The Rawlings Research Group -- Rahul Bindlish

James B. Rawlings Research Group

Rahul Bindlish
Ph.D., University of Wisconsin-Madison, 2000

Publications

1: Rahul Bindlish and James B. Rawlings.
Target linearization and model predictive control of polymerization processes.
AIChE J., 49(11):2885-2899, November 2003.
2: Rahul Bindlish, James B. Rawlings, and Robert E. Young.
Parameter estimation for industrial polymerization processes.
AIChE J., 49(8):2071-2078, August 2003.
3: Rahul Bindlish and James B. Rawlings.
Model predictive control of a prototypical copolymerization process.
Annual AIChE Meeting, Dallas, Texas, November 1999.
4: Rahul Bindlish, James B. Rawlings, and Robert E. Young.
Parameter estimation in a dynamic model of a copolymerization process.
In Proceedings of the 1997 American Control Conference, pages 2424-2428, 1997.

Thesis Abstract

Modeling and Control of Polymerization Processes

This dissertation presents modeling and control of prototypical industrial polymerization processes in the presence of disturbances and plant-model mismatch. The process model consists of the material and energy balances for the reactor, and simplified dynamic model for the downstream separator. The kinetic expressions are simplified by using the quasi-steady-state assumption for live polymer chains and the moments of chain length distribution for the live and dead polymer chains. The kinetic parameters in the process model are estimated from industrial data sets using the maximum likelihood method to give a validated fundamental model.

The process model is used to develop a model predictive controller (MPC) for the copolymerization process with the following typical characteristics

Plant operations: plant start-up, product grade change, regulation around a setpoint
Regulated outputs: polymer production rate, copolymer composition and viscosity
Manipulated variables: vinyl acetate feed rate, acetaldehyde feed rate, coolant temperature
Measured disturbances: feed inhibitor, feed temperature

A target linearization model predictive controller gives superior performance compared to a linear model predictive controller during different plant operations.

The performance of the model predictive controllers is further evaluated using the theoretical benchmark of a minimum variance controller based on the process model. The modeling inadequacies that impose limitations on the performance of the feedback controllers are greater for the linear model in comparison to the non-linear model based on the correlation analysis of the process data. Correlation patterns also indicate that better feedback controller performance may be obtained by incorporating a dynamic disturbance model for feed inhibitor and feed temperature in the target linearization model predictive controller.

Personal Web Page: jbrwww.che.wisc.edu/~bindlish

University of Wisconsin
Department of Chemical Engineering
Madison WI 53706