基于灰箱模型的聚合反应分子量分布形状控制

利用离散正交多项式组合神经网络对聚合反应分子量分布(Molecular weight distri-bution,MWD)进行建模,由MWD矩向量线性表示网络权向量,从而实现了灰箱建模。首先,推导了作为网络基函数的正交多项式需满足的约束条件,扩展了基函数的选择范围,为建模方法提供了理论基础。然后,利用最优化方法推导了MWD控制序列,利用权向量可以由矩向量代替的性质,对控制系统性能指标函数进行了修正,实现了对分子量分布形状的控制,提高了算法鲁棒性。最后,通过实验室规模的苯乙烯聚合试验,证明了该方法的可行性。

Control method of molecular weight distribution in polymerization reaction based on grey-box model

The control of molecular weight distribution(MWD) is one of important object of polymerization reaction product design.In orthogonal polynomial neural network model,the weights of neural network can be substituted by the moments of MWD,and the weights in neural network have a clear physical meaning,so the model of neural network can be called grey-box model.Based on these,this paper deduced the constraints of orthogonal polynomial,so the theory of the modeling method has been developed and been more general.Based on the modeling result,the optimization method is utilized to get the control sequence.The control performance function is developed to making the control strategy more robust.The control method is tested on styrene polymerization reacted in CSTR,and simulation results proved the effectiveness of the method.