采用BP神经网络预测乳聚丁苯橡胶的门尼黏度  

Prediction of Mooney viscosity of emulsion polymerized styrene-butadiene rubber by back-propagation neural network

作　　者：金彦江[1,2] 沈本贤[1] 杨磊[2] 隋军[2] 赵基钢[1] 

机构地区：[1]华东理工大学化学工程联合国家重点实验室,上海200237 [2]中国石油吉林石化公司,吉林吉林132021

出　　处：《合成橡胶工业》2013年第1期36-40,共5页China Synthetic Rubber Industry

摘　　要：在工业生产乳液聚合丁苯橡胶配方的基础上,于实验室聚合釜中考察了增加引发剂和乳化剂用量对聚合速率的影响以及补加相对分子质量调节剂及其加入时机对聚合产物门尼黏度的影响。结果表明,增加引发剂和乳化剂用量可以加快丁苯乳液聚合的速率,配合补加相对分子质量调节剂的手段可以使单体转化率达到70%时丁苯橡胶生胶的门尼黏度达到国家标准的要求。同时以原配方及其调整数据为基础,采用Levenberg-Marquardt算法对所建立的BP神经网络模型进行训练,仿真结果显示该网络的仿真数据与实验数据的误差小于1%,具有较好的一致性,可以用于判断丁苯乳液聚合不同配方在特定反应条件下产物的门尼黏度。Based on the industrial original formula, the effects of increasing initiator and emulsifier on the polymeriza- tion rate and replenishing relative molecular mass regulator and its adding time on the Mooney viscosity of emulsion poly- merized styrene-butadiene rubber （ESBR） were investigated in a polymerization reactor. The results showed that the in- crease of initiator and emulsifier, which accelerated the poly- merization rate, together with the replenishment of relative molecular mass regulator made the Mooney viscosity of the rubber met the need of national standard when the conversionreached up to 70%. The back-propagation neural network es- tablished was trained by Leyenberg-Marquardt algorithm on the basis of the original formula and ameliorated formula, and the relative error between the simulative results and experi- mental values was less than 1%. The good consistency showed that the BP neural network could predict the Mooney viscosity of ESBR with different formulas.

关 键 词：丁苯橡胶 BP神经网络 乳液聚合 门尼黏度 预测 

分 类 号：TQ333.1[化学工程—橡胶工业]