Synthesis of a Neuro-Fuzzy Model of Radical Polymerization Process Kinet-Ics

Abstract

There was considered the kinetic scheme of the processes of methyl methacrylate and styrene polymers synthesis by the radical polymerization method. It is shown that the main difference in the above kinetic schemes is observed in the last stage of the reaction and lies in the mechanisms of breaking the polymer chain – by recombination for polystyrene and by disproportionation for polymethyl methacrylate. There were given the well-known mathematical models describing the kinetic scheme for the polymerization of methyl methacrylate and styrene and showing the dependence of the degree of monomer to polymer conversion on the time. It is shown that each of the presented models contains a large number of empirical constants, the values of which differ by orders of magnitude among different researchers, which greatly complicates the use of these models and often leads to confusion and incorrect results of simulation modeling. In addition, some of the known models do not take into account the phenomenon of auto-acceleration of the process - the gel effect inherent in the radical polymerization of styrene and, to a greater extent, methyl methacrylate. Thus, the necessity of developing a simpler in its structure adequate mathematical model of the processes under consideration based on the methods of neuro-fuzzy modeling was substantiated. The results of studying the rate of polymerization of methyl methacrylate and styrene in a laboratory reactor under different initial conditions were used as initial data in the development of a fuzzy model. With the help of an adaptive neuro-fuzzy system, there were obtained primary models of the Takagi-Sugeno type, after processing which, we obtained the final fuzzy model. The accuracy of the obtained model was verified by conducting a comparative analysis of the results of process simulation with different initial data and results obtained in a laboratory reactor. The good data convergence were shown and recommendations were given for using the model.