中文摘要：

准确监控和预测系统所处状态,在危险发生之前发出告警对于防止飞机进入危险状态具有重要意义。针对飞行安全系统复杂性特点,以飞行失控为例建立了系统状态空间模型,用主元表征系统状态简化了系统知识表达。通过判别分析确定系统状态点与失控区域重心的距离,判断系统所处状态,并设计了飞行失控告警的软件框架。结果表明,多元统计分析能够准确判断飞机是否可能进入飞行失控状态,可以推广到其他安全状态的预测。

英文摘要：

The paper is about the author＇ s idea on how to apply the multivariate statistical analysis method to the flight safety monitoring activities under the aircraft landing and leaving monitoring conditions. Since the flight safety system is a highly complicated system including the flight control and instruction workers, as well as the aircraft themselves and the surrounding environment involved, each kind of the aforementioned factors is likely to lead to fight accidents due to the failure of the integrated communications and interaction of man, aircraft and environment. In this ease, LOC （Loss of Control in flight） and CFIT （Controlled Flight into Terrain） should be taken as the two primary threats contrary to the flight safety rules. Seeing the extremely complicated situation of the flight safety system, we have built a current space status in-situ model in reference to the examples listed in the flight regularity of the Loss of Control in flight, the primary constituent elements which are used to express the system state as a result of calculation of the correlation matrix. And, then, we have developed a new safety space model which can help to simplify the data and information presentation, whose discriminating function has been established via the Bayesian discriminance model. And, the distance of the system state position to barycenter of LOC area can be conveniently worked out by the discrimination function to determine the system＇s in-situ working condition. Taking the critical parameters of a transporting system that was entering the Loss of Control in flight for example, it is possible to determine the discriminative results of each group of the sample by means of the original data validation （discriminant function can be gained by all the 51 groups of samples） and the intersectional data validation （discriminant function can be worked out by 50 groups of samples except for the sample tO be deter- mined） . The final results show that all the 51 groups of samples prove to be consistent