Large eddy simulation of pulverised coal combustion

Combustion

von der

Fakultät Energie-, Verfahrens- und Biotechnik der Univerät Stuttgart

zur Erlangung der Würde eines Doktors der Ingenieurwissenschaften

(Dr.-Ing.) genehmigte Abhandlung

vorgelegt von Gregor Olenik aus Erfurt, Deutschland

Hauptberichter: Prof. Dr. Andreas Kronenburg Mitberichter: apl. Prof. Dr.-Ing. Uwe Schnell Tag der mündlichen Prüfung: 28. Mai 2020

Institut für Technische Verbrennung der Universität Stuttgart

Hiermit versichere ich:

1. dass ich meine Arbeit selbständig verfasst habe,

2. dass ich keine anderen als die angegebenen Quellen benutzt und alle wörtlich oder sinngemäss aus anderen Werken übernommenen Aussagen als solche gekenn-zeichnet habe,

3. dass die eingereichte Arbeit weder vollständig noch in wesentlichen Teilen Gegen-stand eines anderen Prüfungsverfahrens gewesen ist,

4. dass ich die Arbeit noch nicht vollständig veröffentlicht habe und,

5. dass das elektronische Exemplar mit den anderen Exemplaren übereinstimmt.

Stuttgart, November 2020

Gregor Olenik

Parts of this thesis have been presented at conferences and published in the archival literature. The background and theoretical developments given in chapters 2-6.3 have been modified with respect to the originally published texts and been significantly extended to ensure completeness, coherence and consistency of the present manuscript.

The relevant papers are:

1. G. Olenik, O.T. Stein, A. Kronenburg, LES of a piloted pulverized coal jet, 1st ERCOF-TAC Conference on Simulation of Multiphase Flows in Gasification and Combustion, Dresden, Germany (2011)

• This paper has been published in the conference proceedings. Data and results discussed therein are presented in chapter 7.

• Author’s contribution: Programming (100%), data generation (100%), scientific originality (50%)

2. O. T. Stein, G. Olenik, A. Kronenburg, F. Cavallo, Marincola, B. M. Franchetti, A. M. Kempf, M. Ghiani, M. Vascellari, C. Hasse, Towards comprehensive coal combustion modelling for LES, Flow, Turbulence and Combustion, Volume 90, Issue 4, pages 859–884, (2013)

• Data and results discussed in this paper are presented in chapter 7.

• Author’s contribution: Programming (30%), data generation (25%), scientific originality (30%)

3. G. Olenik, O.T. Stein, A. Kronenburg, LES of an industrial scale swirled pulverised coal furnace, 26th German Flame Day, VDI-Tagungsband 2161, pages 235-242 (2013). • This paper has been published in the conference proceedings. Data and results

discussed therein are presented in chapter 8.

• Author’s contribution: Programming (100%), data generation (100%), scientific originality (50%)

vi

4. G. Olenik, O. T. Stein, A. Kronenburg, LES of swirl-stabilised pulverised coal com-bustion in IFRF furnace No. 1, Proceedings of the Comcom-bustion Institute, Volume 35, Issue 3, pages 2819-2828, (2015)

• Data and results discussed in this paper are presented in chapter 8.

• Author’s contribution: Programming (100%), data generation (100%), scientific originality (60%)

Die vorliegenden Arbeit entstand an der Fakultät Energie-, Verfahrens- und Biotechnik der Universität Stuttgart. Hier durfte ich unter der Betreuung von Professor Dr. Andreas Kro-nenburg und Dr. Oliver Stein an meinem Forschungsprojekt arbeiten. Ich möchte daher meine große Dankbarkeit Professor Dr. Andreas Kronenburg für die Ermöglichung dieser Arbeit und das mir entgegengebrachte Vertrauen danken. Meine besondere Dankbarkeit gilt auch Dr. Oliver Stein für zahllose Stunden spannender wissenschaftlicher Diskussio-nen, seiner großen Hilfe das vorliegende Schriftstück in eine lesbare Form zu bringen und seinen Einsatz bei der Heranführung an die gewissenhafte wissenschaftliche Arbeit. Wei-terhin möchte ich auch Professor Dr.-Ing. Uwe Schnell für seine zahlreichen Rätschläge zur Verbesserung dieser Arbeit danken. Auch möchte ich Ricarda Schubert, der guten Seele des ITV danken, insbesondere für ihre Hilfe bei allen formalen Fragen.

Mein besonderer Dank gilt meiner Frau Lene Ersfeld und meinen Kindern Paula, Greta und August. Ohne ihre Geduld und Liebe wäre die vorliegende Arbeit unmöglich gewesen. Daher möchte ich auch um Verzeihung bitten, für die ungezählten Stunden, die ich als Mann und Vater fern von euch war. Mein Dank gilt auch meinen Eltern Dagmar und Uwe Olenik, ohne sie und ihre Unterstützung wäre ich heute nicht, wo ich bin. Ebenso danke ich der Familie Ersfeld für ihre zuverlässige Unterstützung.

Niko Seubert möchte ich danken, dass er mir stets eine Couch zum Schlafen bot, ein of-fenes Ohr hatte und überhaupt einer der besten Freunde ist, die man sich wünschen kann. Gregor Neuber möchte ich für die vielen Gespräche danken, die mir Kraft und Inspira-tion brachten. Mein Dank gilt auch der gesamten Mann/Frauschaft und Weggefährten am ITV: Anna Stahl, Jesus Rivas, Satoshi Ukai, Giovanni Luigi Tufano, Dirk Dietzel, Carmen Straub, Milena Smiljanic, Jonas Kirchmann, Bosen Wang, Son Vo, Santanu De, Papakorn Siwaborworn, Jung Choi, Benedikt Heine, Marvin Sontheimer, Jesus Rivas, Francisco Car-rasco Maldona, Shreya Shekar, Anup Date und Tyler Pelkey. Dank gilt auch meinen Freunden, die mich in dieser Zeit begleitet haben: Samuel Braun, Georg Blesinger, Oliver Kolbe, Jan Köpsel, Thomas Walter, Henrike Polek, Theresa Ziegenhorn, Catalina Schmitt, Marie Falke, Arthur de Conihout, Daniel Fritzsche, Xavier Kleitz, Matthias Pürzel, Peggy Achsnick, Michael (Michi) and Christina (Resi) Meuser, Johannes Hauer, Johannes Ott, Gäa and Florian Ziemann und natürlich den Zugspitz-Bergsteigern: Volker Bartenbach, Franz Pütz, Tobias Fickenscher, Thorsten Schulz und Tobias Brack.

viii

Erklärung iii Preface v Danksagungen vii Table of Contents ix Summary xxix Zusammenfassung xxxi Nomenclature xxxiii 1 Introduction 1 1.1 Motivation . . . 1

1.2 State of the Art . . . 2

1.3 Project Focus . . . 4

1.4 Outline . . . 4

I Theoretical Background

7

2.2 Coal Combustion . . . 12

2.2.1 Coal Drying and Particle Heat-up . . . 12

2.2.2 Pyrolysis . . . 13

2.2.3 Volatile Combustion . . . 22

2.2.4 Char Combustion . . . 23

2.2.5 Radiation in Coal Flames . . . 26

CONTENTS x

3 Fundamentals of Reacting Two Phase Flows 27

3.1 Governing Equations . . . 27

3.1.1 Gas Phase . . . 27

3.1.2 Particles . . . 29

3.1.3 Thermal Radiation . . . 30

3.2 A Brief Overview of Turbulence . . . 32

3.3 Phenomenology of Turbulence-Chemistry Interactions . . . 33

3.4 Turbulence Particle Interaction . . . 34

4 Computational Fluid Dynamics 37 4.1 Overview of Methods to Transform PDEs to AEs . . . 37

4.2 An Overview of the RANS, LES, and DNS Methods . . . 38

4.3 Introduction to the Filtering Procedure for CFD . . . 39

4.4 Transformation of Transport Equations to a System of Algebraic Equa-tions . . . 41

II Model Formulation

45

5.1.1 LES Transport Equations for the Gas-phase . . . 48

5.1.2 Turbulence Model . . . 51

5.1.3 Turbulence Chemistry Interaction Models . . . 53

5.2 Solid Phase Modelling . . . 54

5.3 Interphase Transfer Modelling . . . 56

5.3.1 The Parcel Approach . . . 57

5.3.2 Particle Momentum Transfer Models . . . 57

5.3.3 Particle Heat Transfer Model . . . 59

5.3.4 Radiation Modelling . . . 60

5.3.5 Mass Transfer Models . . . 61

6 Pyrolysis Modelling 63 6.1 Mass Basis and Q-factor Definitions of Coal Compositions . . . 63

6.2 Implemented Pyrolysis Models . . . 64

6.2.1 Constant Rate Model . . . 65

6.2.2 Single Kinetic Rate Model . . . 65

6.2.3 Competing Rates Model . . . 67

6.2.4 Richards & Fletcher Pyrolysis Model . . . 68

6.3.1 The Pyrolysis Kinetic Preprocessor . . . 72

6.3.2 The CPD Preprocessor . . . 75

6.3.3 Calibration of Target Model Properties . . . 80

6.3.4 Postulate Substance Compositions . . . 83

6.3.5 Postulate Substance Enthalpies . . . 86

6.3.6 Effects of the two Postulate Substance Composition Correction . . . 89

6.4 Optimisation of Pyrolysis Models for Specific Combustion Conditions . . . 90

6.4.1 Constant Rate Pyrolysis Model . . . 93

6.4.2 Single Kinetic Rate Model . . . 94

6.4.3 The Kobayashi Pyrolysis Model . . . 95

6.4.4 Richards and Fletcher Pyrolysis Model . . . 96

III Results

99

7.2 Simulation Strategy . . . 103

7.3 CRIEPI Isothermal Case . . . 106

7.4 Comparison of Isothermal LES Results with Experimental Data . . . 107

7.5 General Flow and Flame Structure . . . 109

7.6 Sensitivity to the Pyrolysis Model . . . 111

7.7 Comparison of LES Results and Experimental Data . . . 113

7.8 Conclusion . . . 124

8 The IFRF Test Case 125 8.1 Experimental Case Setup . . . 125

8.2 Computational Case Setup . . . 127

8.2.1 Inflow Data Generation . . . 136

8.3 General Flow and Flame Structure . . . 137

8.4 Sensitivity to the Pyrolysis Model . . . 140

8.5 Comparison of LES Results and Experimental Data . . . 146

8.6 Conclusions . . . 155

9 Conclusions and Outlook 157 Appendices 175 Appendix A Additional Theoretical Background 177 A.1 Frobenius Inner Product . . . 177

CONTENTS xii

A.3 Boussinesq Approximation . . . 178

A.4 Properties of Density Weighted Conservation Equations for RANS and LES 178 A.5 The Closure Problem of the Homogeneous LES Reaction Term . . . 182

A.6 Molecular Models of Coal . . . 184

A.7 Further Bethe Lattice Properties . . . 187

A.8 Boundary Conditions . . . 188

A.8.1 LES Inflow Conditions . . . 188

A.8.2 Wall Models . . . 191

Appendix B Implementation Details of OpenFOAM 195 B.1 Overview of OpenFOAM’s Solution Procedure for Transport Equations . . 196

B.2 Spatial Interpolation and Integration Schemes . . . 197

B.3 Numerical Instabilities of Low Order Spatial Interpolation Schemes . . . . 199

B.4 High Resolution Spatial Interpolation Schemes . . . 200

B.5 Pressure Correction . . . 201

B.5.1 OpenFOAM Compressible PIMPLE Implementation . . . 202

B.5.2 Rhie and Chow Interpolation . . . 206

B.6 Explicit and Implicit Discretisation of Transport Equations . . . 207

B.7 Incident Radiation Transport Equation . . . 210

B.8 Gas-phase Submodels . . . 212

Appendix C Implementation Details of CoalFOAM 213 C.1 Package Structure . . . 213

C.2 Computational Procedure . . . 213

C.3 Lagrangian Classes . . . 216

C.3.1 Integration of the Particle Motion Equation . . . 220

C.3.2 Integration of the Particle Temperature Equation . . . 221

C.3.3 TheCOxidationKineticDiffusionLimitedRateChar Conversion Model . . . 222

C.4 Transport Equations . . . 223

C.4.1 Continuity Equation . . . 223

C.4.2 Incompressible Pressure Correction Equation . . . 223

C.4.3 Enthalpy Transport Equation . . . 223

C.4.4 Species Transport Equation . . . 224

C.5 Particle Sub-models . . . 224

C.5.1 Devolatilisation and Thermolysis Models . . . 225

Appendix D Additional Results 229

D.1 Further Pyrolysis Model Parameter Study Results . . . 229

D.1.1 Single Kinetic Rate Model . . . 229

D.1.2 The Competing Rates Model by Kobayashi et al. . . 234

D.2 CRIEPI Isothermal Inflow Conditions . . . 238

D.3 CRIEPI Reacting Case Parameter Studies . . . 239

D.3.1 Grid Independence Study . . . 240

D.3.2 Statistical Convergence Study . . . 246

D.3.3 Influence of the Time Step Size and Time Integration Schemes . . . 248

D.3.4 Lagrangian Parcel Injection Rate . . . 250

D.3.5 Particle Dispersion Model . . . 251

D.3.6 Particle Heat Capacity Influence . . . 253

D.3.7 Thermolysis Enthalpy . . . 255

D.3.8 Particle Sampling Probe Size . . . 257

D.3.9 Turbulent Inflow Sensitivity . . . 259

D.3.10 TCI Sensitivity . . . 262

D.3.11 Particle Size Distributions . . . 265

D.3.12 Comparison of the OpenFOAM Char Conversion and the Baum and Street Model . . . 266

D.3.13 JANAF Polynomials . . . 267

D.4 IFRF Preliminary Studies . . . 269

D.4.1 Grid Independence Study . . . 270

D.4.2 Statistical Convergence Study . . . 278

D.4.3 Inflow Conditions . . . 283

D.4.4 Dispersion Model . . . 287

2.1 Example apparatus for different coal analysis techniques: a)

thermogravi-metric analysis, b) drop tube furnace, c) heated grid. . . 17 2.2 Different types of lattices used to model the molecular structure of coal.

On the left side a linear chain (DISCHAIN), in the middle a Bethe lattice (DISARAY and CPD), and on the right side a two dimensional lattice

(FG-DVC). . . 19 2.3 Different types of lattices after some pyrolyis progress. The linear chain

(DISCHAIN, left) and a two dimensional lattice model (FG-DVC, right). . 21 6.1 Profile of the inverse error function erfinv versus the pyrolysis progress

(1 −YInt) for different coefficientsσi. . . 70 6.2 Volatile yield (top) and its time derivative (bottom) over temperature for

different coefficientsσ_i. Low heating rates (L) are drawn in blue, medium

(M) in red, and high (H) in orange. The base parameter set is drawn with solid lines. Coarse and fine dashed lines are for σ_{i= 2000 and −2000,}

respectively. . . 71 6.3 Inverse temperature Θ at which a certain volatile yield percentage is

emit-ted over a range of the coefficientsσ_i. Low heating rates (L) are drawn in

blue, medium (M) in red, and high (H) in orange. The 5% volatile yield profile set is drawn with solid lines, the 95% volatile yield profile with fine

dashed lines, and the 50% volatile yield with coarse dashed lines. . . 71 6.4 Basic PKP workflow. . . 74 6.5 Implementation of the fitting procedure for the pyrolysis model

coeffi-cients. . . 81 6.6 Elemental composition of the total volatile yield of Newlands coal over the

Q-factor. . . 89

LIST OF FIGURES xvi

6.7 Carbon content (left column) and heating value (right column) of the total volatile yield for different coefficientsσ₂versus temperature. Low heating

rates (L) are drawn in blue, medium (M) in red, and high (H) in orange. In the top row results forσ₂= 1, in the middle row σ₂= −2000, and the

bottom rowσ₂= 2000 shown. . . 91 6.8 Particle temperature versus particle residence time t, color-coded with

cor-responding particle mass fraction of ash, solid lines indicate the extracted

high, medium, and low particle temperature profiles for the CRIEPI case. 92 6.9 Volatile yield over residence time computed by CPD (top) and the

corre-sponding fitted particle temperatures which served as input values

(bot-tom). . . 93 6.10 Volatile yields over time (left, top) and temperature (right, top) for the

Constant Rate pyrolysis model (CR, solid lines) and CPD results (CPD, dashed lines). The corresponding rates are shown in the bottom row. The low, medium, and high temperature profile data are indicated by the L, M,

and H suffix. . . 94 6.11 Volatile yields over time (left, top) and temperature (right, top) for the

Arrhenius rate pyrolysis model (SKR, solid lines) and CPD results (CPD, dashed lines). The corresponding rates are shown in the bottom row. The low, medium, and high temperature profile data are indicated by the L, M,

and H suffix. . . 95 6.12 Volatile yields over time (left, top) and temperature (right, top) for the

Kobayashi pyrolysis model (KOBA, solid lines) and CPD results (CPD, dashed lines). The corresponding rates are shown in the bottom row. The low, medium, and high temperature profile data are indicated by the L, M,

and H suffix. . . 96 6.13 Volatile yields over time (left, top) and temperature (right, top) for the

Richards & Fletcher pyrolysis model (RF, solid lines) and CPD results (CPD, dashed lines). The corresponding rates are shown in the bottom row. The low, medium, and high temperature profile data are indicated by

the L, M, and H suffix. . . 97 7.1 Schematic diagram of the CRIEPI burner, dimensions given in mm. . . . 102 7.2 Computational grid with boundaries. . . 105 7.3 Mean axial velocity component of the gas-phase (left) and the coal particle

7.4 Axial mean velocities (top) and velocity RMS (bottom) in comparison to experimental LDV data (EXP). Here, LES-G denotes the gas-phase only, LES-GP the gas-phase data of the EUL-LAG, and LAG the particle data of

the EUL-LAG simulation. . . 108 7.5 Radial profiles of mean velocities (left) and velocity RMS (right) in

compar-ison to experimental LDV data. Here, LES-G denotes the gas-phase only, LES-GP the gas-phase data of the EUL-LAG, and LAG the particle data of

the EUL-LAG simulation. . . 109 7.6 Mean axial velocity component of the gas-phase (left) and the coal particle

cloud colored by its instantaneous axial velocity component (right). . . 110 7.7 Mean temperature of the gas-phase (left) and the coal particle cloud

col-ored by its instantaneous temperature (right). . . 111 7.8 Axial profiles of mean axial velocities (top), velocity RMS (middle), and

mean temperature (bottom) along the centerline for the SKR, the KOBA,

and the RF model with identical postulate substances. . . 113 7.9 Profiles of mean axial velocities (top), velocity RMS (middle), and mean

temperature (bottom) along the centerline. . . 117 7.10 Profiles of particle mean axial velocities (top), velocity RMS (middle), and

mean temperature (bottom) along the centerline. . . 118 7.11 Profiles of mean species concentrations along the centerline. . . 120 7.12 Radial profiles of mean particle velocities (left) and corresponding particle

velocity RMS (right). The suffixes L,S are used to indicate LDV and SPDA

data respectively. . . 121 7.13 Temperature scatter plot of instantaneous particle data (TP) and profiles

extracted by rolling mean (RMT), weighted rolling mean (RMT4), and

ex-perimental data (EXP). . . 122 7.14 Particle mean diameter along the centerline for the RF2S case (LES) and

experimental data (EXP). . . 123 7.15 Probability density function of particle diameters extracted from

instanta-neous particle data. . . 123 8.1 Computational domains for the IFRF furnace simulation, dimensions

given in [mm] . . . 126 8.2 Schematic of the computational grid. Shown is a close-up of the first 2 m

of the coarse mesh, level 2 (top), a close-up of the quarl of the fine mesh,

level 10 (bottom, left), and a 3D view of the coarse mesh (bottom, right). . 132 8.3 LES and experimental profiles of particle size distribution (CDF) . . . 132

LIST OF FIGURES xviii

8.4 Snapshots of the instantaneous axial velocity component (left) and the swirling velocity component (right) of the unswirled and the superimposed swirled inflow within the annulus as calculated by means of DES. The axis

of symmetry of the inflow annulus is indicated by the dashed lines. . . 136 8.5 Snapshots of the instantaneous axial velocity component (left) and the

swirling velocity component (right). For better visibility a close-up of the

quarl region is shown. Dimensions given in [m]. . . 138 8.6 Snapshot of the instantaneous temperature field. Dimensions given in [m]. 139 8.7 Snapshots of the instantaneous species mass fraction fields. Dimensions

given in [m]. . . 140 8.8 PKP results of volatile yields versus time for the Knill dataset and the

PKP optimised SKR and RF models for the lower temperature (top), the medium temperature (middle), and the higher (bottom) temperature

pro-file. . . 142 8.9 Snapshots of the mean volatile yield mass fraction for the Knill-SKR (left),

the PKP-SKR (middle), and the PKP-RF (right) coefficient sets in the

vicin-ity of the quarl. Dimensions given in [m]. . . 143 8.10 Mean gas-phase temperatures over radial position at the first four axial

locations for the Knill-SKR, PKP-SKR, and the PKP-RF coefficient sets. . 144 8.11 Mean gas-phase temperatures over radial position at the three

down-stream axial locations for the Knill-SKR, PKP-SKR, and the PKP-RF

coef-ficient sets. . . 144 8.12 Snapshots of the mean CO2 mass fraction for the Knill-SKR (left), the

PKP-SKR (middle), and the PKP-RF (right) coefficient sets. Dimensions

given in [m]. . . 146 8.13 LES and experimental radial profiles of mean gas-phase axial velocities

(left column) and corresponding velocity RMS (right column). . . 147 8.14 Radial profiles of mean gas-phase (EUL) and particle (LAG) axial

veloc-ities (left column) and corresponding swirling velocity components (right

column) for the RF case. . . 148 8.15 LES and experimental radial profiles of mean gas-phase swirling velocities

(left column) and corresponding velocity RMS (right column). . . 149 8.16 LES and experimental radial profiles of mean gas-phase temperatures for

the upstream measurement locations. . . 150 8.17 LES and experimental radial profiles of mean gas-phase temperatures for

the downstream measurement locations. . . 150 8.18 Axial profiles of mean temperature (top), CO₂ (middle), and O₂ (bottom)

8.19 LES and experimental radial profiles of the mean species concentrations of CO₂(left column) and O₂(right column) for the upstream measurement

locations. . . 153 8.20 LES and experimental radial profiles of the mean species concentrations of

CO₂(left column) and O₂(right column) for the downstream measurement

locations. . . 153 8.21 Profiles of char burnout along the z-direction for different radial positions.

The radial distance from the centerline in [mm] is indicated by the trailing

number in the corresponding legend label . . . 154 A.2 Probability that a given site is part of a cluster of size n as a function of the

coordination number with p = 0.5 (top) and the fraction of intact bridges

withσ_{= 2 (bottom) . . . .} 188

B.1 OpenFOAM PIMPLE procedure. . . 203 B.2 Pseudo call graph for the implementation of the spatial discretisation

pro-cedure. The scope of a function is depicted as a solid box. Indentation

indicates the call of a sub-function. . . 208 C.1 Excerpt of the CoalFOAM directory structure. . . 214 C.2 Schematic of the default implementation of devolatilisation and char

con-version. . . 225 C.3 Schematic of the implementation of the pyrolysis procedure with the

addi-tional thermolysis process and subsequent char conversion. . . 227 D.1 Input temperature profiles for different constant heating rates. . . 230 D.2 Volatile yield (top) and its time derivative (bottom) over temperature for

different pre-exponential constants A (left column) and activation tem-peratures T_Ac (right column). Low heating rates (L) are drawn in blue, medium (M) in red, and high (H) in orange. The base parameter set is drawn with solid lines, the upper parameter bound with fine dashed lines,

and the lower parameter bound with coarse dashed lines. . . 231 D.3 Temperature limits at which 5% (solid lines), 50% (coarse dashes), and

95% (fine dashes) of the final yield are reached over the pre-exponential constant A (left) and the inverse of the activation temperature TAc (right). Low heating rates (L) are drawn in blue, medium (M) in red, and high (H)

in orange. . . 233 D.4 Temperature limit at which 50% is reached over the pre-exponential

LIST OF FIGURES xx

D.5 Different volatile yield components (top) and their time derivatives (bot-tom) over temperature. Low heating rates (L) are drawn in blue, medium (M) in red, and high (H) in orange. The total volatile yield is drawn with solid lines, the first yield component with coarse dashed lines, and the

sec-ond volatile yield component with fine dashed lines. . . 234 D.6 Volatile yield ratios (left) and the corresponding volatile yield profiles

(right) over temperature for different activation temperature ratios (top r=1, middle r=1.25, and bottom r=1.66). Low heating rates (L) are drawn

in blue, medium (M) in red, and high (H) in orange. . . 236 D.7 Final volatile yield versus heating rate for different pre-exponential

coef-ficient ratios (top), activation temperature ratios (middle), and conversion

factor ratios (bottom). . . 237 D.8 Axial mean velocities and velocity RMS for different inflow variances of

0.03, 0.1, and 0.3 [m2/s2] and an integral length of L = 0.3D. . . 239 D.9 Grid density influence on mean velocities (top), velocity RMS (center), and

mean gas-phase temperatures (bottom) of the CRIEPI case. . . 241 D.10 Axial profiles of the mean species mass fractions for different

computa-tional grid densities. . . 243 D.11 Radial profiles of mean velocities (left) and velocity RMS for different

com-putational grid densities. . . 244 D.12 Radial profiles of mean gas-phase temperature for different computational

grid densities. . . 245 D.13 Grid density influence on mean SGS viscosity (top) and mean molecular

viscosity (bottom). . . 246 D.14 Convergence of statistical data over different sampling time lengths with

0.25 s difference in sampling time length between profiles. The blue line

represents the first sampling period after 0.1 s. . . 247 D.15 Convergence of statistical data over different sampling time lengths with

0.25 s difference in sampling time length between profiles . . . 248 D.16 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

gas-phase temperature (bottom) for different CFL numbers. . . 249 D.17 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

gas-phase temperature (bottom) for different time integration schemes. . 250 D.18 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

gas-phase temperature (bottom) for different parcel injection rates. . . 251 D.19 Axial profiles of mean particle velocities (top), velocity RMS (centre), and

mean temperature (bottom) of the Lagrangian particles with and without

D.20 Radial profiles of mean particle velocities (left) and corresponding veloc-ity RMS (right) of the Lagrangian particles with and without dispersion

modelling. . . 253 D.21 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

temperature (bottom) for different particle heat capacities. . . 254 D.22 Radial profiles of mean temperature for different particle heat capacities. 255 D.23 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

temperature (bottom) for different values of the heat of pyrolysis. . . 256 D.24 Dependency of the mean species mass fractions on the enthalpy of

ther-molysis. . . 257 D.25 Axial profiles of mean particle velocities (top), particle velocity RMS

(cen-tre), and mean particle temperature (bottom) for different particle

sam-pling probe sizes. . . 258 D.26 Radial profiles of particle mean velocities (left) and velocity RMS (right)

for different particle sampling probe sizes. . . 259 D.27 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

temperature (bottom) for different inflow variances, L = 0.3D. . . 260 D.28 Radial profiles of mean velocities (left) and velocity RMS (right) for

differ-ent inflow variances, L = 0.3D. . . 261 D.29 Axial profiles of particle mean velocities (top), velocity RMS (centre),

and mean temperature (bottom) for different inflow length scales, V =

0.15m2s−2_{. . . . 262} D.30 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

temperature (bottom) for different EBU coefficients C₁. . . 263 D.31 Radial profiles of mean temperatures for different EBU coefficients C1. . 264 D.32 Axial profiles of mean species mass fractions for different EBU coefficients

C₁. . . 265 D.33 Axial profiles of mean velocities (top), velocity RMS (centre), and mean

temperature (bottom) for different particle size distributions. . . 266 D.34 Influence of the char conversion model on mean velocities (top), velocity

RMS (center), and mean temperatures (bottom) of the CRIEPI case. Com-paring the Baum and Street model (BS) to the OpenFOAM

COxidationKi-neticDiffusionLimitedRate (OF) model. . . 267 D.35 Axial profiles of mean velocities (top), velocity RMS (center), and mean

temperatures (bottom) for different CaHbOcNdJANAF polynomials. . . 268 D.36 Species mean concentrations along the centreline for different JANAF

LIST OF FIGURES xxii

D.37 Profiles of mean axial velocities (top), velocity RMS (centre), and mean temperatures (bottom) along the centreline for different computational

do-main lengths. . . 270 D.38 LES radial profiles of mean gas-phase axial velocities (left column) and

corresponding velocity RMS (right column). . . 272 D.39 LES radial profiles of mean gas-phase swirling velocities (left column) and

corresponding velocity RMS (right column). . . 273 D.40 LES radial profiles of mean gas-phase temperatures for the upstream

mea-surement locations. . . 274 D.41 Snapshots of the instantaneous axial velocity component on a level 2 (top

left), a level 5 (top right), a level 8 (bottom left), and a level 10 (bottom

right) grid. . . 275 D.42 Snapshots of the mean axial velocity component on a level 2 (top left), a

level 5 (top right), a level 8 (bottom left), and a level 10 (bottom right) grid. 275 D.43 Snapshots of the instantaneous swirling velocity component on a level 2

(top left), a level 5 (top right), a level 8 (bottom left), and a level 10 (bottom

right) grid. . . 276 D.44 Snapshots of the mean swirling velocity component on a level 2 (top left), a

level 5 (top right), a level 8 (bottom left), and a level 10 (bottom right) grid. 276 D.45 Snapshots of the instantaneous temperature field on a level 2 (top left), a

level 5 (top right), a level 8 (bottom left), and a level 10 (bottom right) grid. 277 D.46 Snapshots of the mean temperature field on a level 2 (top left), a level 5

(top right), a level 8 (bottom left), and a level 10 (bottom right) grid. . . . 277 D.47 Convergence of mean axial velocity (top), axial velocity fluctuations

(mid-dle) and, mean temperature (bottom) along the centreline with a continu-ous sampling over a 10 s period. Each line represents a time step with the

color changing from black to red and yellow for every 0.25 s. . . 279 D.48 Convergence of mean axial velocity (top), axial velocity fluctuations

(mid-dle) and mean temperature (bottom) along the centreline with a continu-ously restarted sampling window of 0.25 s over a total simulation time of 10 s. Each line represents an independent sampling window with the color

changing from black to red and yellow for every 0.25 s. . . 280 D.49 Convergence of axial mean velocity (top, left), axial velocity fluctuations

(top, right), mean y velocity component (bottom, left), and corresponding velocity fluctuations (bottom right) versus radial position at z = 0.25m. Each line represents an independent sampling window with the color

D.50 Convergence of mean species mass fractions, O₂ (top) and CO₂ (bottom) along the centreline with a continuously restarted sampling window of 0.25 s over a total simulation time of 10 s. Each line represents an inde-pendent sampling window with the color changing from black to red and

yellow for every 0.25 s. . . 282 D.51 Convergence of mean species mass fractions, CO₂(top, left), O₂(top, right),

H2O (bottom, left), and N2 (bottom, right) versus the radial position at z = 0.25m. Each line represents an independent sampling window with

the color changing from black to red and yellow for every 0.25 s. . . 283 D.52 Radial profiles of the mean axial velocity (left column) and the mean

swirling velocity component (right column). . . 284 D.53 Snapshots of the mean axial velocity component for a bulk inflow velocity

of 22 ms−1_{(left), 33 ms}−1_{(middle), and 44 ms}−1_{(right). . . . 284} D.54 Snapshots of the mean swirling velocity component for a bulk inflow

veloc-ity of 22 ms−1_{(left), 33 ms}−1_{(middle), and 44 ms}−1_{(right). . . . 285} D.55 Radial profiles of the mean temperature, upstream. . . 286 D.56 Radial profiles of the mean temperature, downstream. . . 286 D.57 Snapshots of the mean temperature for a bulk inflow velocity of 22 ms−1

(left), 33 ms−1_{(middle), and 44 ms}−1_{(right). . . . 287} D.58 Radial profiles of the mean O₂ (left) and CO₂ (right) mass fractions,

up-stream. . . 287 D.59 Radial profiles of mean axial velocities (left column) and corresponding

ve-locity RMS (right column). Here EBJ refers to the faulty implementation

of the Bini & Jones and BJ to the corrected version. . . 288 D.60 Radial profiles of the mean temperature. Here EBJ refers to the faulty

2.1 Coal constituents determined by the PA. . . 10 2.2 Coal constituents determined by the UA. . . 10 2.3 ASTM coal rank classification, ASTM Standard D388-15. . . 11 5.1 Individual transport equation source terms. . . 50 5.2 Coal particle sub-model terms . . . 57 6.1 Coal components mass basis nomenclature . . . 63 6.2 Base parameters for the Richards & Fletcher pyrolysis model parameter

study. . . 70 6.3 σ_i coefficients for the Richards & Fletcher pyrolysis model parameter

study. . . 70 6.4 Available temperature profiles. . . 73 6.5 PKP main classes. . . 74 6.6 Examples of various 2D-clusters of size n and different coordination

num-bersσ+ 1. Sites are denoted by o. . . 75 6.7 CPD coefficients. . . 79 6.8 Elemental composition of the postulate substances CaHbOcNd1 and

CaHbOcNd2. . . 90 6.9 Q-factor results for different temperature profiles. . . 93 7.1 Experimental conditions, Hwang et al. (2005). . . 102 7.2 Properties of Newlands Bituminous Coal, UA left, PA right,α_{as received,}

β_{dry basis, Hwang et al. (2005). . . . 102}

7.3 Summary of model parameters for the base case, BC, and the parameter

study reference case, RC, setup. . . 104 7.4 Boundary conditions, where NE indicates Neumann and DI Dirichlet

boundaries. . . 106 7.5 Parameters of the reference case for the pyrolysis model study. . . 112 7.6 Elemental composition of the postulate substance C_aH_bO_cN_d. . . 112 7.7 Summary of model parameters for the CRIEPI case. . . 114

LIST OF TABLES xxvi

7.8 Optimised coefficients for the SKR based pyrolysis model . . . 116 7.9 Richards & Fletcher pyrolysis model parameters . . . 116 7.10 Lower heating values of the different postulate substances. . . 116 8.1 Experimental conditions, Weber et al. (1992) . . . 127 8.2 Properties of Göttelborn hvBb Coal, UA (daf) left, PA (as received) right

Weber et al. (1992) . . . 127 8.3 Summary of model parameters for the BC and RC cases. . . 128 8.4 IFRF parameter studies. . . 129 8.5 Number of grid cells and various simulation properties for different mesh

levels for a domain length of 4m. . . 130 8.6 Boundary conditions, where N indicates Neumann and D Dirichlet

bound-aries . . . 131 8.7 Boundary conditions of species mass fractions, where N indicates

Neu-mann and D Dirichlet boundaries . . . 131 8.8 Summary of model parameters for the IFRF case . . . 134 8.9 Elemental compositions and lower heating values of the postulate

sub-stances CaHbOcNd, CaHbOcNd1, and CaHbOcNd2. . . 135 8.10 Computational setup of the inflow generation DES . . . 137 8.11 Optimised coefficients for the SKR based pyrolysis models . . . 141 8.12 Richards & Fletcher pyrolysis model parameters for Eqs. (6.15) to (6.17) . 141 8.13 Elemental composition of the gas-phase at the furnace outlet . . . 155 A.1 Chemical nomenclature of coal constituents and moieties. . . 186 A.2 Common chemical reactions during coalification and pyrolysis. . . 186 B.1 Variable Names in the OpenFOAM implementation of P₁. . . 211 C.1 Computational procedure, where ∗ indicate intermediate values critical to

mass conservativeness based on uncorrected fluxes, n-previous time step values, n +1-current time step values, and i +1 current PIMPLE iteration

values. At the final iteration i +1 values are simultaneously n +1 values. 217 D.1 Base parameters for the SKR parameter study . . . 229 D.2 Pre-exponential constants. . . 230 D.3 Activation temperatures. . . 230 D.4 Base parameters for the Kobayashi parameter study . . . 234 D.5 Activation temperatures for the yield ratio study . . . 235 D.6 Base parameters for the Kobayashi Q-factor parameter study . . . 236 D.7 Parameter studies conducted for the CRIEPI case . . . 240 D.8 Coal particle parameter . . . 240

D.9 Number of grid cells per level . . . 241 D.10 Summary of model parameters for the Merrick case . . . 254 D.11 Coefficients of the char conversion model study . . . 266 D.12 Parameter studies conducted for the IFRF case. . . 269 D.13 Simulation properties for different mesh levels for a domain length of 4m 271 E.1 Archived files for the isothermal CRIEPI parameter studies in the

ITV-DATABASE/ITV/INTERNAL/2018_Olenik_PHD/CRIEPI/ISOdirectory. . 291 E.2 Archived files for the reacting CRIEPI parameter studies in the

ITV-DATABASE/ITV/INTERNAL/2018_Olenik_PHD/CRIEPI/REACTING

directory. . . 291 E.3 Archived files for the IFRF parameter studies in the

ITV-DATABASE/ITV/INTERNAL/2018_Olenik_PHD/IFRFdirectory. . . 292 E.4 Code repositories on thegit.itv.uni-stuttgart.deserver. . . 292

The present work examines the process of pulverized coal combustion by means of numeri-cal simulation. For this purpose the large eddy simulation (LES) technique with an

Euler-Lagrange approach is used. Scope of this work is the modelling of the pyrolysis process during the combustion of coal.

First, the fundamentals of coal formation and its combustion are discussed. Then, the fluid mechanical fundamentals, the necessary numerical methods, and their implementa-tion are presented. Furthermore, an overview of phenomenological pyrolysis models, pyrol-ysis models for the use as coal preprocessors, and pyrolpyrol-ysis models used within the context of numerical simulation using the Euler-Lagrange approach are given. This includes the de-scription of the novel pyrolysis model of Richards and Fletcher (2016), its classification with respect to existing pyrolysis models, implementation into the employed numerical frame-work, and its validation. Here, validation is performed with the help of a coal preprocessor, cross validation with other existing pyrolysis models and through the use within Euler-Lagrange LES. Furthermore, an extension to the model of Richards and Fletcher (2016) is proposed in order to reflect the composition and heating values of the volatile gases, without the need for prior assumptions concerning the temperature profile during pyrolysis.

The behaviour of the Richards & Fletcher model was validated against PKP coal pre-processor data, cf. Vascellari et al. (2013), and standard pyrolysis models such as the single kinetic rate model of Badzioch and Hawksley (1970) and Kobayashi et al. (1977). Here, the model proposed by Richards and Fletcher (2016) performed superior compared to the model of Badzioch and Hawksley (1970) since the former was able to capture the trend of the temperature-dependent final volatile yield. Additionally, compared to the model of Kobayashi et al. (1977) an improvement in the prediction of the volatile yield profiles in the late pyrolysis stages at high temperature is noticeable. The Richards & Fletcher model is able to reflect the reaction-inhibiting trend of an increasing ratio of strong carbon-carbon double bonds remaining inside the coal molecule. The improved predictive performance of the Richards & Fletcher model comes at the price of additional unknown coefficients, for which a computation procedure was proposed by means of a coal preprocessor. This pro-cedure was tested using the PKP preprocessor, but is in principle extendable to other coal preprocessor.

xxx

Further validation work includes LES comparing different pyrolysis models. The re-sults of the simulations of the laboratory scale burner, experimentally investigated by Hwang et al. (2005, 2006a,b), and the semi-industrial scale furnace, experimentally inves-tigated by Weber et al. (1992); Peters and Weber (1997), demonstrate the principle appli-cability of Richards & Fletcher model for the usage as LES pyrolysis model. In both cases, the simulation results based on Richards & Fletcher pyrolysis model are comparable to the results obtained by the model of Badzioch & Hawksley. Additionally, the results of the simulations show a good agreement with the experimental data. However, one important advantage of the Richards & Fletcher model compared to the SKR based model is that the former is independent of an a-priori estimation of a fixed final yield of the volatile species. Furthermore, it is demonstrated that the Richards & Fletcher model can be extended to incorporate the changes of the volatile yield composition and heating values caused by dif-ferent pyrolysis temperature profiles into the postulate substance approach. This is be achieved by employing two separate postulate substances, for which their compositions and heating values are calculated independently. A method to calculate the compositions and heating values of both postulate substances is presented.

Die vorliegende Arbeit untersucht den Prozess der Kohlestaubverbrennung unter Verwen-dung der numerischen Simulation. Zu diesem Zweck wird die "Large Eddy Simulation" (LES) Technik mit einem Euler-Lagrange Ansatz eingesetzt. Der thematische Schwer-punkt der Arbeit liegt insbesondere auf der Modellierung des während der Verbrennung von Kohlestaub einhergehenden Pyrolyseprozesses.

Zunächst werden hierfür die Grundlagen der Entstehung von Kohle sowie der Verbren-nung dieser diskutiert. Danach werden die strömungsmechanischen Grundlagen und die notwendigen numerischen Methoden sowie ihre Implementierung erörtert. Hinzu kommt die Diskussion von einer Reihe von phänomenologischen Modellen zur Beschreibung der Pyrolyse, von Pyrolysemodellen für den Einsatz als Präprozessor als auch Modelle zur Verwendung im Kontext von Euler-Lagrange Simulationen. Dies beinhaltet insbeson-dere die Beschreibung des neuartigen Pyrolysemodells von Richards and Fletcher (2016), der Einordnung in Bezug auf existierende Pyrolysemodelle, der Implementierung innhalb eines bestehenden Simulationsframeworks und der Validierung. Die Validierung er-folgt hierbei mittels Präprozessordaten, durch einen Vergleich mit bestehenden Modellen sowie durch den Einsatz innerhalb von Euler-Lagrange LES. Es wird zusätzlich eine Er-weiterung des Modells von Richards and Fletcher (2016) vorgeschlagen, um vor allem die Zusammensetzung und den Heizwert der freigesetzten volatilen Gase zu reflektieren, ohne vorherige Annahmen über den Temperaturverlauf während der Pyrolyse treffen zu müssen. Zur Validierung der Eigenschaften des Pyrolyse Modells von Richards & Fletcher wird der PKP-Kohlepräprozessors, vgl. Vascellari et al. (2013), sowie verschiedene Standard Pyrolyse Modelle, wie zum Beispiel von Badzioch and Hawksley (1970) und Kobayashi et al. (1977), eingesetzt. Hierbei ist ein vorteilhaftes Verhalten des Modells von Richards and Fletcher (2016) im Vergleich zu dem Modell von Badzioch and Hawksley (1970) zu beobachten. Das Modell von Richards and Fletcher (2016) bietet die Möglichkeit die Temperaturabhängigkeit der Gesamtmenge an freigesetzten volatilen Gasen darzustellen. Weiterhin bietet das Modell von Richards & Fletcher, im Vergleich zu dem Modell von Kobayashi et al. (1977), die Möglichkeit der verbesserten Vorhersage der Freisetzung volatiler Gase in den späteren Phasen des Freisetzungsprozesses bei hohen Temperaturen. Hierbei kommt dem Modell von Richards & Fletcher vorallem die Möglichkeit der

xxxii

dung der zunehmenden Dominanz von Kohlenstoff-Kohlenstoff-Doppelbindungen zugute, die in der devolatilisierenden Kohle verbleiben. Der verbesserten Vorhersagekraft des Mo-dells von Richards & Fletcher steht allerdings eine erhöhte Anzahl unbekannter Modellko-effizienten gegenüber. In der vorliegenden Arbeit wird daher ein Verfahren zur Bestim-mung der unbekannten Modellkoeffizienten unter Zuhilfenahme eines Kohlepräprozessors vorgeschlagen. Dieses Verfahren wird anhand des PKP Präprozessors getestet, ist aber auf beliebige Präprozessoren übertragbar.

Als weiteren Validierungsschritt werden LES unter Verwendung verschiedener Pyro-lyse Modelle durchgeführt. Die Ergebnisse der Simulationen von experimentellen Unter-suchungen eines Brenners im Labormaßstab, untersucht von Hwang et al. (2005, 2006a,b), und einer Brennkammer im semi-industriellen Maßstab, untersucht von Weber et al. (1992); Peters and Weber (1997), zeigen eine grundsätzliche Eignung des Modells von Richards & Fletcher für den Einsatz als LES Pyrolyse Modell. In beiden Fällen liefern die mit dem Modell von Richards & Fletcher durchgeführten Simulationen Ergebnisse, die vergleichbar sind mit denen, die unter Verwendung des Modells von Badzioch & Hawk-sley erhalten werden. Zudem kann unabhängig von der Pyrolysemodellierung generell eine gute Übereinstimmung der Simulationsergebnisse mit den experimentellen Befun-den beobachtet werBefun-den. Ein gewichtiger Vorteil der hier präsentierten Methode gegenüber dem Modell von Badzioch & Hawksley für den Einsatz in strömungsmechanischen Simu-lationen ist allerdings die Unabhängigkeit des Modells von einer a-priori Abschätzung der Gesamtmenge an freigesetzten flüchtigen Pyrolysegasen. In der vorliegenden Arbeit wird weiterhin gezeigt, dass das Modell von Richards & Fletcher unter Beibehaltung der Ver-wendung von postulierter volatiler Spezies erweitert werden kann, so dass Änderungen der Zusammensetzung der volatilen Gase und ihres Heizwertes bedingt durch einen variablen Temperaturverlauf während des Pyrolyseprozesses berücksichtigt werden können. Hierfür wird die Verwendung von zwei separaten postulierten Substanzen vorgeschlagen und eine Methode zur Bestimmung ihrer Zusammensetzung und des Heizwertes vorgestellt.

Roman S ymbols

a_P Central coefficients discretised transport equation ms−2

A Area m2

A Pre-exponential factor s−1

cp Specific heat capacity Jkg−1K−1

c Number of char bridges −

E Radiative emission Wm−2

E Activation Energy Jmol−1

f Mass specific force ms−1

F Cell face mass flux kgs−2

g Gravitational acceleration ms−2

G Incident radiation W

h Specific enthalpy Jkg−1

H Off-central coefficients discretised transport equation ms−2

I Intensity of radiation Wsr−1

k_c Cut-off wave number m−1

k Rate coefficient of reaction s−1

k Specific kinetic energy m2s−2

l Generic length m

L Labile bond −

L Characteristic width of the flow or macroscale m

MW Molar mass kgkmol−1

m Mass kg

n Generic number −

n Rosin Rammler spread parameter −

p Number of labile bridges −

r Generic ratio −

r Flux gradient −

p Pressure Nm−2

xxxiv

q Q-factor −

Ru Universal gas constant Jkmol−1K−1

Si j Strain rate tensor s−1

S Generic discretised source term s−1

t Time s

T Temperature K

ui Velocity vector (ui,uj,uk) or (u, v, w) ms−1

V or∆V Volume m3

V Diffusion velocity ms−1

W Wiener term −

xi Cartesian coordinate vector (xi,xj,xk) m

xi Molar concentration molm−3

y Wall distance m

Y Mass fraction −

Dimensionless numbers

CFL Courant Friedrichs Lewy number

Da Damköhler number Ka Karlovitz number Le Lewis number Ma Mach number Nu Nusselt number Sc Schmidt number Pe Peclet number Pr Prandtl number Re Reynolds number Greek S ymbols α Generic coefficient α Species

α Heat diffusion coefficient kgm−1s−1

β Generic coefficient

β Extinction coefficient −

Γ Diffusion coefficient

δ Dirac delta function ₋

δ Kronecker delta ₋

∆ Filter width m

η Emission intensity Ws2m−1rad−1

η Mass stoichiometric coefficient kg

² Emissivity coefficient −

σ Scattering coefficient −

σ_{+ 1} Coordination number ₋

σ Stefan-Boltzmann constant W/2m/K4

φ Scattering phase function −

ν Radiation wavelength m λ Thermal conductivity Wm−1K−1 κ Absorption coefficient ₋ κ Compressibility coefficient s2m−2 λ Interpolation factor − µ Dynamic viscosity kgm−1s−1 ν Kinematic viscosity m2s−1

ν Stoichiometric factor mol

ρ Density kgm−3

τ Viscous stress tensor Nm−2

τ Timescale s

τ Number of broken bridges ₋

Θ Heat source term Jm−3s−1

Θ Inverse temperature K−1 φ Mechanism factor − φ Generic field ₋ ψ Generic field ₋ ψ Surface factor − Ψ Limiter function −

² Cut off dissipation rate Js−1

² Error ₋

ω Correction factor −

ω Generic specific source term s−1m−3

Ω Mass source term kgm−3_s−1

Ω Solid angle −

Λ Momentum source term kgm−2s−2

Ψ Species mass fraction source term kgm−3_s−1

σ Coefficient −

xxxvi

Superscripts

Average

s _{Spatial average}

Filter operator

e Favre (density-weighed) filtered 0 _{Unconditional fluctuation} 00 _{Conditional fluctuation} 0 _{Initial value} ∗ _{Labile bond} ∗ _{Estimated value} 0 Corrector value

+ _{Non-dimensionalised wall distance value}

Subscripts Ac Activation AR As Received BB Black Body Br Bridge CC Char Conversion Cr Cross

DAF Dry Ash Free

Dev Deviatoric Dev Devolatilisation Disp Dispersion Dr Drag Dry Dry Elem Elemental

E, e Eastern cell, face

Ev Evaporation

Exp Explicit

EYC Excess Yield Carbon

f Face F Formation Fwd Forward FC Fixed Carbon Fu Fuel FM Fixed Matter G Gas

G Gravity Hom Homogeneous i Cartesian direction Imp Implicit Inp Input Int Intermediate j Cartesian direction k Species index La Latent Moi Moisture

Mol Mole, molar

Meta Meta N Neighbour OF OpenFOAM Ox Oxidiser P Central cell Part Particle Parc Parcel

PCC Pulverised coal combustion

PP Pre-processor Pr Product Py Pyrolysis Rad Radiation Re Released Rev Reverse Reac Reaction Ref Reference Rel Relative Si Site Smag Smagorinsky St Stoich Su Surface Therm Thermodynamic Turb Turbulent Tar Tar Vol Volatiles S Sensible Term Terminal Th Thermolysis

xxxviii

W, w Western cell, face

Shorthands

AASB Aerodynamically Air Staged Burner

AE Algebraic Equation

ASTM American Society for Testing and Materials

BC Base Case

BTU British Thermal Unit

BGR Bundesanstalt für Geowissenschaften und Rohstoffe CPD Chemical Percolation Model for Devolatilization CTE Crossing Trajectory Effect

CFD Computational Fluid Dynamics

CG Conjugate Gradients

CR Constant Rate Pyrolysis Model

CRIEPI Central Research Institute of Electric Power Industry CBC Coal and Biomass Conversion

CDF Cumulative Distribution Function CDS Central Differencing Scheme CFL Courant–Friedrichs–Lewy DNS Direct Numerical Simulation DISCHAIN Distributed-energy Chain

DVC Depolymerization Vaporization Crosslinking DOM Discrete Ordinates Methods

DES Detached Eddy Simulation

EBU Eddy Break-Up

EUL Eulerian

ERZ External Recirculation Zone FTIR Fourier Transform Infrared

FTT Flow Through Times

FTAC Flow Turbulence and Combustion

FG Functional Group

FDM Finite Difference Method FVM Finite Volume Method FEM Finite Element Method GPL GNU General Public License

GS Grid Scale

HHV Higher Heating Value hvbB High Volatile Bituminous B IRZ Internal Recirculation Zone

xl

IFRF International Flame Research Foundation IMC Implicit Monte Carlo

KOBA Competing Rates Model of Kobayashi et al. (1977)

LAG Lagrangian

LDV Laser Doppler Velocimetry LES Large Eddy Simulation LIF Laser-Induced Fluorescence MILES Monotonically Integrated LES

MICRO Multicolor Integrated Receiving Optics NVD Normalised Variation Deminishing NMR Nuclear Magnetic Resonance

OH-PLIF OH radical Planar Laser Induced Fluorescence PISO Pressure-Implicit with Splitting of Operators PCC Pulverised Coal Combustion

PKP Pyrolysis Kinetic Preprocessor PDE Partial Differential Equation

PA Proximate Analysis

QUICK Quadratic Upwind

RC Reference Case

RF Richards & Fletcher

RANS Reynolds Averaged Navier Stokes RTE Radiative Transfer Equation SDPA Shadow Doppler Particle Analyser

SIMPLE Semi-Implicit Method for Pressure Linked Equations

SGS Sub-Grid Scale

SKR Single Kinetic Rate

SDPA Shadow Doppler Particle Analyser SPH Smoothed Particle Hydrodynamics TVD Total Variation Diminishing TCI Turbulence Chemistry Interaction TGA ThermoGravimetric Analysis

UA Ultimate Analysis

VLES Very Large Eddy Simulation

URANS Unsteady RANS

I

NTRODUCTION

1.1 Motivation

In 2017, approximately 81% of the global demand for primary energy was covered by fossil fuels. With a share of 27.1%, the combustion of coal plays an important role. Electrical power generation accounted for approximately 60% of the global coal demand. For which, with approximately 38.5%, coal played a significant role, International Energy Agency (2017). According to the “new policies scenario”, out-lined by the International Energy Agency (IEA), a stable consumption of coal and a slightly lower market share of approximately 22% of the projected 2040 global primary energy mix is expected. Even in the case of the “sustainable development scenario” the IEA expects a market share of coal of approximately 13% with a stable total fuel consumption.

Coal resources are abundantly distributed over the whole planet, cf. Andruleit et al. (2016) for a detailed listing of global coal resources. This makes coal usage, in contrast to oil, gas, and nuclear fuels, relatively independent from geopolitical conflicts. Additionally, coal is easy to transport, store, and use. These attributes in combination with the relatively low costs, makes coal a popular source of primary energy.

Concerns about the environmental impact of coal combustion and its associated CO₂emissions necessitate the deployment of “clean coal technologies” such as flue gas desulphurisation, low NO_x burners, and oxy-coal combustion for carbon cap-ture and storage (CCS). The applicability and success of the aforementioned clean coal technologies rely on a detailed understanding of the underlying physical and chemical processes. Additionally, a detailed understanding can be a key factor for a responsible coal utilisation due to improved boiler efficiency.

1.2. STATE OF THE ART 2

Advanced laser diagnostics can deliver detailed insights for gas-phase flows and combustion, Aldén et al. (2011); Kutne et al. (2011); Böhm et al. (2011); Masri et al. (2004). However, the often limited optical access to industrial coal boilers and in-terference of soot complicate reliable laser measurements. Hence, state-of-the-art pulverised coal combustion (PCC) experiments, e.g. Coraggio et al. (2011); Guo et al. (2017) are still mostly based on intrusive experimental techniques, which may interfere with the combustion process.

Progress made in the simulation techniques for combustion in general and ever increasing availability of computational resources open the opportunity of studying coal combustion processes by numerical simulations. This enables one to study combustion phenomena without the constraints of accessibility due to geometrical limits or thermodynamical reasons. Simulation techniques can be used to evaluate critical design decisions even before a prototype of the burner or furnace of interest has been built. In order to increase the reliability and accuracy of simulations, a considerable effort to develop, validate, and test necessary sub-models has to be invested. The present work aims to contribute towards this goal.

1.2 State of the Art

Numerical simulations of combustion processes are generally constrained by their high computational costs. Thus, a number of modelling approaches such as the Reynolds averaged Navier-Stokes (RANS) or the large eddy simulation (LES) have been developed in order to reduce the computational costs. Especially for complex geometries and swirled turbulent flows, LES is a promising technique. Pioneer-ing work utilisPioneer-ing LES for PCC was conducted by Kurose and Makino (2003), who demonstrated the general applicability of LES to coal combustion in furnaces with complex geometries. In the study of Kurose et al. (2009) the authors demonstrated the superiority of LES flow field results in comparison to RANS results for a large laboratory-scale PCC furnace. The superiority of LES results over RANS results was also confirmed by Stein et al. (2011). Moreover, Yamamoto et al. (2011) inves-tigated an Arrhenius rate based, single kinetic rate devolatilisation model with a correction factor for the pre-exponential factor determined by the devolatilisation progress for the LES of a pulverised coal jet flame. A study by Chen and Ghoniem (2012) compared the influence of RANS and LES turbulence models on the simula-tion of a 100 kW oxy-coal furnace. The authors found an improved predicsimula-tion of the species concentration as result of an enhanced mixing for the LES results. Using

a direct quadrature method of moments (DQMOM), Pedel et al. (2013) studied the applicability of LES for the simulation of a PCC jet under oxy-coal conditions. Based on the PCC jet flame of Hwang et al. (2005, 2006a,b), Stein et al. (2013) carried out a comprehensive study of common PCC sub-models. The LES of Franchetti et al. (2013) of the same experimental case, discussed a simple particle radiation model to allow for a comparison of the experimental temperature measurements using La-grangian particle data. A study by Rabacal et al. (2015) presented a detailed LES of a large-scale laboratory furnace. Muto et al. (2015) used LES to investigate the formation of NOx under oxy-coal conditions. Rieth et al. (2016) introduced the ap-plication of the flamelet model in the context of PCC-LES showing an overall good agreement with experimental results. The approach of using flamelets as a way to model turbulence chemistry interaction in the context of PCC-LES was also in-troduced by Wen et al. (2016). Knappstein et al. (2018) successfully demonstrated the application of a flamelet generated manifold (FGM) to model the simultaneous process of homogeneous reactions of the pyrolysis and char conversion products for the simulation of a 20 kW burner.

The aforementioned studies generally relied on an a-priori approximation of the pyrolysis rates and the final volatile yield, making preliminary simulations neces-sary in order to predict typical heating rates of coal particles, which served as input parameter for coal preprocessors. As a result, the studies of Rieth et al. (2017a) and Wan et al. (2017) have successfully demonstrated the direct coupling of the chemical percolation model for devolatilization (CPD) preprocessor to LES.

The pyrolysis process is of great importance for the overall combustion process, since it affects the ignition, flame temperature, and pollutant formation. Hence, the pyrolysis model plays a central role in PCC-LES and affects the accuracy of the simulation. Moreover, the pyrolysis model might determine simulation properties such as computational costs, available homogeneous reaction mechanism, turbu-lence chemistry interaction models, and pollutant formation models. Therefore, the effect of pyrolysis model uncertainties and the resulting uncertainties in the volatile yield composition on the results of LES need to be addressed.

However, to the author’s knowledge, no approach exists to reflect the effects of different heating rates and pyrolysis temperatures on the volatile composition. Ex-isting methods are either computationally too expensive, not suitable to incorporate into an LES framework, or simply assume a constant, a-priori determined compo-sition during pyrolysis and subsequent combustion. Pyrolysis models based on a competing rates approach, in combination with a single a-priori determined final

1.3. PROJECT FOCUS 4

volatile composition, are prone to inconsistencies in the species mass balances, if the final volatile yield of either the single coal particles or the statistical average of the released volatile mass deviates from the a-priori approximated volatile mass with which the volatile composition was derived. Furthermore, to this date no ex-tensive investigation of the effect of the maximum released volatile yield has been conducted.

1.3 Project Focus

The present work attends to the numerical simulation of pulverised coal combustion by means of LES. Here, especially the aforementioned uncertainties of the existing pyrolysis models for LES regarding changes in the volatile yield quantity and com-position at different pyrolysis temperatures and heating rates shall be addressed.

Furthermore, a purpose of this study is to develop an approach to model the volatile yield, its composition, and heating values of the volatile species indepen-dent of a-priori knowledge of the actual heating rates which occur during the sim-ulation. Ideally, necessary enhancements of existing approaches shall be realised without a notable increase of the computational costs.

This study investigates the integration of the novel pyrolysis model of Richards and Fletcher (2016) into a comprehensive PCC-LES model and its impact on laboratory- and semi-industrial scale burner simulations. The performance of the implemented model is tested against two established experimental test cases. These test cases are the laboratory-scale burner of Hwang et al. (2005, 2006a,b) and a semi industrial scale furnace investigated by Weber et al. (1992); Peters and Weber (1997). In addition, an approach to reflect changes in the composition of the volatile species caused by different pyrolysis temperature profiles is discussed. The effect of the single, a-priori determined, volatile species approach and the corre-sponding uncertainties in the global species mass balance on the simulation results are investigated.

1.4 Outline

The present work is subdivided into three parts. Firstly, the theoretical part dis-cusses the fundamentals of coal formation, coal structure, and the relevant pro-cesses during combustion, cf. chapter 2. Subsequently, chapter 3 discusses turbu-lent two phase flows while chapter 4 demonstrates basic numerical solution

strate-gies for the governing reacting equations. The second part, dedicated to modelling, focuses on a detailed discussion of the modelling strategies. In chapter 5, the rele-vant gas- and dispersed-phase sub-models are presented. The subsequent chapter centers on the pyrolysis modelling, and presented here are the most common py-rolysis models for LES, cf. chapter 6. Furthermore, the Richards & Fletcher model is discussed and an extension is proposed. The third part, consisting of chapters 7 and 8, presents and discusses results of numerical simulations of two test cases. In the ensuing appendix, a large amount of supplementary material is available. This includes further theoretical background, cf. appendix A, implementation details, appendices B and C, and further parameter studies for both test cases in appendix D.

Theoretical Background

C

OAL

C

OMBUSTION

F

UNDAMENTALS

2.1 Coal Formation and Characteristics

This section gives a brief overview of the formation of coal and the resulting struc-tural properties. Due to the enormous variety in coal and the corresponding re-search the presented discussion cannot be comprehensive by any means and the interested reader is referred to existing literature, e.g. the review of O’Keefe et al. (2013) which gives a good introduction to the coal formation process and identifies various parameters influencing the composition and structure of coal. Given the multitude of factors influencing the coal formation process, O’Keefe et al. (2013), it is no surprise that Smoot and Smith (2013) state that “of 1200 categorized coals by the Bituminous Coal Research Institute no two had exactly the same composition”. To overcome the problem of finding a suitable definition for coal posed by the large variety of properties from now on coal is broadly referred to as

" [. . . ] an inhomogenous organic fuel, formed largely from partially decomposed and metamorphosed plant materials.", cf. Smoot and Smith (2013)

in the present work.

Coal is comprised of an inhomogenous mixture of organic materials known as macerals (vitrinite, fusinite, exinite etc.). In most cases the dominant maceral is vitrinite, cf. Saxena (1990), which is mainly composed of organic polymers such as cellulose (C₆H₁₀O₅)_nand lignin (C₃₁H₃₄O₁₁)_nstemming from decomposed plant cell walls. Additionally to organic matter, coal consists of inorganic components mainly clays, silica, sulfides and carbonates, Gavalas (1982). Even though inorganic con-stituents are not in the main focus of this work a short discussion is given here

2.1. COAL FORMATION AND CHARACTERISTICS 10

since inorganic constituents might show catalytic activity during pyrolysis, Tsub-ouchi and Ohtsuka (2008). Hence inorganic constituents can influence yield compo-sition and weight loss of the coal particle. Schobert (2013) names three origins of inorganic material in coal:

1. Concentration of inorganic material by plants;

2. Transport of minerals into the swamp from which the coal originates; 3. Water percolating into the coal seam.

Inorganic constituents of coal play an important role in, slagging and fouling in coal furnaces, Reid (1984), pollutant formation as discussed by Tsubouchi and Ohtsuka (2008), and may have possible side effects during pyrolysis, Williams et al. (2000).

The transitional process (diagenesis and metamorphosis) from peat to coal is called coalification, during which the plant material undergoes a

"[. . . ] physical and chemical transformation from peat through lig-nite and subbituminous coal, to bituminous coal, and through bitumi-nous coal to anthracite and meta-anthracite, and approaching graphite." O’Keefe et al. (2013)

The different stages in the coalification process are commonly referred to as coal rank. The rank of the coal is dependent on time, temperature, and pressure dur-ing coalification. It is, however, independent of the original organic and inorganic peat constituents, O’Keefe et al. (2013). Two standard techniques for the classifi-cation of coal, namely the proximate analysis (PA), ASTM Standard D7582-15 and the ultimate analysis (UA), ASTM Standard D3175-15 are defined by the American Society for Testing and Materials (ASTM). The PA is used to determine the compo-sition of the major constituents, whereas the UA determines the major elemental components. The constituents of coal according to the PA are shown in Tab. 2.1 and for the UA in Tab. 2.2

Table 2.1: Coal constituents de-termined by the PA.

Moisture Ash Fixed Carbon Volatile Matter

Table 2.2: Coal constituents de-termined by the UA.

Carbon Hydrogen

Nitrogen Oxygen Sulphur

Table 2.3: ASTM coal rank classification, ASTM Standard D388-15. Class Group Fixed Carbona[-] or HHVb [BTU/lb]c 1. Anthracite 1. Meta-anthracite FC > 98%

2. Normal anthracite 98% > FC > 92% 3. Semi-anthracite 92% > FC > 86% 2. Bituminous 1. Low volatile 86% > FC > 77%d

2. Medium volatile 77% > FC > 69% 3. High volatile A 69% > FC, HHV > 14000 (32.65) 4. High volatile B 14000 (32.65) > HHV > 13000 (30.24) 5. High volatile C 13000 (30.24) > HHV > 11000 (26.75)e 3. Sub-bituminous 1. Sub-bituminous A 13000 (30.24) > HHV > 11000 (26.75)f 2. Sub-bituminous B 11000 (26.75) > HHV > 9500 (22.10) 3. Sub-bituminous C 9500 (22.10) > HHV > 8300 (19.31) 4. Lignite 1. Lignite A 8300 (19.31) > HHV > 6300 (14.65) 2. Lignite B 6300 (14.65) > HHV a_Dry b_Moist

c_{Value in parentheses are given in 1 MJkg}−1 d_{Non-agglutinating}

e_{Either agglutinating or non-weathering} f_{Both weathering and non-agglutinating}

Table 2.3 displays the classification of coal into different coal ranks using British thermal unit (BTU) and fixed carbon contents determined by the PA as proposed by the ASTM. Besides the ASTM coal classification system several others exist e.g. the UNECE, cf. United Nations (1989), or the classification by the German Bun-desanstalt für Geowissenschaften und Rohstoffe (BGR), cf. Cramer et al. (2009). Generally, with increasing coal rank the carbon content increases combined with decreasing moisture and volatile yield. Decrease in oxygen content is particularly pronounced at sub-bituminous C and B rank, Stach (1982), whereas hydrogen is rel-atively unchanged up to low volatile bituminous and anthracite coal, Str ˛apo´c et al. (2011). The calorific value increases with the coal rank up to rank of bituminous high volatile A coal and decreases slightly for higher ranks. Yu et al. (2007) state that with increasing rank the heterogeneity among different coal types diminishes and thus might be less influential during combustion. In summary, O’Keefe et al. (2013) identify three main parameters that determine the properties of coal due to origin and formation process as:

• Organic petrological/geochemical; • Inorganic petrological/geochemical;