Modelling of polymerization reactors: deterministic and by neural networks

Author(s)

Sheila Contant

• Document type: Doctoral Thesis
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Química
• Defense date: 2007-02-23

Examining board members

Liliane Maria Ferrareso Lona; Rubens Maciel Filho; Reginaldo Guirardello; Amilton Martins dos Santos; Roberto de Campos Giordano

• Advisor: Liliane Maria Ferrareso Lona

Abstract

In this work different polymerization processes were studied: (1) styrene homopolymerization and styrene/methyl methacrylate copolymerization in emulsion in the conventional freeradical process, and (2) styrene homopolymerization in bulk in the nitroxidemediated controlled/living freeradical process. Modelling was developed using two different approaches: initially deterministic models were developed in each case, and using results from these models neural networks were trained to the inverse modelling of the processes. In the deterministic modelling, computational programas were developed to the emulsion polymerizations, and simulations were performed for different operating conditions. A modified computational program from the literature was used in the controlled polymerization in bulk. In all cases, large databases of kinetic parameters to all the compounds present were searched. A modified computational program previously developed was used in the work with neural networks. Neural networks were used to the inverse modelling of the processes, and were trained to predict operating conditions that could lead to production of polymers with specific properties. The two methodologies used in the mathematical modelling were able to extract important and different information from the polymerization processes studied, showing its potential to an efficient aplication in the polymerization area (AU)