| 초록 |
In this study, we presents applications of various black-box modeling technologies to industrial polymerization processes for modeling the melt flow indices depending on which the product qualities are mainly determined. Several state-of-the-art modeling methods including support vector machines, which are introduced recently and known as powerful modeling methods, are adopted to model the melt indices of two commercial polymers: polypropylene and styrene-acrylonitrile. The melt flow indices are well estimated using the black-box models constructed using partial least squares, artificial neural networks, and support vector machines. However, the comparison of these models showed that the support vector regression with a radial bias kernel outperformed the other models. The black-box modeling methods for estimation of the melt indices proposed in this study also provide guidance in developing inferential models to estimate melt indices in various polymerization processes. |
