The Application of the PCA Based on Neural Network in Modeling Study of P-xylene Oxidative Side-reaction

The model of p-xylene (PX) oxidative side -reaction was proposed in this paper. Variables that influence the oxidative side-reaction had been studied using principal component analysis method based on the on-line data, the dominant factors were six. The Neural network technique was employed to model the relationship between the main operation conditions (i.e. residence time, reaction temperature and catalyst concentration) and the content of carbon dioxide in the reaction off-gas based on the factors. Furthermore, put forward the combustion expanse model of acetic acid and PX. And the sum squared error of the obtained model is 3.75%.