基于模型诊断的改进贝叶斯方法

Improved Bayesian Method for Model-based Diagnosis

基于模型诊断是针对系统或设备的行为和结构建立模型,从而进行诊断的。但是基于模型诊断的方法存在不确定性问题,诊断的结果可能为一组故障部件。为解决不确定性问题,很多学者在基于模型诊断中使用了概率的方法,利用待诊断设备组成部件的故障概率信息来寻找最可能的诊断。通过对模型诊断中存在的不确定性问题的深入研究,在基于模型诊断中提出了概率的贝叶斯解释,从而利用后验概率形式量化了元件故障的可能性的衡量标准,并且改进了计算元件后验概率的方法,分析了改进后算法的复杂性和完备性,证明了改进后的方法降低了时间和空间的复杂性。实验结果表明,改进后算法的执行效率较原有的算法有明显的提高,且有些问题可以提高两个数量级。

Model-based diagnosis concerns using a model of the structure and behavior of a system or device in order to establish why the system or device is faulty. But the fact is that determining a diagnosis for a problem always involves uncertainty. This situation is not entirely satisfactory. This paper built upon and extended previous work in model-based diagnosis by supplementing the model-based framework with probabilistic sound ways for dealing with uncertainty. This was done in a mathematically way in Bayesian theory to compute the posterior probability that a certain component is not working correctly given some diagnosis. And in this paper we proposed a general method to increase efficiency. The complexity and the completeness of the method were analyzed. The time complexity and the space complexity were reduced in the improved method. The experimental results illustrate that the improved method has a better executive efficiency than the traditional method in general. In fact, the executive efficiency may be improved up to two orders of magnitude in some cases.