Computational fluid dynamics (CFD), artificial neural network (ANN) and genetic algorithm (GA) as a hybrid method for the analysis and optimization of micro-photocatalytic reactors: NOx abatement as a case study

In this work the hybrid CFD-ANN-GA method is proposed as a tool for the analysis and optimization of micro-photocatalytic reactors, taking NOx abatement as a case study. Initially, a 3D CFD model of the microreactor allowed the investigation of the effects of residence time, light intensity, relative humidity and initial NO concentration on the performance of the photocatalytic reaction. Then, an artificial neural network (ANN) was implemented to predict the overall conversion of NO in the micro device. Different ANN structures were developed using data from 256 CFD simulations, and the best structure was chosen based on the performance factors MSE, RMSE and R2. Moreover, a genetic algorithm (GA) was used to find the optimal operating conditions that maximize the NO conversion. The best ANN model consisted of a feed-forward back-propagation structure with three layers and 11 neurons in the hidden layer (4:11:1), logsig-logsig transfer function and training through the Levenberg-Marquardt algorithm. This network presented a high predictivity (R2 = 0.9997), and it was used for optimization by GA to determine the optimum conditions. Based on the optimization results, full NO conversion (100%) was achieved when the residence time, light intensity, relative humidity and initial concentration were 2.12 s, 10 W·m−2, 10%, and 2.09 × 10−8 kmol·m−3, respectively. Furthermore, the most influential variable on the NO conversion prediction was the residence time, with a relative importance of 48.97%. The ANN was then modified to yield two outputs: NO consumption rate and pressure drop. All parameters were kept the same, except the number of neurons in the hidden layer (17). GA was then applied to a multi-objective optimization, aiming to maximize the NO consumption rate while minimizing the pressure drop in the system. The optimal set of operating conditions in this scenario was found based on a Pareto front analysis.