神经网络用于无磷洗涤助剂合成条件的预测与优化

Prediction and Optimization of Technological Conditions for Synthesis of Phosphorus-free Builder by Neural Network Method

以丙烯酸和马来酸酐为单体,采用溶液聚合法合成了一种三聚磷酸钠的无磷替代品丙烯酸与马来酸酐共聚物。以L16(45)正交实验的结果,采用误差反向传播BP(backpropagation)算法,完成了神经网络的学习,据此预测了MA与AA质量比、引发剂与单体质量比、反应温度、反应时间、滴加单体时间对助洗性能的影响。在给定误差范围内,预测结果表明,最佳产品的合成条件是:MA与AA质量比为0 25,引发剂与单体质量比为0 045,反应温度为85℃,反应时间1 5h,单体滴加时间为20min;实验结果表明,BP神经网络在合成无磷助洗剂中有较好的离线学习功能,可以较少的实验次数而获得更多的信息,指导合成工作。

The synthesis of phosphorus-free builder was studied using acrylic acid and maleic anhydride as polymerization monomers in aqueous system.The neural network is trained by the results of orthogonal design.The influences of monomer ratio,initiator amount,reaction time,reaction temperature and monomer dripping time on the performance of the detergent are predicted by the neural networks.It is shown that the best synthesis conditions are as follows.The weight ratio of maleic anhydride to acrylic acid was 25%,weight of initiator was 4.5% the weight of monomers, monomer dripping time was 20 min,and the reaction was carried out at 85 ℃ for 1.5 h.The results show that back propagation neural network has better off-time learning performance for the synthesis of phosphorus-free builder,thus more informations are obtained by fewer experiments to guide the technological process.