Melt index modeling with support vector machines, partial least squares, and artificial neural networks

Abstract

This article presents the application of three black-box modeling methods to two industrial polymerization processes to predict the melt index, which is considered an important quality variable determining product specifications. The modeling methods covered in this study are support vector machines (SVMs; known as state-of-the-art modeling methods), partial least squares (PLS), and artificial neural networks (ANNs); the processes are styrene-acrylonitrile (SAN) and polypropylene (PP) polymerizations currently operated for commercial purposes in Korea. Brief outlines of the modeling procedure are presented for each method, followed by the procedures for training and validating the models. The SVM models yield the best prediction performances for both the SAN and PP polymerization processes. However, the ANN models fail to accurately predict the melt index when sufficient data are not available for model training in the PP polymerization process. The PLS models are not effective either when applied to the SAN polymerization process, for which the melt index has strong nonlinear functionality with the process variables. The good prediction performance that the SVM models show despite the insufficient data or strong process nonlinearity suggests that SVMs can be effectively used as alternative to PLS or ANNs for modeling the melt indices in other polymerization processes as well.

Synthesis and characterization of a large heterometallic alkoxide cation: Self-assembly and rational route to

“…Nonetheless, in spite of the variety of applications for these species, the possibility of existence of geometrically isolated alkoxide cations, not containing cyclopentadienyl or alkenyl ligands, has been envisaged only very recently. The first representatives of this class were discovered accidentally and they usually contained bulky anions like those in [{LiWO 2 (OC 2 H 4 OMe) 3 [12] was reported. It has also been noticed quite recently that the application of ionogenic complexes, derivates of [Na 2 Fe 2 (O t Bu) 6 (THF) 2 ], is able to promote lactides polymerizations under very mild conditions and with high turnover [13,14].…”

“…The large interest in the chemistry of alkoxides is due to their widespread application in both materials science and organic synthesis [1][2][3]. Metal alkoxides have been applied successfully in particular to promote several organic reactions as catalysts in the polymerization of olefins and lactones [4][5][6].…”

Optimal online soft sensor for product quality monitoring in propylene polymerization process

“…Whereas in our paper, the RMSE value of the Sys-LS-SVM model is 0.0872, with an obviously huge percentage decrease. Next, the standard SVM model was recommended by Han et al [9] from its compared PLS and ANN models. This was quantitatively supported by the smallest RMSE value of 1.51 on the testing dataset.…”

“…The simulation results showed that PLS and support vector regression both exhibited excellent predicting performances even for operating situations accompanying severely frequent grade changes. Han et al [9] introduced three different approaches, including supported vector machine, PLS and back propagation neural network, to estimate the MI and concluded that the standard support vector machine yielded the best prediction. Nowadays, neural networks have been widely applied to model and control dynamic processes because of their extremely powerful adaptive capabilities in response to the nonlinear behaviors [10,11].…”

Melt Index Prediction Based on Adaptive Particle Swarm Optimization Algorithm‐Optimized Radial Basis Function Neural Networks

“…Neural networks have been widely applied to model and control dynamic processes because of their extremely powerful adaptive capabilities in response to nonlinear behaviors [7,8]. Han et al [9] employed three approaches for the MI estimation of the styrene-acrylonitrile (SAN) and propylene polymerization (PP) processes: supported vector machines (SVM), partial least squares (PLS) and artificial neural networks. Shi and Liu [10][11][12] developed several soft-sensor models for MI prediction based on independent component analysis (ICA), multi-scale analysis (MSA) and radial basis function neural networks (RBFN) (ICA-MSA-RBFN), MSA-principal component anylysis (PCA)-RBFN, and weighted least squares SVM (LS-SVM).…”