인공신경망을 이용한 폴리에틸렌 테레프탈레이트 공정의 모델링

Abstract

Polyethylene terephatalte (PET) is most widely used man made polymer. The diverse uses of PET have made its manufacturing process very attractive to researchers around the world. Solid state polycondensation (SSP) is the most accepted and respected method used for the production of PET. Modeling of PET manufactured through SSP is a complex process involving intensive mathematical modeling. Thus, a need for a model with the use of new computational techniques is indentified in this work. Among the emerging techniques available for modeling complex non-linear chemical processes artificial neural network is found to be the most powerful computational technique for modeling of PET. In this work a feedforward back-propagation artificial neural network (BP-ANN) model is developed for the estimation of the intrinsic viscosity in the production of PET which is manufactured by the SSP. It is shown that the proposed model can predict intrinsic viscosities at various temperatures. The results of numerical simulations obtained by the BP-ANN model show excellent agreements compared with the experimental data. Furthermore, the immense potential of neural networks in chemical industry is shown by providing their applications briefly in distillation and fault diagnosis.