A simulation study on NOx reduction efficiency in SCR catalysts utilizing a modern C3-CNN algorithm

•The simulation of De-NOx system by SCR catalyst is realized by AI technology.•BPNN model exhibits high predictive capability for the SCR simulation.•C3-CNN can improve computational accuracy without adding more model parameter scale further. The simulation of De-NOx system by selective catalytic reduction (SCR) catalyst is very important in industrial application, however, the simulation is always highly time-consuming. In this work feed-forward back-propagation artificial neural network (BPNN) and Cross-Channel Communication Convolutional Neural Network (C3-CNN) algorithm are first proposed as a tool for numerical simulation on De-NOx system by selective catalytic reduction (SCR) catalyst. Initially, one-dimensional Computational Fluid Dynamics (CFD) model allowed the analysis of the contribution of several parameters in SCR reaction (gas velocity, ammonia-to-nitrogen ratio, temperature, and channel length) to DeNOx efficiency. Then, 3600 derived data samples are trained by BPNN neural network which shows a high predictivity (R2 = 0.95542). Additionally, the influence on simulation results of algorithm parameters is analyzed. Furthermore, the introduced Cross-Channel Communication Convolutional Neural Network (C3-CNN) algorithm enhanced the accuracy, efficiency and reduced training time for the De-NOx system simulation.