WANG Jing, LI Xiao-lin. Application of RBFNN in outliers filtering of velocity sensor[J]. 2014, 33(11): 153-155. DOI: 10.13873/J.1000-9787(2014)11-0153-03.
RBFNN在速度传感器异常值滤除中的应用
摘要
基于速度的延滞特性即不会瞬间突变
利用径向基函数(RBF)优秀的预估和拟合逼近能力
通过对已知时刻的速度值进行网络训练和学习
可以很好地预测下一刻的值及其变化趋势。根据是否落入由训练误差所确定的置信区间
判定异常值并进行异常值的滤除。经实验测试
95%的置信区间能够完全满足剔除异常数据、保留正常数据的功能。
Abstract
Based on arrearage character of speed
use excellent ability of forecast and fitting approximation of radial basis function( RBF)
through network training and learning of speed values of known moment value of the next moment and it’s change trend can be well predicted. According to whether the value falling into the confidence interval caused by training error
judge and filter the outliers. Through experimental tests
95 %confidence interval can completely meet the function of eliminating abnormal data and keeping normal data.