Modelling and control of different types of polymerization processes using neural networks technique: A review

Abstract

Polymerization process can be classified as a nonlinear type process since it exhibits a dynamic behaviour throughout the process. Therefore, it is highly complicated to obtain an accurate mechanistic model from the nonlinear process. This predicament always been a "wall" to researchers to be able to devise an optimal process model and control scheme for such a system. Neural networks have succeeded the other modelling and control methods especially in coping with nonlinear process due to their very conciliate characteristics. These characteristics are further explained in this work. The predicament that is encountered by researchers nowadays is lack of data which consequently lead to an imprecise mechanistic model that scarcely conforms to the desired process. The implementations of the neural network model not only restrict to polymerization reactor but to other difficult-to-measure parameters such as polymer quality, polymer melts index and mixture of initiators. This work is aimed to manifest ascendancy of neural networks in modelling and control of polymerization process.Le processus de polymérisation peutêtre catégorisé comme un processus de type non linéaire, puisqu'il affiche un comportement dynamique tout au long du processus. Par conséquent, il est très compliqué d'obtenir un modèle mécaniste précis du processus non linéaire. Cette situation fâcheuse a toujours représenté un « mur » pour les chercheurs qui souhaitent concevoir un schéma de contrôle et de modélisation du processus optimal pour un tel système. Les réseaux neutres ont succédé aux autres méthodes de modélisation et de contrôle, surtout pour s'occuper du processus non linéaire en raison de leurs caractéristiques très conciliantes. Ces caractéristiques sont expliquées plus en détail dans ce travail. La situation fâcheuse rencontrée par les chercheurs aujourd'hui est le manque de données, ce qui mène par conséquentà un modèle mécaniste imprécis qui està peine conforme au processus souhaité. Les mises en oeuvre du modèle des réseaux neutres se limitent non seulement au réacteur de polymérisation, mais aux autres paramètres difficilesà mesurer, comme la qualité du polymère, l'indice de fluidité du polymère et le mélange des initiateurs. Ce travail chercheà manifester l'ascendance des réseaux neutres en modélisant et contrôlant le processus de polymérisation.

Industrial melt index prediction with the ensemble anti‐outlier just‐in‐time Gaussian process regression modeling method

“…In contrast to traditional neural networks and SVR‐based deterministic modeling methods, the GPR‐based methods provided probabilistic information for prediction. Comparisons of the predicted variance values for the test samples ( ) with two online local modeling methods, ESVC–JGPR and SVC–JGPR, are shown in Figure .…”

“…It should be noted that the soft‐sensor modeling method can also be applied to predict other difficult to measure parameters, such as the polymer quality, polymer melt index, and mixture of initiators. These polymerization processes include the polymerization of methyl methacrylate, the polymerization of nylon 6,6, and rubber‐mixing processes …”

Models in the Polymer Industry: What Present? What Future?

“…This designation refers to models that combine a first principles part and an empirical part (e.g., artificial neural network); the latter usually represents a portion of the process that is incompletely elucidated or that is difficult to describe. These methods date from the mid‐90s and there are by now several reports of successful application to polymerization processes . So far, they have been employed mostly for process modeling and control, but we think that there is an opportunity for product development as well.…”

Melt index prediction with a mixture of Gaussian process regression with embedded clustering and variable selections

“…To this end, nonlinear modeling methods are used. Among other nonlinear modeling methods in MI prediction are artificial neural networks (ANNs), support vector regression, and Gaussian process regression (GPR) …”