间歇过程的神经网络模型

Modeling a batch process based on a neural network

间歇过程通常具有非线性、时变和易燃易爆的特点 ,用常规的建模方法建立起模型比较困难。本文针对间歇聚丙烯过程 ,利用前馈神经网络建立其数学模型。首先根据实际系统的输入输出建立网络的结构 ,再用经验数据对网络进行训练 ,并用未参加训练的数据对网络进行测试。测试的最大误差是 0 0 3MPa 。

It is difficult to model a batch process just by using general modeling techniques,for the process is time varying,nonlinear and dangerous.In this paper a feedforward neural network is used to model the reaction phase of a batch polypropylene according to the real inputs and outputs .Then some empiric data are used to train the neural network,and some other data are used to test the trained network.The maximal error is 0.03MPa,which is satisfying.