›› 2015, Vol. 43 ›› Issue (3): 615-619.DOI: 10.3969/j.issn.0372-2112.2015.03.031

• 科研通信 • 上一篇    下一篇

非线性系统故障诊断的粒子滤波方法

张玲霞1, 刘志仓1, 王辉1, 齐会云1, 胡旦2   

  1. 1. 西安电子科技大学空间科学与技术学院, 陕西西安 710126;
    2. 电子科技大学自动化工程学院, 四川成都 611731
  • 收稿日期:2013-01-13 修回日期:2014-06-23 出版日期:2015-03-25 发布日期:2015-03-25
  • 作者简介:张玲霞 女,1965年生,陕西乾县人.西安电子科技大学空间科学与技术学院副教授,硕导,博士,主要从事综合测试与故障诊断、多传感器信息融合、可靠性理论与应用等方面的研究工作. E-mail:zlxnpu@163.com;刘志仓 男,1989年生,江西瑞金人.西安电子科技大学空间科学与技术学院硕士生,主要研究方向:综合测试与故障诊断. E-mail:liuzhicang@126.com
  • 基金资助:

    西安电子科技大学临近空间飞行器测控与特种测量创新基金(No.20140106)

Particle Filter Method for Fault Diagnosis in Nonlinear System

ZHANG Ling-xia1, LIU Zhi-cang1, WANG Hui1, QI Hui-yun1, HU Dan2   

  1. 1. School of Aerospace Science & Technology, Xidian University, Xi'an, Shaanxi 710126, China;
    2. College of Automation Engineering, UESTC, Chengdu, Sichuan 611731, China
  • Received:2013-01-13 Revised:2014-06-23 Online:2015-03-25 Published:2015-03-25

摘要:

针对粒子滤波存在粒子退化问题,提出一种基于无迹卡尔曼滤波和部分重采样的改进的粒子滤波算法.通过无迹卡尔曼滤波产生重要性分布函数和使用部分重采样算法进行重采样,以丰富粒子的多样性.并针对非线性系统故障诊断中非高斯背景下,似然函数检测量难以导出的问题,提出一种基于多模型和似然函数值的诊断方法.仿真结果表明:改进的滤波算法的估计精度优于标准的粒子滤波算法及其现有的两种改进算法,提出的故障诊断方法能够做到快速检测与准确隔离.

关键词: 故障诊断, 粒子滤波, 无迹卡尔曼滤波, 非线性系统, 似然函数

Abstract:

In order to solve particle degeneracy problem, we present an improved particle filter algorithm based on unscented Kalman filter and partial resampling algorithm.By using unscented Kalman filter to generate importance distribution function and partial resampling algorithm to resample particles, the method enriches the diversity of the particles.Furthermore, to solve the problem which likelihood detection statistics is obtained with difficulty in typically nonlinear and non-Gaussian, a fault diagnosis method based on the multiple model and the likelihood is proposed.Simulation results show the precision of the presented filter algorithm outperforms that of the standard particle filter and the improved particle filter existed in the filter system, and the proposed fault diagnosis method can detect fault quickly and isolate accurately.

Key words: fault diagnosis, particle filter, unscented Kalman filter, nonlinear system, likelihood

中图分类号: