Optimization for Preparation Conditions of Mn-Ce Catalyst Based on BP Artificial Neural Network Model

Three influencing factors (roasting temperature, roasting time, and metal ratio) which affect the preparation conditions of Mn-Ce catalysts for catalytic wet air oxidation was investigated. A BP artificial neural network model was established, in which the input conditions were selected as roasting temperature, roasting time, and metal ratio, and the output condition was TOC removal of n-butyric. The highest TOC removal was regarded as the optimization aim, along with constraints of each factor's bounds. The model validation results showed that there was only less than 5% of average relative deviation existed between the values of BP model predicted and experimental ones. The determination coefficient between the fitting curve and the Nash-Suttcliffe simulation efficiency coefficient (NSC) were 0.8324 and 0.8116 (>0.80) respectively, indicating the model predicted well. Meanwhile, two-factor and three-factor optimization of Mn-Ce catalyst preparation was executed through genetic algorithms, and the value of TOC removal over catalytic wet air oxidation of n-butyric could increased by more than 10% compared to the experimental one under the optimal reaction conditions.