基于因果图的分布式诊断模型研究及应用

为了降低系统多故障诊断的计算复杂性,研究了一种基于因果图的分布式诊断模型.首先用因果网络图建立集中式模型,然后通过有向树分解技术把集中式模型转化为分布式模型,并离线计算集群表,得到局部诊断解后再采用诊断融合算法实现全局诊断.对电源系统的应用结果表明,分布式模型简化了集中式模型的计算复杂性,提高了效率,并且通过计算后的集群表及融合算法可以实现耦合故障的诊断.此种方法适用于卫星这类复杂系统的故障诊断.

Research and Application of Distributed Diagnosis Model Based on Causal Graph

For reducing computational complexity of complex system multiple fault diagnosis, a diagnosis strategy is researched in the paper. Firstly, centralized model is founded using causal networks. Then, centralized model is conversed to several distributed models through directed tree decomposition technology, and the clan table is calculated off line. Finally, global diagnosis is realized by using a synthesized diagnosis algorithm after the local diagnosis is finished. The application of a satellite power system proves that computational complexity of centralized model is reduced through distributed model, efficiency is increased and coupling faults can be diagnosed by using calculated clan table and synthesized diagnosis algorithm, which are suitable in fault diagnosis for complex satellite system.