基于函数链神经网络的非线性系统稳态模型辨识及应用

Steady-state Medel Indetification for Nonlinear System Based on Neural Network of Function-link

利用函数链神经网络的非线性映射能力和快速收敛特性，将良好的推广能对视为网络评价函数的约束条件，以改善网络的泛化特性，提高系统的稳态辨识精度，仿真和实际应用结果表明，具有收敛快，辨识精度高，所需样本少等优点。该方法在SO3磺化过程非线性系统建模中得到应用。

On the basis of good approximation and fast leaming of the function-link neural netwoks,this peper meke the generalization performance of the net as constraint's condition of the evaluating function for improving the genalization performence,and proposes a new algorithm. Using this neural network model for nonlinear modeling,the error can be effectively redued. The simutation and practical application results show that this approach is high efficiency and high precision with less samples. This method has been applied to nonlinear systeim modeling for SO3 sulphonation process.