基于贝叶斯估计的特征事件判别方法研究

针对利用遥测指令参数进行特征事件自动判别时，现有方法忽略了数据源之间相互印证的条件，仅依据测量值相同来判别特征事件的发生时间，存在一定的误判概率，提出了一种基于贝叶斯估计的特征事件判别方法。方法不仅利用了当前数据源的测量值，同时还兼顾特征事件固有的先验分布信息，能够有效提高自动判别的准确率。首先，通过对特征事件数据源的内涵进行分析，得出了不同数据源相互印证的条件；其次，通过对历史数据进行统计分析，建立了特征事件发生时间的正态分布概率模型，并以此为基础提出了基于贝叶斯最大后验估计算法，设计了完整的特征事件判别方法；最后，通过工程数据的仿真计算和结果分析，验证了方法在特征事件判别中的实用性和有效性。

Research on discriminant method of characteristic events based on Bayesian estimation

For the automatic discrimination of characteristic events using telemetry instruction parameters, the existing methods ignore the condition of mutual verification between data sources, and only judge the occurrence time of characteristic events based on the same measured values, which has a certain misjudgment probability. A discriminant method of characteristic events based on Bayesian estimation is proposed. The method not only takes advantage of the current measured value but also takes into account the inherent prior distribution information of characteristic events, which can effectively improve the accuracy of automatic discrimination. Firstly, by analyzing the connotation of data sources for characteristic events, the conditions of mutual corroboration of diffe-rent data sources are obtained. Firstly, through the analysis of the connotation of characteristic event data sources, the conditions of mutual verification between data sources are obtained. Secondly, through the statistical analysis of the historical data, the normal distribution probability model of the occurrence time for characteristic events is established. The algorithm based on Bayesian maximum posterior estimation is proposed, and the discriminant method of characteristic events is designed. Finally, through the simulation calculation of engineering data and the analysis of results, the effectiveness and practicability of this method are verified.