Premium
Neuro‐Fuzzy Prediction of Fe‐V 2 O 5 ‐Promoted γ‐Alumina Catalyst Behavior in the Reverse Water–Gas–Shift Reaction
Author(s) -
Takassi M. A.,
Gharibi Kharaji A.,
Esfandyari M.,
KoolivandSalooki M.
Publication year - 2013
Publication title -
energy technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.91
H-Index - 44
eISSN - 2194-4296
pISSN - 2194-4288
DOI - 10.1002/ente.201200012
Subject(s) - water gas shift reaction , mean squared error , catalysis , approximation error , fuzzy logic , gradient descent , chemistry , range (aeronautics) , analytical chemistry (journal) , mathematics , materials science , algorithm , computer science , artificial neural network , statistics , chromatography , organic chemistry , machine learning , artificial intelligence , composite material
The application of an Fe‐V 2 O 5 ‐promoted γ‐alumina catalyst was studied for the reverse water–gas–shift reaction. The reaction was performed under a wide range of synthesis conditions, temperatures, and CO 2 /H 2 ratios in a batch reactor. The experimental results were compared with neuro‐fuzzy simulation results. The Mamdani algorithm (a gradient descent algorithm) was applied to train the fuzzy system, and a test set was used to evaluate the performance of the system by applying the efficiency coefficient ( E ), root‐mean‐square error (RMSE) and mean absolute error (MAE). The predicted values from the model are in good agreement with the experimental data. The outcome of this study demonstrates how the neuro‐fuzzy method, as a promising prediction technique, can be effectively applied to the reverse water–gas–shift reaction. This study applies neuro‐fuzzy modelling to predict the product composition of CO, CO 2 and CH 4 in the reverse water–gas–shift reaction, for which the input vector was three‐dimensional (including the variables of operating temperature, time, and CO 2 /H 2 ratio) for 34 different experiments, and the output vectors consisted of CO, CO 2 and CH 4 conversions.