The application of hybrid DOE/ANN methodology in lumped kinetic modeling of Fischer–Tropsch reaction

In this article the application of design of experiment coupled with artificial neural networks in kinetic study of CO hydrogenation reaction and a method of data collection/fitting for the experiment are presented. The kinetic experimental data has been collected from two factors of central composite designs, pressure and feed ratio in different temperatures. Response surface and artificial neural network models have been constructed based on the DOE points and used for generating simulated data. Then different data sets were used to fit LHHW kinetic rates of CO disappearance. The application of the neural networks to solve the problem of correlation of the rate equation parameters has been addressed. The results of kinetic modeling with simulated data sets from ANN and RSM models were compared with randomly collected experimental and the DOE data sets. It was observed that three rate equations were able to fit the data sets and kinetic modeling with the proposed DOE-based and simulated data, are very efficient for assessing the quality of the models. However, conducting kinetic modeling with ANN simulated data presented the best results.

► Lumped kinetic study of Fischer Tropsch on Fe-Co catalysis has been presented.
► Kinetic study has been carried out with experimental data collected according to the proposed DOE and also RSM and ANN simulated data.
► A strategy of curve fitting with ANN simulated data were proposed.
► Three LHHW rate equations were able to fit the data.
► It was concluded that our proposed DOE is efficient and curve-fitting with ANN simulated data presented the best results.