Modeling and Optimization of Propane Selective Oxidation to Acrylic Acid Over Mo 1 V 0.3 Te 0.23 NB 0.12 O X Catalyst Using Artificial Neural Network and Box-Behnken Design

Abstract

The prediction capability of response surface methodology (RSM) and artificial neural network (ANN) models for propane selective oxidation to acrylic acid (AA) over Mo₁V_0.3Te_0.23Nb_0.12O_x catalyst was investigated in this work. 15 experimental runs based on the Box-Behnken design (BBD) were employed to study the effects of temperature (380 to 500 °C), superficial velocity (33.3 to 66.7 mL (min g_cat)^-1), (O₂)/(C₃H₈) ratio (1 to 3) and their interactions on propane conversion, AA selectivity and CO_x selectivity. The quadratic polynomial BBD equations and the feed-forward back propagation ANN models were developed based on the designed experimental data. Statistical analysis; coefficient of determination (R2), mean absolute error (MAE) and analysis of variance (ANOVA) illustrated that there was acceptable adjustment between BBD and ANN predicted responses as compared to experimental data. While, the ANN model showed a clear preference and generalization capability over BBD model in the case of experimental data set which were not used to training the models. In addition the optimum conditions were found to be temperature (461.7 °C), GHSV (51.9 mL (min g_cat)^-1) and (O₂)/(C₃H₈) ratio (2.1) which were determined by desirability function approach. In these conditions, propane conversion of 15.2%, AA selectivity of 32% and CO_x selectivity of 44% which obtained experimentally were in reasonable agreement with predicted responses.

Chemical Engineering Research Bulletin 21(2019) 1-19