基于遗传算法的故障诊断规则提取方法的研究

Research on Fault Diagnosis Rule Extraction Approach Based on Genetic Algorithm

针对粗糙集应用于故障诊断过程中存在的决策表属性连续值离散化和属性约简之间分离的问题,根据粗糙集的约简思想,本文研究了一种新的基于遗传算法的规则自动提取方法,可以在知识获取过程中同时实现属性连续值离散和属性约简。将该方法用于卫星电源系统故障诊断,可以从故障决策表中直接提取规则,得到超过90%的故障诊断准确率,进一步验证了这种方法的可行性和有效性。

In the application of rough set on fault diagnosis,there is a problem to be solved in which decision table attributes reduction is separated from discretization of continuous value attributes. According to the idea of attribute reduction in rough set, a new genetic algorithm based automatic rule extraction approach is studied. By this method, both continuous value attributes discretization and attributes reduction can be done during knowledge acquirement. Furthermore, it is used to extract rules for fault diagnosis of a satellite power system. Rules can be derived directly from its decisions tables, and more than 90% accuracy can be achieved, which shows the feasibility and validity of this approach.