MODEL DISCRIMINATION VIA DESIGNED EXPERIMENTS - DISCRIMINATING BETWEEN THE TERMINAL AND PENULTIMATE MODELS ON THE BASIS OF COMPOSITION DATA

In this paper, computer simulations have been used to study the application of statistical model discrimination methods to the modeling of copolymerizations. Statistical model discrimination methods describe how experiments should be designed and analyzed to obtain the maximum possible information on the strengths of competing models. The potential benefits of applying model discrimination methods are described by contrasting them with past work on the modeling of copolymerizations. Three model discrimination techniques (Buzzi-Ferraris and Forzatti,(15) exact entropy,(14) and Hsiang and Reilly(21)) have been applied to the systems styrene-acrylonitrile, styrene-methyl methacrylate, and styrene-butyl acrylate. Simulated copolymer composition data were used to discriminate between the terminal and penultimate models for these systems. Data are presented to show that the computer simulations are capable of accurately predicting copolymer composition, sequence distribution, and rate data. The results of the simulations show that model discrimination techniques, particularly the Buzzi-Ferraris and Forzatti(15) method, are capable of discriminating between the terminal and penultimate models on the basis of copolymer composition data. The Buzzi-Ferraris method is capable of deciding between the two models in fewer experiments than have been performed in past research and is capable of detecting explicit penultimate effects which are smaller than that found by Hill et al.(30) for the system styrene-acrylonitrile. This work on copolymer composition suggests that model discrimination methods will improve the modeling of copolymer systems and that composition data may be more useful in discriminating between copolymer models than previously thought.