用人工神经网络预测催化精馏塔开工过程的研究  

Predictions Catalytic Distillation Column Start-up Processes Via Artificial Neural Network

作　　者：胡晖[1] 邬慧雄[2] 徐世民[3] 李鑫钢[3] 

机构地区：[1]福州大学化学化工学院,福州350002 [2]清华大学化学工程系,北京100084 [3]天津大学精馏技术国家工程研究中心,天津300072

出　　处：《分子催化》2006年第4期360-362,共3页Journal of Molecular Catalysis(China)

摘　　要：The time consumed in starting up the distillation unit with appreciable holdups can be an important fraction of the total distillation time,particular for catalytic distillation systems with large holdups.To optimize the whole process,the start-up period has to be considered as a part of the complete catalytic distillation process.In this paper,BP artificial neural network model was presented as a tool to estimate the start-up process for a given catalytic distillation system.It can been seen that through the examination of the case studied in this work,a good start-up policy can reduce both the energy and time requirements in the start-up phase of catalytic distillation processes.The results based on 20 start-up policies showed that the time consumed in start-up period with an average error of 4.140% and a maximum error of 10.291% for the case studied in this work.The accuracy of the model will depend upon the data available and the type of model.The time consumed in starting up the distillation unit with appreciable holdups can be an important traction of the total distillation time, particular for catalytic distillation systems with large holdups. To optimize the whole process, the start-up period has to be considered as a part of the complete catalytic distillation process. In this paper, BP artificial neural network model was presented as a tool to estimate the start-up process for a given catalytic distillation system. It can been seen that through the examination of the case studied in this work, a good start-up policy can reduce both the energy and time requirements in the start-up phase of catalytic distillation processes. The results based on 20 start-up policies showed that the time consumed in start-up period with an average error of 4. 140% and a maximum error of 10.291% for the case studied in this work. The accuracy of the model will depend upon the data available and the type of model.

关 键 词：人工神经网络 催化精馏 开工 预测模型 

分 类 号：O643.3[理学—物理化学]