基于BP神经网络的MCM-22分子筛催化性能预测模型研究  

Performance prediction model for a MCM-22 zeolite catalyst based on a back-propagation neural network

作　　者：唐艳[1] 李建伟[1] 孙晓岩[1] 李英霞[1] 陈标华[1] 

机构地区：[1]北京化工大学化工资源有效利用国家重点实验室,北京100029

出　　处：《北京化工大学学报（自然科学版）》2010年第1期13-18,共6页Journal of Beijing University of Chemical Technology(Natural Science Edition)

基　　金：国家自然科学基金(20576012)

摘　　要：基于人工神经网络建模方法,建立了MCM-22分子筛催化剂在苯与丙烯液相烷基化过程中性能预测的BP神经网络模型,该模型关联了催化剂本征性能、工艺条件和反应产物分布之间的相互关系。对不同成型条件下得到的16种结构性能各异的MCM-22分子筛催化剂性能进行实验评价,将所获得的数据用于模型训练和预测结果检验。结果表明,所建立的模型对MCM-22分子筛催化性能具有较高的预测精度,预测平均相对误差为4.21%。因此,将BP神经网络作为MCM-22分子筛催化剂的性能预测和苯与丙烯液相烷基化过程的定量描述模型,是适宜和可靠的。Based on the artificial neural networks （ANN） modeling method, a back-propagation （BP） neural networks model was developed for the performance prediction of an MCM-22 zeolite catalyst in the liquid alkylation of benzene with propylene. The intrinsic performance of the MCM-22 zeolite, the process conditions, and the product distribution were included in the model. By means of the extrusion molding method, 16 kinds of MCM-22 zeolite catalysts with different textural properties were prepared, and their catalytic performances were also experimentally investigated. The ANN model developed here was trained and tested according to the experimental data. The results showed that the ANN model had a very good simulation capability and the predicted results agreed closely with the experimental data, with an average relative prediction error of only about 4.21%. Therefore, the ANN model established here is suitable and reliable for prediction of the catalytic performance of MCM-22 zeolites and simulation of the results of the alkylation of benzene with propylene using this material as a catalyst.

关 键 词：BP神经网络 MCM-22分子筛 催化性能 预测 烷基化 

分 类 号：TQ032[化学工程]