基于粒子滤波与线性自回归的故障预测算法  

Fault Prediction Algorithm Based on Particle Filter and Linear Autoregressive Models

作　　者：龙凤[1] 薛冬林 陈桂明[1] 杨庆[1] 

机构地区：[1]第二炮兵工程学院装备管理工程系 [2]第二炮兵96604部队

出　　处：《计算机技术与发展》2011年第11期133-136,140,共5页Computer Technology and Development

基　　金：军队科研项目(2010066)

摘　　要：粒子退化是粒子滤波在故障预测应用中存在的主要问题。针对粒子滤波算法样本贫化问题,提出一种基于粒子滤波与线性自回归的故障预测算法。在算法的状态估计阶段,使用混合状态系统模型和粒子滤波算法对系统状态的概率密度函数进行估计,并实时给出故障发生概率;在算法的状态预测阶段,采用线性自回归模型对故障征兆随时间的演化情况进行估计及修正,同时给出剩余使用寿命的概率密度函数。故障预测仿真实验结果证明了算法的有效性。Particle degeneracy is the main problem when a particle filter is applied to fault prediction. Focusing on the problem of sample impoverishment of particle filter algorithm, the fault prediction algorithm based on particle filter and finear autoregressive models is proposed, At the state estimation stage, the algorithm uses hybrid system state models and particle filter to estimate probability density function of the system state and support the real-time fault prognosis. At the state prediction stage, the algorithm estimates and corrects the system fault evolution process using linear autoregressive models. Simulation results demonstrate that the fault prediction algorithm based on particle filter and linear autoregressive models is feasible.

关 键 词：粒子滤波 线性自回归模型 故障预测 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]