基于小波网络的外测缺失数据插补方法  

Trajectory Missing Data Interpolation Method Based on Wavelet Neural Network

作　　者：李振兴 张必彦 黄晓冬 LI Zhenxing;ZHANG Biyan;HUANG Xiaodong(No.92124 Troops of PLA,Dalian 116023)

机构地区：[1]中国人民解放军92124部队,大连116023

出　　处：《舰船电子工程》2020年第5期157-160,共4页Ship Electronic Engineering

摘　　要：针对飞行器试验中外测测量数据在某些条件下可能产生的数据丢失问题,提出了一种基于小波网络的缺失数据插补方法。利用外测缺失数据段落两段的有效测量数据构建学习样本,对小波网络进行训练并进行双向预测,对缺失数据进行预测插补。为进一步提高预测插补数据精度,首先对学习样本进行降采样处理,粗略预测缺失段落数据,逐步提高采样率至与原始测量数据采样率一致,对预测数据进行逐步修正,实现缺失段落数据的高精度预测。实测数据处理结果证明了利用小波网络对外测缺失数据进行预测插补可获得满足精度要求的插补结果。The trajectory data of vehicle test may encounter missing data problem under some conditions.Hereby a missing da⁃ta interpolation method based on wavelet neural network(WNN)is proposed.The valid data between the missing data of the trajecto⁃ry is designed as the training sample for WNN,and the bidirectional prediction for the missing data can be carried on based on the WNN after training.To further improve the precision of the interpolation of the missing data,the training sample carry on down sam⁃pling at first to obtain the cursory prediction data,and the sampling frequency can be increased gradually till it is consistence with the original trajectory data.Therefore,the precision of the prediction data can be improved gradually to implement the high preci⁃sion interpolation results.The simulation and the test data processing results show the effectiveness of the proposed method.

关 键 词：飞行器试验 外测 小波网络 采样率 

分 类 号：TN911[电子电信—通信与信息系统]