基于改进Elman神经网络的石膏纤维板厚度控制建模

针对石膏纤维板厚度控制系统的非线性、时变性及不确定性，采用改进的Elman神经网络对其建立动态模型，介绍了改进Elman网络的结构及学习算法；分析了影响石膏纤维板厚度控制精度的主要因素，并根据实际情况确定了输入层和中间隐层的维数，从而确定了模型的结构.由于改进的Elman网络具有适应时变特性的能力，而且学习精度高、学习速度快，与传统的BP网络相比，收敛速度有明显提高.通过对现场采集的数据进行仿真，得到了令人满意的结果.

Thickness-control modeling of gypsum-fibre board based on modified Elman neural network

For the nonlinear, time varying and uncertainty in thickness-control system of gypsum-fibre board, the dynamic modeling adopting modified Elman neural network was set up. The structure and learning arithmetic of the modified Elman neural network were introduced. The main factors influencing thickness-control precision of gypsum-fibre board were also analyzed. The dimension of input and middle layers as well as the structure of modeling were determined. Because the modified Elman neural network has the ability of adapting time varying, and also offers high learning precision and speed, the converging speed of the network gets significantly enhanced, compared with the traditional BP neural network. The simulating result on data collected from locale is satisfying.