利用RBF神经网络实现聚合反应的内模控制

文中研究基于径向基(RBF)神经网络算法的内模控制策略在苯乙烯本体聚合反应相对分子质量分布控制领域的应用。利用神经网络对非线性系统的逼近能力，把内模控制推广到聚合反应过程质量指标控制这一非线性系统中。针对建模过程中存在的稳态误差，在训练数据中增加了部分静态数据，有效的提高了模型的验证精度，大大改善了由神经网络构成的内模控制器的控制精度，消除了系统余差。仿真结果证明，基于神经网络算法的内模控制策略达到了较好的控制质量。

Internal model control via RBF neural network in polymerization

A nonlinear internal model control (IMC) strategy based on radial basis function(RBF) network models was proposed for bulk polymerization of styrene. Taking advantage of the neural network's approximate ability to any nonlinear system, the internal model control strategy was extended to the quality control of polymerization. Aiming at removing those static errors in modeling, some static process operating data were added in training samples. Therefore the validation accuracy of the model, as well as the accuracy of the internal controller based on RBF neural network, is raised, so that the offset of the system is removed. The simulation outcomes show that the strategy has realized a good control quality.