Particle Filtering Algorithm for Fault Diagnosis of Multiple Model Hybrid Systems with Incomplete Models
-
摘要: 针对模型不完备的混合动态系统故障诊断问题提出了一种粒子滤波算法. 系统未建模动态利用未知故障模式描述, 当存在未知模式时, 常规的粒子滤波器算法存在发散现象. 本文分析了常规粒子滤波器发散的原因, 提取了两个基于粒子集合的统计量: 粒子集的规格化因子 W 以及最大后验概率估计状态的信度 B. 在此基础上设计了检测未知故障模式的阈值逻辑, 即当 W 几乎为0且 B 较小时离散状态为未知故障模式. 在一定假设下从理论上证明了算法的正确性. 通过不完备的非线性混合系统诊断问题验证了算法的有效性.Abstract: A particle filtering algorithm is presented for fault diagnosis of hybrid dynamic systems with incomplete models. Un-modeled dynamics of complex systems is denoted as "unknown-fault". Firstly, the divergence of general particle filter (GPF) for imperfect systems is discussed. Secondly, two kinds of statistics are put forward, i.e., normalization factor of particles (W), and belief of maximal a-posteriori probability estimation state (B). Thirdly, threshold logic is presented to detect unknown-faults, and its correctness is proved under some reasonable assumptions. The efficiency of the method is testified by the simulated results.
-
Key words:
- Hybrid system /
- fault diagnosis /
- incomplete model /
- particle filtering /
- unknown fault
计量
- 文章访问数: 2995
- HTML全文浏览量: 66
- PDF下载量: 1524
- 被引次数: 0