Analysis of selective CO oxidation over promoted Pt/Al2O3 catalysts using modular neural networks: Combining preparation and operational variables

In this work, modular neural network modeling was applied to analyze the experimental data for the selective CO oxidation over promoted Pt/Al2O3 catalysts. The effects of preparation (Pt wt.% and promoter type) and operational (reaction temperature, feed composition and time on stream) variables on CO conversion were modeled. As a novel approach to neural network modeling of the catalytic performance, the preparation and the operational variables were used together but processed differently in the network so that the accuracy of the model could be improved since the nature of these two variables are quite different although they are both important. Similarly, the continuous (like Pt wt.%) and categorical (like promoter type) variables were also treated in different manner. After the most successful network structure was determined, the relative importances of the preparation and the operational variables as well as their effects on CO conversion were analyzed in detail. It was found that modular neural networks used this way are quite successful in predicting and explaining the experimental results, and they are superior to the monolithic neural network models.

The effects of preparation (Pt wt.%, promoter type) and operational (reaction temperature, feed composition, time on stream) variables on CO conversion were modeled by using modular neural networks for the selective CO oxidation over promoted Pt/Al2O3 catalysts. As a novel approach to modeling of the catalytic performance, the preparation and the operational variables were processed differently to improve the accuracy of the model.Figure optionsDownload high-quality image (84 K)Download as PowerPoint slide