多模型粒子滤波的故障诊断方法研究  

Fault Diagnosis Based on Multiple-models Particle Filter

作　　者：张三同[1] 郝晶晶[1] 

机构地区：[1]北京交通大学电子信息工程学院,北京100044

出　　处：《控制工程》2012年第5期864-869,共6页Control Engineering of China

基　　金：国家自然科学基金资助项目(60776831)

摘　　要：近年来,非线性非高斯系统的故障诊断问题一直是热点问题也是研究难点。针对非线性非高斯系统的故障检测与定位问题,运用基于粒子滤波(Particle Filter)的似然概率密度函数值的方法来检测系统故障,在检测到故障发生时,运用多模型方法与粒子滤波相结合,进行故障定位,从而进一步改善故障诊断的效果。仿真结果表明,该方法可以对非线性系统进行快速、准确的故障诊断。For recent decades, fault diagnosing of non-linear non-Guassian system is a hotspot as well as a difficult problem. In order to solve the issue of fault detection and isolation（FDI） in non-linear non-Guassian system, in this paper, likelihood probability is pro- posed to solve the detection problem, and particle filter combining with multiple models is used for fault isolation. The simulation results show that this method could detect the fault immediately and diagnose the fault correctly.

关 键 词：故障诊断 非线性非高斯系统 粒子滤波 多模型方法 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]