NAR模型的Korenberg算法及其应用

函数逼近是许多工程领域，乃至经济领域中的一个基本问题，本文深入研究了NAR（Nonlinear Autoregresive)模型的Korengberg函数逼近算法，并从实时控制角度提出了一种可行的、新的模型优化方法。在仿真实验中，对混沌信号和火炮弹道函数逼近时，得到了较好的效果。实实验结果表明，本算法具有以下4个优点；1）对不同结构的系统具有较好的适应性。2）具有较高的逼近精度。3）具有较短的计算时间。4）具有较强的抗噪声能力。可见，本算法是一种有效高实际应用价值的方法。

Korenberg's Algorithm for NAR Model and Its Applications

The basic problem in many engineering fields, and even in economy field, is function approximation. In this paper we study NAR (Nonlinear Autoregressive) model and Korenberg's algorithm for function estimation deeply, and from the point of view of real time control we propose a new feasible method of optimally selecting the terms in the model. The new method has gained good results in fitting chaotic systems and a function of gun ballistics. The experimental results show that the algorithm has four advantages: 1) It can adapt to the differently structural systems better. 2) It has a higher accuracy for function approximation. 3) It requires a shorter time of calculation. 4) It is not greatly effected by noise. Apparently, the method has great practical value.