基于聚类模糊系统的动态数据野值剔除方法  

A New Method for Outlier Removal from Dynamic Data Based on Clustering Fuzzy System

作　　者：朱学锋[1] 

机构地区：[1]92941部队

出　　处：《飞行器测控学报》2011年第5期81-84,共4页Journal of Spacecraft TT&C Technology

摘　　要：针对影响遥测参数处理和分析的野值问题,提出了基于聚类法实时设计模糊系统实现动态数据野值辨识和剔除的新方法。该方法能够自适应跟踪不同变化特性的遥测参数,基于聚类法实现模糊系统的动态建模并获得预测值与观测值的残差序列,再按照狄克松准则实现野值的快速剔除。对实测数据的仿真实验表明：该方法能够显著降低动态建模的复杂度,快速跟踪信号变化,方法可行且有效。To eliminate the influence of outliers in telemetry data processing and analysis,the paper presents a new method for identification and elimination of outliers in dynamic measurement data by realtime design of a clustering fuzzy system.The method adaptively tracks telemetry parameters of different variation characteristics and dynamically establishes fuzzy system modeling by clustering to obtain residual error series between forecast value and observation value.Then,it identifies and removes outliers in accordance with Dickson criterion.Simulation and computation with actual measurement data show that the method remarkably reduces complexity of system modeling and tracks signal variation promptly.

关 键 词：聚类法 模糊系统 野值 辨识 数据处理 

分 类 号：V557[航空宇航科学与技术—人机与环境工程]