论文部分内容阅读
针对分子筛上苯与丙烯烷基化反应速度快,难以获得准确动力学方程的特点,通过探索构建人工神经网络模型来关联分子筛催化剂本征性能、反应条件和催化性能之间的相互关系。研究分别以单输出和多输出参数为预测目标分别建立2种不同形式的BP神经网络模型,从预测数据和实验结果的比较上可以看出,无论单输出还是多输出网络,两者之间的平均相对误差均较小并且具有较高的相关系数,说明所建立的神经网络模型可以较准确的预测苯与丙烯烷基化反应性能。比较的结果还表明,单输出网络由于其针对性较强,较多输出网络具有更加精确的预测能力。该神经网络的建立可以为实际生产提供理论指导,也可以应用于催化剂的设计与开发,确定适用于反应最佳的催化剂织构特性和反应条件。
In view of the fast reaction of benzene and propylene on molecular sieve, it is difficult to obtain accurate kinetic equations. The relationship between intrinsic performance, reaction conditions and catalytic performance of molecular sieve catalyst is discussed by exploring and building an artificial neural network model. Two different forms of BP neural network model are established respectively based on single output and multiple output parameters as prediction targets. It can be seen from the comparison between the predicted data and experimental results that no matter single output or multiple output network, The average relative error is small and has a high correlation coefficient, indicating that the established neural network model can more accurately predict benzene and propylene alkylation reaction performance. The results of the comparison also show that the single-output network has more accurate prediction ability because of its more pertinence. The establishment of the neural network can provide theoretical guidance for practical production, and can also be applied to the design and development of catalysts to determine the optimal catalyst texture and reaction conditions for the reaction.