一种基于贝叶斯网络的模型诊断方法

Method of Model-based Diagnosis Founded on Bayesian Network

提出一种结合贝叶斯网络进行基于模型诊断的方法。在基于模型诊断的基础上,建立了元件状态模型,并将诊断模型转换为贝叶斯网络,利用团树算法求解征兆产生时系统状态的后验概率,再通过计算边缘分布获得元件故障概率。最后给出一个数字故障电路的实例,在Matlab上进行推理,得到了精确的概率值,验证了该方法的有效性。

Put forward a method which applies bayesian network to the model-based diagnosis. Building a componentstate model which is founded on the model-based diagnosis, authors got the posterior probability of system state in case the appearance of certain symptoms through junction tree algorithm. The component fault probabilities were obtained by the calculation of marginal distribution. An example of fault digital circuit was given. Making use of Matlab, the accurate probabilities were got. Therefore, the method was proved out.