基于RBF神经网络的飞参数据预处理

Preconditioning of Flight Data with RBF Neural Network

由于存在干扰,飞参系统记录的发动机参数中,经常会有不少间断点和奇异值。为了利用数据对发动机性能趋势进行预测,必须对数据进行预处理。发动机作为一个系统,其各主要输入和输出参数之间必然存在着一定的函数关系。本文研究了利用RBF神经网络和参数之间的关系对数据进行预处理,得到了较为正常的数据,结果表明该方法是有效的。

The FDR（flight data recorder） records some parameters of the aeroengine. Because of the jamming, there are some abnormal data in these parameters. The data must be preconditioned so that it can be used to forecast the performance trend of the aeroengine. As a system, there must be some functional relation between the input and output of the aeroengin. This paper researches the preconditioning of the flight data with the RBF neural network and relation of the parameters. Normal data is worked out in this way. So this method is available.