| 聚合物分子量分布的多变量动态系统模型 |
| |
| 引用本文： | 姜尔超,曹柳林.聚合物分子量分布的多变量动态系统模型[J].控制工程,2005(Z1). |
| |
| 作者姓名： | 姜尔超 曹柳林 |
| |
| 作者单位： | 北京化工大学信息科学与技术学院 北京100029
(姜尔超)，北京化工大学信息科学与技术学院 北京100029(曹柳林) |
| |
| 摘 要： | 为了解决聚合产品分子量分布控制的难题,将神经网络引入对其进行了无需任何系统内部先验知识的黑箱建模。所使用的神经网络是由B样条神经网络和非线性递归神经网络(DRNN)组合而成,并使用误差反传算法对网络进行训练和学习,从而建立了多变量动态系统的分子量分布模型。在模型建立中将控制变量与分布参数的函数关系利用非线性递归神经网络描述,分子量分布函数使用B样条神经网络表示,仿真研究结果证明该方法取得了预期的建模效果,具有一定的推广实用价值。
|
| 关 键 词： | 非线性回归神经网络 B样条神经网络 分子量分布 建模 |
Modeling Multi-variable Molecular Weight Distribution for Polymerization |
| |
| JIANG Er-chao,CAO Liu-lin.Modeling Multi-variable Molecular Weight Distribution for Polymerization[J].Control Engineering of China,2005(Z1). |
| |
| Authors: | JIANG Er-chao CAO Liu-lin |
| |
| Abstract: | For solving the difficulty in the control of molecular weight distribution (MWD) of polymer product, MWD via neural network without any transcendental information is modelled.During the model, the neural network cansist of the B-spline neural network and the nonlinear diagonal recurrent neural network (DRNN ). The combined network was trained with error back-propagation algorithm. A model of multi-variable dynamic MWD with combined network is built up.The relation between the control variable with the distributed parameters was described with nonlinear diagonal recurrent neural network, and the molecular weight distribution function was described with B-spline neural network . The simulation results proved the feasibility of the method. |
| |
| Keywords: | nonlinear-recurrent neural network B-spline neural network molecular weight distribution modeling |
| 本文献已被 CNKI 等数据库收录！ |
