Design of an intelligent system for modeling and optimization of perovskite-type catalysts for catalytic reduction of NO with CO

Abstract

In this paper, a hybrid artificial neural network-genetic algorithm (ANN-GA) method was applied to design and optimize a perovskite catalyst for the reduction of NO with CO. A series of perovskite-type oxides with the general formula of La_1-xSr_x(Cu_1-yMn_y)_1-αPd_αO₃ were investigated. Catalysts were synthesized via the sol-gel auto-combustion method. The effects of four design parameters (x, y, α, and calcination temperature) and reaction temperature as an operational variable on NO conversion were investigated by modeling the experimental data obtained in the experimental design. Based on the results, the optimum neural network architecture predicted NO conversion data with an acceptable level of correctness. The optimum neural network architecture was used as a capability function for the genetic algorithm to find the optimal catalyst. For catalyst optimization, the Pd mole fraction was set to 0.02. The values of other parameters in the optimum catalyst were as follows: Sr mole fraction of 0.175, Mn mole fraction of 0.596, and calcination temperature of 674.89°C. To investigate the structure, morphology, specific surface area, and reducibility, the catalysts were characterized by XRD, BET, H₂-TPR, XPS, and SEM.