, MODELLING AND CONTROL OF SUSTAINED OSCILLATIONS IN THE CONTINUOUS EMULSION POLYMERIZATION OF VINYL ACETATE by Mark James Pollock A Thesi s Submitted to the Schoo~ of Graduate Studies in Partial Fulfilment of the ReQui~ments f or the Degree Doctor of Philosophy c, McMaster University tee), 1983.. )
- MODELLING AND CONTROL OF SUSTAINED OSCILLATIONS IN THE CONTINUOUS EMULSI~ POLYMERIZATION OF VINYL ACETATE It
To my wife, Fran for her support and encouragement ~. (
\ \ \" ''I l DOCTOR OF PHILOSOPHY (Chemical Engineering) McMaster University Hamilton, Ontario Title: Modelling and Control of Sustained Oscillations in the Continuous Emulsion Polymerization of Vinyl Acetate r, Author: Mark Ja~es Pollock, B.Eng. (McMaster uni~'s,(ty) \ \ Supervisors: Dr. A. E. Hamielec, Dr. J. F. MacGregor Number of Pages: :~i,258 i i
\ ( ABSTRACT The population balance model of Kiparissides (1978) for the continuous emulsion polymerization of vinyl acetate, has been extended to " predict molecular weight moments and has been corrected,fcr induction ) time and particle shrinkage due to dens'ity changes. -- - was successf~lly The developed model able to simulate experimental results from Kiparissides (1978) and Greene et al.(1976) for the continuous emulsion polymerization of vinyl acetate. Application of the model to the batch data of Keung (1974) had reasonable access as well. The model was ext~nded to the continuous emulsion polymerization of styrene which follows different nucleation kinetics and was able to predict'the average conversions and particle diameters for the data of Brooks et al.(1978). Little success was achieved in predicting the small experimentally observed oscillations for the styrene system. ;;,~, To eliminate the sustained property oscillations in the vinyl acetate system Linear-Quadratic stochastic optimal control theory was applied to the model. Due to extreme non-linearities inherent in the system, this approach was shown to be inadequate. Instead, the reaction system was redesigned to include~ small con!inuous seeding reactor with monomer and water bypass. Model predictions indicated that the redesigned, system eliminated the oscillations. Experimental testing of the redesigned system verified that a dramatic improvement in stability was possible. The redesigned reactor configuration was also shown to be more iii
flexible in controlling particle sizes and conversion th~ough of ~~ypass. use A method for selecting the optimum sensors was developed. It was shown that the measurement combination providing the most information would be conversion (as currently available from an on-line density meter) and some measure of weight average molecular weight..... "\ I i I ".. iv \..
I ACKNOWLEDGEMENTS contributions: The author would like to thank the following people for their.his research directors, Drs. A. E. Hamielec and J. F. MacGregor for their direction, experience and encouragement throughout this work. Ms. Stephanie Lowe for her patience in typing of this manuscript. My wife, my parents and my wife's parents for their moral support and encouragement during this work. Finally, I wish to thank the Department of Chemical Engineering and NSERC for their financial support during this research. ) v
TABLE OF CONTENTS DESCR IPTI VE NOTE ABSTRACT ACKNOWLEDGEMENTS TABLE O~ONTENTS LIST OF FIGURES LIST OF TABLES ii iii 'r.l xi CHAPTER 1: INTRODUCT ION 1 CHAPTER 2: OVERVIEW OF EMULSION POLYMERIZATION THEORY 2.1 Introduction 2.2 Physical Picture 2.3 Emulsion Polymerization Mechanism 2.4 Smith-Ewart Theory 2.4.1 St age 2.4.2 Stage II 2.4.3 Modified Smith-Ewart Theory 2.5 Nucleation Mechanisms 2.6 Overall Kinetic Picture for Particle Size Development f J.1 4 <1 '7 10 10 13 10 H< 20 2.7 Overall Kinetic Picture for 22 Molecular Weight Development CHAPTER 3: MATHEMATICAL MODELLING OF CONTINUOUS EMULSION 25 POLYMERIZATION REACTORS 3.1 Introduction 25 3.2 Dynamic Phenomena in Emulsion Polymerization CSTR's 27 vi
TABLE OF CONTENTS (CONT.) 3.3 Modelling of Continuous Reactor1 3.4 Model Oevelopment for Vinyl Acet;re 3.4.1 Balances for Initiator, Emulsifier, Monomer 31 and Oligomeric Radicals 3.4.2 General Property Balances 37 3.4.3 Particle Size Oeve1opment 40 3.4.4 Molecular Weight Oeve1opment 47 3.4.5 Summary of Model Equations ~i 28 30 ~" 3.5 Model Applications 3.5.1 3.5.2 3.5.3 Parameter Estimation for PSD I States Off-Line Prediction of Particle Size Distribution Batch Reactor Operation 3.5.4 Reactor Start-up and Multiple Seed Reactors 75 3.6 Extension of Model ~o Styrene Kinetics 79 3.6.1 Introduction 3.6.2 Styrene CSTR Modelling 3.6.3 Styrene Model Development 3.6.4 Simulation Results for Styrene CHAPTER 4: APPLICATION OF ADVANCED CONTROL THEORY TO CONTINUOUS EMULSION POLYMERIZATION 4.1 Introduction 4.2 Control of Polymerization Reactors 4.2.1 Batch Reactor Control 95 57 57 65 69 79 79 00 85 95 95 95 4.2.2 Control of Continuous Polymerization Reactors '98 vii
TABLE OF CONTENTS (CONT.) Page 4.3 Application of Linear Quadratic Optimal Stochastic Cont ro 1 Theory 4.3.1 Introduction 4.3.2 Model Linearization and Discretization ~ 4.3.3 State Estimation by Kalman Filtering 4.3.4 General Linear Quadratic Stochastic Optimal Control 4.3.5 Application of Theory to Vinyl Acetate Model 4.3.6 Summary 4.4 Simulation Results.;". 4.5 Conclusions CHAPTER 5: CHOICE OF OPTIMUM SENSORS., 5.1 Introduction 5.2 Observability 5.3 5.4 Application of Kalman Filter Theory Simulation Results 5.4.2 Sensor Combinations 5.4.3 Weight Average Molecular Weight Sensor 5.5 Conclusions 6.1 Introduction t> 5.4.1 Individual Sensors CHAPTER 6: REDESIGN OF LATEX REACTOR SYSTEM 6.2 System Redesign 6.2.1 Problems and Goals 6.2.2 Literature Survey 101 101 102 106 108 1 111.120 120 134 135-135 138 139 142 145 151 154 157 159 159 159 159 161 viii