Predicting carbon formation in a methane tri-reformer using artificial neural network

Tri-reforming is a process that combines steam reforming, dry reforming, and partial oxidation of methane to produce syngas. If the methane to steam, CO۲, and H۲O ratios are not properly adjusted, coke formation occurs. In this article, the equilibrium thermodynamic amount of coke formation has been calculated through the simulation using the Gibbs reactor at different conditions by changing variables of temperature (۵۰۰-۱۰۰۰℃), pressure (۱-۲۹ bar), O۲/CH۴(۰-۰.۵), H۲O/CH۴(۰-۱) and CO۲/CH۴(۰-۱). Then, the simulation results were used for training an artificial neural network (ANN), Initially, this modeling investigated the impact of varying the number of hidden layers and neurons in each layer, with the results indicating that a configuration of three hidden layers, each with ۱۲ neurons, yielded the best performance. and some parts of the data were used for testing. The coefficient of determination (R۲) for the train, test, and validation sets are ۰.۹۹۹۸۵, ۰.۹۹۹۸۴, and ۰.۹۹۹۸۴, respectively. Consequently, the results demonstrate coke formation can be predicted using the ANN with high precision.

Keywords:

Tri-reforming , Coke formation , Process simulation , Artificial neural network