论文部分内容阅读
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。
Aiming at the problem of outliers affecting the processing and analysis of telemetry parameters, a new method of real-time fuzzy system design based on clustering is proposed to identify and remove the outliers of dynamic data. The method can adaptively track the telemetry parameters with different variation characteristics, realize the dynamic modeling of the fuzzy system based on the clustering method, obtain the residual sequence of the predicted value and the observed value, and then eliminate the outliers according to Dixon criterion. Simulation results of measured data show that this method can significantly reduce the complexity of dynamic modeling and track the signal changes rapidly. The method is feasible and effective.