Melt Index Prediction Based on Two Compensation by Compound Basis Function Neural Network and Hidden Markov Model

DOI

10.2991/meic-14.2014.192How to use a DOI?

Keywords

integrating data processing; compound basis function Neural Network; HMM; exponential smoothing; melt index prediction Of polypropylene

Abstract

To estimate the Melt Index(MI) value accurately and quickly in the quality control process of polypropylene (PP), the paper proposed a forecast model of MI (PHKT-GRBFNN-CRBFNN-HMM) integrating the technologies of data mining, model constructing and two error compensation based on the data. Firstly applied the integrating data processing algorithm including PCA, Holt Exponential Smoothing, Kernel Density Estimation and Time-Variable scale Weighting to mine the data deeply to extract the useful information of the data; Then constructed the MI NARMA prediction model based on Gaussian Radial Basis Function Neural Network and Compound RBFNN on the basis of the data mining; Due to Markov property of the error sequence, used Hidden Markov Model to predict the error as the second compensation for the MI prediction values. The proposed model has been checked based on a real plant history data and the MRE(%), RMSE, STD and TIC of the generalization database is respectively 1.40, 0.045, 0.0457 and 0.0088. The results indicate that the proposed model has better comprehensive characteristics and is worth popularization and application in the PP industry process.

Copyright

© 2014, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Hongmei Chen
AU  - Xinggao Liu
PY  - 2014/11
DA  - 2014/11
TI  - Melt Index Prediction Based on Two Compensation by Compound Basis Function Neural Network and Hidden Markov Model
BT  - Proceedings of the 2014 International Conference on Mechatronics, Electronic, Industrial and Control Engineering
PB  - Atlantis Press
SP  - 856
EP  - 863
SN  - 2352-5401
UR  - https://doi.org/10.2991/meic-14.2014.192
DO  - 10.2991/meic-14.2014.192
ID  - Chen2014/11
ER  -