环氧乙烷反应器工况模拟

由于乙烯氧化生成环氧乙烷的反应过程非常复杂 ,建立机理模型对其生产工况进行模拟存在许多困难。以热油进口温度、原料乙烯流量、原料氧气流量为输入节点 ,以生成物中环氧乙烷体积分数、氧气转化率、乙烯转化率为输出层节点 ,建立 3层BP神经网络模型。该模型不仅具有较强的回忆能力 ,且对生成物中环氧乙烷体积分数和氧气转化率的预测效果比较好 ,平均相对误差分别为 1.2 971%和 2 .99% ,乙烯转化率的预测值相对误差为4 .5 2 4 5 %。神经网络输出层各节点平均相对误差均小于 5 % ,能够满足实际生产需要。该模型用于环氧乙烷反应器工况模拟与预测是可行的 ,对实际生产在线控制具有一定的指导意义

The Simulation of the Oxirane Reactor by Using Artificial Neural Network

The reaction that ethylene is oxidized to produce oxirane is very complex and there are many difficulties to simulate the mode by building a mechanic model. In this paper, a three layers BP neural network was built: the heat oil temperature; the ethylene flow; oxygen flow as input nodes and the oxirane content in the product; the ethylene change ration and the oxygen change ration as output nodes. Using this model to simulate the oxirane mode, it has not only better recalling value but also better predicating value. The average relative error of the each output nodes in the predicating result is 1.297 1%, 2.99% and 4.524 5% respectively, which are all less than 5% and can meet the need of the production and has some guidance to the line control. On the whole, it is feasible to apply the model into the simulation and the prediction of the oxirane reactor, it can meet the need of the prodwotion and has sone guidance to the line control. It can meet the need of the production and has some guidance to the line control.