Hybrid Model of Industrial p-Xylene Oxidation Incorporated Fractional Kinetic Model with Intelligent Models

作者：Dong, Yaming; Yan, Xuefeng^*

来源：Industrial & Engineering Chemistry Research, 2013, 52(7): 2537-2547.

DOI：10.1021/ie3024033

• PHASE CATALYTIC-OXIDATION
• ARTIFICIAL NEURAL-NETWORK
• SELECTIVE CO OXIDATION
• GLOBAL OPTIMIZATION

A hybrid modeling approach for the oxidation of p-xylene (PX) to terephthalic acid (TA) in an industrial continuously stirred tank reactor (CSTR) was proposed. First, considering that the reaction factors have a highly nonlinear effect on the kinetic rate constants, support vector regression (SVR) was used to construct the rate constant model based on the data from a laboratory semibatch reactor (SBR). Second, due to the significant difference between the nature of PX oxidation in the laboratory SBR and that in the industrial CSTR, this model was extended via artificial neural network (ANN) to correlate the rate constants in the industrial CSTR with that in the laboratory SBR, and a genetic algorithm is employed to obtain the optimal weights and thresholds by direct fitting of the difference between the hybrid model results and the real industrial data. Further, the reliability of the hybrid model was investigated and satisfactory results were obtained.

• 出版日期2013-2-20
• 单位华东理工大学