modelling and control of a spray drying process
TRANSCRIPT
-
8/20/2019 Modelling and Control of a Spray Drying Process
1/225
Shayantharan Sivarajalingam
Modelling and Control of a Spray
Drying Process
Master’s Thesis, November 2009
-
8/20/2019 Modelling and Control of a Spray Drying Process
2/225
-
8/20/2019 Modelling and Control of a Spray Drying Process
3/225
Shayantharan Sivarajalingam
Modelling and Control of a Spray
Drying Process
Master’s Thesis, November 2009
-
8/20/2019 Modelling and Control of a Spray Drying Process
4/225
-
8/20/2019 Modelling and Control of a Spray Drying Process
5/225
Modelling and Control of a Spray Drying Process,
This report was prepared by
Shayantharan Sivarajalingam
Supervisors
Hans Henrik Niemann-Associate Professor, DTU Electrical EngineeringOle Ravn -Associate Professor, DTU Electrical EngineeringChrister Utzen- GEA Niro,GEA Process Engineering A/S,Process Control
Release date: 30 November 2009Category: 1 (public)
Edition: First
Comments: This report is part of the requirements to achieve the Master of Science in Engineering (M.Sc.Eng.) at the Technical Universityof Denmark. This report represents 30 ECTS points.
Rights: cShayantharan, 2009
Department of Electrical EngineeringAutomation and ControlTechnical University of DenmarkElektrovej building 326DK-2800 Kgs. LyngbyDenmark
http://www.dtu.dk/centre/autTel: (+45) 45 25 35 50Fax: (+45) 45 88 12 95
-
8/20/2019 Modelling and Control of a Spray Drying Process
6/225
Preface
This master’s thesis documents the process and results of the project Mod-elling and Control of a Spray Drying Process by Shayantharan Sivarajalingam.
The project has been conducted at the Technical University of Denmark(DTU) at the Department of Electrical Engineering, Automation and Con-trol in the period from June till November 2009 and represents a workloadof 30 ECTS points.
I would like to express my sincere thanks and appreciation to my supervisorsAssociate Professor Hans Henrik Niemann and Associate Professor Ole Ravnfor their competent guidance and support throughout the development of the ideas in my thesis.
All at once I would like to say thanks to Christer Utzen, for his assistanceduring the thesis work and GEA Niro for giving me the opportunity to workon this real system, which has given me a lot of experience.
Last but not least, I will express my gratitude to my friends: Kristian, Lars,Soaban, Malcolm, Mickey, Varun and my Chellams for their patience andgreat support during the project.
-
8/20/2019 Modelling and Control of a Spray Drying Process
7/225
-
8/20/2019 Modelling and Control of a Spray Drying Process
8/225
Abstract
This Master thesis is about modelling of a spray drying process. In a spraydrying process a liquid feedstock is dried by spraying the feed into heated
air. This process is utilized in the dairy industry, where milk is dried intomilk powder. Moreover, this process is also applied in the chemical andmedical industries.
The quality of the final product can vary, depending on how the system iscontrolled. The purpose of modelling the spray dryer is to use the model fordevelopment and testing of control strategies to the dryer.
The point of reference is a steady state model, which has been developedto a dynamic model of a multi stage dryer with a mixed air flow. Themodel is based on mass and energy equations, as well as product dryingcharacteristics. The model describes the drying conditions in the dryingchamber with regards to temperature and humidity.
The developed model is verified with data from a genuine Multi stage dryerat GEA Niro’s test station. Furthermore, it has been demonstrated thatthe model can be applied in connection with a controller and thus examinevarious control systems.
-
8/20/2019 Modelling and Control of a Spray Drying Process
9/225
-
8/20/2019 Modelling and Control of a Spray Drying Process
10/225
Dansk Resumé
Denne Master speciale omhandler modellering af et tørsprayingsanlæg. Iet tørspraying anlæg bliver et givent flydende stof tørret ved at sende det
flydende stof gennem opvarmet luft. Det er anvendt i blandt andet mejeriindustrien, hvor blandt andet mælk tørres til mælkepulver. Dette er ligeledesanvendt i kemi og medicinal industrien.
Kvaliteten p̊a det færdigtørrede product kan variere alt efter hvordan an-lægget bliver styret. Målet med modelleringen af anlægget er at kunneanvende modellen til videre udvikling og test af reguleringssystemer til tør-spraying anlæg.
Med udgangspunkt i en steady state model, er der blevet en udviklet endynamisk model af en ”multi stage dryer med mixed air flow”. Denne erbaseret på masse og energi balance ligninger og produkt tørrings karakter-stikker. Modellen beskriver tørringsforholdene i tørspraying kammeret medhensyn til temperatur og luftfugtighed.
Den udviklede model er blevet verificeret med data fra Multi stage dryeranlæg hos GEA Niro’s test station. Desuden er det blevet demonstreret atmodellen kan anvendes til at undersøge reguleringssystemer.
-
8/20/2019 Modelling and Control of a Spray Drying Process
11/225
-
8/20/2019 Modelling and Control of a Spray Drying Process
12/225
Contents
List of Figures xiii
List of Tables xviii
1 Introduction 3
1.1 Description of the Problem . . . . . . . . . . . . . . . . . . . 5
1.2 The objective of the project . . . . . . . . . . . . . . . . . . . 6
2 Introduction to the Spray Drying Process and Spray Dryers 9
3 Modeling 133.1 Black Box model . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 White Box model . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Grey Box model . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Modelling a Spray Dryer 17
4.1 Preparations and Assumptions for Modelling . . . . . . . . . 17
4.2 Steady State Mass and Energy Balances for Spray Dryers . . 204.2.1 Mass Balance . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.2 Equilibrium Moisture Content . . . . . . . . . . . . . 25
4.2.3 Steady State Solution . . . . . . . . . . . . . . . . . . 29
-
8/20/2019 Modelling and Control of a Spray Drying Process
13/225
4.3 Test on a Multi Stage Dryer- MSD20 . . . . . . . . . . . . . . 29
4.3.1 The Test on MSD-20 . . . . . . . . . . . . . . . . . . . 31
4.4 Steady State Calculations . . . . . . . . . . . . . . . . . . . . 33
4.4.1 Steady State Results . . . . . . . . . . . . . . . . . . . 354.4.2 Effect of varying Operation Variables . . . . . . . . . 37
5 Dynamic Modelling of a Mixed Flow Spray Dryer 43
5.1 Total mass of air in the Spray Dryer . . . . . . . . . . . . . . 44
5.2 Drying Kinetic Mechanism . . . . . . . . . . . . . . . . . . . . 47
5.2.1 Mass Transfer Rate . . . . . . . . . . . . . . . . . . . 49
5.2.2 Mass and Heat Transfer Coefficients . . . . . . . . . . 51
5.2.3 Droplet size . . . . . . . . . . . . . . . . . . . . . . . . 53
5.3 Matlab Simulink Implementation . . . . . . . . . . . . . . . . 54
5.4 Test:Dynamic model . . . . . . . . . . . . . . . . . . . . . . . 54
5.4.1 Step: Temperature of main inlet air . . . . . . . . . . 57
5.4.2 Step: Feed Flow rate . . . . . . . . . . . . . . . . . . . 58
5.4.3 Step: Main inlet air flow rate . . . . . . . . . . . . . . 60
5.5 Test:Drying Time for particle . . . . . . . . . . . . . . . . . . 62
5.5.1 Test 1: particle sizes . . . . . . . . . . . . . . . . . . . 655.5.2 Test 2: Temperature . . . . . . . . . . . . . . . . . . . 66
5.5.3 Test 3: Effective Diffusivity . . . . . . . . . . . . . . . 67
5.5.4 Test 4: Critical Moisture Content . . . . . . . . . . . . 67
5.6 Summary: tests . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.7 Modifications in the Dynamic Model . . . . . . . . . . . . . . 69
5.7.1 Implementation of longer drying times and change inevaporation rate . . . . . . . . . . . . . . . . . . . . . 70
5.8 Summary: modifications . . . . . . . . . . . . . . . . . . . . . 73
6 Linearisation Analysis 75
6.1 Operating Point . . . . . . . . . . . . . . . . . . . . . . . . . 76
-
8/20/2019 Modelling and Control of a Spray Drying Process
14/225
6.2 Linearised results . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.4 Comparison of the linear model with the non-linear model . . 81
7 System Identification of Residual Moisture Content 85
7.1 Applied Identification Methods . . . . . . . . . . . . . . . . . 86
7.2 Estimation Data and Validation Data . . . . . . . . . . . . . 87
7.3 System identification Results . . . . . . . . . . . . . . . . . . 89
7.3.1 ARX model . . . . . . . . . . . . . . . . . . . . . . . . 90
7.3.2 ARMAX Model . . . . . . . . . . . . . . . . . . . . . . 91
7.3.3 State Space model . . . . . . . . . . . . . . . . . . . . 93
7.4 Summary: System identification . . . . . . . . . . . . . . . . 93
8 Control of spray dryers 95
8.1 Control Strategy . . . . . . . . . . . . . . . . . . . . . . . . . 95
8.1.1 PI controller . . . . . . . . . . . . . . . . . . . . . . . 97
8.2 PI controller for disturbance rejection of solids content variation 98
8.2.1 PI controller design . . . . . . . . . . . . . . . . . . . 98
8.3 Results from PI controller implementation in dynamic model 101
8.3.1 Step on reference temperature . . . . . . . . . . . . . 101
8.3.2 Step on Solids content . . . . . . . . . . . . . . . . . . 101
8.4 Possible control strategies . . . . . . . . . . . . . . . . . . . . 102
9 Conclusion 105
Nomenclature 110
References 111
Appendix 114
A Appendix A 115
-
8/20/2019 Modelling and Control of a Spray Drying Process
15/225
A.1 Desorption Isotherm at low and high humidity level . . . . . 116
A.2 General moisture Characteristic and food microbiology . . . . 117
B Appendix B 119
B.1 Modelling Variables . . . . . . . . . . . . . . . . . . . . . . . 120
B.2 Steady State Calculation . . . . . . . . . . . . . . . . . . . . . 121
B.3 Results from the Dynamic Model before modification . . . . . 123
B.3.1 Main inlet air temperature step up . . . . . . . . . . . 123
B.3.2 Feed Flow step down . . . . . . . . . . . . . . . . . . . 124
B.3.3 Temperature SFB step down . . . . . . . . . . . . . . 125
B.3.4 Temperature SFB step up . . . . . . . . . . . . . . . . 126
B.3.5 SFB air flow step down . . . . . . . . . . . . . . . . . 127B.3.6 SFB air flow step up . . . . . . . . . . . . . . . . . . . 128
B.3.7 Main air flow step down . . . . . . . . . . . . . . . . . 129
B.4 Results from the Dynamic Model after modification . . . . . 130
B.4.1 Feed Flow step down with modefication . . . . . . . . 130
B.4.2 Temperature MAIN step down with modification . . . 130
B.4.3 Air flow MAIN step up with modification . . . . . . . 131
B.4.4 Absolute Humidity in Dryer For feed step up with
modification . . . . . . . . . . . . . . . . . . . . . . . . 132
B.4.5 Response for the system G(s)evap . . . . . . . . . . . . 132
B.4.6 Simulink model . . . . . . . . . . . . . . . . . . . . . . 133
C Appendix C 135
C.1 MSD-20 test 24-7-2009 . . . . . . . . . . . . . . . . . . . . . . 135
C.2 Test Step & Responses . . . . . . . . . . . . . . . . . . . . . . 136
C.2.1 Test Program MSD-20 week 30 2009 . . . . . . . . . . 136
C.2.2 Test Step & Results for the entire test on MSD-20 . . 138
C.2.3 Moisture content of the particle from the SFB dis-charge for the entire test on MSD-20 . . . . . . . . . . 139
C.2.4 Test Step & Results for change in feed rate on MSD-20 140
-
8/20/2019 Modelling and Control of a Spray Drying Process
16/225
C.2.5 Feed flow rate and Nozzle pressure results from teston MSD-20 . . . . . . . . . . . . . . . . . . . . . . . . 141
C.2.6 Test Step & Results for change in Main inlet air tem-perature on MSD-20 . . . . . . . . . . . . . . . . . . . 142
C.2.7 Test Step & Results for change in Main inlet air flowon MSD-20 . . . . . . . . . . . . . . . . . . . . . . . . 143
C.2.8 Test Step & Results for change in SFB inlet air tem-perature on MSD-20 . . . . . . . . . . . . . . . . . . . 144
C.2.9 Test Step & Results for change in SFB inlet air flowon MSD-20 . . . . . . . . . . . . . . . . . . . . . . . . 145
C.2.10 Ambient Air Conditons At AIR intake (21/7-2009) . . 146
C.3 Logbook for MSD-20 test 24-7-2009 . . . . . . . . . . . . . . 147
D Appendix D 151
D.1 Humidity Calculation . . . . . . . . . . . . . . . . . . . . . . 151
D.2 Droplet calculations . . . . . . . . . . . . . . . . . . . . . . . 152
D.3 Thermal Conductivity Air . . . . . . . . . . . . . . . . . . . . 153
D.4 Thermal Diffusivity Air . . . . . . . . . . . . . . . . . . . . . 154
D.5 Kinematic Viscosity Air . . . . . . . . . . . . . . . . . . . . . 154
D.6 Mean Residense Time for the particle in the Spray dryer . . . 155
E Appendix E 159
E.1 Air and Particle Trajectory in Chamber . . . . . . . . . . . . 160
F Appendix F 161
F.1 Mass transfer and Drying time appendix . . . . . . . . . . . . 161
F.1.1 Default state operation condtion . . . . . . . . . . . . 161
F.1.2 T Out varied . . . . . . . . . . . . . . . . . . . . . . . . 161
G Appendix G 165
G.1 1st order system . . . . . . . . . . . . . . . . . . . . . . . . . 165
H Appendix H 167
-
8/20/2019 Modelling and Control of a Spray Drying Process
17/225
xii
H.1 System Identification . . . . . . . . . . . . . . . . . . . . . . . 168
H.2 ARMAX models . . . . . . . . . . . . . . . . . . . . . . . . . 168
H.1.1 Model Misfit Vs Number parameters for ARX model . 169
H.1.2 Zero Pole plot for the ARX model . . . . . . . . . . . 170
H.2.1 ARMAX simulations . . . . . . . . . . . . . . . . . . . 171
H.2.2 ARMAX Zero-pole plot for 6th order model . . . . . 174
H.2.3 State Space- continous time zero-pole plot . . . . . . . 178
I Appendix I 179
I.1.1 Linearized model- State Space(Jacobians) . . . . . . . 179
I.1 Linearization of the dynamic Model - open loop . . . . . . . . 180
I.1.2 Frequency response - from inputs to output . . . . . . 181
I.1.3 Zero-Pole plot for the transfer functions - from eachinputs to output . . . . . . . . . . . . . . . . . . . . . 185
I.2 Comparison of linear and Non linear model . . . . . . . . . . 187
I.2.1 Feed flow . . . . . . . . . . . . . . . . . . . . . . . . . 188
I.2.2 Main inlet air flow . . . . . . . . . . . . . . . . . . . . 190
I.2.3 Main inlet air temperature . . . . . . . . . . . . . . . 193
I.2.4 Solids Content . . . . . . . . . . . . . . . . . . . . . . 196
I.2.5 Relative Humidity of Ambient air . . . . . . . . . . . . 197
-
8/20/2019 Modelling and Control of a Spray Drying Process
18/225
List of Figures
1.1 Basic Spray drying process . . . . . . . . . . . . . . . . . . . 4
2.1 Spray drying process stages . . . . . . . . . . . . . . . . . . . 9
2.2 Multi Stage Dryer . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Agglomeration Process Schematic . . . . . . . . . . . . . . . . 11
4.1 Schematic of a CSTR and a plug flow reactor. . . . . . . . . . 18
4.2 Temperature profile for a mixed flow spray dryer . . . . . . . 20
4.3 Morphology of Particle with maltodextrin . . . . . . . . . . . 21
4.4 Basis Blockdiagram- variable description to the system . . . . 22
4.5 Desorption Isotherm Maltodextrin DE12 . . . . . . . . . . . . 27
4.6 MSD-20 Test Station Setup . . . . . . . . . . . . . . . . . . . 40
4.7 Test Centre at GEA Niro - Multi stage Dryer MSD-20 . . . . 41
5.1 Basic Dynamic Model . . . . . . . . . . . . . . . . . . . . . . 46
5.2 Particle Morphology . . . . . . . . . . . . . . . . . . . . . . . 47
5.3 Schematic of Drying mechanism . . . . . . . . . . . . . . . . . 48
5.4 Drying Proces of a Particle with a Shrinking Model . . . . . . 49
5.5 Dynamic model blockdiagram . . . . . . . . . . . . . . . . . . 55
5.6 Dynamic step response of the T OutAir for decrease in maininlet air temperature . . . . . . . . . . . . . . . . . . . . . . . 58
-
8/20/2019 Modelling and Control of a Spray Drying Process
19/225
xiv LIST OF FIGURES
5.7 Dynamic step response of the T OutAir for a increase in feedflow rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.8 Dynamic step response of the T OutAir for increase in Maininlet air flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.9 Mass tranfer rate and drying time test setup . . . . . . . . . 62
5.10 Drying time for a particle at default operation state . . . . . 64
5.11 Drying time for various particle sizes at default operation state 65
5.12 Drying time for various for a single particle for various effec-tive diffusivity coefficients . . . . . . . . . . . . . . . . . . . . 67
5.13 Drying time for various for a single particle for various criticalmoisture contents . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.14 Drying Time for single Particle of different sizes for Def f =8e − 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.15 Temperature response of the outlet air for the modified model(Feed step) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.1 Basic Dynamic Model . . . . . . . . . . . . . . . . . . . . . . 75
6.2 Bode plot- example: minimum phase & non minimum phase 79
6.3 Zero-Pole plot for the linearized model . . . . . . . . . . . . . 80
6.4 Comparison of linear and Non-linear model: Main inlet airtemperature. Step change= 10 from linearised input. No
change difference is observed . . . . . . . . . . . . . . . . . . 826.5 Comparison of linear and Non-linear model: Feed step : 2
& 10. For the small step no difference is observed. For thelarger step a small deviation is noted. . . . . . . . . . . . . . 82
6.6 Comparison of linear and Non linear model: Solids contentstep from 50 % to 80 % Such a large change is not possiblein reality. Small difference between the non- linear and linearmodel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.1 Simulated ARX model output and measured output . . . . . 90
7.2 ARMAX model: 6th order & 10th order . . . . . . . . . . . . 91
7.3 State Space model : 4th order . . . . . . . . . . . . . . . . . . 93
8.1 Input and output variables for process control . . . . . . . . . 96
-
8/20/2019 Modelling and Control of a Spray Drying Process
20/225
LIST OF FIGURES xv
8.2 Feed rate control . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.3 The control system . . . . . . . . . . . . . . . . . . . . . . . . 99
8.4 Bode plot for the open loop transfer function . . . . . . . . . 100
8.5 PI controller:step on reference temperatur . . . . . . . . . . . 1028.6 PI controller:step on solids content . . . . . . . . . . . . . . . 103
8.7 PI controller:Continuous Solids content Disturbance . . . . . 104
A.1 Desorption Isotherm Maltodextrin DE12 . . . . . . . . . . . . 116
A.2 General moisture Characteristic and Food microbiology . . . 117
B.1 Dynamic step response of the T OutAir for increase in maininlet air temperature . . . . . . . . . . . . . . . . . . . . . . . 123
B.2 Dynamic step response of the T OutAir for decrease in maininlet air temperature . . . . . . . . . . . . . . . . . . . . . . . 124
B.3 Dynamic step response of the T OutAir for decrease in SFBinlet air temperature . . . . . . . . . . . . . . . . . . . . . . . 125
B.4 Dynamic step response of the T OutAir for decrease in SFBinlet air temperature . . . . . . . . . . . . . . . . . . . . . . . 126
B.5 Dynamic step response of the T OutAir for increase in SFB inletair flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
B.6 Dynamic step response of the T OutAir for decrease in SFBinlet air flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
B.7 Dynamic step response of the T OutAir for decrease in Maininlet air flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
B.8 Dynamic step response of the T OutAir for decrease in maininlet air temperature . . . . . . . . . . . . . . . . . . . . . . . 130
B.9 Dynamic step response of the T OutAir for decrease in MAINinlet air temperature(modified) . . . . . . . . . . . . . . . . . 130
B.10 Dynamic step response of the T OutAir for decrease in MAINinlet air flow(modified) . . . . . . . . . . . . . . . . . . . . . . 131
B.11 Absolute Humidity in Dryer For feed step up with modification 132
B.12 Response for the system G(s)evap . . . . . . . . . . . . . . . . 132
B.13 Simulink dynamic model . . . . . . . . . . . . . . . . . . . . . 133
-
8/20/2019 Modelling and Control of a Spray Drying Process
21/225
xvi LIST OF FIGURES
C.1 Test Step & Results for the entire test on MSD-20 . . . . . . 138
C.2 Moisture content of the particle from the SFB discharge . . . 139
C.3 Test Step & Results for change in feed rate on MSD-20 . . . 140
C.4 Feed flow rate and Nozzle pressure results from test on MSD-20141C.5 Test Step & Results for change in Main inlet air temperature
on MSD-20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
C.6 Test Step & Results for change in Main inlet air flow on MSD-20143
C.7 Test Step & Results for change in SFB inlet air temperatureon MSD-20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
C.8 Test Step & Results for change in SFB inlet air flow on MSD-20145
C.9 Ambient Air Condition at Intake . . . . . . . . . . . . . . . . 146
C.10 Logbook From Test on MSD-20 week 30 2009 . . . . . . . . . 147C.11 Logbook From Test on MSD-20 week 30 2009 . . . . . . . . . 148
D.1 Thermal Conductivity of Air vs. Temperature . . . . . . . . . 154
D.2 Thermal Diffusivity of Air vs. Temperature . . . . . . . . . . 155
D.3 Kinematic Viscosity of Air vs. Temperature . . . . . . . . . . 156
E.1 Air Stream and particle trajectory in mixed flow chamber . . 160
F.1 Mass evaporated for various particles sizes . . . . . . . . . . . 161
F.2 Mass Transfer coefficient for various particles sizes . . . . . . 162
F.3 Crust resistance f for various particle sizes . . . . . . . . . . . 162
F.4 Drying time for particle -various feed flow . . . . . . . . . . . 163
F.5 Crust resistance f for various feed flow rates . . . . . . . . . . 163
F.6 Drying time for particle -various Main inlet air flow . . . . . 164
G.1 Control variable as a first order system . . . . . . . . . . . . . 165
H.1 Model Misfit Vs Number parameters for ARX model . . . . . 169
H.2 Zero-Pole plot for the 10th order ARX model . . . . . . . . . 170
H.3 ARMAX model 1 . . . . . . . . . . . . . . . . . . . . . . . . . 171
-
8/20/2019 Modelling and Control of a Spray Drying Process
22/225
LIST OF FIGURES xvii
H.4 ARMAX model 1 . . . . . . . . . . . . . . . . . . . . . . . . . 172
H.5 ARMAX model 1 . . . . . . . . . . . . . . . . . . . . . . . . . 172
H.6 ARMAX model 1 . . . . . . . . . . . . . . . . . . . . . . . . . 173
H.7 ARMAX model 1 . . . . . . . . . . . . . . . . . . . . . . . . . 173
H.8 Zero-Pole plot for ARMAX 6th model . . . . . . . . . . . . . 174
H.9 Zero-Pole plot for ARMAX 6th model 1 . . . . . . . . . . . . 175
H.10 Zero-Pole plot for ARMAX 6th model 2 . . . . . . . . . . . . 176
H.11 Zero-Pole plot for State space model 4th order . . . . . . . . 177
H.12 Zero-Pole plot for State space model 4th order continuos time 178
I.1 Frequency response of the linearized model process input tooutput . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
I.2 Frequency response of the linearized model- Feed flow inputto output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
I.3 Frequency response of the linearized model- disturbance input(Solids content) to output . . . . . . . . . . . . . . . . . . . . 183
I.4 Frequency response of the linearized model- disturbance input(Ambient Relative humidity) to output . . . . . . . . . . . . . 184
I.5 Zero-Pole plot for the transfer functions - from each processinput to output . . . . . . . . . . . . . . . . . . . . . . . . . . 185
I.6 Zero-Pole plot for the transfer functions - from disturbanceinputs to output . . . . . . . . . . . . . . . . . . . . . . . . . 186
I.7 Comparison of linear and Non linear model: feed flow . . . . 188
I.8 Comparison of linear and Non linear model: feed flow 1 . . . 189
I.9 Comparison of linear and Non linear model: feed flow 2 . . . 189
I.10 Comparison of linear and Non linear model: Main inlet air flow190
I.11 Comparison of linear and Non linear model:Main inlet air flow 1191
I.12 Comparison of linear and Non linear model: Main inlet airflow 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
I.13 Comparison of linear and Non linear model: Main inlet airt e m p e r a t u r e . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
I.14 Comparison of linear and Non linear model: Main inlet airt e m p e r a t u r e . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
-
8/20/2019 Modelling and Control of a Spray Drying Process
23/225
xviii LIST OF FIGURES
I.15 Comparison of linear and Non linear model: Main inlet airtemperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
I.16 Comparison of linear and Non linear model: Main inlet airtemperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
I.17 Comparison of linear and Non linear model: Solids content . 196
I.18 Comparison of linear and Non linear model: Solids content . 196
I.19 Comparison of linear and Non linear model: Relative humid-ity of amb ie n t air . . . . . . . . . . . . . . . . . . . . . . . . . 197
I.20 Comparison of linear and Non linear model: Relative humid-ity of amb ie n t air . . . . . . . . . . . . . . . . . . . . . . . . . 197
-
8/20/2019 Modelling and Control of a Spray Drying Process
24/225
List of Tables
3.1 Modelling Approach . . . . . . . . . . . . . . . . . . . . . . . 15
4.1 The manipulated variables in test of the dynamic process(MSD-20) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2 The manipulated range of the variables in test of the dynamicp r oc e s s (MS D - 20) . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.3 Default Operation Variable Values for Testing . . . . . . . . . 34
4.4 Steady State Results for the drying air temperature T OutAir . 36
4.5 Steady State Results of absolute humidity and equlibriummoisture content . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.6 Energy Level of the components . . . . . . . . . . . . . . . . 38
5.1 Ranz-Marshall correlation Parameter Values . . . . . . . . . . 53
6.1 Operating Point for linearization: Stationary state . . . . . . 78
7.1 Sensor Description for Test system(MSD-20) . . . . . . . . . 88
B.1 Default Operation Variable Values for Testing APPENDIX . 120
B.2 Steady State Results for the drying air tempereture T OutAirAPPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
B.3 Steady State Results of absolute humidity and equlibriummoisture content . . . . . . . . . . . . . . . . . . . . . . . . . 122
-
8/20/2019 Modelling and Control of a Spray Drying Process
25/225
xx LIST OF TABLES
C.1 Sensor Description for Test system(MSD-20) . . . . . . . . . 135
C.2 Test Program for Test On MSD-20 . . . . . . . . . . . . . . . 136
H.1 ARMAX models . . . . . . . . . . . . . . . . . . . . . . . . . 168
I.1 Operating Point for linearization: Stationary state . . . . . . 180
-
8/20/2019 Modelling and Control of a Spray Drying Process
26/225
LIST OF TABLES 1
0
-
8/20/2019 Modelling and Control of a Spray Drying Process
27/225
2 LIST OF TABLES
-
8/20/2019 Modelling and Control of a Spray Drying Process
28/225
Chapter 1
Introduction
Baby food, milk powder, coffee whitener, flavours, and various other prod-ucts are produced in a spray drying process.
Spray drying is a process that transforms a given feedstock from a fluid stateinto a dried particulate form by spraying the feed into a gaseous hot dryingmedium. The feedstock can be a solution, emulsion or fluid paste, though itis required that the feed is pumpable so it can be atomised into droplets. Theextensive contact between the droplets and the drying medium is the mainprinciple of the spray drying process, where the drying medium provides theenergy for the evaporation of the solvent in the feed. The resulting dried
product is conformed to a powder of either single particles, agglomerates,granules, or pellets. The shape and structure of the particles depends onthe physical and chemical properties of the feed, spray dryer design, and theoperation conditions used.
The spray drying process, compared to other drying processes, is unique inits ability to dry liquid feedstock to powder with specific physical properties,particle morphology, and moisture content. The broad range of spray dryerdesigns available fulfils the specifications stipulated by various industriesboth in terms of product properties and production capacity. The spraydried product reduces transportation cost, as there is less liquid to transport,but also simplifies storage and handling.
Furthermore the spray dryer has the great advantage due to the ability todry both non-heat sensitive and heat sensitive materials. Thus the productcan be dried without any loss or changes in the volatile compounds of theproduct. These compounds can be the aromatic characteristics of the prod-
-
8/20/2019 Modelling and Control of a Spray Drying Process
29/225
4 Introduction
Dryingchamber
Stac
fluid bed
Parculatesseparaon /
collecon
Feed
Air Air
Total dried
product disharge
Figure 1.1: Spray drying process is characterised by a liquid feedstock that issprayed into a chamber in which contact with the heated dryingmedium(air) results in evaporation of moisture from the droplets.Drying takes place as the droplet moves through the drying cham-ber. If the particle reaches the bottom of the chamber and the dryingis complete the product is discharged. Some feedstock requires posttreatment of the particles to reach a specific powder characteristic. Toreduce the amount of particles in the exhaust air, a particle collectoris used to ’clean’ the air before leaving the drying system.
uct or some proteins in the product, which is especially important in thefood and dairy industries. Here spray drying helps with preservation of theproduct which gives product stability and extended shelf life by reducingthe moisture content to levels where microbiological growth is not possible.
In addition to this the spray dryer can handle materials under aseptic and
hygienic drying condition which makes it applicable in the pharmaceuticalindustry. It has also proved its worth in the protection of the environment asit is capable of evaporating organic solvents which are potentially explosiveor have toxic risks. Thus spray dryers have a variety of industrial andcommercial uses and just a few examples were named here.
-
8/20/2019 Modelling and Control of a Spray Drying Process
30/225
1.1 Description of the Problem 5
1.1 Description of the Problem
The objective of the spray drying process is to produce a product of a desiredquality at minimum cost, with maximum throughput and sustain the desired
dried product quality regardless of the disturbances in the drying operationand variations in feed supply. Appliance of automatic control systems offersan opportunity to improve the dryer operation and its efficiency. However,the product quality, which is the parameter that is wished to be controlled,consists of sub parameters such as moisture content, thermal degradation,aroma retention, and structure and size of the particles. These parametersare difficult to predict, but also to measure online in the system, with theaim to be used as a control variable in the controller.
By experience it is identified that the most effective parameter to controlquality is the moisture content in the product. This can be measured witha suitable moisture sensor, but these are not used for controller purposes in
spray dryers yet as these are expensive and have a low reliability at presenttime. It is recognised that the quality of the final product is the outcomeof the drying conditions in the chamber, with respect to temperature andhumidity. Therefore the spray dryer uses the outlet drying air leaving thechamber to describe the drying condition and uses it as a control variablein a feedback control. So the moisture content is controlled indirectly bymaintaining a specific outlet drying air temperature by varying either thefeed flow into the dryer or the inlet drying air temperature.
So far the main spray drying operation controller is based on a single controlvariable and a single manipulated variable, a SISO control system, for whicha PI controller is utilised. The proportional gain and integral time constantsare selected from the experience of setting up previous spray dryers. Thisis often considered to be sufficient when the product moisture content canbe kept within a narrow limit. In case large variations or fluctuations areobserved these parameters are adjusted, which is an accepted and easy pro-cedure.
However, there is obviously potential in optimising the existing controlsystem or developing a more advanced control system by using analyticalmethods(MIMO- or Model Predictive Control System). This could minimisethe variations in moisture contents and improve the product quality. At thesame time this can lead to a more cost-effective drying, either by increasing
the throughput or reducing the energy consumption in the dryer, since spraydryers are known to have a relatively poor thermal efficiency compared toother drying processes.
With the purpose of being able to develop and analyse a control system, amodel of the spray drying system is required. There are two types of models:
-
8/20/2019 Modelling and Control of a Spray Drying Process
31/225
6 Introduction
• An equipment model, which combines the factors that affects the spraydrying process and describes the environment the particles are expe-riencing as drying takes place.
• A particle model, which describes how the particles respond to the
drying environment.
There are several papers 1 on the study of how a single particle of a specificproduct reacts in certain conditions. There are also very detailed ’Compu-tational Fluid Dynamic’ (CFD) models that are able to illustrate the flowsof the particles and the air in the spray dryer very precisely. Common forboth cases are that the models have not been studied in a view to be usedin the development of the control systems in the spray dryer. But are moreused to examine the product for chemical purposes or in the developmentof the spray drying design. For the CFD models the calculation times arevery long, which is not acceptable.
1.2 The objective of the project
The object of this project is to construct a general dynamic equipment modelthat is fundamental and is able to predict and estimate the drying conditionsin the spray drying chamber as this is the factor that affects the quality of the powder. The model is developed in preparation to be further used inthe development of control systems for the spray dryer. However, the mainfocus in this project is the modelling and development process.
• The spray dryer that is modelled is of the type: multi stage dryer witha mixed air flow
– The focus is on the spray drying chamber and the condition in-side this for a running process. The start-up of the process isneglected.
– Modelling the temperature of the drying air is the main target.Subsequently the humidity level inside the chamber is of interest.These parameters have a significant influence on the drying of theproduct.
• The aim is to model the temperature and humidity levels in the cham-ber by using first principle methods.
1(Langrish and Kockel (2001)),( Shabde (2006)),( Ireneusz Zbicinski and Delag (2002)),(Ruud E.M. Verdurmen and JONG (2002)),( Lixin Huang and Mujumdar (2005)), (Kieviet(1997))
-
8/20/2019 Modelling and Control of a Spray Drying Process
32/225
1.2 The objective of the project 7
– Mass and Energy balance equations are applied.
– Initially a steady state model of the spray dryer is examined.
– With the steady state model as the underlying basis a dynamicmodel is developed .
• The drying characteristics of the feed is examined.
– The equilibrium moisture content and the drying kinetics is stud-ied for maltodextrin which has been used as an example duringthe modelling.
– The particle behaviour, such as the agglomeration process andfines are neglected.
• For test and verification of the model, experiments on a multi stagespray dryer (MSD-20) at GEA Niro’s test station is conducted.
– step responses for this system was examined and data recorded.
– Data is used to compare steady state model results and the thedynamic response times for the dynamic model.
– The default operating settings for this dryer has been applied inthe models
• The dynamic model is linearised in order to analyse the system.
• System identification principles has been tried on the data collectedfrom the experiment on the MSD-20.
– A simple linear model of the moisture content in the particles as
a function of the main process variables is tried to be estimated.• In the last part of the project, the developed model is used to demon-
strate that the model can be used for development and test of con-troller.
– A PI controller is applied to control the outlet temperature of thespray drying model.
-
8/20/2019 Modelling and Control of a Spray Drying Process
33/225
8 Introduction
-
8/20/2019 Modelling and Control of a Spray Drying Process
34/225
Chapter 2
Introduction to the Spray
Drying Process and Spray
Dryers
The spray drying process can be described by four stages: Atomization,spray and air contact, evaporation of the moisture from the droplets andproduct discharge.
ATOMIZATION SPRAYAIR CONTACT DRYING OF SPRAY PRODUCT DISCHARGE
Liquid feed into
a spray of drop-
lets
Mixing and flow
paern
Moisture
evaporaon
Separaon of
dried product
from the air
STAGE 1 STAGE 2 STAGE 3 STAGE 4
Figure 2.1: Spray drying consists of four process stages. This involves sprayingof liquid, contact and mixing with drying air, droplet drying parti-cle formation and at last powder collection. These four stages areillustrated in figure 2.2
The plant in which the spray drying process takes place is a spray dryer.Spray dryers exists in various designs with regards to size, type(conventional,compact, tall-form, multistage etc), air flow characteristics (Co-current,counter, and mixed air flow) and mobility. The spray drying design andmode of operation, together with the physical and chemical properties of
-
8/20/2019 Modelling and Control of a Spray Drying Process
35/225
10 Introduction to the Spray Drying Process and Spray Dryers
the feed determines the final characteristics of the dried product (particlesize and structure). The spray dryer used in this project is a Multi StageDryer (MSD) as illustrated in figure 2.2. The multi stage refers to the factthat post-treatment equipment for the powder is a part of the spray dryingsystem, which will be elaborated below.
Fan Heater
FeedExhaust air
Fines
Air outDrying air - Main
Drying air - SFB
Powder out
Cyclone /
Bag filter
Vibro Fluidizer
SFB
MULTI STAGE DRYER
STAGE 1
STAGE 2
STAGE 3
STAGE 4
Figure 2.2: Multistagedryer with mixed air flow. Feed enters the spray dryingchamber from the top. Air is drawn from atmosphere by a fan andheated with a heater. The heated air is mixed with the feed, whichfalls down while it dries. The base of the chamber is the static fluidbed(SFB) and is used for agglomeration and to finalize the drying of thepowder. Air outlet is at the top of the chamber. A cyclone and bagfilter is to filter the exhaust air for fine particles. The vibro fluidizer
is for post treatment of powder [GEA Niro].
At the first stage the feed is pumped from the feed tank to the atomizer1.
1In some cases the feed is send through a preheater/evaporator with the intention of
-
8/20/2019 Modelling and Control of a Spray Drying Process
36/225
11
The atomizer is either a rotary atomizer (rotating disc) or a nozzle whichmakes a spray of droplet. The nozzle atomizer which is operated in thesystem that is examined in this project utilizes pressure to create dropletsthrough an orifice. The atomizer is located at the ceiling of the chamber.The formed droplets are mixed with the drying air and evaporation com-
mences.
The air is drawn from the atmosphere by a fan and passed through a heater.In this spray dryer setup the air flow is mixed, which means the drying airenters the dryer both from the top of the chamber and from the bottom.The drying air entering from the ceiling of the chamber is denoted as theMain inlet air. The air from the base of the chamber is the Static Fluid Bed(SFB) inlet air. The air leaves the chamber from the top of the chamber aswell, which is mentioned as the outlet air.
Following the evaporation of moisture from the droplets the majority of thedried particles fall down to the base of the drying chamber. A Multi stage
dryer has a Static Fluid Bed at this place which serves to finalize the dryingof the product and for agglomeration. Agglomeration is the process wherewet or partially dried droplets get in contact with dry particles and formslarger particles. This is illustrated in figure 2.3. The air leaves the chamberfrom the top of the chamber as well, which is mentioned as the outlet air. Acyclone and/or a bag filter are used to filter out particles from the exhaustair. The filtered particles are referred to as fines and these are led back intothe chamber to increase the agglomeration process.
Figure 2.3: Agglomeration Process Schematic. Dry particles(fines)collision withwet particles, and thus gets into a structure or increase its size bygetting more layers[GEA Niro].
increasing the viscosity of the feed or increasing the solids content in the feed.
-
8/20/2019 Modelling and Control of a Spray Drying Process
37/225
12 Introduction to the Spray Drying Process and Spray Dryers
The choice of spray drying setup, which includes drying chamber design,atomizer, air inlet and disperser has an influence on the resulting powdersize and how the product reacts to the temperature and humidity profilesexisting in the dryer due to the selected operation conditions. Besides thedesign of the spray dryer and its equipment, the chemical composition of
the solids, affects the particle shape and formation during the drying.
Particles are continuously discharged from the SFB. These are led to a vi-brating fluid bed, which is equipment for post treatment of the powder.However, the focus in this project has only been on the spray drying cham-ber.A more detailed description of the spray dryer design and its equipmentis found in (Masters (2002)).
The terms droplet and particle has been alternately used throughout thereport to describe the element which is dried. To elaborate this: an elementwhich enters the drying chamber is at the outset a droplet and turns into aparticle as it solidifies.
-
8/20/2019 Modelling and Control of a Spray Drying Process
38/225
Chapter 3
Modelling
With the intention to be able to evaluate more advanced or novel controlstrategies, an understanding of the static and dynamic properties of thespray drying process is needed. This can be obtained with an estimatedmathematical model of the plant. Describing and developing a completelyaccurate mathematical model of a spray drying process is a complicatedassignment due to high complexity of the physical, chemical, and mechanicalproperties in such a system. This embraces for instance the heat and masstransfer both within the particle at the boundary between the solid phaseand liquid phase, but also the particle and surroundings. Another aspect
of this multifaceted process is the accounting for the various entrances of the drying medium into the spray dryer and the resulting flow patterns forthe gas and the particles. However, in all models some uncertainty in theprocess behaviour will arise due to unmeasured disturbances, unmodelleddynamics and nonlinearities. Although the mathematical model only willbe an approximation of the real process, it will be acceptable if it is capableof giving a practically realistic representation of the process and thus satisfythe previously defined objective.
The use of steady state models are well established in chemical engineeringfor plant analysis that can be used to calculate the necessary process condi-tions for an optimal exploitation of the system concerning powder properties
and energy consumption. However, a dynamic model is an important partof operability study, both in assessing the consequences of plant malfunctionand in the mitigation of possible effects. Moreover it gives a better under-standing of process performance and is therefore a significant instrument forprocess optimisation.
-
8/20/2019 Modelling and Control of a Spray Drying Process
39/225
14 Modeling
To decide which approach to use for a modelling task the required level of flexibility, time frame or validity goal, available resources , and the numberof approximations that is acceptable, has to be considered. In the forthcom-ing section different methods to develop or estimate a dynamic model areexplained (Labspace (2009)),(O’Callagan and Cunningham (2005)).
3.1 Black Box model
The black box modelling strategy is used for investigating a complex systemwith no or minimal knowledge and assumptions about the process and theinternal structure. Such a model is represented by an empirical descriptionor a set of transfer parameters that relate the output of the model to a setof inputs. With the sufficient data available, containing the significant dy-namics of the system, an estimate of a model is achievable and is known assystem identification. Thus the need of experimental data for this methodinvolves data collection, determination of model structure, parameter esti-mation, and model validation. However correct the dynamics are revealed inthe model, the physical details of the process are excluded. The determinedmodel is specific to the system, operating region, and the product whichdata is extracted from.
This lack of flexibility is the main disadvantage of black box modelling, sincethe effect of changes in any of the process conditions outside of those metduring the structuring of the model cannot be concluded. Another constrainton this type of model, is the lack of any form of physical meaning, whichmakes it difficult to relate it to the real object being modelled. Nonetheless
it has proven its effectiveness in situations where important parameters arecomplicated to identify and measure online, such as the residual moisturecontent in the final powder. This subject will be elaborated in section 7.
3.2 White Box model
A white box model is the most detailed and comprehensive category withinmodelling. It is based on a first principle approach, which describes thephysical processes at the lowest level. The result will be a true nonlineardynamic model and as close as possible to the true description of the plant.
In contrast to the black box model, this type of model will be fully predictive,even in the situation when changes in process conditions are outside thenormal operating conditions. In spite of the fact that the model is flexibleand realistic, the outcome could be a model of great complexity. The morecomplex a model is, the more difficult it will be to identify the increased
-
8/20/2019 Modelling and Control of a Spray Drying Process
40/225
3.3 Grey Box model 15
number of parameter values. A pure white box model cannot exist as it isessentially a copy of the reality. So what is needed, is a model with a simpleapproach but which demonstrates realistic process phenomena. Thus themajority of simulation models are grey box models.
3.3 Grey Box model
A grey box model provides a physical representation of the system, thoughsome of the physical parameters are simplified or approximated by an em-pirical model. This hybrid model structure is the result of the combinationof the best properties from the white and black box model, and the methodthat is used in this project: the flexibility enables one to model the plantdesign and determine the effect of variations in chamber size or changes inmaterial parameters. In addition to this it is physically close to the real pro-cess, however, a convection heat transfer coefficient is utilized to describe the
heat changes around a particle instead of a model of the actual laminar flowthat requires an airflow model. An empirical model is chosen for determi-nation residual powder moisture content. The reasonable trade-off betweencomplexity and performance is suitable from a control point of view.
3.4 Chapter Summary
Modelling ApproachType Advantage Disadvantage Time Frame
White box Extremely flexible High Complexity LongRealistic Large computer power
slowBlack box Parameter identification Low flexibility Short
Minimal computer power Non physicalFast
Grey Flexibility Error checking mediumphysicality
Table 3.1: Summing up the pros, cons and the time frame for the three modelingapproaches. The time frame is connected to flexibilty requirement.
-
8/20/2019 Modelling and Control of a Spray Drying Process
41/225
16 Modeling
-
8/20/2019 Modelling and Control of a Spray Drying Process
42/225
Chapter 4
Modelling a Spray Dryer
In this chapter the white box modelling method described in chapeter 3 isapplied to model the spray drying process. Explaining the reality completelyby physical equations is difficult. For that reason some simplification aremade which is elaborated in the follwing section.
4.1 Preparations and Assumptions for Modelling
Modelling the spray dryer can be done at various levels and degrees of details,
from describing the flow, reaction rate and the effect on circumstances of droplet of a liquid to the overall energy flow and mass flow for the totalspray dryer. The purpose with modelling the spray dryer in this project isto be capable of estimating and predicting the temperature of the drying airin the spray drying chamber, which is to be used further in the control of the moisture content of the final product. There are four main phenomenain a spray drying operation:
1. Atomisation of the liquid feed
2. Drying of the droplets once they are formed
3. Motion of the droplet in the spray drying unit4. Product discharge
The region of interest is the spray drying chamber and on the drying of the droplets once they are formed. Modelling the motion of the droplet is
-
8/20/2019 Modelling and Control of a Spray Drying Process
43/225
18 Modelling a Spray Dryer
recognised as being dependent on the geometry of the chamber and the me-chanical setup of the system. Thus it is more of a mechanical problem, sincethe motion of the particles cannot be directly controlled with the flow andtemperature parameters we have at hand. To describe this class of prob-lems, ”‘rate based models”’, which are dynamic models that describe the rate
at which the solvent removed from the droplets as they travel through thespray drying chamber can be used(Gauvin and Katta (1976)). Otherwise the”‘particle-source-in-cell”’ models, that assumes the droplets to be a sourceof mass, energy and momentum in a grid of the drying gas can also be usedPapadakis (1988). Due to the complexity of calculating the heated dryinggas flow and particle motion, it usually requires ”‘Computational Fluid Dy-namic”’ (CFD) techniques, which are tools that use numerical methods andalgorithms, to solve these models. The disadvantages using this approachare the long calculation times and model parameter values that may haveno physical meaning.
Therefore a simpler mass and energy balance model, incorporating equi-
librium relationships on the amount of moisture in the particle is utilized.Chemical reaction engineering techniques have been considered and used tomodel the spray dryer, in view of the fact that the drying process can beviewed as a reaction (mixing) between gas and liquid/solids (vapour). Thereaction process can be represented either as in a ”‘Continuously StirredTank Reactor”’ (CSTR), a plug flow reactor or a sequence of these.(4.1)
Figure 4.1: Schematic of a CSTR and a plug flow reactor. The effluent com-position of the CSTR is identical to the conditions that exist in thereactor. For the plug flow reactor the outlet condition varies along
the length of the tube.
In the CSTR the contents of the reactor are assumed to be ideally well mixedand the reactants and products flow into and out of the reactor continuously.This means that the temperature, pressure and concentration levels are in-
-
8/20/2019 Modelling and Control of a Spray Drying Process
44/225
4.1 Preparations and Assumptions for Modelling 19
dependent of spatial position within the reactor. Accordingly it also impliesthat the composition and the temperature of the effluent flow are identicalto the gas in the chamber.
The plug flow reactor is an ideal flow assumption in a tube in which the fluidis well mixed in the radial and angular directions. The velocity, compositionand temperature of the fluid are functions of the axial position (along thelength of the tube) only. The plug flow can also be described as an infinitenumber of CSTR’s in a cascade connection. To model a co-current spraydryer (Air inlet from top and outlet air in the bottom), it is anticipated thatthis reactor type will give the best representation. But due to the fact that amodel of a mixed flow spray dryer is wanted, which has an inlet air flow fromthe top and the bottom of the chamber, it is expected that the well mixedCSTR reactor model is most suitable and will illustrate the inlet air mixingbest. Moreover the air leaving the dryer is from the top of the chamber andthus using a CSTR model, this temperature will not dependent on the travellength of the air flow.
Additionally, when looking at the temperature profile for a mixed flow spraydrying chamber (figure 4.2), it is observed that the chamber temperature toa great extent is the same and well mixed, except at the air inlet entranceswhere the temperature is higher due to the heated air. Therefore it is as-sumed that the spray dryer can be modelled as a CSTR process.
The model prepared is based on the following assumptions and simplifica-tions:
1. The Spray drying process is modelled as Continuously Stirred TankReactor (CSTR). In this the drying gas and the feed are continuously
injected into the chamber at uniform flow rates. The state of the gasin the chamber is identical to the state of the gas leaving the chamber.
2. The model will be based on mass and energy balance with equilibriumrelationship incorporated.
3. The gas is assumed to be a composition of dry air and vapour, whichbehaves as an ideal gas and flows as a perfect mixture. This has aninfluence on the calculations of the gas density and as well on therelation between absolute humidity and the partial vapour pressure.
4. The liquid feed is assumed to be completely atomised, that is all the
droplets are of uniform size and homogeneous. In figure 4.3 particleswith various amount of maltodextrin are depicted and as expectedthe particles are not perfectly spherical. However for simplicity theparticles are modelled as having a spherical shape. They are all wellmixed in the chamber and do not interact with one another. Because of
-
8/20/2019 Modelling and Control of a Spray Drying Process
45/225
20 Modelling a Spray Dryer
Figure 4.2: Temperature profile for a mixed flow spray dryer. The figure to theleft is the temperature profile given by Masters (2002). The arrowsindicate air flow. The dotted arrows point towards product flow di-rection. The figure to the right is a temperature profile with relativetemperatures(GEA Niro). Red is hot. Blue is less warm. It is seenon the figures that a great part of the chamber, except from the airinlet entrances, has the same temperature. Due to this information itis assumed that the spray dryer can be modelled as a CSTR process.
this simplification, the agglomeration process is disregarded, as therewill be no small particles (fines) nor will there be created any, due tono collisions of particles.
4.2 Steady State Mass and Energy Balances forSpray Dryers
The spray dryer operating requirements are found by solving mass and heatbalance calculations in steady state which is very common in chemical en-gineering. With the production rate requests, feedstock, dried product, andambient air properties at hand, the air flow rate requirements can be es-timated. Correspondingly the moisture content of the final product, for apresented drying air flow, can be calculated at certain conditions. For this
to succeed it is presumed that the dryer is well mixed and hence the gascondition is uniform inside the drying chamber. It is expected that the out-let gas and outlet particles are in equilibrium, such that the temperaturesof these elements are equal. Accordingly the solids moisture content of theoutlet product is in equilibrium with the gas temperature and humidity. The
-
8/20/2019 Modelling and Control of a Spray Drying Process
46/225
4.2 Steady State Mass and Energy Balances for Spray Dryers 21
Figure 4.3: Morphology of rice starch with little maltodextrin in particle on theleft picture and increased amount of maltodextrin on the right picture.Though the particles are not completely spherical, this is assumed inthe modelling. The density of the structure is also dependent on thedrying temperature.[GEA Niro]
equilibrium moisture content is the moisture content at which the product isneither gaining nor losing moisture; this however, is a dynamic equilibriumand changes with relative humidity and temperature.
In the following the equations of the mass and energy balances, for an opencycle mixed flow spray dryer with an aqueous feedstock, are illustrated. Theconservation of mass and energy in a steady state flow process is expressedas: the rate of mass/energy flow into the system is equal to the rate of mass/energy flow out of the system. First the energy balance is consideredin (4.1). In figure 4.4 a diagram illustrating input and output to the systemand a variable description for the following equations.
F Maindry H airIn + F SFBdry H airIn + F feedH feedIn
= F OutdryH airOut + F powder H powderO ut (4.1)
Where F Maindry is the Main inlet air flow, F SFBdry is the inlet air flow from
the SFB while F Out is the outlet air flow, all of them in dry form(Kg
s ).The dry components are easier to handle in the equations and later in thischapter it is shown, how these are determined from the true humid air flow.The humid air is a mixture of mv mass of water vapour and mass mdryAirof dry air. F Feed is the flow rate of the feed in (
Ls ) and F powder is the flow
of powder out of the system( Kgs ). The enthalpy H is a composite energy of the internal energy of the constituent atoms and the flow work associated
with forcing streams in and out of a system against a pressure. It has theunit energy per unit mass( J Kg ). The enthalpy of mixtures such as the humiddrying medium, the feedstock etc. is the sum of the partial enthalpies of the components and a residual enthalpy term which for example takes intoaccount the heat of mixing. However, in this report the influence of the
-
8/20/2019 Modelling and Control of a Spray Drying Process
47/225
2 2
M o d e l l i n g a S p r a
y D r y e r
SPRAY
DRYER
WEEL MIXED
MAIN INLET DRYING AIR
Flow
Absolute humidity
Enthalpy
Temperature
Specific heat capacity dry air
SFB INLET DRYING AIR
OUT LET
FEED IN
Moisture content
Feed (liquid)
AMBIENT AIR
Relav humidity
Temperature
OUT LET PRO
Specific heat capacity vapour
Flow
Absolute humidity
Enthalpy
Temperature
Specific heat capacity dry air
Specific heat capacity vapour
Flow
Enthalpy
Temperature
Specific heat capacity
Specific heat capacity
Flow
Absolute humidity
Enthalpy
Temperature
Specific heat capacity
Specific heat capacity
Moisture content
Flow
Enthalpy
Temperature
Specific heat capacity
Specific heat capacity
Kg/s
Kgmoist
/Kgdry air
KJ/Kg
°C
J/KgK
J/KgK
Kg/s
Kgmoist
/Kgdry air
KJ/Kg
°C
J/KgK
J/KgK
Fmain
Yin
Hmain
Tmain
Cdry airC
vapour
FSFB
Yin
HSFB
TSFB
Cdry air
Cvapour
Ffeed
Xin
Hfeed
Tfeed
CsolidC
water
Fout
Yout
Hout
Tout
Cdry air
Cvapour
Fout
Xout
Hout
Tout
Csolid
Cwater
Powder out
Figure 4.4: Basis Blockdiagram- variable description for the system used in energy and mass balance equation. Tinput and output for air and product. The block shows the true flows of the air flow and feed flow. Inare modified into dry flows for easier use.
-
8/20/2019 Modelling and Control of a Spray Drying Process
48/225
4.2 Steady State Mass and Energy Balances for Spray Dryers 23
residual enthalpy is neglected as this value often is very small comparedto the enthalpy of the main components (≈ 1%). With this definition thehumid air enthalpy for both inlet and outlet airflow is defined in 4.2:
H humAir = H dryAir + Y H vapor (4.2)
Y is the absolute humidity or moisture content in the air expressed by the
relation between mass of vapour and dry air mvmdryAir ,
KgvapourKgdryAir
. From being
Y In at the inlet it increases during the spray drying operation to Y out. Interms of specific heat the enthalpy is given by:
H humAir = C dryAir(T air − T ref ) + Y (λ + C vapor(T air − T ref )) (4.3)
where C dryAir is the specific heat capacity of dry air (
Kj
Kg ·K ), C vapor is thespecific heat capacity for water vapor KjKg ·K . The heat capacity is definedas the energy required to raise the temperature a unit mass of a substanceby a unit temperatur. The specific heat capacity is temperature dependent,however, it is convenient to use mean values for this parameter, which isthe heat capacity evaluated at the arithmetic mean temperature for a giventemperature range. This has been used through the entire project. T air isthe air temperature(oC ), T ref is the reference temperature. (0
oC ), is usedas the reference temperature at which there is zero enthalpy. λ is the latentheat of vaporization, which is the heat required for water to change fromliquid- to gas phase (vaporize).
(4.3) is a simplification since it is assumed that the final enthalpy is inde-pendent of the vaporisation path, accordingly the vaporisation is assumedto take place at (0oC ) at which the enthalpy is chosen to be zero and thensuperheated to the air temperature T air . Originally to reach vapour state,the vaporisation occurs at the dew point temperature, which is the temper-ature at which the air become saturated and then heated up to the final airtemperature. This becomes of practical importance for the calculations if the absolute humidity is above 0.05 Kgwaterkgdry (Mujumdar (2007)). The en-
thalpy of the feed entering the dryer is the sum of the enthalpy of the drysolid and the moisture liquid in the product
H feed = C solid(T feed − T ref ) + X In C water(T feed − T ref ) (4.4)H powder = C solid(T powder − T ref ) + X outC water(T powder − T ref ) (4.5)
where C solid and C water are the specific heat capacity of dry solid and water.X In/out is the solids moisture content and is based on a unit weight of dry
-
8/20/2019 Modelling and Control of a Spray Drying Process
49/225
24 Modelling a Spray Dryer
product ( KgwaterKgsolids ) . The reason for using a dry basis in the equations abovefor the air and powder moisture content is that the flow rates of the dryair and the dry solids is the same at both the inlet and outlet, which makesthe calculations more straightforward as the moisture now is directly relatedto the dry substance. (4.5) for the enthalpy of powder leaving the dryer is
similar to the enthalpy of feed (4.4). It is assumed that all the moistureevaporated from the feed is absorbed by the outlet air and taken out of thedryer. Hence the moisture content in the final powder can be related to theoutlet humidity of the dryer by a mass balance.
4.2.1 Mass Balance
The mass balance over the spray dryer relates the moisture entering thedryer with the outgoing moisture and gives (4.6). Due to the assumption of a well mixed dryer and equilibrium state between outlet air and solids, theoutlet moisture content of the powder X 0 is expected to be the equilibriummoisture content of the solid in the respective air conditions. The outletabsolute humidity Y out is then isolated.
F MainDryY In + F SFBDryY In + M sIn X In = F OutDryY Out + M sOutX Out
(4.6)
F MainDry + F SFBDry = F OutDry
M sIn = M sOut
M s(X In − X out) = F Outdry (Y Out − Y In )
Y Out = Y In + M sF Outdry(X In − X out) (4.7)
(4.7) is inserted into the previously stated energy balance equation, whichresults in (4.8). Hereafter the unknown and unspecified parameters are:Primary and secondary inlet airflows, F Main and F SF B respectively and thebelonging air temperatures, T Main and T SF B , the moisture content of theair going into the system Y In , the moisture contents of the feed X In and thefinal product X Out, and the in- and outlet solids rate M sIn and M sOut.
F MainDry(C dryAirT Main + Y In (λ + C vaporT Main))
+ F SFBDry(C dryAirT SF B + Y In (λ + C vaporT SF B))
+ M s(C solidT feed + X In C waterT feed)
= F OutDry(Y In + M s
F OutDry(X In − X out))(λ + C vaporT Outair)
+ C dryAirT Outair) + M s(C solidT powder + X outC waterT powder ) (4.8)
-
8/20/2019 Modelling and Control of a Spray Drying Process
50/225
4.2 Steady State Mass and Energy Balances for Spray Dryers 25
T Outair = T powder (4.9)
In our present situation, the temperature of the outlet drying air and theresulting moisture content of the final product are the variables that arerequired to be estimated. As it can be seen from the two equations abovethese variables are influenced by the input operational variables and materialparameters outlined in the block diagram in figure 4.4. In the calculationprocess the input operational variables to the plant are assumed to be knownand predetermined. Thus by first determining the outlet powder moisturecontent, (4.7) can be solved for the outlet drying air humidity. Taking theassumption into account that the temperature of the gas and the product issimilar, this temperature can be computed from (4.8).
4.2.2 Equilibrium Moisture Content
The moisture content of the outlet powder is approximated to be the equi-librium moisture content. The equilibrium moisture content is the result-ing state of an interaction between the environment and the substance, towhich the moisture content of the substance converges to either by moistureuptake(adsorption) or by drying(desorption). It is noted that equilibriummoisture content may vary depending on whether the substance is exposed toadsorption or desorption. So changes in the moisture content of a substanceare dependent on the surrounding partial vapour pressure and temperaturecondition, but also on the nature of the solids. After an adequate amount of time has passed with steady state condition an internal moisture diffusionbalance takes place until the equilibrium moisture content is attained. Thusfor the vapour pressure at a given temperature the substance will have astate where it will neither gain nor lose any moisture.
This relationship between the equilibrium vapour pressure and the moisturecontent in the substance can be presented by a moisture sorption isothermfunction. This sorption isotherm designates the equilibrium moisture con-tent for a certain humidity value, at a constant temperature and herebygives a description of a products ability to bind water. Due to the complex-ity of the sorption process, the isotherm cannot be determined analytically,but instead measured experimentally. Different products and materials havedifferent hygroscopic properties, which is affected by their molecular struc-
ture and their solubility. There are various empirical relations describingthe sorption characteristics for food ingredients using different models inliterature.
The desorption isotherm for maltodextrin, which is the test material that hasbeen used in this project, is determined in (Jesús M. Fŕıas and Schittkowski
-
8/20/2019 Modelling and Control of a Spray Drying Process
51/225
26 Modelling a Spray Dryer
(2001)). Maltodextrin is a polysaccharide and consist of dextrose(glucose)molecules connected in a chain of variable length. The length of the moleculechain is described by a DE(Dextrose equivalent) number and explains itsproperties(flavour). The sorption isotherm found in (Jesús M. Fŕıas andSchittkowski (2001)) is for maltodextrin DE 12 and the maltodextrin used
in our test setup is of type DE 10. The difference was discussed with achemist from GEA NIRO and it was found acceptable to use the sorptionisotherm for maltodextrin DE 12.
As explained there will be a slight dissimilarity in moisture binding proper-ties though it is insignificant and can be neglected. The equilibrium mois-ture content model is based on Guggenheim-Anderson-de Boer (GAB) modelequation, as recommended by the ”‘European Project Group COST 90 onthe Physical Properties of Foods”’ 1 for the characterisation of water sorptionin food materials.
X eq(T, aw) = C eqK eqW eqaw
(1 − K eqaw)(1 − K eqaw + C eqK eqaw) (4.10)
Where the model parameters W m, C eq and K eq are determined by (JesúsM. Fŕıas and Schittkowski (2001)). All the parameters are dependent on thetemperature of the solid in celsius.
C eq = 0.04exp( 1257.14
T solid + 273) (4.11)
K eq = 0.65exp( 144.57
T solid + 273 ) (4.12)
W eq = 0.05exp( −99.27
T solid + 273) (4.13)
To characterise equilibrium vapour pressure in the sorption isotherm therelative humidity content of the drying air is applied, as the vapour pressurein the solid is equal to the partial vapour pressure in the drying air whenno more moisture can be lost to the surroundings and thus be equilibrium.This is also known as the water activity of the product aw. The equilibriummoisture content model is derived from desorption isotherm measurements
performed at four different temperatures (4
o
C , 25
o
C , 37
o
C , 50
o
C ). Sinceit is expected that the spray dryer will work at higher temperature rates,the behaviour of the model is tested at higher temperatures. In figure (4.5)a graph is produced in which the equilibrium moisture content is plotted
1http://www.esf.org/ (2/10-09)
-
8/20/2019 Modelling and Control of a Spray Drying Process
52/225
4.2 Steady State Mass and Energy Balances for Spray Dryers 27
against the relative humidity for various temperatures. If the relative hu-midity of the surrounding air is close to zero, then the equilibrium moistureinside the dry product also is nearly zero independent of the temperature.At higher temperatures a larger variation in equilibrium moisture contentis noted. The model is unrealiable for water activities above 0.9. For large
values of water activities values (4.10) gets negative.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Water activity aw
E q u i l i b r i u m M o i o s t u r
e C o n t e n t ( K g H
2 O
/ K g s o l i d
)
Desorption Isotherm for Maltodextrin DE12 at various Temperatures
4oC
25oC
37oC
50oC
65oC
85oC
100oC
115oC
increasing T
Figure 4.5: Desorption Isotherm Maltodextrin DE12: Equilibrium moisture con-
tent as function of water activity for temperatures between 40C and1150C . If the relative humidity of the surrounding air is close tozero, then the equilibrium moisture inside the dry product also isnearly zero independent of the temperature. At higher temperaturesa larger variation in equilibrium moisture content is noted. The modelis unrealiable for water activities above 0.9
The relative humidity ψ of the vapour gas mixture is measured as the frac-tional saturation with moisture and is defined as the ratio of the partialvapour pressure P v to the saturated pressure P sat at the same temperature:
ψ = P v
P sat(4.14)
For drying to take place the relative humidity of the surrounding drying airmust be lower than the water activity of the product. The partial vapour
-
8/20/2019 Modelling and Control of a Spray Drying Process
53/225
28 Modelling a Spray Dryer
pressure is related to the absolute humidity of the surrounding air and isrecalled to be the ratio of vapour mass mv to mass of dry air mdryAIR. Usingthe gas law for the two fractions at constant temperature T and total volumeV results in (4.15). R is the universal gasconstant (8.314 J (mol·K )
mv = P vV
RT M w
mdryAir = P dryAirV
RT M dryAir
Y = P vP dryAir
M wM dryAir
(4.15)
Dalton’s law (4.16) states that the total pressure exerted by mixture of gasis equal to the sum of the partial pressures of each fraction in the mixture
and knowing the molar mass of water is M w = 18.01
g
mol and and dry airM dryAir = 28.96 gmol (4.15) becomes:
P total = P v + P dryAir (4.16)
Y = 0.622 P v
P total − P v(4.17)
The outcome of rearranging the above written equation is the partial vapourpressure as function of absolute humidity Y, since P total is assumed to beequal to the standard atmospheric pressure : 101325 Pa(Langrish (2008)).
P v = ( Y 0.622 )P atm
1 + ( Y 0.622 )(4.18)
At 100 % relative humidity, the partial vapour pressure equals the vapourpressure of liquid and the drying air or the surface of the substance is saidto be saturated with vapour. There are many formulations to calculate thesaturation vapour pressure.(4.19) (Richard Shelquist (2009)) used here issimple with only 3 parameters and still offers good results when compared
to the Smithsonian reference table for vapour pressure found in (Wiederholt(1997)). A deviation of 1 percent at high temperatures (above 100oC ) andmuch less for lower temperatures from reference value is acceptable, as thiswill only have a very small effect on the relative humidity calculation andequilibrium moisture content (see appendix A).
-
8/20/2019 Modelling and Control of a Spray Drying Process
54/225
4.3 Test on a Multi Stage Dryer- MSD20 29
P sat = 100 · C 0 · 10C 1T
C 2+T (4.19)
C 0 = 6.1078
C 1 = 7.5
C 2 = 237.3
With the system of equations put forward in this section, the temperatureand moisture content of the outlet air, T out and Y out respectively can bedetermined. To that, the moisture content of the final product is estimated.The procedure is written in a Matlab script and solved with Matlab.
4.2.3 Steady State Solution
There are six unknown variables (X out, Y out, ψ , P v, P sat,and T outair) and six
corresponding equations ((4.19), (4.18), (4.14), (4.10), (4.8) , and (4.7))which means the system has a unique solution that is found by the followingiterative process.
1. X 0 is initialised to have the same value as X In , moisture in the feed
2. Y out is calculated using equation (4.7)
3. T outAir is solved for equation (4.8), where T powder is set to be equal toT outAir.
4. P v is calculated using equation (4.18)
5. P sat is calculated using equation (4.19)
6. ψ is calculated using equation (4.14)
7. X out is calculated using equation (4.10). Hereafter the process is re-turned to the second step and a new outlet moisture value Y out iscalculated. This is reiterated until the process converges to the finalvalues. In the Matlab script the process stops when the differencebetween the previous determined temperature and the presently cal-culated temperature is less than (1/1000).
4.3 Test on a Multi Stage Dryer- MSD20
With the purpose of being able to validate the correctness of the dynamicsand the steady state values of the created model, a test has been prepared
-
8/20/2019 Modelling and Control of a Spray Drying Process
55/225
30 Modelling a Spray Dryer
and completed on a real spray dryer system. The tests was basically aboutputting in step changes at the most important input operation variables andexamine the resulting step responses for some significant output variables.Such a test on an actual system would depict the known but also the hiddendynamics of the system and thus explain its behaviour for certain changes
in the system.
The test was completed on a Multi Stage Dryer(MSD)-20 open cycle2 systemat GEA NIRO’s test station in Soeborg, Denmark. Maltodextrin DE 10 wasused as the material to be spray dried. As drying gas atmospheric air wasapplied with the use of an electric heater. The system setup was as shownin figure 4.6.
The drying chamber has a diameter of 2 m and a height of 2.30 m, witha total volume of approximately 10.3 m3. Compared to the largest spraydryers used in the dairy industry which can be up to 16 m in diameter witha total volume of 1920 m3, this is a small one. This spray dryer can produce
approximately 70 kg
hour of powder.The input variables which are chosen to be manipulated are the most signif-icant ones and have the greatest effect on the spray drying process. That isthe feed flow rate, given by sensor 1626, the Main inlet airflow and temper-ature, (sensor 1701) and 1702 respectively. Similarly the SFB inlet airflow(1703) and temperature (1704) are controlled. In the end the airflow intothe Vibrofluidizer is operated, but this is only commented superficially giventhat the focus in this project has been on the drying chamber and the processin it. The central output variables are the outlet air temperature (1709),outlet powder particle size and residual moisture content. (see table in app.C.1 and figure 4.6)
Furthermore in the figure the variables are marked as either manually (M ) orautomatically (A) controlled. Automatical control indicates that a variableis controlled by a PI regulator when the specific variable value is set. Forthe manually controlled variable an operator sets the controller output value.As described previously, under normal circumstances the moisture contentof the powder leaving the dryer, is controlled indirectly by maintaining aconstant outlet drying air temperature through the regulation of the speedon the feed rate pump.
Throughout the test this feedback loop was disconnected, so the main sys-tem functioned as an open loop and the feed rate was controlled manually
with pumps 1639 and 1606. The inlet air flows and temperatures were au-tomatically controlled, likewise the pressure inside the chamber. This isto prevent the chamber from crumpling up due to the air flows. The PI
2In a open cycle spray dryer the drying medium is not reused. Air enters one placeand exits at another place
-
8/20/2019 Modelling and Control of a Spray Drying Process
56/225
4.3 Test on a Multi Stage Dryer- MSD20 31
parameters are shown in the logbook from the test in appendix C.3. The
Manipulated variablesSensor Name Description Control Unit
MAINKGH Main air flow into chamber Auto Kg/h
T1702 Temperature of MAINKGH Auto o
CSFBKGH SFB air flow into chamber Auto Kg/h
T1704 Temperature of SFBKGH Auto oCF1626 Feed flow into chamber Manual L/h
Table 4.1: The manipulated variables in test of the dynamic process(MSD-20).Auto= automatic control (PI), manual: Manual control
blue lines in figure 4.6 depict air flows, while the yellow lines show productflow. A cyclone was used to filter out the fines from the outlet air. Thesewere returned at the top of the chamber to be applied in the agglomerationprocess. The bag filter shown was disengaged.
Each