基于神经网络的多组分相平衡常数预测模型  

MODELS FOR PREDICTING MULTI COMPONENT PHASE EQUILIBRIUM RATIO BASED ON ARTIFICIAL NEURAL NETWORKS

作　　者：仲卫涛[1] 陈曦[1] 钱积新[1] 

机构地区：[1]浙江大学工业控制技术国家重点实验室

出　　处：《石油学报（石油加工）》1999年第3期29-33,共5页Acta Petrolei Sinica(Petroleum Processing Section)

基　　金：国家自然科学基金

摘　　要：以抚顺石油二厂乙苯车间的精馏工段为背景，以相平衡常数的简化计算公式为依据，利用多层前馈网络（ＭＮＮ）和径向基函数网络（ＲＢＦＮ）分别建立了各自的相平衡常数神经网络预测模型，并对网络训练过程中所应注意的问题进行了讨论。通过将预测结果与ＰＲＯ／Ⅱ的模拟结果进行分析比较。Under several rational suppositions,a simplified formula for phase equilibrium ratio was achieved.Two models based on mutilayer neutral networks(MNN)and radial basis function networks(RBFN)were employed to predict the value of the phase equilibrium ratio,respectively.Both of them can be applied to the process covering wide ranges of temperature,pressure and composition.Compared with the simulative data derived from PRO/Ⅱ,the results of the two models were satisfactory.

关 键 词：相平衡常数 精馏 分离 神经网络 模型 

分 类 号：O642.42[理学—物理化学] TQ028.31[理学—化学]