Desenvolvimento de modelo de inferência de CCl4 em uma unidade de produção de MVC

Abstract

In the production of Vinyl Monochloride (VMC), 1,2-Dichloroethane (1,2-DCE) must undergo a purification process, as the 1,2-DCE cracking process is highly susceptible to inhibition and deposition for traces of impurities, it must be maintained at high levels of purity. One of the impurities present in DCE is carbon tetrachloride, which is an important promoter of the cracking reaction, due to the pyrolysis of DCE being catalyzed mainly by the chlorine present in the CCl4 molecule, however, its presence also accelerates the formation of coke, which causes undesirable effects in the process such as the reduction of the operation cycle. Therefore, online monitoring of the presence of CCl4 at the output of the 1,2-DCE purification process enables greater control of the cracking reaction, prevents production losses, enables process optimization, and generates an increase in campaign time of the cracking furnaces. The importance of this realtime analysis, the need to control this impurity and the difficulties in maintaining the analytical instruments motivated this work to develop an inference model for the analysis of CCl4. The statistical model was developed using the supervised machine learning method, using Aspen and Python tools. The results obtained indicate paths of input variables for inferring the composition of CCl4, in addition to a viable virtual analyzer model.