一种基于案例推理的动态故障集诊断算法

摘要

针对单纯静态集的故障诊断算法存在诊断准确率低、效率差等缺点,文中提出了一种基于案例推理的动态故障集诊断算法( CBR-DFDA)。 CBR-DFDA算法根据故障与症状依赖的不确定性,采用二分贝叶斯网络建立依赖模型,在故障持续时间统计的基础上修正先验故障概率；并引入动态故障集,给出了故障案例的表示、案例属性约简、案例属性权重的分配及相似算法。实验结果表明,CBR-DFDA算法可以有效地针对动态故障集中的故障,改善内存的存储空间,提高诊断效率和准确率。%For the shortcomings existing in the fault diagnosis algorithm of simple dynamic set,such as low efficiency and poor accuracy, propose a new algorithm to handle the fault diagnosis problem in the condition of dynamic fault set,called Dynamic Fault Diagnosis Algo-rithm based on Case-Based Reasoning ( CBR-DFDA for short) . Considering the uncertainty of dependency between failure and symp-toms,CBR-DFDA algorithm builds dependency model using bipartite Bayesian network,and corrects priori probability of failure on the basis of analyzing fault duration. Also,introduce a set of dynamic fault and give the way to express the failure,simplify the attributes and assign attribute weights in a specific case. Then the optimal solution will be gained through running the algorithm. Experimental results show that CBR-DFDA algorithm can improve utilization of the memory and obtain a higher diagnostic efficiency and accuracy for the failure in a dynamic fault set.