Melt flow index estimation using neural network models for propylene polymerization process

Thesis (Sarjana Pendidikan (Pengajaran Bahasa Inggeris sebagai Bahasa Kedua)) - Universiti Teknologi Malaysia, 2013One of the major challenges in polymerization industry is the lack of online instruments to measure polymer end-used properties such as melt flow index (MFI). As an alternative to the online instruments and conventional laboratory tests, these properties can be estimated by using a model based-soft sensor. This research presents models for soft sensors to measure MFI in industrial polypropylene loop reactors by using the artificial neural network (ANN), hybrid FP-ANN (HNN) and stacked neural network (SNN) models. The ANN model of the two loop reactors was developed by employing the concept of Feed-Forward Back Propagation (FFBP) network architecture using Levenberg-Marquardt training method. Serial hybrid FP-ANN (HNN) models were developed in this study. The error between actual MFI and simulation MFI from FP model was fed into the HNN model as one of the input variables. To construct the stacked neural network (SNN) model, two layers were needed: 1) level-0 generalizer output comes from a number of diverse ANN models and 2) level-1 generalizer was developed using the results of level-0 generalizer with additional input variables. All models were developed and simulated in MATLAB 2009a environment. The simulation results of the MFI based on the ANN, HNN, and SNN models were compared and analyzed. The HNN model is the best model in predicting all range of MFI with the lowest root mean square error (RMSE) value, 0.010848, followed by ANN model (RMSE=0.019366) and SNN model (RMSE=0.059132). When these three models (ANN, HNN, and SNN) were compared, the SNN model shows the lower RMSE for each type of MFI studied