Data mining macrokinetic approach based on ANN and its application to model industrial oxidation of p-xylene to terephthalic acid

Considering that kinetics and thermodynamics are coupled with heat and mass transfer effects being present in the liquid-phase catalytic oxidation of p-xylene to terephthalic acid (OXTA) in an industrial type of continuous stirred tank reactor (CSTR) and that the time evolution of the concentration of all the interest intermediate and final products of OXTA in the industrial CSTR cannot be obtained to estimate macrokinetic parameters, a novel data mining macrokinetic approach based on artificial neural network (ANN) was proposed to develop the macrokinetic model of OXTA in the industrial CSTR, which mines intrinsic kinetics and transport phenomena information from the sample data collected from OXTA in the industrial CSTR. Firstly, the reaction orders of OXTA are estimated by the mass transfer-free experiment data in the laboratory semi-batch stirred tank reactor. The kinetics of OXTA in the industrial CSTR is assumed to be zeroth-order with respect to gaseous reactants, 0.65-order with respect to p-xylene, and first-order with respect to the other liquid reactants, respectively. Secondly, ANN is employed to model the influence of the reaction parameters on the rate constants of OXTA in the industrial CSTR. Based on the sample data collected from OXTA in the industrial CSTR, heuristic differential evolution algorithm is employed to adjust the weights and thresholds of the rate constant ANN in such a way that it minimizes the prediction error of the macrokinetic model, and thus the optimal weights and thresholds are obtained and the macrokinetic model of OXTA in the industrial CSTR is developed. The reliability of the macrokinetic model was investigated and the satisfactory results were obtained. Further, a generalized macrokinetic approach for multi-phase reaction in industrial reactor was suggested too.