Department of Applied Chemistry, Faculty of Chemistry, University of Tabriz, Iran.
Reactor & Catalyst Research Lab., Department of Chemical Engineering & Petroleum, University of Tabriz, Iran.
J Environ Manage. 2019 May 15;238:360-367. doi: 10.1016/j.jenvman.2019.03.018. Epub 2019 Mar 8.
In the present study, two statistical methods including the response surface method (RSM) and artificial neural network (ANN), were employed for modeling and optimization of selective catalytic reduction of NOx with NH (NH-SCR) over VO/TiO nanocatalysts. The relationship between catalyst preparation variables, such as metal loading, impregnation temperature, and calcination temperature on NO conversion were investigated. The R value of 0.9898 was obtained for quadratic a RSM model, which proves the high agreement of the model with the experimental data. The results of Pareto analysis revealed that three factors including calcination temperature, V loading, and impregnation temperature have a considerable impact on the response. Deduced from the established RSM model, the order of influence on the NO conversion was as follows: calcination followed by V loading and impregnation temperature. The optimum condition of catalyst preparation for maximum NO conversion over VO/TiO nanocatalysts was predicted to be at 0.0051 mol of V loading, an impregnation temperature of 50 °C and a calcination temperature of 491 °C. Moreover, an ANN model was created by a feed-forward back propagation network (with the topology 4, 12 and 1) to model the relation between the selected catalyst preparation variables and NH-SCR process temperature. The R values for training, validation as well as test sets, were 0.99, 0.9810 and 0.9733. These high values proved the accuracy of the AAN model in modeling and estimating the NO conversion over VO/TiO nanocatalysts. According to the ANN model, the relative significance of each variable on NO conversion is calcination temperature, process temperature loading, and impregnation temperature from high to low importance, respectively, corroborating the obtained results from RSM.
在本研究中，采用响应面法（RSM）和人工神经网络（ANN）两种统计方法，对 VO/TiO 纳米催化剂上 NH（NH-SCR）选择性催化还原 NOx 的模型和优化进行了研究。考察了催化剂制备变量（如金属负载量、浸渍温度和煅烧温度）与 NO 转化率之间的关系。二次 RSM 模型的 R 值为 0.9898，证明了模型与实验数据具有高度的一致性。Pareto 分析结果表明，煅烧温度、V 负载量和浸渍温度这三个因素对响应有较大影响。根据建立的 RSM 模型，对 NO 转化率的影响顺序如下：煅烧、V 负载量和浸渍温度。预测 VO/TiO 纳米催化剂上最大 NO 转化率的最佳催化剂制备条件为 V 负载量为 0.0051 mol、浸渍温度为 50°C 和煅烧温度为 491°C。此外，还通过前馈反向传播网络（拓扑结构为 4、12 和 1）创建了一个 ANN 模型，以模拟所选催化剂制备变量与 NH-SCR 工艺温度之间的关系。训练、验证和测试集的 R 值分别为 0.99、0.9810 和 0.9733。这些高值证明了 AAN 模型在建模和估计 VO/TiO 纳米催化剂上的 NO 转化率方面的准确性。根据 ANN 模型，各变量对 NO 转化率的相对重要性依次为煅烧温度、工艺温度加载和浸渍温度，这与从 RSM 得到的结果一致。
