基于相关参数的自修正飞行数据估计方法

A Self-emendation Algorithm for Flight Data based on Interrelated Parameters

提出了自修正的数据估计方法。用待拟合参数及相关参数组成多阶数据矩阵,利用递归神经网络对数据矩阵中缺失数据进行初次多阶拟合;利用滑动窗口,在历史数据矩阵中进行相似模式搜索,以相似度作为权重系数,对初次拟合结果进行加权修正。分析证明,该算法的最差、最好复杂度分别达到了理论最好结果。从仿真结果可以看出,自修正算法数据拟合结果误差较之单纯神经网络拟合更小,估计精度更高。

Put forward self-emendation algorithm for flight data. The algorithm constructs multi-rank data matrix with interrelated parameters. Firstly, the algorithm estimate missing data in data matrix by Recvrrent Neuron Network （RNN）, searchs similar models in history data matrix,emendates the results from RNN based on similar coefficient. Theoretical analysis shows that the algorithm has optimal time complexities in the worst ,best cases. The simulation result reveals that compared with results with RNN, self-emendation algorithm not only has smaller error, but also has higher accuracy.