Research Information System
Computer Aided Chemical Engineering, vol.21, no.C, pp.1617-1622, 2006 (Scopus)
Acrylic fiber is commercially produced by free radical polymerization, initiated by a redox system. Industrial production of polyacrylonitrile is a variant of aqueous dispersion polymerization, which takes place in homogenous phase under isothermal conditions with perfect mixing. The fact that the kinetics is a lot more complicated than that of ordinary polymerization systems makes the problem of controlling molecular weight a difficult one. On the other hand, abundant data is being gathered in industrial polymerization systems, and this information makes the neural network based controllers a good candidate for a difficult control problem. In this work, neural network based control of continuous acrylonitrile polymerization is studied, based on our previously developed new rigorous dynamic model for the polymerization of acrylonitrile. Two typical neural network controllers are investigated: model predictive control and NARMA-L2 control. These controllers are representative of the variety of common ways in which multilayer networks are used in control systems. As with most neural controllers, they are based on standard linear control architectures. The concentration of bisulfite fed to the reactor as the manipulated variable and weight average molecular weight of the polymer as an output function are used in control studies. The results present a comparison of two common neural network controllers, and indicate that the model predictive controller requires larger computational time. Furthermore, the model predictive controller involves difficulties in determining the values for the weighting factor and the prediction horizons. The NARMA-L2 controller requires minimal online computation. © 2006 Elsevier B.V. All rights reserved.