Dry reforming of methane is a potentially important process to convert the greenhouse gases carbon dioxide and methane simultaneously to syngas (CO + H2). The most serious problem with the dry reforming of methane is carbon deposition, so preparation parameters of the citric acid method were surveyed to prepare an active Co _ MgO catalyst with low carbon deposition using design of experiment, artificial neural network and grid search. The preparation parameters such as Co loading, amount of citric acid, calcination temperature, and pelletization pressure were determined according to an L9 orthogonal array. After 9 data sets of the parameter activity were designed and measured in a conventional pressurized fixed bed reactor, an artificial neural network was constructed. The optimum composition was determined by a grid search and verified experimentally to be active with a small amount of carbon deposition. Design of experiment combined with an artificial neural network and grid search was useful for catalyst development.

No abstract

Optimization of catalyst composition using a genetic algorithm (GA) is intended to increase the activity in a series of repetitive steps consisting of determination of the composition, catalyst preparation, activity test and feedback to the program. The laborious steps of catalyst preparation and activity test can be replaced by calculation provided that a radial basis function network (RBFN) trained using experimental results is used to evaluate the fitness of the catalyst code.Optimization of the Cu/Zn/Al/Sc ratio of mixed oxide catalyst for methanol synthesis from syngas was simulated. In the simulation, activity was calculated by equations fitted to some experimental results to evaluate the fitness for use in the genetic algorithm program. Data of catalyst composition for input and the STY for output totalling 69-92 points were necessary for successful mapping of the catalytic activity. The network then was trained using 92 experimental results. The highest activity of the catalyst optimized by GA and RBFN was higher than that optimized by GA only. The combination of catalyst design by genetic algorithm and the activity evaluation by RBFN is promising for highly efficient catalyst screening.

Preparation conditions of Co _ MgO catalyst for methane dry reforming were optimized to maximize the CO yield. Response surface method, composed of design of experiment and regression method, was applied. For the best regression, genetic programing, radial basis function network, and polynomial equation were compared. As a result, genetic programing succeeded to express explicitly the non-linear relationship between catalyst preparation parameters and catalyst performance, and the prediction error was the least. Genetic programing modifies and optimizes many functions by genetic algorithm to fit the experimental results. Radial basis function network could express the non-linear relationship, but the complexity of the function hindered the understanding of the importance of catalytic parameters. While polynomial equation proposed the explicit equation, the accuracy was not sufficient to show the non-linear relationship. Genetic programing is the most suitable regression method of response surface for the catalyst development. KeywordsGenetic programing, Cobalt _ magnesia catalyst, Methane dry reforming, Catalyst optimization, Response surface method, Design of experiment