ARPN Journal of Science and Technology Logo

ARPN Journal of Science and Technology >> Volume 7, Issue 2, November 2017

ARPN Journal of Science and Technology


System Identification of Nylon-6 Caprolactam Polymerization Process

Full Text Pdf Pdf
Author D.O. Araromi, B. Adegbola
ISSN 2225-7217
On Pages 99-107
Volume No. 4
Issue No. 2
Issue Date March 01, 2014
Publishing Date March 01, 2014
Keywords NARX, System identification, nonlinear dynamics, ASPEN Tech, MATLAB, Hammerstein and Weiner



Abstract

The first Engineering Plastics are Polyamides or nylons and still represent the biggest and most significant class of these types of material. Nylon-6 and nylon-6, 6 are the first two commercial polyamides and are still the most important polyamides with respect to their production volumes. Modeling Nylon 6 polymerization process from first principle constitutes serious computational challenges due to the non-linear nature of the process. This work therefore developed simplified model for the process using time series modeling techniques. The steady state simulation was done using ASPEN polymer plus to measure the effect of the input variable on the outputs. The input variable was the temperature of water feed and the output variable were Caprolactam conversion, mass fraction of nylon produced in the product stream and mass flow rate of nylon 6. The steady state flow sheet was then exported to ASPEN dynamics to generate dynamic data needed for the process identification generated. The data generated from ASPEN dynamics were exported into the MATLAB environment in order to identify the process model. The two time series modeling techniques used are Nonlinear Auto Regressive with exogenous input (NARX) model and Hammerstein and Weiner model. The models developed using NARX model of the structure (10, 7, 1) for conversion gave a best fit of 71.78% while the one for mass flow rate of Nylon produced using structure (2,2,1) gave a best fit of 88.36% and that of produced Nylon 6 mass fraction using the structure (5,5,1) gave a best fit of 75.33% . The models developed using Hammerstein and Weiner for the Nylon mass fraction in the product stream using input nonlinear estimator of four (4) and output non linear estimator of two (2) gave a best fit of 77.98%, while the one of conversion of caprolactam using input nonlinear estimator of three (3) and output non linear estimator of three (3) gave a best fit of 88.67% and the produced Nylon 6 mass flow rate using input nonlinear estimator of four (4) and output non linear estimator of two (3) gave a best fit of 87.21%. The residual analysis results show that the model is a good model because the residual autocorrelation function falls within the confidence interval of 99%. The results prove that the nonlinear ARX model and Hammerstein-Weiner model are able to capture the nonlinear dynamics behavior of the process. Nonlinear ARX model is performs better than Hammerstein-Weiner for Nylon 6 mass flow rate while Hammerstein -Weiner model performs better Nonlinear ARX for Caprolactam conversion and Nylon 6 mass fraction.


Back
Seperator
    Journal of Computing | Journal of Systems and Software     
Copyrights
2014 ARPN Publishers