Abstract: Researches on characteristics of QAR data and rough set theory, combined with knowledge of information decision tables are used in the paper to inspect the QAR outliers in the data mining to support aircraft fault detection and exclusion. It utilizes the application of rough set theory characteristics to discretize the QAR data, establishes decision tables for discretized data, performs attribute reduction and rule extraction for the decision tables. According to the particularities of the QAR data, improved algorithms are given for both of discretization of the QAR data and decision table attribute reduction in this article. Finally, the paper compares the results of project trials with the data given by experts, it proves that improved algorithms are feasible and efficient, improving the efficiency of the aircraft troubleshooting, having very important practical significance.

Key words: Quick Access Recorders（QAR） data, rough set, discretization, attribute reduction, rule extraction

摘要： 通过研究飞机快速存取记录器（Quick Access Recorders，QAR）数据和粗糙集理论的特点，结合信息决策表的相关知识，对QAR数据中的异常数据进行检测挖掘，以辅助飞机故障检测及排除。主要工作是：应用粗糙集理论特点对QAR数据进行离散化，并建立离散化后数据的决策表，然后对决策表进行属性约简和规则提取。根据QAR数据的特殊性，给出了数据离散化和决策表属性约简的改进算法。最后通过对比项目实验及专家给出的数据证明了其可行性和有效性，提高了飞机排故效率，具有很重要的现实意义。

关键词: 快速存取记录器数据, 粗糙集, 离散化, 属性约简, 规则提取