The role of wave-net models in emulsion polymerisation

As the first step in a study to assess the impact of modern control techniques on the performance of polymerisation reactors, an existing kinetic model (written in Fortran) was used to predict the particle size distribution (PSD), particle number and amount of secondary nucleation in an emulsion polymerisation system. However, this model proved far too slow for studying the dynamic behaviour (and advanced control) of such reactors. To circumvent this problem, two approaches were taken. Firstly, the relevant population balance and kinetic equations were solved using the commercial gPROMS package—this gave a very significant reduction in model solution time. Secondly, two input–output wave-net models were developed—the first to estimate an “averaged” particle size distribution, the second to estimate the full particle size distribution. These wave-net models used simulation data obtained from the gPROMS model for the case of styrene polymerisation, and covered a range of initiator and surfactant concentrations, as well as different reaction temperatures. The wave-net models were acceptably accurate (relative to the full kinetic model), although the study showed that extrapolation beyond the available training data can yield poor results.