基于模型诊断的贝叶斯解释及应用

在基于模型诊断的基础上，建立了贝叶斯模型．该模型首先假设系统的所有状态都是可能发生的，并对系统中的各个元件设定一个先验概率．当系统的观测与预期不符合时，利用最小冲突集的概念得到系统可能的状态．之后根据各个系统状态的对应概率计算出元件发生故障的后验概率，该后验概率本质上是一个边缘概率分布，它在数值上表示了元件发生故障的可能性，确定了故障查找顺序．用一个数字故障电路的实例验证了该方法的有效性，井对其存在的问题进行了讨论．

Bayesian Interpretation of Model-based Diagnosis and Its Application

A Bayesian model was introduced into model based diagnostics. First, the occurrence of all the system states is considered to have an equal possibility and for each component in the diagnosis system set a prior probability. Then, the impossible system states can be determined when the observation is conflict with the system's prediction. Through the idea of minimal conflict sets, the probability of all the possible system states can be obtained and the posteriori probabilities of a component in certain state can be calculated from the marginal probability distribution. A real circuit diagnosis was used to illustrate the method's validity. The result is useful to determine the order of detecting fault components.