This paper discusses the techniques of soft-sensor and strategy of advanced control of the oxidation reaction in PTA process. At first, the properties and improved methods of regression analysis and neural networks which are the common methods of getting the soft-sensing models are discussed. After choosing the right methods, this paper gives some data to prove that the generalization performance of ANN in 4-CBA simulation has been greatly improved. Based on the former research, a kind of close-loop model rectification way is advanced and testified effective in industrial application. At last, the APC system project is introduced, the on-line 4-CBA soft-sensing model is proved to have high accuracy, and the new control strategy shows the improvement in the oxidation process. |