计算机辅助甲烷氧化偶联催化剂设计研究

Computer-aided catalystdesign for oxidative coupling of methane

制备了一种多组分的甲烷氧化偶联催化剂 .为了找到较优的催化剂配方 ,应用人工神经网络辅助设计催化剂 ,并确定了辅助设计的步骤及网络结构 ;将 L evenberg- Marquardt法用于网络的训练 ,改进了网络的收敛特性 ;以训练好的网络为目标函数 ,应用 SWIFT方法优化得到了较优的甲烷氧化偶联催化剂配方 ,实验表明 ,该催化剂的C2 收率可达 2 0 .77% 。

After considering many catalysts used in oxidative coupling of methane, a catalyst containing six main elements, including Mn, Zr, S, W, P and Na were prepared. An improved Back-Propagation network, its structure, training method, and general ability are described. A general computerized method was developed to aid design of the catalysts. The application of this method in design of catalysts for oxidative coupling of methane was discussed. An optimal catalyst was found by SWIFT method, in which the trained network was used as the objective function. The yield of C2 could be 20.77% when reacting on the optimal catalyst, and the result was better than that of any catalyst in the training group and the testing group.