NUMERICAL AND EXPERIMENTAL

STUDY ON CONCRETE SUBJECTED TO

ELEVATED TEMPERATURE

WU Zhuoya

Ph.D. Thesis

THE UNIVERSITY OF HONG KONG

February 2021
 Abstract of thesis entitled

NUMERICAL AND EXPERIMENTAL

STUDY ON CONCRETE SUBJECTED TO

ELEVATED TEMPERATURE

Submitted by

WU Zhuoya

for the degree of Doctor of Philosophy

at The University of Hong Kong

in February 2021

Fire-induced spalling has caused primary concern in the application of concrete,

especially the high-strength and high-performance concrete with more compact
microstructure, subjected to elevated temperature. Though decades of experimental and
numerical research have been concentrated on the performance of concrete at high
temperature, fire-induced spalling remains a challenging topic in the design of building
structures. Reliable and practical model is therefore highly desirable and critical to the
assessment and prediction of spalling.

The first contribution of the thesis involves the development of a coupled thermo-
hygro-mechanical (THM) model to study the behaviour of concrete subjected to
elevated temperature. Macroscopic governing equations of the numerical model are
established based on mass, enthalpy and momentum conservations, supplemented with
relevant constitutive laws and state equations. A two-step spalling criterion is
developed combining the assessment of mechanical damage and strain energy density.
Validated against experiments conducted by various researchers, the model is
demonstrated to show good correlations with the experimental data of temperature and
pore pressure profiles. An algorithm is proposed to account for the stochastic nature of
concrete behaviour with steps of implementation to incorporate the algorithm into the
model.

The second part of the thesis investigates the early residual splitting tensile strength
(ERSTS) of concrete at elevated temperature, which could serve as an important
parameter in determining the occurrence of fire-induced spalling. Experimental studies
are designed and performed to obtain data of ERSTS while the concrete specimens are
still at high temperature, and to compare with the residual splitting tensile strength
(RSTS) of concrete measured when specimens have cooled down and resumed to
ambient temperature. Differences have been shown to exist between the ERSTS and
RSTS of concrete. Numerical analyses are also conducted to simulate the thermo-
mechanical process and reproduce the experimental results.

The third and final part of the thesis proposes an innovative protective measure to
prevent fire-induced spalling. A bilayer column system, consisting of a high strength
concrete (HSC) inner core and a normal strength concrete (NSC) outer layer, is
introduced to examine the effect of thermal barrier on columns. The behaviour of
bilayer column is evaluated by compression and fire tests conducted at ambient and
elevated temperatures respectively. Preliminary experimental results show that the
bilayer columns properly cast and reinforced are generally effective in preventing
spalling. However, further study on thickness of the outer layer and integrity of the
whole column is needed to improve the mechanical and thermal performance of bilayer
columns.

(403 words)
 DECLARATION

I declare that the thesis thereof represents my own work, except where due
acknowledgement is made, and that it has not been previously included in a thesis,
dissertation or report submitted to this University or to any other institution for a degree,
diploma or other qualifications.

Signature:

WU Zhuoya

i
 ACKNOWLEDGEMENT

First and foremost, I would like to express my earnest gratitude and appreciation to my
supervisors, Prof. S.H. Lo and Dr. K.L. Su, for their guidance and supervision
throughout the period of my candidature. Special thanks are credited to my primary
supervisor Prof. Lo, who has unreservedly assisted me in writing the finite element
computer program. The high standards that Prof. Lo and Dr. Su espouse, together with
their stimulating suggestions, have continuously contributed to my abilities as an
independent researcher. I feel much privileged to study under their mentorship.

I’m genuinely grateful to Prof. K.H. Tan for his generous support in my studies at
Nanyang Technological University. Prof. Tan showed sincere care and made time to
review my research despite his tight schedule. Many thanks are also owed to Prof. Tan’s
research group, particularly Dr. S.X. Fan, Dr. P.W. Du and Dr. Y.H. Ng, for their kind
help in the experiments and insightful advice.

I would like to thank the University of Hong Kong and the Department of Civil
Engineering for providing scholarship and financial assistance during the study.
Appreciation also goes to the administrative staff in the department and technicians in
the structural laboratory for their valued help whenever necessary. Many thanks to my
friends and colleagues at HKU for their company when I was going through the most
trying period.

Last but not least, the deepest gratitude and a huge thank you to my beloved parents.
Your unconditional and selfless love has always supported me to overcome any
obstacle. Thank you for so firmly believing in me in every aspect of life, this thesis is
dedicated to you.

ii
TABLE OF CONTENTS

DECLARATION ............................................................................................................ i

ACKNOWLEDGEMENT .............................................................................................ii

TABLE OF CONTENTS ............................................................................................. iii

LIST OF FIGURES ...................................................................................................... ix

LIST OF TABLES ..................................................................................................... xiii

LIST OF SYMBOLS .................................................................................................. xiv

1 INTRODUCTION ................................................................................................. 1

1.1 Background ..................................................................................................... 1

1.2 Objectives ........................................................................................................ 2

1.3 Scope of Thesis ............................................................................................... 4

2 FIRE-INDUCED SPALLING – A STATE OF THE ART REVIEW .................. 6

2.1 Introduction ..................................................................................................... 6

2.2 Types and Mechanisms of Spalling ................................................................ 6

2.2.1 Types of Spalling ..................................................................................... 6

2.2.2 Pore Pressure-Induced Spalling and Moisture Clog Theory.................... 8

2.2.3 Thermal Stress-Induced Spalling ............................................................. 9

2.2.4 Combined Pore Pressure and Thermal Stress Spalling .......................... 11

2.3 Methods of Assessment ................................................................................. 12

2.3.1 Fire Testing ............................................................................................ 12

2.3.2 Prescriptive Methods ............................................................................. 13

iii
 2.3.3 Performance-Based Methods ................................................................. 14

2.4 Protective Measures ...................................................................................... 21

3 CONCRETE EXPOSED TO ELEVATED TEMPERATURE ........................... 35

3.1 Introduction ................................................................................................... 35

3.2 Thermal Properties of Concrete .................................................................... 35

3.2.1 Thermal Conductivity ............................................................................ 35

3.2.2 Density ................................................................................................... 39

3.2.3 Specific Heat Capacity ........................................................................... 41

3.2.4 Porosity .................................................................................................. 44

3.2.5 Permeability ........................................................................................... 46

3.2.6 Relative Permeability ............................................................................. 47

3.3 Mechanical Properties of Concrete ............................................................... 49

3.3.1 Compressive Strength ............................................................................ 49

3.3.2 Tensile Strength ..................................................................................... 50

3.3.3 Elastic Modulus ..................................................................................... 51

3.3.4 Thermal Expansion ................................................................................ 52

3.4 Material Data Related to Water Species and Air .......................................... 53

3.4.1 Material Data Related to Water ............................................................. 53

3.4.2 Material Data Related to Water Vapour ................................................ 55

3.4.3 Material Data Related to Air .................................................................. 56

4 COUPLED THERMO-HYGRO MODEL .......................................................... 58

4.1 Introduction ................................................................................................... 58

4.2 Mass and Enthalpy Conservation .................................................................. 58

4.3 Constitutive Laws and State Equations ......................................................... 61

iv
 4.3.1 State Equations of Gaseous Mixture ...................................................... 61

4.3.2 Young-Laplace Equation ....................................................................... 62

4.3.3 Kelvin-Laplace Equation ....................................................................... 63

4.3.4 Degree of Saturation .............................................................................. 65

4.3.5 Darcy-Buckingham Law ........................................................................ 66

4.3.6 Fick’s First Law of Diffusion ................................................................ 67

4.3.7 Fourier’s Law ......................................................................................... 69

4.4 Numerical Solution ....................................................................................... 69

4.4.1 Initial and Boundary Conditions ............................................................ 71

4.4.2 Spatial Discretization by Finite Element ............................................... 74

4.4.3 Temporal Discretization by Finite Difference ....................................... 76

5 COUPLED THERMO-MECHANICAL MODEL .............................................. 80

5.1 Introduction ................................................................................................... 80

5.2 Momentum Conservation .............................................................................. 80

5.3 Constitutive Laws .......................................................................................... 81

5.3.1 Bishop Stress.......................................................................................... 82

5.3.2 Decomposition of Total Strain ............................................................... 84

5.3.3 Damage of Concrete .............................................................................. 87

5.4 Numerical Solution ....................................................................................... 92

5.5 Spalling Criterion .......................................................................................... 96

6 THERMO-HYGRO-MECHANICAL MODEL ANALYSIS ........................... 102

6.1 Introduction ................................................................................................. 102

6.2 Experiment of Ichikawa and England (2004) ............................................. 102

6.2.1 Model Description ............................................................................... 102

v
 6.2.2 Model Parameters ................................................................................ 103

6.2.3 Results and Discussions ....................................................................... 103

6.3 Experiment of Mindeguia et al. (2009) ....................................................... 105

6.3.1 Model Description ............................................................................... 105

6.3.2 Model Parameter .................................................................................. 106

6.3.3 Results and Discussions ....................................................................... 106

6.4 Experiment of Ozawa et al. (2012) ............................................................. 108

6.4.1 Model Description ............................................................................... 108

6.4.2 Model Parameters ................................................................................ 109

6.4.3 Results and Discussions ....................................................................... 109

6.5 Algorithm to Predict Fire-Induced Spalling ................................................ 111

7 EARLY RESIDUAL SPLITTING TENSILE STRENGTH AT ELEVATED

TEMPERATURE ...................................................................................................... 125

7.1 Introduction ................................................................................................. 125

7.2 Experimental Program................................................................................. 126

7.2.1 Concrete Mix and Test Specimens ...................................................... 126

7.2.2 Heating Regime ................................................................................... 127

7.2.3 Instrumentation and Test Setup ........................................................... 127

7.2.4 Test Procedure ..................................................................................... 128

7.3 Experimental Results................................................................................... 129

7.3.1 Temperature Distribution ..................................................................... 129

7.3.2 Measured Early Residual Splitting Tensile Strength ........................... 130

7.3.3 Spalling Behaviour............................................................................... 130

7.3.4 Failure Mode ........................................................................................ 130

7.4 Numerical Simulation ................................................................................. 131

vi
 7.4.1 Temperature Distribution ..................................................................... 131

7.4.2 Predicted Early Residual Splitting Tensile Strength ............................ 132

7.5 Discussions .................................................................................................. 132

7.5.1 Measured Early Residual Splitting Tensile Strength at Elevated

Temperature ....................................................................................................... 132

7.5.2 Predicted Early Residual Splitting Tensile Strength at Arbitrary High

Temperature ....................................................................................................... 134

7.6 Conclusions ................................................................................................. 134

8 PROTECTIVE MEASURE – BILAYER COLUMN TEST ............................. 148

8.1 Introduction ................................................................................................. 148

8.2 Experimental Program................................................................................. 149

8.2.1 Concrete Mix and Test Specimens ...................................................... 149

8.2.2 Test Setup and Procedures ................................................................... 150

8.3 Compression Test Results and Discussions ................................................ 151

8.3.1 Pexp/Ppre Ratios ................................................................................ 152

8.3.2 Axial Deformations .............................................................................. 152

8.3.3 Failure Modes and Strains of Concrete ................................................ 153

8.4 Fire Test Results and Discussions ............................................................... 154

8.4.1 Temperature Distribution ..................................................................... 155

8.4.2 Failure mode and Spalling Analysis .................................................... 156

8.4.3 Minimum Effective Thickness of Outer Layer .................................... 159

8.5 Conclusions ................................................................................................. 161

9 CONCLUSIONS AND RECOMMENDATIONS ............................................ 181

9.1 Conclusions ................................................................................................. 181

vii
 9.1.1 Modelling of Concrete at Elevated Temperature ................................. 181

9.1.2 Experiment of Early Residual Splitting Tensile Strength .................... 182

9.1.3 Experiment of Bilayer Column ............................................................ 184

9.2 Recommendations for Further Study .......................................................... 185

Appendix A Derivation of Thermo-Hygro Equations .......................................... 187

Appendix B Finite Element Solution .................................................................... 193

Appendix C Derivation of Thermo-Mechanical Equation ................................... 197

Appendix D Pioneer Bilayer Cylinder Test .......................................................... 200

D1) Test Specimens and Test Setup ................................................................... 200

D2) Test Procedures ........................................................................................... 200

D3) Results and Discussions .............................................................................. 201

D4) Post-Cooling Behaviour .............................................................................. 203

Reference ................................................................................................................... 211

viii
LIST OF FIGURES
Figure 2.1 Pore pressure induced-spalling by Liu et al. (2018a) ................................. 26

Figure 2.2 Thermal stress-induced spalling by Liu et al. (2018a) ............................... 27

Figure 2.3 Combined pore pressure and thermal stress spalling by Klingsch (2014) . 28

Figure 2.4 Fire assessment methods and design process, after Khoury (2000) ........... 29

Figure 2.5 Various fire scenarios reproduced from curves by Lottman (2017)........... 30

Figure 2.6 Comparison between nomograms for assessment of spalling .................... 30

Figure 2.7 Nomograms of moisture content and applied level of stress by (a)
Sertmehmetoglu (1977) and (b) Zhukov (1994) ....................................... 31

Figure 2.8 Nomograms of compressive stress and element thickness by

Sertmehmetoglu (1977) ............................................................................ 32

Figure 2.9 Idealized sketch of evaporable water ......................................................... 32

Figure 2.10 Decomposition of concrete as multiphase medium .................................. 33

Figure 2.11 Representative elementary volume of porous medium by Gawin and

Pesavento (2012) ...................................................................................... 34

Figure 2.12 Hollow spherical model of solid skeleton by Ichikawa (2000) ................ 34

Figure 4.1 Curved interface between gas and liquid by Bažant and Jirásek (2018) .... 79

Figure 5.1 Concrete element subjected to combined stresses .................................... 101

Figure 6.1 Modelling of concrete wall ....................................................................... 113

Figure 6.2 Comparison of temperature ...................................................................... 113

Figure 6.3 Comparison of pore volume ..................................................................... 114

Figure 6.4 Moisture content comparison: (a) experimental data from Ichikawa and
England (2004) and (b) numerical analysis ............................................ 115

Figure 6.5 Pore pressure comparison: (a) experimental data from Ichikawa and England
(2004) and (b) numerical analysis .......................................................... 116

Figure 6.6 Experimental setup, after Mindeguia et al. (2009) ................................... 117

ix
Figure 6.7 Locations of gauges, by Mindeguia et al. (2009) ..................................... 118

Figure 6.8 Comparison of furnace temperature ......................................................... 119

Figure 6.9 Comparison of internal temperature ......................................................... 119

Figure 6.10 Comparison of pressure at depth of: (a) 10 mm, (b) 20 mm and (c) 30 mm
................................................................................................................ 121

Figure 6.11 Locations of thermal couple and pressure gauges by Ozawa et al. (2012)
................................................................................................................ 121

Figure 6.12 Experimental setup of heating test, after Ozawa et al. (2012)................ 122

Figure 6.13 Comparison of wet specimen: (a) temperature and (b) pressure ............ 123

Figure 6.14 Comparison of air-dried specimen: (a) temperature and (b) pressure .... 124

Figure 7.1 Residual tensile strength of HSC (C60-C76) ........................................... 139

Figure 7.2 Residual tensile strength of HSC (C80-C100) ......................................... 139

Figure 7.3 Arrangement of thermal couples in a specimen ....................................... 140

Figure 7.4 Location of thermal couple inside furnace ............................................... 140

Figure 7.5 Schematic representation of heating regime............................................. 140

Figure 7.6 Control panel and chamber of furnace ..................................................... 141

Figure 7.7 Trolley installed with a jig ........................................................................ 141

Figure 7.8 Comparison of temperature distribution of C60 specimens: (a) Centre and
(b) 50 mm from centre ............................................................................ 142

Figure 7.9 Comparison of temperature distribution of C90 specimens: (a) Centre and
(b) 50 mm from centre ............................................................................ 143

Figure 7.10 Failure mode of specimens ..................................................................... 144

Figure 7.11 Temperatures before and after transportation......................................... 144

Figure 7.12 Early residual tensile strengths compared with simulation results ......... 145

Figure 7.13 Relative residual tensile strength of HSC (C60-C76) ............................ 145

Figure 7.14 Relative residual tensile strength of HSC (C80-C100) .......................... 146

x
Figure 7.15 Average early residual tensile strengths compared with residual strengths
................................................................................................................ 146

Figure 7.16 Comparison between average relative strengths .................................... 147

Figure 8.1 Types of specimen .................................................................................... 165

Figure 8.2 Before casting the outer layer of bilayer and sealed bilayer column ....... 165

Figure 8.3 Compression test setup ............................................................................. 166

Figure 8.4 Configuration of LVDTs and strain gauges ............................................. 167

Figure 8.5 Fire test setup............................................................................................ 168

Figure 8.6 ISO 834 temperature-time curve .............................................................. 168

Figure 8.7 Effect of outer layer thickness of reinforced specimens .......................... 169

Figure 8.8 Effect of outer layer thickness of plain specimens ................................... 169

Figure 8.9 Effect of the presence of reinforcement ................................................... 170

Figure 8.10 Failure mode of plain specimen ............................................................. 171

Figure 8.11 Failure mode of reinforced specimen ..................................................... 172

Figure 8.12 Top view of a failed specimen................................................................ 173

Figure 8.13 HSC core of a failed specimen ............................................................... 173

Figure 8.14 Measured furnace temperatures versus standard curve .......................... 174

Figure 8.15 Temperature distributions of unsealed specimens: (a) One-hour exposure

and (b) two-hours exposure .................................................................... 175

Figure 8.16 Temperature distribution of sealed specimens: (a) 2L174 specimens and (b)
2L190 specimens .................................................................................... 176

Figure 8.17 Failure modes of specimens ................................................................... 177

Figure 8.18 Schematic representation of outer layer cracking .................................. 177

Figure 8.19 Temperatures at interface of bilayer columns ........................................ 178

Figure 8.20 Pressure in bilayer column: (a) 2L190 and (b) 2L174 ........................... 179

xi
Figure 8.21 Pressure in bilayer column of 50-mm thickness..................................... 180

Figure 8.22 Pressure in bilayer column of 100-mm thickness................................... 180

Figure D.1 Cylindrical specimens for experimental tests .......................................... 205

Figure D.2 Electric furnace ........................................................................................ 206

Figure D.3 Steel cage protecting the specimen .......................................................... 206

Figure D.4 Specimen on a steel cube base................................................................. 207

Figure D.5 Furnace temperatures versus standard curve ........................................... 207

Figure D.6 Comparison of HSC200 specimen .......................................................... 208

Figure D.7 Comparison of 2L200 specimen .............................................................. 208

Figure D.8 Comparison of HSC250 specimen .......................................................... 209

Figure D.9 Comparison of 2L250 specimen .............................................................. 209

Figure D.10 Comparison of HSC300 specimen ........................................................ 210

Figure D.11 Comparison of 2L300 specimen ............................................................ 210

xii
LIST OF TABLES
Table 2.1 Types and characteristics of spalling ........................................................... 25

Table 7.1 Concrete mix proportions .......................................................................... 136

Table 7.2 Specimen sizes of different groups ............................................................ 136

Table 7.3 Concrete properties .................................................................................... 136

Table 7.4 Splitting tensile strengths at elevated temperatures ................................... 136

Table 7.5 Average splitting tensile strengths at elevated temperatures ..................... 137

Table 7.6 Temperature-dependent thermal properties ............................................... 137

Table 7.7 Thermal Properties ..................................................................................... 137

Table 7.8 Simulated splitting tensile strengths (𝑓𝑡,𝑇 ) with or without transportation 138

Table 7.9 Temperature-dependent properties of HSC ............................................... 138

Table 7.10 Temperature-dependent CDP parameters ................................................ 138

Table 8.1 Details of specimens .................................................................................. 163

Table 8.2 Concrete mix proportions .......................................................................... 164

Table 8.3 Averaged properties of concrete ................................................................ 164

Table 8.4 Results of compression tests ...................................................................... 164

Table D.1 Concrete mix proportions ......................................................................... 204

Table D.2 Compression Test Results ......................................................................... 204

xiii
LIST OF SYMBOLS

𝐴𝑐 area of high strength concrete

𝐴𝑠 area of longitudinal reinforcement

𝑏 Biot’s coefficient

𝑩 strain matrix

Ca specific heat capacity of dry air

Cfw specific heat capacity of free water

Cp specific heat capacity of concrete

Cv specific heat capacity of water vapour

d average measured diameter of the cylindrical specimen

difference between the previous and current estimate of solution x θ

𝑑x
at (n + 1)th time step

norm on the integration error (difference between the predicted and

𝑑 𝑛+1
corrected solution)

𝐷 mechanical damage parameter

𝐷𝐵 coefficient of adsorbed water diffusion

𝐷𝜋𝑖 diffusion coefficient of phase π diffusing in phase i (πi = av, va)

𝐄𝟎 initial elastic stiffness matrix (fourth order tensor)

𝐄𝒔𝒆𝒄 secant stiffness matrix (fourth order tensor)

Young’s modulus at temperature T of the mechanically damaged

𝐸(𝑇), 𝐸0 (𝑇)
and undamaged material

𝐸0 (𝑇𝑎 ) mechanically undamaged Young’s modulus at room temperature

̇
𝐸𝑓𝑤 rate of evaporation of free water (including desorption)

𝑓𝑐 cylinder compressive strength of high strength concrete

𝑓𝑐𝑢 cube compressive strength of high strength concrete

𝑓𝑐0 initial compressive strength at room temperature

xiv
𝑓𝑠𝑦 yield strength of longitudinal reinforcement

𝑓𝑡 splitting tensile strength of concrete at room temperature

𝑓𝑡,𝑇 splitting tensile strength of concrete at elevated temperature T

F maximum load in splitting tensile test

g ratio of compressive to tensile strength

𝑔 gravity acceleration

𝐺𝑓 (𝑇) fracture energy release rate

ℎ𝑞 convective heat transfer coefficient on the boundary

ℎ𝑟 radiative heat transfer coefficient on the boundary

combined convection-radiation heat transfer coefficient on the

ℎ𝑞𝑟
boundary

h𝜋 specific free enthalpy of phase π (π = l, v)

𝐻𝑔 enthalpy of gaseous mixture

𝐻 𝑖𝑑 terms associated with damage evolution (id = md, td)

𝐼1 , 𝐼2 first and second invariant of the strain tensor

𝐈, 𝟏 second-order unit (identity) tensor

𝐽2 second invariant of the deviatoric strain tensor

𝐽𝜋 mass flux of phase π (π = a, fw, g, l, v)

𝑘 thermal conductivity of concrete

𝐾 intrinsic permeability of dry concrete

K bulk modulus of the skeleton

Ks bulk modulus of the solid phase (grains)

𝐾𝜋 relative permeability of phase π (π = g, l)

𝑙𝑐 proportionality factor

L average measured length of the cylindrical specimen

xv
𝑳 differential operator

Mπ molar mass of constituent π (π = a, g, w)

𝒏 unit vector normal to the boundary

N total number of nodes in the problem domain

𝑃𝑐 capillary pressure

𝑃𝑐𝑟𝑖 critical pressure of water (22.064 MPa)

Pexp experimentally obtained axial load capacity

𝑃𝑛 pressure in the non-wetting phase

Ppre predicted axial load capacity

Ps pressure exerted on solid phase by the surrounding fluids

𝑃𝑠𝑎𝑡 saturated vapour pressure

𝑃𝜋 partial pressure of phase π (π = a, g, l, v)

𝑞 heat flux

𝑟 radius of capillary menisci

𝑅 universal gas constant

RH relative humidity

𝑅𝜋 specific gas constant of constituent π (π = a, v)

𝑠 increase in rate of stress

derivative of the equivalent strain measure with respect to the strain

𝒔
tensor

𝑆 degree of saturation with free water

𝑆𝐵 degree of saturation with adsorbed water

𝑆𝑛 degree of saturation with the non-wetting phase (𝑆𝑛 = 1 − 𝑆)

𝑆𝑠𝑠𝑝 solid saturation point

t time

xvi
𝒕̅ imposed tension force

𝐭π macroscopic partial stress tensor for phase π (π = g, l, s)

𝑇 absolute temperature (in K)

𝑇𝑐 temperature (in ℃)

𝑇𝑐𝑟𝑖 critical temperature of water (647.096 K)

𝑇𝑟𝑒𝑓 reference temperature (in K)

𝑢 displacement vector of the solid skeleton

𝑣 velocity of fluid transported within concrete

𝑉 thermo-chemical damage parameter

𝑣𝐵 diffusion velocity

𝑣𝜋 volume flux of phase π (π = g, l)

𝑊𝑑 strain energy density

𝑊𝑠𝑝 critical energy required to initiate spalling

independent, uniformly distributed random numbers on unit interval

𝑥1 , 𝑥2
(0, 1)

y1 , y2 independent, normally distributed random numbers

Greek symbols

α thermal diffusivity (α = 𝑘/𝜌Cp )

𝛼𝑎 thermal diffusivity of dry air

𝛼𝑓𝑡 coefficient of free thermal strain

𝛽 coefficient of water vapour mass transfer on the boundary

𝛽𝑡𝑚 coefficient of uniaxial thermo-mechanical strain

𝛾1 , 𝛾2 experimentally determined coefficients in the formulation of

uniaxial thermo-mechanical strain rate

𝛾𝑔𝑙 surface tension of the curved gas-liquid interface

xvii
𝛾(𝑇) ductility parameter at temperature T

Γπ boundary corresponding to state variable π (π = T, g, v, u)

𝛿𝑖𝑗 Kronecker delta

𝛥𝑡 time step

Δu axial deformation corresponding to the ultimate load

𝜖𝑡 user-defined error tolerance

𝜀 total strain

𝜀̅ non-local equivalent strain measure

𝛆̂ strain subjected to undamaged effective stress

⟨𝜀𝑖 ⟩+ positive components of the principal strains

𝜀𝑐 creep strain

𝜀𝑒 elastic strain

𝜀 𝑓𝑡 free thermal strain

𝜀 𝑙𝑖𝑡𝑠 load induced thermal strain

𝜀𝑝 plastic strain

𝜀 𝑡𝑚 thermo-mechanical strain

𝜀𝜋 volume fraction of phase π (π = a, cem, D, fw, g, l, s, v)

θ parameter for temporal approximation

𝜅 𝑚𝑑 history parameter for mechanical damage

𝜅 𝑡𝑑 history parameter for thermo-chemical damage

𝜅0𝑚𝑑 (𝑇) threshold strain measure at temperature T

𝜆𝐷 specific heat of dehydration of chemically bound water

𝜆𝐸 specific heat of evaporation (and of desorption when appropriate)

𝜇, 𝜎 expectation and standard deviation of the generated normal

distribution

xviii
𝜇𝜋 dynamic viscosity of phase π (π = a, g, l, v)

ν Poisson’s ratio

𝜈tc Poisson’s ratio of transient creep

𝜌 averaged density of the multiphase porous medium

𝜌𝑐𝑟𝑖 critical density of water (322 kg/m3)

𝜌𝜋 density of phase π (π = cem, l, s)

𝜌̃π mass of phase π per unit volume of gaseous material (π = a, g, v)

𝜌𝐶 effective heat capacity of multiphase medium (concrete)

(𝜌𝐶)𝑎𝑖𝑟 heat capacity of air

𝜌𝐶𝑣 energy transported by fluid flow (i.e., convection)

σl external load induced stress

σ𝑝 pore pressure induced stress

σ𝑡 thermal load induced stress

𝛔 total stress tensor

𝛔′ mechanical effective stress tensor/ Bishop stress tensor

̂′
𝛔 undamaged effective material stress

𝜙 porosity

χ Bishop’s parameter

Ω problem domain

Subscripts

a dry air within the gaseous phase

c capillary

cem cement

D dehydrated water released from chemically bound water

xix
 fw free water phase (capillary water)

g gaseous phase (mixture of dry air and water vapour)

𝑖 the 𝑖 th node

k iteration index (number of iterations)

l liquid water phase (capillary and adsorbed water)

p predicted solution

s solid phase

u displacement of the solid skeleton

v water vapour within the gaseous phase

w water

∞ atmospheric environment beyond the boundary

Superscripts

0 initial ambient conditions

1/2/3 first / second / third type of boundary condition

D diffusion-induced

e element matrix

md mechanical damage

n the nth time step

P pressure-induced

td thermo-chemical damage

xx
1 INTRODUCTION

1.1 Background

Concrete is generally perceived as fire-resistant material owing to it low thermal

conductivity and incombustibility. The conventional normal strength concrete (NSC)
exhibits good thermal insulating properties at elevated temperatures. Thanks to the
development in concrete technology and construction needs, soaring use of concretes
with increased strength and reduced permeability, such as the high strength concrete
(HSC), ultra-high-performance concrete (UHPC) and self-compacting concrete (SCC),
has been witnessed in recent years. Compared to NSC, these new types of concrete have
advantages in not only the structural but also the economic and aesthetic aspects. Yet,
under fire conditions, these modern concretes may not be able to provide the same level
of fire performance as that of NSC and has been reported to have problems such as
faster degradation of strength, stiffness and overall integrity (Phan, 1996, Kodur, 2000).
Among the various problems reported, fire-induced spalling is of the most interests
because of its potential risks to the concrete structures and structural elements. In
practical engineering construction, the uncertainty in behaviour and mechanism of fire-
induced spalling at elevated temperature is the critical reason why HSC is hesitated to
be used in locations such as basement and foundation.

Spalling refers to the falling off or breaking away of surface layers (pieces) of concrete
from the structural elements exposed to high and rapidly rising temperatures, and
explosive spalling typically involves a more sudden and violent manner (Phan, 2008b,
Kodur, 2000, Hertz, 2003). Traditionally, the extensive and predominant use of
ordinary NSC in concrete type structures has made the possibility of fire-induced
spalling relatively low. Under some extreme scenarios, however, the likelihood of
spalling may be significantly increased by the direct exposure to fires or fast rising
temperatures. These scenarios include, but are not limited to, the threats of natural
disasters, terrorist attacks, accidental fires and explosions. Moreover, some specific
infrastructures, such as the electricity power plants, pressure vessels used in

1
petrochemical industry, nuclear reactors, and storage tanks for hot materials and liquids,
are more prone to spalling due primarily to their regular subjection to elevated
temperatures (ElMohandes, 2013). And the consequences of spalling in these structures
could be catastrophic considering the nature of their unique functionalities.

It is therefore of major importance to study the behaviour of concrete at elevated

temperatures and the associated possibility of fire-induced spalling. As will be
discussed in details in Chapter 2, intensive past research has been focused on this topic,
including the mechanisms (Ozawa et al., 2012, Jeongwon et al., 2011, Kodur, 2000,
Mindeguia et al., 2010, Liu et al., 2018a, Khoury, 2000), contributing factors (Phan,
2008b, Shah et al., 2017, Kodur and Phan, 2007, Kodur, 2000) and predictive models
(Dwaikat and Kodur, 2009, Gawin et al., 1999, Bažant and Thonguthai, 1979,
ElMohandes, 2013, Burgh, 2016, Tenchev et al., 2001, Davie et al., 2006, Davie et al.,
2010, Ichikawa and England, 2004) of spalling. While it is desirable to establish a
robust and practical methodology as well as numerical model to predict the occurrence
and extent of spalling, no single methodology has manged to accurately and
conclusively explain the phenomenon of explosive spalling and predict its occurrence,
resulting from the uncertainties introduced by the stochastic nature of spalling. Hence,
effective practical measures to mitigate and prevent the fire-induced spalling, especially
the explosive one, are also in great need of investigation.

1.2 Objectives

For the reasons mentioned above, it is obvious that research work has to be dedicated
to both the numerical and experimental aspects to better understand, predict and prevent
the occurrence of fire-induced spalling. The first objective of the research is to develop
a practical thermo-hygro-mechanical model to describe the behaviour of concrete at
elevated temperatures as well as the corresponding changes of the internal liquid and
gaseous phases. In particular, the stochastic nature associated with the spalling
phenomenon will be considered by an algorithm predicting the position of spalling if it
happens. This algorithm, together with a rigorously derived mathematical model, could
render totally new insights into the study of spalling behaviour.

2
A reliable spalling criterion is critical to the accurate determination of whether spalling
would happen under certain conditions. The second objective is therefore to establish
an appropriate spalling criterion and then incorporate it into the numerical model. The
criterion is developed in a two-step manner. To begin with, a suspected spalling zone
will be determined by evaluation of the mechanical damage of concrete, using an
equivalent stain mapping of the strain tensor. For areas with high mechanical damage,
the second step is then carried out to compare the strain energy density and the required
fracture energy of spalling. If the strain energy density stored in the element exceeds
the energy needed for spalling, spalling is assumed to be initiated. Through this two-
step criterion, the pattern of spalling zone could be approximated, and most importantly,
effects of both the pore pressure and thermal stress are properly accounted for.

One important parameter required for the development of the thermo-hygro-mechanical

model and the spalling criterion is the tensile strength of concrete at elevated
temperatures. Owing primarily to the difficulties of conducting tensile concrete tests,
especially at high temperatures, the tensile strengths of concrete obtained by most of
the research to date are residual strengths measured at room temperature after the
specimens have been cooled down. It is necessary to design a tailored apparatus capable
of conducting such tests while specimens are still at high temperatures. The third
objective of this research is to obtain experimental data of concrete tensile strengths at
high temperatures, termed as the early residual tensile strengths. The results are then
analysed and compared with existing data to find the difference between residual and
early residual tensile strengths of concrete. Finally, the experimental data is applied to
the numerical model for simulation.

Last but not least, this research aims to investigate a protective practical measure to
prevent the concrete elements from fire-induced spalling. A novel bilayer system has
been proposed to improve the fire performance of high strength concrete at elevated
temperatures. The bilayer system consists of a normal strength concrete outer layer and
a high strength concrete core. The NSC outer layer acts as an insulation to reduce the
thermal gradient in HSC core and, at the same time, provides lateral confinement to the

3
HSC core. Mechanical and thermal behaviours of the bilayer system are studied and the
results are discussed.

1.3 Scope of Thesis

This thesis consists of nine chapters, followed by four appendices. Chapter 2 presents
a review regarding the fire-induced spalling of concrete at elevated temperature. A
categorization of spalling is given based on the main characteristic of their behaviour,
and the corresponding mechanisms are subsequently explained. Three major methods
for the assessment of fire resistance are reviewed. Finally, the existing protective
measures from design codes or literature are summarized.

Chapter 3 mainly provides the thermal and mechanical properties of concrete subjected
to elevated temperature. These properties are evaluated to be employed in the numerical
model. In addition, some of the critical parametric relationships associated with water
species and air are also provided, which play an important role in the modelling of
thermo-hygro behaviour of concrete.

Chapter 4 introduces the formulation, implementation and solution for the thermo-
hygro model, describing the heat and mass transfer process within concrete. The
macroscopic governing equations and their related constitutive laws are discussed. The
detailed derivation of the governing equations and the finite element solution procedure
is given respectively in Appendix A and Appendix B.

Chapter 5 describes the second component, i.e. the thermo-mechanical component, of

the numerical model. Similar to Chapter 4, the macroscopic governing equations and
constitutive laws are discussed in like manner, and extended details for derivation are
presented in Appendix C. Eventually, a criterion is developed and incorporated into the
model for prediction of spalling.

Chapter 6 mainly consists of three numerical validations of the HTM model against
published experiments. Each example of validation provides description of the problem

4
and discussions on the results. Lastly, an algorithm is proposed to account for the
stochastic nature of concrete behaviour at elevated temperature. Steps of
implementation to incorporate the algorithm to the model are also given.

Chapter 7 presents the experimental studies on the early residual tensile strength of high
strength concrete at elevated temperature. Numerical simulations were conducted to
analyse the temperature distribution and reproduce the test results. The early residual
tensile strength could serve as an input parameter in the thermo-hygro-mechanical
model discussed in Chapter 4 and Chapter 5.

Chapter 8 introduces a protective measure to prevent spalling. A bilayer column system

was designed, adopting an outer layer of normal strength concrete as thermal barrier.
Experimental investigation, including compression and fire tests at ambient and
elevated temperature, was conducted on the bilayer specimens. Experimental results
were analysed and discussed to evaluate and improve the bilayer design.

Chapter 9 ends this thesis by concluding the key outcomes made in this study. Last but
not least, recommendations are given for further study.

5
2 FIRE-INDUCED SPALLING – A STATE OF THE ART
REVIEW

2.1 Introduction

Although documentations of “the falling away of concrete” (Himmelwright, 1906,

Woolson, 1905) could trace back to the early years of twentieth century, it was not until
the 1970s that the phenomenon of concrete spalling was explicitly summarized and
investigated (Meyer-Ottens, 1972, Hertz, 1984). Since then studies on the effects of
high temperature on the fire performance of concrete surged, especially that of the high
strength concrete. Among these early studies were mostly material tests (Felicetti et al.,
1996, Furumura et al., 1995, Sullivan and Sharshar, 1992, Hammer, 1995, Diederichs
et al., 1988, Hertz, 1992, Castillo, 1987) and element tests (Diederichs et al., 1995,
Hansen and Jensen, 1995, Sanjayan and Stocks, 1993, Shirley et al., 1988). The one
problem that early studies commonly address is the spalling or explosive behaviour of
HSC at elevated temperatures, though inconsistencies are also observed from the test
results. While some researchers reported spalling phenomenon in HSC structural
members, there were a few experimental studies showing little or no obvious spalling.
Possible reasons for this conflicting picture on the occurrence of spalling may be
attributed to the massive number of factors affecting spalling and their interdependency.
To better understand and predict the fire-induced spalling, and to satisfy the fire safety
requirements in practical construction projects, more recent research has been done
regarding the behaviour and mechanism of spalling and the performance of concrete at
elevated temperatures.

2.2 Types and Mechanisms of Spalling

2.2.1 Types of Spalling

Based on the findings of fire test series performed by Gary more than a century ago,
Meyer-Ottens (1972) summarized the spalling phenomenon into four categories with

6
slight modifications to Gary’s terminology (Jansson, 2013). Khoury (2008) has more
recently extended the types of spalling to six categories. The main characteristics of
each type of spalling behaviour when exposed to standard ISO fire, are presented in
Table 2.1 (Khoury et al., 2007). It should be noted that, in a single fire case, multiple
types of spalling could happen simultaneously, depending on the properties of concrete
itself as well as the fire characteristics.

It can be concluded from Table 2.1 that the types of spalling could also be grouped
using different parameters, such as the time of occurrence, the intensity of spalling
behaviour, etc. Among these six types of spalling, the influence of aggregate spalling
is the most superficial and can be mitigated by appropriate aggregate choice under non-
repetitive fire scenarios (Hertz, 2003). Since aggregate spalling is generally of
negligible structural significance, it will not be discussed in details in the thesis.

Sloughing-off and post-cooling spalling are also termed as thermo-chemical spalling

by Liu et al. (2018a), resulting primarily from the break-down of aggregate cement
bond, such as calcium silicate hydroxide and calcium hydroxide (Schneider, 1988b).
Sloughing-off spalling occurs at extreme high temperatures (usually higher than 800 ℃)
when the concrete has weakened and the strength has decreased to low levels. Post-
cooling spalling occurs during or after the cooling of concrete, usually associated with
the re-absorption of moisture in the air. If the ambient environment is sufficiently humid,
then post-cooling spalling could even occur in a progressive manner. Similar to the
sloughing-off spalling, post-cooling spalling is assumed to occur at high temperatures
around 750 ℃. Since the threshold temperature of the occurrence of thermo-chemical
spalling is very high (larger than 750 ℃), its influence is considered to be less
detrimental compared to the remaining three types of spalling.

Corner spalling is characterized by the phenomenon of section loss from convex beam
or column corners, typically happens at a later stage of fire (Khoury and Anderberg,
2000). It is mainly caused by the faster heat penetration due to the multiple heat
exposures of element surfaces (Burgh, 2016). Once corner spalling occurs, corner
reinforcement could be directly exposed to the heat source, leading to potentially

7
serious structural consequences. The final two types of spalling, namely the explosive
and surface spalling, pose the greatest threat to reinforced concrete structures due to
their violent nature and early occurrence. A detailed review on the mechanisms and
theories related to the two spalling types is necessary for better understanding of this
violent yet critical phenomenon. However, the distinction between these two categories
is not so clear (though strictly speaking the explosive spalling do experience a more
violent expulsion of concrete chunks) because of the similar concrete spalling
behaviours. Therefore, explosive and surface spalling will be examined as one category
in the following subsection as the fire-induced spalling.

2.2.2 Pore Pressure-Induced Spalling and Moisture Clog Theory

Pore pressure-induced spalling is also called the thermal-hygro (hydro) spalling by

some researchers (Mindeguia et al., 2010, Liu et al., 2018a). It was first introduced by
Shorter and Harmathy (1961) and later Harmathy (1965) proposed the “moisture clog”
to refine this theory. Pore pressure build-up due to evaporation of free water and
increased moisture content near the heated surface has been hypothesized as the main
trigger of spalling.

In the course of heat transfer, a thermal gradient is formed through the thickness of
heated concrete and free water starts to evaporate when temperature exceeds 100 ℃.
The evaporation of water adds to the vapor pore pressure and subsequently forms a
pressure gradient, which drives the moisture from high pressure zone towards low
pressure zone. In this sense, the moisture will not only be forced outwards to the heated
surface, where pore pressure could be released to the surrounding atmosphere, but also
to the cooler, inner area of the concrete under the combined effect of thermal and
pressure gradient. As vapor migrates deeper inwards, the hot vapor condenses when
passing through the cooler region. Moisture content is gradually increased in the
neighboring region until the concrete pores are fully saturated. As a result, three zones
are generated during the vapor migration. The schematic representation of this
phenomenon by Liu et al. (2018a) is depicted in Figure 2.1. Dry zone is the zone of
dehydration on the outermost heated region, where the concrete dries out and only dry

8
air exists. Adjacent to the dry zone is the zone of mixed vapor and liquid water,
recognized as the wet zone, whose total pore pressure should be equal to the moist air
pressure consisting of the saturated vapor pressure and dry air pressure. Since dry air
pressure is often of negligible magnitude, the peak pressure in the wet zone is then
approximated as the saturated vapor pressure (Liu et al., 2018a). Beyond the wet zone
is the saturated zone, where further vapor migration is restricted by the low permeability.
Vapor pressure is therefore rapidly accumulated in front of the saturated zone, which
acts as a “moisture clog”. Fire-induced spalling of the pressure front layer occurs when
the tensile strength of concrete is exceeded by the peak pore pressure.

Though the phenomenon of moisture clog was successfully visualized and confirmed
by Jansson and Boström (2009), some experimental studies have shown to be
inconsistent with this mechanism. Bažant and Cusatis (2005) claimed that pore
pressures exceeding 20 atm (2 MPa) has never been measured, meaning that no pore
pressure exceeding the tensile strength of concrete have never been observed. Similar
experimental observations were reported by Li et al. (2019), where the measured
maximum pore pressure were far less than the tensile strength in spalled specimens.
The experimental results of Jansson and Boström (2012) described in Van der Merwe
(2019)’s work demonstrated that the restrain form associated with different stress
distributions may be a critical factor to the occurrence of spalling. Klingsch (2014)
concluded that the phenomenon of gradual pressure rises within the dry zone remained
unanswered and that the “vapor drag forces” proposed by Meyer-Ottens (1972) was
never verified by experimental tests. Lottman (2017) pointed out that substantial drop
of pore pressure due to the formation of cracks prior to spalling could not be the main
cause of explosive spalling.

2.2.3 Thermal Stress-Induced Spalling

The concept of thermal stress-induced spalling (also called the thermo-mechanical

spalling) was first introduced by Saito (1966) and later improved by Dougill (1972).
This mechanism attributes the main cause of spalling to the thermal buckling instability
of the weakened surface layer. It is supposed that sufficiently steep thermal gradient

9
resulting from the rapid heating rate could lead to significant thermal stresses. The
thermal deformations (longitudinal thermal expansion, curvature and thermal strain) of
heated region is restrained by the cooler region, therefore, compressive stresses parallel
to the heated face are generated, which induce microcracks in the weakened surface.
When the maximum thermal stress exceeds the temperature dependent compressive
strength of concrete, thermal instability is said to cause spalling of the weakened surface
layer. A schematic representation of the thermal stress-induced spalling is shown in
Figure 2.2. This mechanism gives plausible explanations for the increased risk of
spalling when compressive loading applied to concrete elements superimposes the
effect of thermal stress. On the other hand, Bažant and Cusatis (2005) rejected the
dominant role of pore pressure developed due to moisture clog in fire-induced spalling.
Their justification was that the volume made available to gaseous and liquid water is
increased by several orders of magnitude once cracks start to propagate. Pore pressure
could be instantaneously reduced through this volume while additional liquid water
cannot suddenly flow into and re-pressurize the crack. The time needed for the
surrounding water to re-fill the crack openings are therefore deemed far longer than the
duration of a fire-induced spalling. Bažant and Cusatis claimed that pore pressure could
only serve as a trigger of fracture, other supply of energy must be provided to drive an
explosion. They further attributed this supply of energy to the potential energy stored
in the thermal stresses parallel to heated surface. The only exception to their
argumentation is when liquid water is pressurized.

While similar thermo-mechanical spalling has been reported and illustrated in brittle
ceramic materials (Khoury et al., 2007), there are still considerable experimental
observations being contradictory with this thermal stress-induced mechanism. A series
of tests on structural elements with varying moisture content performed by Meyer-
Ottens (1972) indicated the relevance of moisture content, where members with low
moisture content (< 3%) did not experience spalling. Ichikawa (2000) argued that
adding PP fibres to concrete mix could successfully mitigate spalling, which cannot be
explained solely by this mechanism. The work of Klingsch (2014) showed that even at
an extremely low heating rate (0.5 ℃/min)，ultra-high strength concrete could still

subject to fire-induced spalling regardless of the negligible thermal stress. Lu (2015)

10
again emphasized the role of moisture content. He pointed out that this mechanism
would imply reduced risk of spalling for highly moisten concrete, yet this implication
opposes the experimental observations of water-cured specimens.

2.2.4 Combined Pore Pressure and Thermal Stress Spalling

Based on the discussions presented above, it is very clear that the effect of pore pressure
and thermal stresses on fire-induced spalling should be combined and considered
simultaneously. Zhukov (1975) was the first to propose the combined mechanisms of
explosive spalling. He assessed the likelihood of spalling with equations predicting the
stresses and a spalling envelope depending on the moisture content and stress level of
concrete. Under the combined mechanism, stresses were considered to be comprised of
the superposition of applied loads, restrained thermal dilatation and pore pressure. It is
said that spalling is unlikely to occur when moisture content is less than 3%. Zhukov’s
idea was later refined and developed by Sertmehmetoglu and Connolly in 1977 and
1995 respectively. Sertmehmetoglu (1977) assumed reduced temperature dependent
tensile strength of concrete perpendicular to the heated surface, parallel to which a
system of cracks is developed. Klingsch (2014) reproduced the schematic
representation of the combined pore pressure and thermal stress-induced spalling
shown in Figure 2.3 based on Zhukov (1975) after Khoury et al. (2007).

The stresses resulting from external loads (σl ) and thermal loads (σt ) are compressive
while the pore pressure induced stress (σp ) is tensile stress. The total strain εx in the
direction of spalling, considering Poisson effect and according to elastic theory, can be
expressed as:

1
εx = [𝜎 − 𝜈(𝜎𝑙 + 2𝜎𝑡 )] (2.1)
𝐸 𝑝

The strain energy density in the direction of spalling (Wx ) is take as the area under the
linear stress-strain curve:

11
 1
Wx = 𝜎𝑝 𝜀𝑥 (2.2)
2

Zhukov suggested that the required energy to initiate spalling (Wspalling ) is the rupture
strain energy computed with the tensile strength (𝑓𝑡 ) of concrete:

1 𝑓𝑡
Wspalling = 𝑓𝑡 (2.3)
2 𝐸

When the strain energy density Wx is larger than Wspalling , spalling is assumed to occur.
Equating (2.2) and (2.3) yields the critical tensile strength at which spalling happens:

ft = √𝜎𝑝2 − 𝜈𝜎𝑝 (𝜎𝑙 + 2𝜎𝑡 ) (2.4)

From Eq. (2.4) it can be seen that the effect of pore pressure (σp ) and thermal stress (σt )
are taken into account simultaneously. In Zhukov’s original work, however, he assumed
that the transformation of pore pressure into stress was complete, which obviously
overestimated the influence of pore pressure. This limitation is overcome by
introducing the Biot’s coefficient, which will be discussed in details in Chapter 5.

2.3 Methods of Assessment

Based on Khoury (2000)’s summary there are mainly three methods of assessment of
fire resistance to date: fire testing, prescriptive methods and performance-based
methods. Their main application in the assessment of fire resistance as well as design
process are shown in the flow chart of Figure 2.4.

2.3.1 Fire Testing

Fire testing refers to the subjection of structural elements or subassembly to standard

temperature-time curves provided by the design codes. Figure 2.5 shows some of the

12
most used standard fire regimes from different codes around the world. These fire
curves generally represent the worst fire scenario where heating rate and the maximum
obtained temperature are high. Yet, limited by the nature and magnitude of restraint and
continuity of adjoining construction, it is unlikely that single structural elements could
reflect the true fire performance of elements in a building. Due to the large-scale of test
specimens and stringent fire characteristics, the preparation and execution of fire testing
is unsurprisingly lengthy and costly, for which Khoury (2000) concluded that “testing
of a complete construction in a fire is a formidable task”. Investigation into other
methods of assessment are therefore necessary.

2.3.2 Prescriptive Methods

Prescriptive methods involve the predetermined requirements based on design tables or

graphs obtained primarily from the full-scale fire tests. Most available codes guide the
practice of fire safety design by specifying the minimum section dimensions and cover
thickness for a certain fire-resistant class, without comprehensive considerations of
spalling behaviour and its influence. For example, BS 8110: Part 2 (1985) tabulates the
sectional requirements for concrete elements for fire resistances from 30 min to 4 hours.
The U.S. ACI 216R (1989) only provides high-temperature properties of concrete for
the determination of fire endurance of concrete elements. Similarly in the Eurocode 2
Part 1-2 (2004), though spalling is explicitly considered, only rigid and restrictive
measures are provided to prevent spalling without a rationally based model. The
Eurocode 2 only considers normal strength concrete with initial moisture content larger
than or equal to 3% the likelihood of spalling, whose risk is dependent on a) the
moisture content, b) aggregate type, c) permeability and d) fire characteristics. For high
strength concrete exceeding grade C80/95 or content of silica fume exceeding 6% by
weight of the total cementitious content, four recommended measures are provided to
reduce the risk of spalling, which will be disclosed in full details in Section 2.4. The
Hong Kong Code of Practice for Structural Use of Concrete (2013) provides basically
the same guidelines as Eurocode 2. The only modification is that it demands a
mandatory fire testing for high strength concrete exceeding grade C80, to verify the
main reinforcement would not be exposed under the designated fire resistance rating.

13
According to Khoury (2000), though cheap and easy to implement, this method can
sometimes be conservative while at other times unsafe. Some standards even make
erroneous reference in the classification of concrete regarding the temperature-
dependent properties. A superior alternative method is needed for a more accurate and
cost-effective assessment.

2.3.3 Performance-Based Methods

With the development of computer technology and numerical techniques, performance-

based method is acquiring increasing interest in the studies of fire-induced spalling.
Over the past decades, numerous efforts have been dedicated to the establishment of
analytical and numerical models, for a more comprehensive and robust description of
concrete behaviour at elevated temperatures.

2.3.3.1 Simplified Predictive Models

Early formulations of such models were mainly in the form of simplified analytical
expressions or nomograms (Van der Merwe, 2019). As discussed previously in Section
2.2.2, Harmathy (1965) proposed the moisture clog theory, based on which he derived
analytical expressions and developed a nomogram to assess the risk of spalling
graphically. Harmathy’s limit criterion for pore saturation was subsequently modified
by Sertmehmetoglu (1977) to include permeability, viscosity and pressure difference,
narrowing the permeability boundary.

Figure 2.6 shows the comparison between the two saturation-permeability nomograms.
Based on the combined thermo-hygro and thermo-mechanical mechanism discussed in
Section 2.2.4, Sertmehmetoglu (1977) and Zhukov (1994) developed nomograms
(Figure 2.7) to account for the relaitonship between moisture content and level of
applied stress. However, the temperature-dependent tensile strength of concrete was not
included. The suceptibility of spalling was also assessed through the relationship
between level of applied stress and element thickness in the same work of
Sertmehmetoglu (Figure 2.8), which clearly neglected the influence of moisutre content

14
and pore pressure. Though served as the earliest attempts for prediction of fire-induced
spalling, the methodologies behind these models are rather superficial. Their indication
is consequently over simplified and restrictive to a limited scope of experimental results.

2.3.3.2 Comprehensive Numerical Models

Later, more sophisticated thermo-hygro and thermo-hygro-mechanical models were

proposed by various researchers. Damages and nonlinearities of materials on
temperature are incorporated in the models. Chemical-physical phase changes, for
instance, the hydration-dehydration, evaporation-condensation and adsorption-
desorption, are also directly applied in the transport process for the first time (Khoury,
2000). Most of these models fundamentally treated the concrete as partially-saturated
multiphase porous medium, consisting of solid skeleton, water and dry air. The
definition of water within concrete in various publications differentiates from one
another, representing basically the same idea with different terminology. The most
thorough and comprehensible (in the author’s opinion) definition by far is provided by
Bažant and Jirásek (2018). Referencing also the views of other researchers, the
following paragraph briefly introduces the constituents of water in concrete.

Water essentially exists as evaporable and non-evaporable water in the harden cement
pastes. Non-evaporable water is also called the chemically bound water resulting from
the hydration process, which will not be liberated upon heating to higher than 550 ℃
(Bažant and Jirásek, 2018). Evaporable water could be categorised into three phases,
namely the capillary water, water vapor and adsorbed water. The mixture of gaseous
and liquid water (vapor and capillary water) is referred to as the moisture in concrete
pores, while free pore water refers to the capillary and adsorbed water. Adsorbed water
could be further classified as free adsorbed water and hindered adsorbed water, which
is also referred to respectively as physically bound water (Davie et al., 2006) and gel
water (Ichikawa, 2000, Liu et al., 2018a) by some researchers. The release of adsorbed
water is initiated from around 175 ℃ (Lu, 2015). The distribution of temperature and
moisture content is assumed to be uniform initially. The idealized sketch of evaporable
water by Bažant and Jirásek (2018) could be found in Figure 2.9. Based on this

15
definition, the multiphase medium (concrete) could be decomposed in a way described
in Figure 2.10. For the purpose of spalling analysis, however, the decomposition of
concrete in the numerical models to be covered next is slightly different. Non-
evaporable water is considered part of the solid skeleton until released upon heating.
Evaporable water comprises only of liquid phase (capillary and adsorbed water) where
water vapour is excluded and instead grouped into the gaseous phase with dry air. The
concept of representative elementary volume (REV) of porous medium (Figure 2.11)
is introduced (Lewis and Schrefler, 1998, Gawin and Pesavento, 2012).

Bažant and Thonguthai (1979) laid the first theoretical foundation of a thermo-hygro
model, considering only the coupled heat and moisture transfer in porous solids. The
mass flux of moisture, 𝒥, is in the same form of Darcy’s Law, driven by P, the pore
pressure. They then justified when P is interpreted as the pressure of water vapour
rather than the liquid (capillary) water, Darcy’s Law could be extended to the heated
non-saturated concrete. The model is then analysed with finite element method to obtain
the moisture state of nuclear containment vessels. Bažant and Thonguthai reported in
their work that irregular moisture movement was expected in region where pressure
gradient was opposite to the temperature gradient. These irregular movement of
moisture may give rise to oscillations in pore pressure and render convergence hard to
achieve. Though being relatively simple (no consideration of mechanical effect) and
having limitations (treatment of water as a single-phase fluid, no incorporation of
spalling), Bažant and Thonguthai’s fomulation has been the pioneer work and inspired
numerous upcoming progress in numerical modelling of concrete behaviour at elevated
temperatrues.

Ichikawa and England (2000, 2004) later proposed a one-dimensional heat and moisture
transfer model to predict pore pressure and spalling. The release of gel water and
chemically bound water due to dehydration mainly depends on temperature. Their
model is validated with both long and short-term experimental tests. However, the
governing equations of this model are uncoupled, treating the effect of heat and mass
transfer separately. By assuming rigid solid skeleton and negligible elastic strain, the
mechanical effect (tensile stress) is indirectly computed with the hollow spherical

16
model shown in Figure 2.12. The maximum tensile stress (σt,max ) could be evaluated
by pore pressure and pore volume at the radius of a void. Spalling is assumed to occur
when σt,max exceeds the tensile strength of concrete. This model fails to consider the
effect of thermal stress, which has already demonstrated to play a significant role in
fire-induced spalling. The numerical solution scheme adopted by Ichikawa is the
explicit finite difference scheme. Sufficiently small temporal and spatial steps are
required to attain convergence and stability. Based on the author’s reproduction,
however, oscillations of pore pressure may result even using the exact same parameters
provided in the publication. This phenomenon may be explained by the findings of
Bažant and Thonguthai (1979) discussed above. Liu and Zhang (2019) presented a
simplified version of Ichikawa’s model, with minor modification to the cementitious
content containing silica fume. Still, no additional coupling or mechanical effects are
considered in Liu and Zhang’s model.

Lewis et al. (1986, 1987, 1998) were among the earliest to establish a full numerical
model of heat and moisture transfer in partially saturated multiphase medium, taking
advantages of the hybrid mixture theory (HMT) originally proposed by Hassanizadeh
and Gray (1979a, 1979b, 1980). Their work was originally designated for the analysis
of static and dynamic deformation and consolidation of porous media, mainly the
geomaterials. Gawin et al. (1999) was the first, to the best of the author’s knowledge,
to apply the model into the analysis of concrete (building materials) behaviour at
elevated temperatures and prediction of spalling. Comprehensive thermo-hygro-
mechanical (THM) models have since then been developed. Most of such models either
evolved from the “Padua model” or “Glasgow model”. The author here continued the
reference adopted by Burgh (2016) in his dissertation, reasoning that the models have
been mainly developed at the University of Padua and the University of Glasgow.

The Padua model, originally called HITECOSP (high temperature concrete spalling)
then later the HITMCOSP2 and COMES-HTC (Gawin and Pesavento, 2012), is the
product of a multinational programme of research (Khoury, 2000). This model is
characterized by four macroscopic governing equations, expressed with four chosen
primary state variables, i.e., gas pressure 𝑃𝑔 , capillary pressure 𝑃𝑐 , temperature 𝑇, and

17
displacement vector of the solid matrix u. The first two governing equations are the
mass conservation of the dry air and water species (liquid water and vapour),
incorporated with the mass balance of solid skeleton. The third equation is the energy
conservation of the multiphase system, taking into account the process of vaporization
and dehydration. The final governing equation is the linear momentum balance of the
multiphase medium, expressed in terms of total stresses. Material deterioration and
non-linearities are also accounted for. The final set of governing equations is discretised
with finite element in space and finite difference in time and solved for numerical
solutions. The achievements of Gawin and the co-workers are without doubt
considerably substantial (Majorana et al., 2010, Gawin et al., 2011a, Gawin et al., 2011b,
Dal Pont et al., 2011, Gawin et al., 2005a, Gawin et al., 2005b, Gawin et al., 2006a,
Gawin et al., 2006b, Witek et al., 2007, Pesavento et al., 2008, Khoury et al., 2002). It
remains to be one of the most, if not the only, well-studied and used models.

Another numerical model that has aroused great interest is the Glasgow model, which
is built by Davie et al. (2006) on the basis of Tenchev and his co-workers (2001, 2005).
The original Glasgow model was a coupled thermo-hygro model considering the mass
conservation of dry air and moisture, and the energy conservation of the system. The
chosen primary state variables were temperature T, gas pressure PG , and mass of vapour
~
per unit volume of gaseous material ρ𝑣 . The most critical limitation of Tenchev’s model
was that it assumed the pore pressures of liquid and gases were equal, i.e., the capillary
pressure was neglected. The Glasgow model extended and improved the original
formulation by explicitly accounting for the capillary pressure as well as the adsorbed
free water. It was later updated by its developer (Davie et al., 2010) to include the
mechanical effect (linear momentum balance), following the work of Gray and
Schrefler (2001). The Glasgow model then investigated a continuum damage model to
predict spalling (Davie et al., 2012).

Chung et al.(2003, 2005, 2006) presented a THM model largely based on the Padua
model. The coupled heat and mass transfer was solved with a finite difference model
and the resulting state variable fields were then mapped onto the finite element mesh
for a thermo-mechanical stress analysis. Chung assumed that temperature changes

18
could induce thermal-dilatational (mechanical) deformations, but not vice versa. Zeiml
(2008) also developed a thermo-hygro model for the coupled analysis of transport
process in heated concrete based on the Padua model, mechanical analysis was
separately applied to assess the structural safety of tunnel linings subjected to fire
loading. Zhang et al. (2014) proposed a similar model with the temperature-dependent
Biot’s coefficient introduced. Later a fast-estimation method for spalling assessment
was suggested (Zhang et al., 2017), employing a 2D axisymmetric model to represent
the plate-like structures (walls, slabs, tunnel linings). ElMohandes (2013) implemented
an algorithm for coupled heat and moisture transfer in a finite element (FE) programme
previously developed by the University of Toronto. This algorithm, varied marginally
from the Glasgow model, served extended analysis capabilities under extreme loading
conditions (thermally-induced spalling) for the FE programme. Burgh (2016)
developed a pseudo 1D coupled THM model incorporating a mechanical hybrid fibre
model to evaluate spalling risk. The Padua model, once again, was chosen as the starting
point of thermo-hygro formulation in Burgh’s study.

The aforementioned models are some of the numerous descendants of the Padua and
Glasgow model, among which they are relatively well-deduced. There have been,
needless to say, plenty of models that deviate from the Padua and Glasgow model
arising from various works. However, most of them are not sufficiently rigorous. For
example, Dwaikat and Kodur (2009) proposed a simplified 1D hydrothermal model to
predict fire-induced spalling for practical purposes, which was incorporated in a FE
model to trace the fire response of reinforced concrete elements. This model treated the
pore pressure as the main trigger of spalling, ignoring the effects of thermal stress. The
applicability of this model was further hampered by the many simplified assumptions
involved, such as the presence of dry air and mobility of liquid water were totally
ignored. Lu (2015) later in his dissertation elaborated on this hydrothermal model by
introducing a spalling criterion to indicate fire-induced spalling. The formulation of
governing equations accounting for heat and moisture transfer remained largely
unchanged. However, the results of case studies conducted by Lu have shown
insensitive responses of pore pressure to heating rate, which was in contrast to the
conclusion drawn by Dwaikat and Kodur and the experimental observations by

19
Klingsch (2014). The reliability of this model was questioned to be restricted to well-
callibratd scenarios (Burgh, 2016). A simpliefied approach built solely depending on
experimental observations was proposed by Kodur et al. (2004). Though the approach
was easy to apply in modelling (Kodur and Dwaikat, 2008c, Dwaikat and Kodur, 2008),
the prediciton of spalling was, in a sense, very crude. It worked basically in like manner
as the design codes, therefore, the efficacy was as rigid and resctritive as the
prescriptive methods.

So far, all the models reviewed are formulated in macro-level, meaning the concrete is
studied as a homogeneous continuum. A minority of researches have attempted to
tackle the problem from a different prospective, by means of the localised description
of concrete components. In a meso-level numeircal investigaiton of spalling conducted
by Zhao (2012), concrete was modelled as a two-phase composite, i.e., aggregate and
matrix, neglecting the interfacial transistion zone. Her work was hightlighted by the
prediction of thermal decomposition of hardened cement paste, based on which the
permeability of silica fume blended system was determined by a three-step prediction.
Details missing in a macro-level model, such as the aggregate influence on moisture
transport, damage induced by the volume mismatch between aggregates and matrix,
could be addressed in a meso-level model. Lottman (2017) developed an conceptual
FEM model focusing on the coupled temperature and pore pressure, and fracture
mechanics. Concrete was explicitly modelled as aggregate particals, surrounding
mortar and the interfacial transistion zone. Inclusion of local material directions
rendered the description of concrete material anisotropic and heterogeneous. At the
same time, the orientation of crack pattern obtained by the fracture mechanics model
was also included. Nonetheless, it can also be seen that the development of meso-level
model suffers from a lot of aspects, including the lack of experimental data (Zhao,
2012), the unclear origin of various phenomena (Burgh, 2016), high numerical costs
and instability (Lottman, 2017).

20
2.4 Protective Measures

As has been briefly mentioned in Section 2.3.2, for high strength concrete, Eurocode 2
Part 1-2 (2004) recommends at least one of the following four protective measures
should be provided to prevent spalling:

• Method A: A reinforcement mesh with a nominal cover of 15 mm. This mesh

should have wires with a diameter ≥ 2 mm with a pitch ≤ 50 x 50 mm. The
nominal cover to the main reinforcement should be ≥ 40 mm.
• Method B: A type of concrete for which it has been demonstrated (by local
experience or by testing) that no spalling of concrete occurs under fire exposure.
• Method C: Protective layers for which it is demonstrated that no spalling of
concrete occurs under fire exposure.
• Method D: Include in the concrete mix more than 2 kg/m3 of monofilament
propylene fibres.

In addition to the general codes of practice, many other protective measures have been
proposed throughout the history of study in spalling. The following paragraphs present
concisely the most suggested measures, commenting on the pros and cons that each
measure possibly have.

Addition of Fibres

The addition of fibres to concrete mix is of no doubt the most extensively investigated
measure of all, especially the polypropylene (PP) fibres. PP fibres melt at a relatively
low temperature between 160 ℃ to 170 ℃ (Khaliq and Kodur, 2013, Kodur and Khaliq,
2011, Ali et al., 2004). When temperature is above this range, the volume provided by
melted PP fibres could serve as “channels” through which vapour pressure would be
dissipated. Therefore, PP fibres mainly take effect in preventing the pore pressure-
induced spalling by increasing the permeability. It was also reported that PP fibres may
potentially improve tensile strength by minimizing cracking in concrete (Khaliq and
Kodur, 2012). But some researchers argued that PP fibres may not prevent spalling in

21
expansive UHSC (Khoury, 2000), structural elements subjected compressive loads or
drilled tunnels (Hertz, 2003).

The melting point of steel fibres is at around 1300 ℃, considerably higher than PP fibres,
therefore, its role in releasing pore pressure is minimal. Instead, steel fibres are believed
to increase the tensile strength and ductility of concrete by restricting the initiation and
propagation of cracking due to their high tensile resistance (Chen and Liu, 2004).
However, Schneider and Horvath (2003) deemed it a “unreliable method”. Though steel
fibres are beneficial in delaying the process in time, the explosion would be much more
violent when spalling actually happens (Jansson, 2013). Khoury et al. (2007) attributed
this phenomenon to the release of energy build-up resulting from the increased tensile
strength. Fibres of less reported benefits in preventing spalling when used alone are not
discussed in this thesis (Klingsch, 2014, Liu et al., 2018b, Lee et al., 2012).

The effectiveness of hybrid fibres is studied with different fibre choices and numerous
combinations. Generally, hybrid fibres have been reported to behave better than the
single-fibres counterpart in reducing the risk of spalling (Liu et al., 2018b, Chen and
Liu, 2004, Peng et al., 2006, Lee et al., 2012). The influence of fibres on various
concrete types is also examined, including but not limited to, the self-consolidating
concrete (SCC), fly ash concrete (FAC), high strength fly ash concrete (HFAC), ultra-
high performance strain hardening cementitious composite (UHP-SHCC), lightweight
concrete and engineered cementitious composite (ECC) (Kodur and Khaliq, 2011,
Khaliq and Kodur, 2012, Liu et al., 2018b, Bilodeau et al., 2004, Liu et al., 2018c).
Though having been extensively investigated, there are still two key areas that remain
uncertain within the method of fibre addition (Khoury, 2000). The first is the exact
mechanism of the influence from fibres on concrete; the other is the optimization
problem in practical application (type, dosage and sizes of fibres with regard to the
applied loading and concrete strength).

Air-Entraining Agent

Another measure that has proved to be effective in preventing spalling is to employ air-
entraining agent, which increases the total pore volume and reduces the degree of

22
saturation in pores by generating air voids of size greater than the typical capillary pores
(Connolly, 1995). Additional pore volume facilitates the increase of permeability,
therefore, mitigating pore pressure and lowering the risk of spalling. Nonetheless, the
diameter of air voids is approximately the same with silica fume, which is a common
constituent of concrete material. This hindered the application of air-entraining agent
in high and ultra-high-performance concrete (Klingsch, 2014). Moreover, the
introduction of air voids is prone to reduce the strength and durability (Khoury, 2000,
Van der Merwe, 2019). Adopting such an agent potentially compromises the expected
performance of concrete.

Concrete Mixes

Evidence has shown that the likelihood and extent of spalling are related to the type and
size of aggregate (Noumowe et al., 2009, Xing et al., 2011, Kodur and McGrath, 2003,
Kodur et al., 2005). Carbonate aggregate is recommended due to higher specific heat
and lower thermal expansion at elevated temperature compared to siliceous aggregate.
Lightweight concrete may suffer from higher thermal gradient and be more susceptible
to violent spalling when the moisture content is high. The combined use of larger-size
aggregates and fibres could lead to synergy effect in increasing permeability and
preventing spalling (Li et al., 2019). But this observation is in contradiction to Khoury
(2000)’s suggestion to choose small-size aggregate. Curing condition of fresh concrete
could be an influential factor because significant spalling was reported when relative
humidity is higher than 80%. Concrete cured in a more humid environment would have
a higher initial moisture, inducing a higher vapour pressure during heating.

Structural Design

Presence of main reinforcement can obviously limit the spread of spalling. Furthermore,
the provision, configuration and spacing of transverse reinforcement all play a role in
enhancing the fire resistance of structural elements (Kodur et al., 2005, Kodur et al.,
2004, Kodur and McGrath, 2006, Shah et al., 2017). It is recommended to provide
lapped cross ties with the ends bent 135 ℃ into the section, instead of the normal 90 ℃
hook. Reduce the tie spacing by using confinement of smaller diameter rather than

23
larger spacing with larger diameter. However, this sometimes could be challenging to
implement in small and narrow sections. Last but not least, thicker concrete covers, and
thicker sections for I-beams or ribbed sections, also help to minimize spalling.

Thermal Barriers

Thermal barriers, or protective barriers, refer to the panels, plates, boards, mortar or
insulating coating covering on the external surface of the concrete element (Van der
Merwe, 2019). They could prevent the protected concrete element from direct contact
with the heat source. Hence, thermal barriers serve to delay the heat transfer at elevated
temperature, reducing the heating rate, thermal gradient and deterioration of concrete
material. It should be noted, however, that currently no standard specifications are
available for the design of thermal barriers. Experimental investigations on HPC slabs
protected with thermal barriers were carried out by Lu (2015). Different thicknesses
(10 mm, 20 mm and 30 mm) of protective mortar were tested under the standard ISO
fire scenario for two hours, but no reasons were given for the choice of the such
thicknesses. Explosive spalling was reported at a deeper depth than the unprotected slab
for the case of 10 mm thickness, though the time of occurrence was indeed postponed.
Besides, the unprotected slab spalled in a gradual, layer-by-layer manner, while the slab
with a 10 mm protective mortar spalled with a single violent explosion. On the other
hand, thicknesses of 20 mm and 30 mm proved to be adequate and no spalling was
observed. A spalling model was then applied to simulate the situations where slabs were
subjected to hydrocarbon fire for two hours. The predicted results showed that the
thickness of 20 mm may just provide a minimum safety margin. As such, a slightly
larger value of 25 mm was suggested to be used in a hydrocarbon fire. Based on the
experimental and simulated results, Lu concluded that the thickness of protective
mortar was critical to the spalling prevention. He further commented that thickness of
30 mm may be too conservative for a two-hour exposure of ISO fire. Similar
investigations were also conducted on protective plates, which are easy to install and
more applicable to larger area of flat concrete. In practical applications, the protection
level (choice of thickness) should depend correspondingly on the project requirements
and budget.

24
 Table 2.1 Types and characteristics of spalling
Type Occurrence time Nature Sound Influence Governing factors
Material and
Aggregate 7 - 30 min Splitting Popping Superficial
thermal properties
Material, thermal
Potentially
Corner 30 - 90 min Non-violent None and structural
serious
properties
Potentially Material and
Surface 7 - 30 min Violent Cracking
serious thermal properties
Material, thermal
Loud
Explosive 7 - 30 min Violent Serious and structural
bang
properties
Material, thermal
When concrete Potentially
Sloughing-off Non-violent None and structural
weakens serious
properties
Material, thermal
During or after Potentially
Post-cooling Non-violent None and structural
cooling serious
properties

25
Figure 2.1 Pore pressure induced-spalling by Liu et al. (2018a)

26
Figure 2.2 Thermal stress-induced spalling by Liu et al. (2018a)

27
Figure 2.3 Combined pore pressure and thermal stress spalling by Klingsch (2014)

28
Figure 2.4 Fire assessment methods and design process, after Khoury (2000)

29
Figure 2.5 Various fire scenarios reproduced from curves by Lottman (2017)

Figure 2.6 Comparison between nomograms for assessment of spalling

30
(a)

(b)

Figure 2.7 Nomograms of moisture content and applied level of stress by (a)
Sertmehmetoglu (1977) and (b) Zhukov (1994)

31
Figure 2.8 Nomograms of compressive stress and element thickness by
Sertmehmetoglu (1977)

Figure 2.9 Idealized sketch of evaporable water

32
Figure 2.10 Decomposition of concrete as multiphase medium

33
Figure 2.11 Representative elementary volume of porous medium by Gawin and
Pesavento (2012)

Figure 2.12 Hollow spherical model of solid skeleton by Ichikawa (2000)

34
3 CONCRETE EXPOSED TO ELEVATED TEMPERATURE

3.1 Introduction

Prior to the development of the thermo-hygro-mechanical (THM) model presented in

Chapter 4 and Chapter 5, the thermal and mechanical properties of concrete subjected
to high temperature are first discussed in this chapter. Most of these properties are
temperature-dependent and are evaluated to be employed in the THM model. In
addition, some of the thermal properties of water, vapour and air are also discussed in
this chapter, as they play an important role in the modelling of the thermo-hygro
behaviour of concrete at elevated temperature.

3.2 Thermal Properties of Concrete

3.2.1 Thermal Conductivity

The thermal conductivity, 𝑘, is the measure of a material’s ability to conduct heat. It’s
the main material-related coefficient in the heat conduction transfer (Fourier’s law).
Concrete, having a relatively low thermal conductivity compared to steel, is therefore
treated as a protective cover for the reinforcement. The thermal conductivity of concrete
at elevated temperature has been extensively investigated in the past decades, with
design codes as well as the academic publications providing both the experimental data
and analytical models for evaluation. Most of the studies calculate the thermal
conductivity based on the types of aggregate, because the majority of concrete volume
is occupied by aggregate. Following are a few of the most commonly used models from
design codes or literature.

Eurocode 2 (ENV1992-1-2:1996, 1996, BSEN1992-1-2:2004, 2004)

In the earlier version of Eurocode 2, the thermal conductivity is determined according

to aggregate type:

35
Siliceous aggregate concrete:

𝑇𝑐 𝑇𝑐 2
𝑘 = 2.0 − 0.24 ( ) + 0.012 ( ) 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 1200℃ (3.1)
120 120

Calcareous aggregate concrete:

𝑇𝑐 𝑇𝑐 2
𝑘 = 1.6 − 0.16 ( ) + 0.008 ( ) 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 1200℃ (3.2)
120 120

Light-weight aggregate concrete:

𝑇𝑐
1.0 − 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 800℃
𝑘={ 1600 (3.3)
0.5 𝑓𝑜𝑟 800℃ < 𝑇𝑐 ≤ 1200℃

In the later (current) version, the upper and lower limit values of thermal conductivity
are given:

𝑇𝑐 𝑇𝑐 2
2.0 − 0.2451 ( ) + 0.0107 ( ) 𝑢𝑝𝑝𝑒𝑟 𝑙𝑖𝑚𝑖𝑡
𝑘= 100 100 (3.4)
𝑇𝑐 𝑇𝑐 2
1.36 − 0.136 ( ) + 0.0057 ( ) 𝑙𝑜𝑤𝑒𝑟 𝑙𝑖𝑚𝑖𝑡
{ 100 100

ASCE Manual of Practice (Lie, 1992)

Siliceous aggregate concrete:

−0.625 × 10−3 𝑇𝑐 + 1.5 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 800℃

𝑘={ (3.5)
1.0 𝑓𝑜𝑟 𝑇𝑐 > 800℃

Calcareous aggregate concrete:

1.355 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 293℃

𝑘={ −3 (3.6)
−1.241 × 10 𝑇𝑐 + 1.7162 𝑓𝑜𝑟 𝑇𝑐 > 293℃

36
Pure quartz aggregate concrete:

−0.850 × 10−3 𝑇𝑐 + 1.9 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 800℃

𝑘={ (3.7)
1.22 𝑓𝑜𝑟 𝑇𝑐 > 800℃

Light-weight aggregate concrete:

−0.39583 × 10−3 𝑇𝑐 + 0.925 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 600℃

𝑘={ (3.8)
0.6876 𝑓𝑜𝑟 𝑇𝑐 > 600℃

Models of Lie and Kodur (1996)

The following models are given for the steel-fibre reinforced concrete, which has a
slightly higher thermal conductivity than the normal concrete due to the presence of
steel fibre.

Steel-fibre reinforced concrete with siliceous aggregates:

3.22 − 0.007𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 200℃

𝑘 = {2.24 − 0.0021𝑇𝑐 𝑓𝑜𝑟 200℃ < 𝑇𝑐 ≤ 400 (3.9)
1.4 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 1000℃

Steel-fibre reinforced concrete with carbonate aggregates:

2.000 − 0.001775𝑇𝑐 𝑓𝑜𝑟 0℃ < 𝑇𝑐 ≤ 500℃

𝑘={ (3.10)
1.402 − 0.000579𝑇𝑐 𝑓𝑜𝑟 500℃ < 𝑇𝑐 ≤ 1000℃

Models of Kodur and Sultan (2003)

Siliceous aggregate high strength concrete:

𝑘 = 2.00 − 0.0011𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 1000℃ (3.11)

Carbonate aggregate high strength concrete:

37
 2.00 − 0.0013𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 300℃
𝑘={ (3.12)
2.21 − 0.0020𝑇𝑐 𝑓𝑜𝑟 300℃ < 𝑇𝑐 ≤ 1000℃

Steel fibre-reinforced high strength with siliceous aggregates:

2.50 − 0.0034𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 200℃

𝑘 = {2.24 − 0.0021𝑇𝑐 𝑓𝑜𝑟 200℃ < 𝑇𝑐 ≤ 400 (3.13)
1.4 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 1000℃

Steel fibre-reinforced high strength with carbonate aggregates:

1.80 − 0.0016𝑇𝑐 𝑓𝑜𝑟 0℃ < 𝑇𝑐 ≤ 500℃

𝑘={ (3.14)
1.20 − 0.0004𝑇𝑐 𝑓𝑜𝑟 500℃ < 𝑇𝑐 ≤ 1000℃

Models of Kodur and Khaliq (2011)

High strength concrete:

2.5 − 0.0033𝑇𝑐 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 400℃

𝑘={ (3.15)
2.3 − 0.0020𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 800℃

High-strength self-consolidating concrete (SCC):

3.12 − 0.0045𝑇𝑐 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 400℃

𝑘={ (3.16)
3.00 − 0.0025𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 800℃
High strength fly ash concrete (FAC):

3.0 − 0.0045𝑇𝑐 𝑓𝑜𝑟 20℃ < 𝑇𝑐 ≤ 400℃

𝑘={ (3.17)
2.6 − 0.0025𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 800℃

Model of Shin et al. (2002)

Shin et al. measured the thermal conductivity of concrete used in Korea (similar in
chemical composition to the US basaltic concrete) and proposed a parabolic expression:

38
 𝑘 = 1.36469 × 10−6 𝑇𝑐 2 − 0.00256908𝑇𝑐 + 2.24266 (3.18)

3.2.2 Density

The temperature dependent density of concrete is mainly used for the evaluation of
volumetric heat capacity in heat and moisture transfer analysis. The density of concrete
would normally decrease with the increase of temperature. In the earlier stage of heating,
the decline of density is primarily attributed to the loss of evaporable free water between
100 ℃ and 150 ℃ , and therefore is proportional to the initial moisture content of
concrete. Later, the decrease is mainly accounted for by the loss of chemically bound
water. At temperatures up to around 600℃, the subsequent loss results mainly from the
dehydration of hydration products, and so far, the change of density is relatively minor.
But once the temperature goes beyond 600℃, the density would experience a sharp
decrease. For concrete with calcareous aggregate, the abrupt change is caused by the
decarbonization of calcium carbonate, while for the siliceous aggregate this is due to
the vast thermal expansion that quartz undergoes at high temperatures. Some of the
temperature dependent density models are summarized in the following.

Eurocode 2 (BSEN1992-1-2:2004, 2004)

For all aggregate types:

1 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 115℃

(𝑇𝑐 − 115)
1 − 0.02 𝑓𝑜𝑟 115℃ < 𝑇𝑐 ≤ 200℃
𝜌(𝑇𝑐 ) 85
= 𝑇𝑐 − 200 (3.19)
𝜌(20℃) 0.98 − 0.03 𝑓𝑜𝑟 200℃ < 𝑇𝑐 ≤ 400℃
200
𝑇𝑐 − 400
{0.95 − 0.07 800 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 1200℃

where 𝜌(20℃) is the density at room temperature.

Models of Lie and Kodur (1996)

39
Instead of directly measuring the density of concrete, Lie and Kodur chose to measure
the loss of concrete mass and gave the following formulations:

Steel-fibre reinforced concrete with siliceous aggregates:

𝑀(𝑇𝑐 )
= 0.9987 − 3.992 × 10−5 𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 1000℃ (3.20)
𝑀0

Steel-fibre reinforced concrete with carbonate aggregates:

1 − 6.5 × 10−5 𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 700℃

𝑀(𝑇𝑐 )
= {2.6 − 0.002235𝑇𝑐 𝑓𝑜𝑟 700℃ < 𝑇𝑐 ≤ 800℃ (3.21)
𝑀0
0.72 − 1.5 × 10−5 𝑇𝑐 𝑓𝑜𝑟 800℃ < 𝑇𝑐 ≤ 1000℃

where 𝑀(𝑇𝑐 ) and 𝑀0 is the concrete mass at temperature 𝑇𝑐 and room temperature.

Model of Shin et al. (2002)

Again, Shin et al. measured the density of concrete used in Korea (similar in chemical
composition to the US basaltic concrete) and proposed a parabolic expression:

𝜌 = 1.89575 × 10−4 𝑇𝑐 2 − 0.39802𝑇𝑐 + 2259.62 (3.22)

Models of Kodur and Sultan (2003)

Siliceous aggregate high strength concrete:

𝑀(𝑇𝑐 )
= 1 − 5 × 10−5 𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 1000℃ (3.23)
𝑀0

Carbonate aggregate high strength concrete:

1.003 − 6 × 10−5 𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 600℃

𝑀(𝑇𝑐 )
= {2.551 − 0.00264𝑇𝑐 𝑓𝑜𝑟 600℃ < 𝑇𝑐 ≤ 700℃ (3.24)
𝑀0
0.71 − 1 × 10−5 𝑇𝑐 𝑓𝑜𝑟 700℃ < 𝑇𝑐 ≤ 1000℃

40
Steel fibre-reinforced high strength with siliceous aggregates:

𝑀(𝑇𝑐 )
= 1 − 4 × 10−5 𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 1000℃ (3.25)
𝑀0

Steel fibre-reinforced high strength with carbonate aggregates:

1.003 − 6 × 10−5 𝑇𝑐 𝑓𝑜𝑟 0℃ ≤ 𝑇𝑐 ≤ 700℃

𝑀(𝑇𝑐 )
= {2.214 − 0.00179𝑇𝑐 𝑓𝑜𝑟 700℃ < 𝑇𝑐 ≤ 785℃ (3.26)
𝑀0
0.817 − 1 × 10−5 𝑇𝑐 𝑓𝑜𝑟 785℃ < 𝑇𝑐 ≤ 1000℃

3.2.3 Specific Heat Capacity

Specific heat capacity is defined as the amount of energy needed to cause a unit increase
of temperature of a unit mass substance. In heat and mass transfer analysis, the specific
heat of concrete is usually present together with the density as the volumetric heat
capacity, ρCp . Therefore, some researchers attempted to investigate the temperature
dependence of their product as a whole rather than the separate influences. Opposite to
the trend of density, specific heat tends to increase with temperature. This could be
explained by fact that as temperature rises, the physical and chemical reactions within
concrete would consume energy in the form of heat. Most of the reactions, such as the
evaporation of free water between 105 ℃ and 150 ℃ , followed by the release of
chemically bound water (dehydration of hydrated products), and later the
decarbonization of carbon hydroxide over 800℃, are all endothermic. It is therefore no
surprise to find that the specific heat capacity is influenced by the moisture content and
types of aggregate.

Eurocode 2 (BSEN1992-1-2:2004, 2004)

The Eurocode 2 evaluated the specific heat of concrete separately from the density, with
regard to moisture content (MC). Different peak values at 115℃ corresponding to
different moisture content are considered to account for the evaporation of free water.
The expression is the same for both siliceous and calcareous aggregate:

41
 900 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 100℃
900 + (𝑇𝑐 − 100) 𝑓𝑜𝑟 100℃ < 𝑇𝑐 ≤ 200℃
Cp (𝑇𝑐 ) = 𝑇𝑐 − 200 (3.27)
1000 + 𝑓𝑜𝑟 200℃ < 𝑇𝑐 ≤ 400℃
2
{1100 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 1200℃

900 𝑓𝑜𝑟 𝑀𝐶 = 0%
with Cp,peak = {1470 𝑓𝑜𝑟 𝑀𝐶 = 1.5% 𝑎𝑡 𝑇𝑐 = 115℃ (3.28)
2020 𝑓𝑜𝑟 𝑀𝐶 = 3.0%

ASCE Manual of Practice (Lie, 1992)

Siliceous aggregate concrete:

0.005𝑇𝑐 + 1.7 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 200℃

2.7 𝑓𝑜𝑟 200℃ < 𝑇𝑐 ≤ 400℃
ρCp = 0.013𝑇𝑐 − 2.5 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 500℃ (3.29)
10.5 − 0.013𝑇𝑐 𝑓𝑜𝑟 500℃ < 𝑇𝑐 ≤ 600℃
{2.7 𝑓𝑜𝑟 600℃ < 𝑇𝑐

Carbonate aggregate concrete:

2.566 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

0.1765𝑇𝑐 − 68.034 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 410℃
25.00671 − 0.05043𝑇𝑐 𝑓𝑜𝑟 410℃ < 𝑇𝑐 ≤ 445℃
2.566 𝑓𝑜𝑟 445℃ < 𝑇𝑐 ≤ 500℃
ρCp = (3.30)
0.01603𝑇𝑐 − 5.44881 𝑓𝑜𝑟 500℃ < 𝑇𝑐 ≤ 635℃
0.16635𝑇𝑐 − 100.90225 𝑓𝑜𝑟 635℃ < 𝑇𝑐 ≤ 715℃
176.07343 − 0.22103𝑇𝑐 𝑓𝑜𝑟 715℃ < 𝑇𝑐 ≤ 785℃
{2.566 𝑓𝑜𝑟 785℃ < 𝑇𝑐

Light-weight aggregate concrete:

42
 1.930 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃
0.0772𝑇𝑐 − 28.95 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 420℃
46.706 − 0.1029𝑇𝑐 𝑓𝑜𝑟 420℃ < 𝑇𝑐 ≤ 435℃
ρCp = 1.930 𝑓𝑜𝑟 435℃ < 𝑇𝑐 ≤ 600℃ (3.31)
0.03474𝑇𝑐 − 18.9140 𝑓𝑜𝑟 600℃ < 𝑇𝑐 ≤ 700℃
0.1737𝑇𝑐 − 126.994 𝑓𝑜𝑟 700℃ < 𝑇𝑐 ≤ 720℃
{1.930 𝑓𝑜𝑟 720℃ < 𝑇𝑐

Models of Kodur (2014)

Siliceous aggregate concrete:

0.005𝑇𝑐 + 1.7 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 200℃

2.7 𝑓𝑜𝑟 200℃ < 𝑇𝑐 ≤ 400℃
ρCp = 0.013𝑇𝑐 − 2.5 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 500℃ (3.32)
10.5 − 0.013𝑇𝑐 𝑓𝑜𝑟 500℃ < 𝑇𝑐 ≤ 600℃
{2.7 𝑓𝑜𝑟 600℃ < 𝑇𝑐 ≤ 635℃

Carbonate aggregate concrete:

2.45 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

0.026𝑇𝑐 − 18.25 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 475℃
0.0143𝑇𝑐 − 6.295 𝑓𝑜𝑟 475℃ < 𝑇𝑐 ≤ 650℃
ρCp = (3.33)
0.1894𝑇𝑐 − 120.11 𝑓𝑜𝑟 650℃ < 𝑇𝑐 ≤ 735℃
−0.263𝑇𝑐 − 212.4 𝑓𝑜𝑟 735℃ < 𝑇𝑐 ≤ 800℃
{2 𝑓𝑜𝑟 800℃ < 𝑇𝑐 ≤ 1000℃

Model of Shin et al. (2002)

Instead of formulating the specific heat, Shin et al. gave a parabolic expression for the
thermal diffusivity, which is defined as α = 𝑘/𝜌Cp . Hence, (3.18) and (3.22) must be
used simultaneously with (3.34) to find the specific heat:

α = 9.1639 × 10−7 𝑇𝑐 2 − 0.00136982𝑇𝑐 + 0.909062 (3.34)

Models of Kodur and Khaliq (2011)

43
High strength concrete without PP fibres:

2.4 + 0.0002𝑇𝑐 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

Cp = { (3.35)
1 + 0.0043𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 800℃

High strength concrete with PP fibres:

2.4 + 0.0002𝑇𝑐 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

Cp = {1 + 0.0043𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 650℃ (3.36)
9.1 − 0.009𝑇𝑐 𝑓𝑜𝑟 650℃ < 𝑇𝑐 ≤ 800℃

High-strength self-consolidating concrete (SCC) without PP fibres:

2.4 + 0.0001𝑇𝑐 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

Cp = { (3.37)
0.6 + 0.006𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 800℃

High-strength self-consolidating concrete (SCC) with PP fibres:

2.4 + 0.0001𝑇𝑐 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

Cp = {0.6 + 0.006𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 650℃ (3.38)
10.6 − 0.01𝑇𝑐 𝑓𝑜𝑟 650℃ < 𝑇𝑐 ≤ 800℃
High strength fly ash concrete (FAC):

2.7 + 0.0004𝑇𝑐 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 400℃

Cp = {0.3 + 0.0065𝑇𝑐 𝑓𝑜𝑟 400℃ < 𝑇𝑐 ≤ 600℃ (3.39)
3.6 + 0.0014𝑇𝑐 𝑓𝑜𝑟 600℃ < 𝑇𝑐 ≤ 800℃

3.2.4 Porosity

The porosity of concrete is defined as the fractional ratio of pore volume to the total
volume of concrete. Pore volume is the volume occupied by liquid water and moist air,
and total volume is the sum of pore volume and volume of solid skeleton. The change
of porosity largely results from the released space due to the dehydration and
decomposition of compound at elevated temperature. The significance of determining

44
porosity at high temperature mainly lies in its connection to other properties, such as
the permeability and sorption isotherm (degree of saturation of free water).

Bazant and Thonguthai (1978) estimated the variation of porosity, considering the
release of free water by dehydration at a risen temperature 𝑇:

𝑤𝑑 (𝑇)
𝜙 = (𝜙0 + ) [1 + 0.12(ℎ − 1.04)] (3.40)
𝜌𝑙0

where 𝜙0 is the initial porosity, 𝑤𝑑 (𝑇) is the amount of free water released into pores,
𝜌𝑙0 is the water density at room temperature and atmospheric pressure, and ℎ is the
relative humidity.

Gawin et al. (1999) proposed a linear expression based on experimental results of

various types of concrete, relating temperature to the porosity of concrete at high
temperature:

𝜙 = 𝜙0 + 𝐴𝜙 (𝑇 − 𝑇0 ) (3.41)

where 𝐴𝜙 is a constant parameter dependent on the concrete type.

Feraille-Fresnet et al. (2003), based on their own experimental results, also used a linear
relation to postulate the change of porosity regarding to the dehydrated water content:

𝜙 = 𝜙0 + (0.72 × 10−3 𝑚3 /𝑘𝑔)𝑤𝑑 (𝑇) (3.42)

where 𝑤𝑑 (𝑇) is the dehydrated water content, the same as that in (3.40).

Toumi and Resheidat (2010) developed the following piecewise function to describe
the variation of porosity with respect to temperature:

0.0131 × 𝑇𝑐 + 5.6929 𝑇𝑐 ≤ 500℃

𝜙={ (3.43)
1.3978 × e0.0037𝑇𝑐 𝑇𝑐 ≥ 500℃

45
Tenchev et al. (2001) used a cubic function to represent the threefold increase of
porosity reported in Luo et al. (2000) such that the first derivatives of porosity are
continuous functions:

1 𝑇𝑐 < 100℃
𝜙 = 𝜙0 × {𝑎𝑇𝑐 3 + 𝑏𝑇𝑐 2 + 𝐶𝑇𝑐 + 𝑑 100℃ < 𝑇𝑐 ≤ 800℃ (3.44)
3 𝑇𝑐 > 800℃

where initial porosity 𝜙0 = 0.08, a = −1.166180758 × 10−8 , b = 1.574344023 ×

10−5 , c = −2.79833819 × 10−3 , d = 1.134110787.

3.2.5 Permeability

The permeability of concrete depends largely on the porosity discussed above and the
pore structure of concrete. The increased porosity and formation of microcracks at high
temperature subsequently lead to the increase of permeability. The variation of
permeability is relatively trivial at temperatures below 100℃, when compared to the
abrupt change of several orders of magnitude at temperatures over 100℃.

The original formulation by Gawin et al. (1999) related the permeability to the
temperature and pore pressure. Later, Gawin et al. (2002) extended their original
formulation by including the effect of damage evolution, referring to the expression
developed by Bary (1996):

𝑃𝑔 𝐵𝐾 𝐷 𝐴
𝐾 = 𝐾0 10 𝐴𝐾(𝑇−𝑇0 ) ( ) 10 𝑡 𝐷 (3.45)
𝑃∞

where 𝑃𝑔 and 𝑃∞ is the gaseous and atmospheric pressure, 𝐾0 is the initial permeability
at room temperature 𝑇0 , 𝐴𝐾 and 𝐵𝐾 are constant parameters, 𝐷𝑡 is the total damage
parameter and 𝐴𝐷 is the material constant (𝐴𝐷 = 4 for concrete material).

Tenchev et al. (2001) assumed the permeability was mainly dependent on the porosity:

46
 2
𝜙(𝑇) 3 (3.46)
𝐾 = 𝐾0 ( )
𝜙0

where 𝜙0 and 𝜙 is the porosity at room temperature and temperature 𝑇.

Bazant and Thonguthai (1978) proposed a rather complicated piecewise function to

capture the sudden jump of permeability at temperature around 100℃:

𝐾 𝑓 (ℎ, 𝑇)𝑓2 (𝑇) 𝑓𝑜𝑟 𝑇 ≤ 𝑇𝑡𝑟

𝐾(ℎ, 𝑇) = { 0 1 (3.47)
𝐾0 𝑓2 (𝑇𝑡𝑟 )𝑓3 (𝑇) 𝑓𝑜𝑟 𝑇 > 𝑇𝑡𝑟

1 − 𝛼(𝑇) 𝑄𝑤 1 1
with 𝑓1 (ℎ, 𝑇) = 𝛼(𝑇) + , 𝑓2 (𝑇) = exp [ ( − )] (3.48)
1 + 256⟨1 − ℎ⟩4 𝑅 𝑇0 𝑇

11 1
and 𝑓3 (𝑇) = − 4.5, 𝛼(𝑇) = (3.49)
1 + exp[−0.455(𝑇 − 𝑇𝑡𝑟 )] 1 + 0.253(𝑇𝑡𝑟 − 𝑇)

where (3.49) is the modified formulation by Bažant and Zi (2003). 𝑇𝑡𝑟 is the
temperature corresponding to the beginning of the permeability jump, ℎ is the relative
humidity, 𝛼(𝑇) is a temperature-dependent parameter, 𝑄𝑤 is the activation energy for
water migration and 𝑅 is the universal gas constant.

3.2.6 Relative Permeability

As discussed in Section 4.3, in order to apply the Darcy’s law to partially saturated
medium, the permeability of concrete has to be reduced to reflect the influence of water
and gas flow. This reduction is implemented by the inclusion of a dimensionless factor
ranging from 0 to 1, called the relative permeability. Technically speaking, this relative
permeability is not an intrinsic property of concrete, but rather an adjusting parameter
accounting for a certain phenomenon. Though the relative permeability largely depends
on the degree of saturation of free water, its presence is closely and only connected to

47
the permeability of concrete in the model. Therefore, it is elaborated in this section
other than in Section 3.4.

Luckner et al. (1989) suggested the following approximations for the relative
permeabilities to liquid water and gas (non-wetting fluid):

1 𝑚 2
𝐾𝑙 = √𝑆 [1 − (1 − 𝑆 𝑚 ) ]
(3.50)
1 2𝑚
𝐾𝑔 = √1 − 𝑆 (1 − 𝑆𝑚)

1
where 𝑆 is the degree of saturation of free water, and the parameter 𝑚 = was
2.2748

adopted by Baroghel-Bouny et al. (1999).

Some researchers (Forsyth and Simpson, 1991, Bear, 2013) assumed that the flow of
liquid water would vanish when the degree of saturation falls below certain minimum
value, and similarly, the flow of gas would stop when saturation degree exceeds certain
critical value. A refined model of (3.50) was proposed:

𝑆 − 𝑆𝑖𝑟 𝐴𝑤
𝐾𝑙 = ( )
1 − 𝑆𝑖𝑟
(3.51)
𝑆𝑙 𝐴𝑔
𝐾𝑔 = 1 − ( )
𝑆𝑐𝑟

where 𝑆𝑖𝑟 is the irreducible saturation value below which liquid water flow vanishes,
𝑆𝑐𝑟 is the critical saturation value above which gas flow stops, and 𝐴𝑤 and 𝐴𝑔 are
parameters normally lying between 1 and 3.

48
3.3 Mechanical Properties of Concrete

3.3.1 Compressive Strength

Compressive strength of concrete at high temperature is reasonably well researched due

to its important role in the fire design and association with other concrete properties. At
elevated temperature, it is affected by the rate of heating, initial (room temperature)
strength and types of binders used (silica fume, fly ash and slag, etc.). Following are
several models from the literature.

Eurocode 2 (ENV1992-1-2:1996, 1996)

In the earlier version of Eurocode 2, a conservative guideline for the siliceous aggregate
concrete was proposed:

1.0 𝑇𝑐 ≤ 100℃
1600 − 𝑇𝑐
𝑓𝑐 (𝑇𝑐 ) 100℃ < 𝑇𝑐 ≤ 400℃
= 1500 (3.52)
𝑓𝑐 0 900 − 𝑇𝑐
400℃ < 𝑇𝑐 ≤ 900℃
625
{0.0 900℃ < 𝑇𝑐 ≤ 1200℃

where 𝑓𝑐 (𝑇𝑐 ) and 𝑓𝑐 0 is the peak compressive strength at temperature 𝑇𝑐 and room
temperature.

Model of Bažant and Chern (1987)

Bažant and Chern summarized the experimental resutls from various researchers and
proposed a bilinear model to fit the test results:

𝑓𝑐 (𝑇𝑐 ) 1.0 − 0.1786 × 10−3 𝑇𝑐 𝑇𝑐 ≤ 350℃

={ (3.53)
𝑓𝑐 0 0.9375 − 1.713 × 10−3 (𝑇𝑐 − 350) 350℃ < 𝑇𝑐 ≤ 800℃

Model of Li and Purkiss (2005)

49
Li and Purkiss presented the following cubic function for siliceous aggregate concrete:

𝑓𝑐 (𝑇𝑐 ) 𝑇𝑐 3 𝑇𝑐 2 𝑇𝑐
= 0.00165 ( ) − 0.03 ( ) + 0.025 ( ) + 1.002 (3.54)
𝑓𝑐 0 100 100 100

Model of Hertz (2005)

Hertz developed a common expression for material properties of various types of

concrete corresponding to different parameters in the formula:

𝑓𝑐 (𝑇𝑐 ) 1−k
=k+
𝑓𝑐0 𝑇𝑐 𝑇 2 𝑇 8 𝑇 64 (3.55)
[1 + ( ) + ( 𝑐) + ( 𝑐) + ( 𝑐 ) ]
𝑇1 𝑇2 𝑇8 𝑇64

where k, T1 , 𝑇2 , 𝑇8 , 𝑇64 are parameters describing the curve of various concretes.

3.3.2 Tensile Strength

Tensile strength of concrete is much lower than the compressive strength due to the
easy propagation of microcracks, therefore, it is often neglected in the fire design.
However, in the heat and mass transfer as well as the thermo-mechanical analysis of
concrete, tensile strength is an important parameter in determining the occurrence of
spalling, and should be considered carefully. The measurement of direct concrete
tensile strength is rather difficult in reality due to the restriction from test apparatus.
The majority of experimental results are thus obtained by the splitting tensile tests
conducted on cylindrical concrete specimens. It should also be born in mind that most
of the existing data on the high temperature tensile strength of concrete is the residual
strength measured after the specimens cool down and resume to room temperature.
Such residual strengths, as a result, are more of the representative of post-cooling
behaviour rather than the behaviour at high temperature desired. An experimental
investigation has been carried out to distinguish the difference between the two
behaviours and will be disclosed in full details in Chapter 7. Discussions on the tensile
strength are therefore not repeated in this section.

50
3.3.3 Elastic Modulus

The elastic modulus of concrete, also known as the Young’s modulus, mainly relies on
the properties of the batch mix (water-cement ratio, amount and types of aggregate),
curing condition, and the age of concrete. The elastic modulus decreases with the
increase of temperature and this trend has been reported by various researchers.

Model of Bažant and Chern (1987)

Bažant and Chern again proposed a bilinear experssion to describe the reduced elastic
modulus at elevated temperature fitting various experimental resutls:

𝐸(𝑇𝑐 ) 1.0 − 1.256 × 10−3 𝑇𝑐 𝑇𝑐 ≤ 650℃

={ −3 (3.56)
𝐸0 0.1837 − 0.565 × 10 (𝑇 𝑐 − 650) 650℃ < 𝑇 𝑐 ≤ 800℃

where 𝐸(𝑇𝑐 ) and 𝐸0 is the elastic modulus at temperature 𝑇𝑐 and room temperature.

Model of Lu (1989)

Taking the elastic modulus as the secant modulus at 0.4Fc (𝑇𝑐 ), Lu developed the
following expression:

1.0 − 1.50 × 10−3 𝑇𝑐 𝑇𝑐 ≤ 300℃

𝐸(𝑇𝑐 )
= {0.87 − 0.84 × 10−3 𝑇𝑐 300℃ < 𝑇𝑐 ≤ 700℃ (3.57)
𝐸0
0.28 700℃ < 𝑇𝑐

Model of Li and Guo (1993)

𝐸(𝑇𝑐 ) 1.0 − 1.256 × 10−3 𝑇𝑐 𝑇𝑐 ≤ 650℃

={ −3 (3.58)
𝐸0 0.1837 − 0.565 × 10 (𝑇𝑐 − 650) 650℃ < 𝑇𝑐 ≤ 800℃

Model of Purkiss and Li (2013)

Purkiss and Li presented the following expression based on their own experimental
results and refering to the current Eurcode 2:

51
 1.0 𝑇𝑐 ≤ 60℃
𝐸(𝑇𝑐 )
= {800 − 𝑇𝑐 (3.59)
𝐸0 60℃ < 𝑇𝑐 ≤ 800℃
740

3.3.4 Thermal Expansion

The thermal expansion of concrete is often represented in the form of free thermal strain,
which is determined either by the coefficient of thermal expansion and the temperature
difference, or thermal strain measured as a whole. The thermal expansion is mostly
governed by the type of aggregate used, due to its large proportion of concrete volume
occupied.

ASCE Manual of Practice (Lie, 1992)

For all types of concrete:

𝜀𝑓𝑡 = [0.004(𝑇𝑐 2 − 400) + 6(𝑇𝑐 − 20)] × 10−6 (3.60)

Eurocode 2 (BSEN1992-1-2:2004, 2004)

Siliceous aggregate concrete:

(0.09𝑇𝑐 + 2.3 × 10−7 𝑇𝑐 3 − 1.8) × 10−4 20℃ ≤ 𝑇𝑐 ≤ 700℃

𝜀𝑓𝑡 = { (3.61)
14 × 10−3 700℃ < 𝑇𝑐 ≤ 1200℃

Calcareous aggregate concrete:

(0.06𝑇𝑐 + 1.4 × 10−7 𝑇𝑐 3 − 1.2) × 10−4 20℃ ≤ 𝑇𝑐 ≤ 805℃

𝜀𝑓𝑡 = { (3.62)
12 × 10−3 805℃ < 𝑇𝑐 ≤ 1200℃

Models of Kodur and Sultan (2003)

Kodur and Sultan presented expressions for the coefficient of thermal expasion instead
of the theraml strain as a whole with respect to types of aggregate:

52
Siliceous aggregate concrete:

11.0 × 10−6 𝑇𝑐 − 0.20 × 10−3 20℃ ≤ 𝑇𝑐 ≤ 450℃

𝛼𝑓𝑡 = {36.0 × 10−6 𝑇𝑐 − 11.5 × 10−3 450℃ < 𝑇𝑐 ≤ 650℃ (3.63)
11.9 × 10−3 650 < 𝑇𝑐 ≤ 1000℃

Calcareous aggregate concrete:

8.0 × 10−6 𝑇𝑐 − 0.20 × 10−3 20℃ ≤ 𝑇𝑐 ≤ 450℃

𝛼𝑓𝑡 = {21.0 × 10−6 𝑇𝑐 − 6.1 × 10−3 450℃ < 𝑇𝑐 ≤ 920℃ (3.64)
12.0 × 10−6 𝑇𝑐 + 24.2 × 10−3 920 < 𝑇𝑐 ≤ 1000℃

Model of Kodur and Khaliq (2011)

For high strength concrete:

𝜀𝑓𝑡 = 10.0 × 10−6 𝑇𝑐 − 5.0 × 10−3 𝑓𝑜𝑟 20℃ ≤ 𝑇𝑐 ≤ 1000℃ (3.65)

3.4 Material Data Related to Water Species and Air

In this section, material data related to the liquid water, water vapour and air required
in the modelling of thermo-hygro-mechanical behaviour of concrete are presented. The
temperature used in the following are the absolute temperature (in K) unless specified.

3.4.1 Material Data Related to Water

Diffusion coefficient of adsorbed water (Gawin et al., 1999):

𝑆 𝑇
DB = 𝐷𝐵0 exp (−2.08 ) 𝑓𝑜𝑟 𝑆 ≤ 𝑆𝑠𝑠𝑝 (3.66)
𝑆𝑠𝑠𝑝 𝑇𝑟𝑒𝑓

where 𝐷𝐵0 = 1.57 × 10−11 𝑚2 /𝑠, and reference temperature 𝑇𝑟𝑒𝑓 = 295 𝐾.

Density of liquid water :

53
 𝜌𝑙 1 2 5 16 43 110
= 1 + 𝑏1 𝜃 3 + 𝑏2 𝜃 3 + 𝑏3 𝜃 3 + 𝑏4 𝜃 3 + 𝑏5 𝜃 3 + 𝑏6 𝜃 3 (3.67)
𝜌𝑐𝑟𝑖

𝑇
with θ=1− (3.68)
𝑇𝑐𝑟𝑖

where 𝜌𝑐𝑟𝑖 (= 322𝑘𝑔𝑚−3 ) and 𝑇𝑐𝑟𝑖 (= 647.096 K) is the critical density and
temperature of water, 𝑏1 = 1.99274064, 𝑏2 = 1.09965342, 𝑏3 = −0.510839303,
𝑏4 = −1.75493479, 𝑏5 = −45.5170352, 𝑏6 = −674694.450.

Dynamic viscosity of liquid water (Gawin et al., 1999):

𝜇𝑙 = 0.6612(𝑇 − 229)−1.562 𝑓𝑜𝑟 𝑇 ≤ 𝑇𝑐𝑟𝑖 (3.69)

where 𝑇𝑐𝑟𝑖 = 647.096 𝐾 is the critical temperature of water.

Mass of dehydrated water (Tenchev et al., 2001):

0 𝑇𝑐 ≤ 200℃
𝜀𝐷 𝜌𝐷 7.0 × 10−4 (𝑇𝑐 − 200) 200℃ < 𝑇𝑐 ≤ 300℃
= (3.70)
𝜀𝑐𝑒𝑚 𝜌𝑐𝑒𝑚 0.4 × 10−4 (𝑇𝑐 − 300) + 0.07 300℃ < 𝑇𝑐 ≤ 800℃
{0.09 800℃ < 𝑇𝑐

where 𝜀𝑐𝑒𝑚 𝜌𝑐𝑒𝑚 is the cement content per unit volume of concrete, and 𝑇𝑐 is the
temperature in ℃.

Specific heat capacity of free water (Davie et al., 2006):

𝑎𝑇 𝑏
Cfw = {(2.4768𝑇 + 3368.2) + ( 513.15
) 𝑇 ≤ 𝑇𝑐𝑟𝑖 (3.71)
24515.0 𝑇 > 𝑇𝑐𝑟𝑖

where 𝑎 = 1.08542631988638, 𝑏 = 31.4447657616636.

54
3.4.2 Material Data Related to Water Vapour

Coefficient of mass transfer of water vapor (Çengel, 2003, Tenchev et al., 2001):

2
hq 𝐷𝑎𝑣 (𝑇∞ ) 3 (3.72)
𝛽= [ ]
(𝜌𝐶)𝑎𝑖𝑟 𝛼𝑎𝑖𝑟

where ℎ𝑞 is the coefficient of heat transfer, (𝜌𝐶)𝑎𝑖𝑟 and 𝛼𝑎𝑖𝑟 is the heat capacity and
thermal diffusivity of air.

Diffusion coefficient of water vapour – air (Atkinson and Nickerson, 1984, Çengel,
2003):

𝛿 𝑇 2.072
𝐷𝑣𝑎 = 𝐷 , 𝐷 = 1.87 × 10−5 ( ) (3.73)
𝜏2 𝑃𝑔

where 𝐷 is the atmospheric diffusion coefficient of water vapour in air, 𝛿(= 0.5) and
𝜏(= 3) is the constrictivity and tortuosity factors of concrete, accounting for the
reduction in the rate of diffusion due to the complex pore structure of concrete.

Dynamic viscosity of water vapour (Gawin et al., 1999):

𝜇𝑣 = 𝜇𝑣0 + 𝛼𝑣 (𝑇 − 𝑇0 ) (3.74)

where 𝜇𝑣0 = 8.85 × 10−6 𝑃𝑎 𝑠 , 𝛼𝑣 = 3.53 × 10−8 𝑃𝑎 𝑠 𝐾 −1 and 𝑇0 is the room

temperature.

Saturation vapour pressure (Tenchev et al., 2001):

𝑃𝑠𝑎𝑡 = 𝑎𝑇 6 + 𝑏𝑇 5 + 𝑐𝑇 4 + 𝑑𝑇 3 + 𝑒𝑇 2 + 𝑓𝑇 + 𝑔 𝑓𝑜𝑟 𝑇 ≤ 𝑇𝑐𝑟𝑖 (3.75)

where coefficients 𝑎 = −1.43742221944687 × 10−9 , 𝑏 = 4.42439058302123 ×

10−6 , 𝑐 = −3.92808082125791 × 10−3 , 𝑑 = 1.59103252944303 , 𝑒=
−325.887438504847, 𝑓 = 32147.7952751975, 𝑔 = −1.15466360325087 × 106 .

55
Another formulation is provided by Wagner and Pruß(2002):

𝑃𝑠𝑎𝑡 𝑇𝑐𝑟𝑖
ln ( )= (𝑎𝜃 + 𝑏𝜃 1.5 + 𝑐𝜃 3 + 𝑑𝜃 3.5 + 𝑒𝜃 4 + 𝑓𝜃 7.5 ) (3.76)
𝑃𝑐𝑟𝑖 𝑇

where 𝑃𝑐𝑟𝑖 (= 22.064 MPa) and 𝑇𝑐𝑟𝑖 (= 647.096 𝐾) is the critical pressure and
temperature for water, coefficients 𝑎 = −7.85951783 , 𝑏 = 1.84408259 , 𝑐 =
−11.7866497, 𝑑 = 22.6807411, 𝑒 = −15.9618719, 𝑓 = 1.80122502.

Specific heat capacity of water vapour (Davie et al., 2006):

𝑎𝑇 𝑏
Cv = { (7.1399𝑇 − 443) + ( ) 𝑇 ≤ 𝑇𝑐𝑟𝑖 (3.77)
513.15
45821.04 𝑇 > 𝑇𝑐𝑟𝑖

where 𝑎 = 1.13771502228162, 𝑏 = 29.4435287521143.

3.4.3 Material Data Related to Air

Diffusion coefficient of air – water vapour (Atkinson and Nickerson, 1984, Çengel,
2003):

𝛿 −5
𝑇 2.072
𝐷𝑎𝑣 = 𝐷 2, 𝐷 = 1.87 × 10 ( ) (3.78)
𝜏 𝑃𝑔

where 𝐷 is the atmospheric diffusion coefficient of air in water vapour, 𝛿(= 0.5) and
𝜏(= 3) is the constrictivity and tortuosity factors of concrete, accounting for the
reduction in the rate of diffusion due to the complex pore structure of concrete. It could
be noted that 𝐷𝑎𝑣 = 𝐷𝑣𝑎 is assumed in this study.

Dynamic viscosity of dry air (Gawin et al., 1999):

𝜇𝑎 = 𝜇𝑎0 + 𝛼𝑎 (𝑇 − 𝑇0 ) − 𝛽𝑎 (𝑇 − 𝑇0 )2 (3.79)

56
where 𝜇𝑎0 = 17.17 × 10−6 𝑃𝑎 𝑠 , 𝑇0 is the room temperature, 𝛼𝑎 = 4.73 ×
10−8 𝑃𝑎 𝑠 𝐾 −1 , and 𝛽𝑎 = 2.22 × 10−11 𝑃𝑎 𝑠 𝐾 −2 .

Dynamic viscosity of gaseous mixture (Tenchev et al., 2001):

𝜌̃𝑎 𝜇𝑎 + 𝜌̃𝑣 𝜇𝑣
𝑓𝑜𝑟 (𝜌̃𝑎 + 𝜌̃𝑣 ) > 0.0
𝜇𝑔 = { 𝜌̃𝑎 + 𝜌̃𝑣 (3.80)
0.0 𝑓𝑜𝑟 (𝜌̃𝑎 + 𝜌̃𝑣 ) = 0

Another formulation is provided by Gawin et al. (1999):

0.608
𝑃𝑎
𝜇𝑔 = 𝜇𝑣 + (𝜇𝑎 − 𝜇𝑣 ) ( ) (3.81)
𝑃𝑔

where 𝜌̃𝜋 and 𝑃𝜋 is the mass per unit volume of gaseous material and the pressure of
phase 𝜋.

Specific heat capacity of dry air (Davie et al., 2006):

Ca = 𝑎𝑇 3 + 𝑏𝑇 2 + 𝑐𝑇 + 𝑑 (3.82)

where 𝑎 = −9.84936701814735 × 10−8 , 𝑏 = 3.56436257769861 × 10−4 , 𝑐 =

−1.2161792398775 × 10−1 , 𝑑 = 1012.50255216324.

Density and thermal diffusivity of dry air:

The author developed the following expressions by curve-fitting to the tabulated data
provided by Çengel (2003):

𝜌𝑎 = 352.79𝑇 −1 𝑓𝑜𝑟 𝑇 ≤ 1773.15𝐾 (3.83)

𝛼𝑎 = 𝑎𝑇 3 + 𝑏𝑇 2 + 𝐶𝑇 + 𝑑 𝑓𝑜𝑟 𝑇 ≤ 1773.15𝐾 (3.84)

where 𝑎 = −2 × 10−14 , 𝑏 = 1 × 10−10 , 𝑐 = 9 × 10−8 , 𝑑 = −1 × 10−5 .

57
4 COUPLED THERMO-HYGRO MODEL

4.1 Introduction

In this study, the development of the coupled thermo-hygro-mechanical (THM) model

describing the behaviour of concrete at elevated temperature is presented, based on the
fundamental conservation equations and constitutive relationships. First of all, the
thermo-hygro (TH) component of the model, describing the heat and mass transfer in
heated concrete, is discussed in this chapter.

As mentioned previously in Section 2.3.3, the Padua model and Glasgow model are
developed in a theoretically more rigorous fashion, compared to all other numerical
counterparts. Therefore, they are chosen as the starting basis for the THM model
presented in this study. The TH model is based mainly on the Glasgow model (Tenchev
et al., 2001, Davie et al., 2006).

4.2 Mass and Enthalpy Conservation

Concrete is treated as a multiphase porous medium (system), shown in Figure 2.11, in

the thermo-hygro modelling. The gas phase (also called moist air, denoted by subscript
“g”) consists of non-condensable dry air and condensable water vapour that share a
same volume, denoted by subscripts “a” and “v” respectively. All gases are assumed to
behave as ideal gases. The definition of liquid phase is quite debatable, because it is
still unclear how “liquid-like” certain portions of water behave (Guttag, 2016). In the
Padua and Glasgow model, free water, which comprises liquid water and sometimes
the adsorbed water, is considered as the liquid phase. This study follows the same
definition, where waters other than water vapour and non-evaporable water are grouped
into liquid phase. Subscripts “l”, “fw”, and “D” are used to represent the liquid water,
free water and the dehydrated water phase respectively. Local thermodynamic
equilibrium is assumed, so that the temperatures of all constituents at a point in the
porous medium are the same. The macroscopic, governing conservation equations can
be expressed as:

58
Mass conservation of free water

∂(𝜀𝑓𝑤 𝜌𝑙 ) 𝜕(𝜀𝐷 𝜌𝑙 )
= −∇
⏟ ∙ 𝐽𝑓𝑤 −𝐸
⏟ ̇
𝑓𝑤 + (4.1)
⏟ 𝜕𝑡 ⏟ 𝜕𝑡
𝑎 𝑏 𝑐 𝑑

where 𝜀𝜋 is the volume fraction, 𝜌𝜋 the density, 𝜌̃π the mass per unit volume of gaseous
̇ is the rate of evaporation of free water
material, 𝐽𝜋 the mass flux, of a phase 𝜋 . 𝐸𝑓𝑤
(including desorption), and t is time. As such, term a, b, c and d represents the rate of
change of free water εfw 𝜌𝑙 in a unit volume of porous material, mass of free water
transported by convection (liquid water) and diffusion (adsorbed water), mass loss due
to evaporation and desorption, and mass gain from dehydration (chemically bound
water assumed to be initially released as liquid water), respectively.

Mass conservation of water vapour

∂(𝜀𝑔 𝜌̃𝑣 )
̇
= −∇ ∙ 𝐽𝑣 + 𝐸𝑓𝑤 (4.2)
𝜕𝑡

Mass conservation of dry air

∂(𝜀𝑔 𝜌̃𝑎 )
= −∇ ∙ 𝐽𝑎 (4.3)
𝜕𝑡

Enthalpy conservation of the multiphase system

𝜕𝑇 𝜕(𝜀𝐷 𝜌𝑙 )
(𝜌𝐶) −∇ ∙ (−𝑘∇𝑇) −
=⏟ (𝜌𝐶𝑣) ∙ ∇𝑇 −𝜆
⏟ 𝐸 𝐸̇
𝑓𝑤 −𝜆 𝐷
⏟ 𝜕𝑡 ⏟ ⏟ 𝜕𝑡 (4.4)
𝑏 𝑑
𝑎 𝑐 𝑒

where 𝜌𝐶 and 𝑘 is the effective heat capacity and thermal conductivity of multiphase

medium (concrete), 𝜌𝐶𝑣 the energy transported by fluid flow (i.e., convection), 𝑣 the

velocity of fluid transported within concrete, 𝜆𝐸 the specific heat of evaporation (and
of desorption when appropriate), 𝜆𝐷 the specific heat of dehydration of chemically

59
bound water, and 𝑇 the absolute temperature (in K). As such, term a to term e represents
the rate of energy accumulated in a unit volume of multiphase medium, energy
transferred by heat conduction, energy transported by convection of fluid flow, energy
consumed for evaporation (and desorption when appropriate), energy required for
releasing chemically bound water from dehydration, respectively.

Summing up (4.1) and (4.2) yields the equation of mass conservation of moisture (water
species) as:

∂(𝜀𝑓𝑤 𝜌𝑙 ) ∂(𝜀𝑔 𝜌̃𝑣 ) 𝜕(𝜀𝐷 𝜌𝑙 )

+ = −∇ ∙ 𝐽𝑓𝑤 + − ∇ ∙ 𝐽𝑣 (4.5)
𝜕𝑡 𝜕𝑡 𝜕𝑡

̇ using (4.1) and substitute it into (4.4) yields:

Express 𝐸𝑓𝑤

𝜕𝑇
(𝜌𝐶) = −∇ ∙ (−𝑘∇𝑇) − (𝜌𝐶𝑣) ∙ ∇𝑇
𝜕𝑡
(4.6)
𝜕(𝜀𝐷 𝜌𝑙 ) ∂(𝜀𝑓𝑤 𝜌𝑙 ) 𝜕(𝜀𝐷 𝜌𝑙 )
− 𝜆𝐸 ( − ∇ ∙ 𝐽𝑓𝑤 − ) − 𝜆𝐷
𝜕𝑡 𝜕𝑡 𝜕𝑡

Rearrange (4.5) and (4.6) by moving the terms containing time derivative to left hand
side (LHS) of the equations and all other terms to right hand side (RHS), together with
(4.3), yields:

∂(𝜀𝑔 𝜌̃𝑎 )
= −∇ ∙ 𝐽𝑎 (4.7)
𝜕𝑡

∂(𝜀𝑓𝑤 𝜌𝑙 ) ∂(𝜀𝑔 𝜌̃𝑣 ) 𝜕(𝜀𝐷 𝜌𝑙 )

+ − = −∇ ∙ (𝐽𝑓𝑤 + 𝐽𝑣 ) (4.8)
𝜕𝑡 𝜕𝑡 𝜕𝑡

𝜕𝑇 𝜕(𝜀𝑓𝑤 𝜌𝑙 ) 𝜕(𝜀𝐷 𝜌𝑙 )
(𝜌𝐶) − 𝜆𝐸 + (𝜆𝐷 +𝜆𝐸 )
𝜕𝑡 𝜕𝑡 𝜕𝑡 (4.9)
= ∇ ∙ (𝑘∇𝑇) + 𝜆𝐸 ∇ ∙ 𝐽𝑓𝑤 − (𝜌𝐶𝑣) ∙ ∇𝑇

60
Equations (4.7) - (4.9) form the system of governing differential equations of dry air,
moisture and enthalpy conservation. It may be noted, however, that the term describing

the energy transferred by fluid flow [(𝜌𝐶𝑣) ∙ ∇𝑇] in (4.9) is neglected in the following

discussion. It is assumed that the energy transfer by convection is accounted for in the
thermal conductivity of concrete, 𝑘, whose relationship is empirically determined for
wet concrete.

4.3 Constitutive Laws and State Equations

This section presents the constitutive laws and state equations needed to translate the
system of governing equations into its final form expressed with certain chosen state
variables. Specifically, the effects of capillary pressure and adsorbed water are
explicitly considered. The step-by-step derivations of these laws and relationships
could be found in the books of Lewis and Schrefler (1998) and Bažant and Jirásek
(2018), in full details.

4.3.1 State Equations of Gaseous Mixture

The moist air is assumed to be a perfect binary mixture of the two ideal gases, dry air
and vapour, occupying a same volume. The ideal gas law relating the partial pressure
(𝑃𝑎 and 𝑃𝑣 ) and partial mass density (also called mass concentration, 𝜌̃𝑎 and 𝜌̃𝑣 ) of dry
air and vapour, is applied:

𝑅
𝑃a = 𝜌̃ 𝑇 = 𝑅𝑎 𝜌̃𝑎 𝑇 (4.10)
𝑀𝑎 𝑎

𝑅
𝑃𝑣 = 𝜌̃ 𝑇 = 𝑅𝑣 𝜌̃𝑣 𝑇 (4.11)
𝑀𝑤 𝑣

where 𝑅𝜋 is the specific gas constant, calculated by the universal gas constant 𝑅
divided by the molar mass Mπ , of constituent π, T is the absolute temperature (in K).

Specifically, the molar mass of the moist air is given by:

61
 −1
𝜌̃𝑣 1 𝜌̃𝑎 1
𝑀g = ( + ) (4.12)
𝜌̃𝑔 𝑀𝑤 𝜌̃𝑔 𝑀𝑎

Dalton’s law defines that the pressure and density of the gas mixture is the sum of the
partial pressures and densities of its constituents:

𝑃g = 𝑃𝑎 + 𝑃𝑣 (4.13)

𝑀𝑔 𝑃𝑔
𝜌̃𝑔 = 𝜌̃𝑎 + 𝜌̃𝑣 = (4.14)
𝑅𝑇

Equations (4.10) to (4.14) can be viewed as the state equations of gaseous mixture.

4.3.2 Young-Laplace Equation

Mechanical equilibrium between two planer interfaces requires the pressures on both
sides to be equal. However, when the interface is curved, such as the case for capillary
menisci, the equilibrium equation will have an additional term resulting from surface
tension and curvature. The Young-Laplace equation mathematically describes this
condition as:

2𝛾𝑔𝑙
𝑃𝑔 = 𝑃𝑙 + (4.15)
𝑟

where 𝛾𝑔𝑙 is the surface tension, 𝑟 the radius of menisci, 𝑃𝑔 and 𝑃𝑙 the pressures of gas
and liquid phases. Equation (4.15) is diagrammatically demonstrated in Figure 4.1. The
signs in (4.15) corresponds to the typical case of a concave meniscus, where gas
pressure is greater than the liquid (capillary) pressure. The difference

2𝛾𝑔𝑙
𝑃𝑐 = 𝑃𝑔 − 𝑃𝑙 = (4.16)
𝑟

62
is defined as the capillary pressure (or the suction stress) 𝑃𝑐 . Equation (4.16) is another
form of the Young-Laplace equation.

4.3.3 Kelvin-Laplace Equation

At the gas-liquid interface, the liquid water and vapour can transform into each other
by evaporation or condensation, driven by the difference in the specific1 free enthalpy
(also called specific Gibbs free energy) hl and hv , which are functions of temperature
and pressure. At thermodynamic equilibrium, the specific free enthalpies share the same
value. But since the exact functions of description are different for liquid water and
vapour, even at the same temperature, thermodynamic equilibrium could not imply the
pressures are the same. The general description of thermodynamic equilibrium at
temperature 𝑇 could be attained by:

hv (𝑃𝑣 , 𝑇) = hl (𝑃𝑙 , 𝑇) (4.17)

The saturated vapour pressure 𝑃𝑠𝑎𝑡 2 could be defined by the condition:

hv (𝑃𝑠𝑎𝑡 , 𝑇) = hl (𝑃𝑠𝑎𝑡 , 𝑇) (4.18)

when vapour and liquid water are subjected to the same pressure at thermodynamic
equilibrium. Instead of directly evaluating (4.17) with the specific forms, hv and hl are
rewritten as:

𝑃𝑣
𝜕ℎ𝑣 (𝑃, 𝑡)
hv (𝑃𝑣 , 𝑇) = hv (𝑃𝑠𝑎𝑡 , 𝑇) + ∫ 𝑑𝑃 (4.19)
𝑃𝑠𝑎𝑡 𝜕𝑃

𝑃𝑙
𝜕ℎ𝑙 (𝑃, 𝑡)
h𝑙 (𝑃𝑙 , 𝑇) = h𝑙 (𝑃𝑠𝑎𝑡 , 𝑇) + ∫ 𝑑𝑃 (4.20)
𝑃𝑠𝑎𝑡 𝜕𝑃

1
Quantities denoted as “specific” refer to a unit mass of the substance
2
The saturated vapour pressure, 𝑃𝑠𝑎𝑡 , depends on temperature. For simplicity, this dependence is not
marked explicitly.

63
where P is a dummy integration variable representing the pressure continuously
changing from 𝑃𝑠𝑎𝑡 to 𝑃𝑣 or 𝑃𝑙 . To evaluate the second terms on the RHS of (4.19) and
(4.20), the concept of specific volume is introduced, which is the reciprocal of the mass
density, i.e., mass of the substance occupying a unit volume. Since specific volume is
the thermodynamic variable conjugate to the pressure, it can be derived that the partial
derivative of specific free enthalpy with respect to the pressure is the specific volume.
The second terms are now equal to:

𝑃𝑣 𝑃𝑣
𝜕ℎ𝑣 (𝑃, 𝑡) 𝑑𝑃
∫ 𝑑𝑃 = ∫ (4.21)
𝑃𝑠𝑎𝑡 𝜕𝑃 𝑃𝑠𝑎𝑡 𝜌
̃𝑣 (𝑃, 𝑇)

𝑃𝑙 𝑃𝑙
𝜕ℎ𝑙 (𝑃, 𝑡) 𝑑𝑃
∫ 𝑑𝑃 = ∫ (4.22)
𝑃𝑠𝑎𝑡 𝜕𝑃 𝑃𝑠𝑎𝑡 𝜌𝑙 (𝑇)

Liquid water is considered to be incompressible compared to vapour, i.e., density of

liquid water depends only on temperature, irrespective of pressure. Integrating (4.21) -
(4.22) and substituting into (4.19) - (4.20), with the ideal gas law (4.11), leads to

𝑅𝑇 𝑃𝑣
hv (𝑃𝑣 , 𝑇) = hv (𝑃𝑠𝑎𝑡 , 𝑇) + ln (4.23)
𝑀𝑤 𝑃𝑠𝑎𝑡

𝑃𝑙 − 𝑃𝑠𝑎𝑡
h𝑙 (𝑃𝑙 , 𝑇) = h𝑙 (𝑃𝑠𝑎𝑡 , 𝑇) + (4.24)
𝜌𝑙 (𝑇)

Taking into account (4.17) - (4.18) and (4.23) - (4.24), the rigorously3 derived form of
the Kelvin-Laplace equation is obtained:

𝑅𝑇 𝑃𝑣 𝑃𝑙 − 𝑃𝑠𝑎𝑡
ln = (4.25)
𝑀𝑤 𝑃𝑠𝑎𝑡 𝜌𝑙 (𝑇)

3
Of course, with negligible small errors caused by the assumptions of a) incompressibility of liquid water
and b) water vapour as an ideal gas.

64
Relative humidity (RH) is defined as the ratio between partial vapour pressure and
𝑃𝑣
saturated vapour pressure, i.e., the term . To link the capillary pressure with RH,
𝑃𝑠𝑎𝑡

(4.25) is further rewritten as:

𝑅𝑇 𝑃𝑣 𝑃𝑣 − 𝑃𝑠𝑎𝑡 𝑃𝑙 − 𝑃𝑣
ln = + (4.26)
𝑀𝑤 𝑃𝑠𝑎𝑡 𝜌𝑙 (𝑇) 𝜌𝑙 (𝑇)

Recalling from (4.21) - (4.22), the following relation can be reached:

𝑃𝑣 𝑃𝑣 𝑃𝑙
𝑑𝑃 𝑑𝑃 𝑑𝑃
∫ =∫ +∫ (4.27)
𝑃𝑠𝑎𝑡 𝜌
̃𝑣 (𝑃, 𝑇) 𝑃𝑠𝑎𝑡 𝜌𝑙 (𝑇) 𝑃𝑣 𝜌𝑙 (𝑇)

The first integral on the RHS is much smaller than the integral on LHS, considering the
fact that 𝜌̃𝑣 ≪ 𝜌𝑙 , and thus can be neglected. The same approximation applies to (4.26)
that the first term on the RHS is negligible. Recall that the gas pressure 𝑃𝑔 equals to the
sum of partial pressures of vapour 𝑃𝑣 and dry air 𝑃𝑎 . But since 𝑃𝑎 is normally negligible,
the gas pressure is approximated as the vapour, therefore, 𝑃𝑐 = 𝑃𝑣 − 𝑃𝑙 . Hereto, the
final form of Kelvin-Laplace equation, or the Kelvin equation, reads:

𝑃𝑣
𝑃𝑐 = −𝑅𝑣 𝑇𝜌𝑙 (𝑇) ln ( ) (4.28)
𝑃𝑠𝑎𝑡

4.3.4 Degree of Saturation

The degree of saturation with free water, 𝑆, is defined as the pore volume occupied by
free water with respect to the total pore volume. Therefore, it can be described in terms
of the free water volume fraction 𝜀𝑓𝑤 and the porosity 𝜙 as:

𝜀𝑓𝑤 𝜀𝑓𝑤
𝑆= = (4.29)
𝜙 𝜀𝑓𝑤 + 𝜀𝑔

And the degree of saturation with adsorbed water, 𝑆𝐵 , follows the definition of Gawin
et al. (1999) and Davie et al. (2006) as:

65
 𝑆 𝑓𝑜𝑟 𝑆 ≤ 𝑆𝑠𝑠𝑝
𝑆𝐵 = { (4.30)
𝑆𝑠𝑠𝑝 𝑓𝑜𝑟 𝑆 > 𝑆𝑠𝑠𝑝

where 𝑆𝑠𝑠𝑝 is the solid saturation point, defined as the upper limit of the hygroscopic
moisture range and the maximum degree of saturation with adsorbed water. Equation
(4.30) implies a process that any moisture entering a dry sample of concrete would first
fill the gel pores and adhere to the surface of capillary pores as adsorbed water, until
the solid saturation point is reached. Then the remaining capillary pores will be filled
with liquid water. This process is assumed reversed during drying (Davie et al., 2006).
Since adsorbed water is physically bound to the solid skeleton within gel pores and on
the surfaces of capillary pores, its behaviour is assumed to be not “liquid-like”.
Therefore, the transport of adsorbed water is governed by diffusion resulting from a
concentration gradient while the transport of liquid water is driven by pressure gradient.

4.3.5 Darcy-Buckingham Law

Darcy’s law was initially proposed to describe the flow of a single-phase fluid that
completely fills the pore space in a porous medium. The Darcy velocity, or the volume
flux, is related to the permeability and dynamic viscosity of the porous medium, and
the pressure gradient over a given distance. Buckingham (1907) extended the original
Darcy’s law to include the partially saturated cases, with the hydraulic permeability
being reduced. The reduction of permeability is imposed by a dimensionless factor 𝐾𝜋 ,
called the relative permeability of phase π. The relative permeability depends on the
water content, ranging from 1 at full saturation to 0 at a certain minimum degree of
saturation. For concrete members of usual sizes, the pressure caused by the self-weight
of fluid is negligible compared to the actual pore pressure. Also considering that the
thermo-hygro phenomenon in concrete is relatively low even at high temperature, the
inertial forces (gravity effects) are omitted in the description of fluid velocities in this
study. The Darcy-Buckingham law is applied, and the volume fluxes of the gas and
liquid water are:

66
 𝐾𝐾𝑔
𝑣𝑔 = − ∇𝑃𝑔 (4.31)
𝜇𝑔

𝐾𝐾𝑙
𝑣𝑙 = − ∇𝑃𝑙 (4.32)
𝜇𝑙

where 𝐾 is the intrinsic permeability of dry concrete, 𝐾𝑔 (𝐾𝑙 ), 𝜇𝑔 (𝜇𝑙 ) and 𝑃𝑔 (𝑃𝑙 ) are
the relative permeability, dynamic viscosity and pressure of the gas (liquid) phase. The
pressure-induced mass fluxes of dry air and water vapour are then obtained:

𝐾𝐾𝑔
𝐽a𝑃 = 𝜀𝑔 𝜌̃𝑎 𝑣𝑔 = −𝜀𝑔 𝜌̃𝑎 ∇𝑃𝑔 (4.33)
𝜇𝑔

𝐾𝐾𝑔
𝐽𝑣𝑃 = 𝜀𝑔 𝜌̃𝑣 𝑣𝑔 = −𝜀𝑔 𝜌̃𝑣 ∇𝑃𝑔 (4.34)
𝜇𝑔

For the mass flux of free water, only the liquid component is drive by pressure:

𝑃
𝑆𝐵 𝑆𝐵 𝐾𝐾𝑙
𝐽𝑓𝑤 = (1 − ) 𝜀𝑓𝑤 𝜌𝑙 𝑣𝑙 = − (1 − ) 𝜀𝑓𝑤 𝜌𝑙 ∇𝑃𝑙 (4.35)
𝑆 𝑆 𝜇𝑙

where 𝜀𝑔 𝜌̃𝑎 , 𝜀𝑔 𝜌̃𝑣 and 𝜀𝑓𝑤 𝜌𝑙 represent the mass concentrations of dry air, vapour and
liquid water per unit volume of concrete.

4.3.6 Fick’s First Law of Diffusion

Fick’s first law describes the diffusion process of substance moving from high-
concentration to low-concentration regions. Diffusive flux is therefore related to the
gradient of concentration. For the description of mass fluxes of dry air and water vapour
diffusion, Fick’s law for binary gas mixtures is applied:

𝜌̃𝑎
𝐽𝑎𝐷 = −𝜀𝑔 𝜌̃𝑔 𝐷𝑎𝑣 ∇ ( ) (4.36)
𝜌̃𝑔

67
 𝜌̃𝑣
𝐽𝑣𝐷 = −𝜀𝑔 𝜌̃𝑔 𝐷𝑣𝑎 ∇ ( ) (4.37)
𝜌̃𝑔

where 𝐷𝜋𝑖 is the diffusion coefficient of phase π diffusing in phase i. Taking into
account the fact that 𝐷𝑎𝑣 and 𝐷𝑣𝑎 are usually assumed equal (Lewis and Schrefler, 1998,
Çengel, 2003) and utilizing (4.10) - (4.14), leads to:

𝜌̃𝑎 𝜌̃𝑔 − 𝜌̃𝑣

𝐽𝑎𝐷 = −𝜀𝑔 𝜌̃𝑔 𝐷𝑎𝑣 ∇ ( ) = −𝜀𝑔 𝜌̃𝑔 𝐷𝑎𝑣 ∇ ( ) = −𝐽𝑣𝐷 (4.38)
𝜌̃𝑔 𝜌̃𝑔

For the free water, only the adsorbed water component is subjected to diffusion:

𝐷
𝑆𝐵 𝑆𝐵
𝐽𝑓𝑤 =( ) 𝜀𝑓𝑤 𝜌𝑙 𝑣𝐵 = − ( ) 𝜀𝑓𝑤 𝜌𝑙 𝐷𝐵 ∇𝑆𝐵 (4.39)
𝑆 𝑆

where 𝑣𝐵 = −𝐷𝐵 ∇𝑆𝐵 is the diffusion velocity and 𝐷𝐵 is the coefficient of adsorbed
water diffusion.

Finally, the total mass fluxes of dry air, vapour and free water are attained by adding
the pressure- and diffusion-induced components:

𝐾𝐾𝑔 𝜌̃𝑎
𝐽𝑎 = −𝜀𝑔 𝜌̃𝑎 ∇𝑃𝑔 − 𝜀𝑔 𝜌̃𝑔 𝐷𝑎𝑣 ∇ ( ) (4.40)
𝜇𝑔 𝜌̃𝑔

𝐾𝐾𝑔 𝜌̃𝑣
𝐽𝑣 = −𝜀𝑔 𝜌̃𝑣 ∇𝑃𝑔 − 𝜀𝑔 𝜌̃𝑔 𝐷𝑣𝑎 ∇ ( ) (4.41)
𝜇𝑔 𝜌̃𝑔

𝑆𝐵 𝐾𝐾𝑙 𝑆𝐵
𝐽𝑓𝑤 = − (1 − ) 𝜀𝑓𝑤 𝜌𝑙 ∇𝑃𝑙 − ( ) 𝜀𝑓𝑤 𝜌𝑙 𝐷𝐵 ∇𝑆𝐵 (4.42)
𝑆 𝜇𝑙 𝑆

68
4.3.7 Fourier’s Law

The Fourier’s law, also called the law of heat conduction, describes that the rate of heat
transfer through a material is proportional to the negative temperature gradient in the
direction of heat flow. In isotropic material, the heat flux is expressed as:

𝑞 = −𝑘∇𝑇 (4.43)

where 𝑘 is the effective thermal conductivity of the porous medium and ∇𝑇 is the
temperature gradient. In anisotropic materials, 𝑘 is a second-order tensor varying with
directions.

4.4 Numerical Solution

The primary state variables are chosen to be the temperature 𝑇, gas pressure 𝑃𝑔 , and
mass of water vapour per unit volume of gaseous material 𝜌̃𝑣 , expressed as a vector x:

𝑇
x = (𝑇, Pg , 𝜌̃𝑣 ) (4.44)

Governing equations (4.7) - (4.9) can be rewritten with respect to the chosen variables,
exploiting the constitutive laws and state equations developed in Section 4.3, leading
to a system of coupled partial differential equations:

𝐂ẋ − ∇ ∙ (𝐊∇x) = 0 (4.45)

where

𝐶𝑇𝑇 𝐶𝑇𝑃 𝐶𝑇𝑉 𝐾𝑇𝑇 𝐾𝑇𝑃 𝐾𝑇𝑉

𝐂 = [ 𝐶𝐴𝑇 𝐶𝐴𝑃 𝐶𝐴𝑉 ] , 𝐊 = [ 𝐾𝐴𝑇 𝐾𝐴𝑃 𝐾𝐴𝑉 ] (4.46)
𝐶𝑀𝑇 𝐶𝑀𝑃 𝐶𝑀𝑉 𝐾𝑀𝑇 𝐾𝑀𝑃 𝐾𝑀𝑉

69
Expanding (4.45) with the notation in (4.46) gives:

𝜕𝑇 𝜕𝑃𝑔 𝜕𝜌̃𝑣
𝐶𝑇𝑇 + 𝐶𝑇𝑃 + 𝐶𝑇𝑉 = ∇ ∙ (𝐾𝑇𝑇 ∇𝑇 + 𝐾𝑇𝑃 ∇𝑃𝑔 + 𝐾𝑇𝑉 ∇ρ̃𝑣 ) (4.47)
𝜕𝑡 𝜕𝑡 𝜕𝑡

𝜕𝑇 𝜕𝑃𝑔 𝜕𝜌̃𝑣
𝐶𝐴𝑇 + 𝐶𝐴𝑃 + 𝐶𝐴𝑉 = ∇ ∙ (𝐾𝐴𝑇 ∇𝑇 + 𝐾𝐴𝑃 ∇𝑃𝑔 + 𝐾𝐴𝑉 ∇ρ̃𝑣 ) (4.48)
𝜕𝑡 𝜕𝑡 𝜕𝑡

𝜕𝑇 𝜕𝑃𝑔 𝜕𝜌̃𝑣
𝐶𝑀𝑇 + 𝐶𝑀𝑃 + 𝐶𝑀𝑉 = ∇ ∙ (𝐾𝑀𝑇 ∇𝑇 + 𝐾𝑀𝑃 ∇𝑃𝑔 + 𝐾𝑀𝑉 ∇ρ̃𝑣 ) (4.49)
𝜕𝑡 𝜕𝑡 𝜕𝑡

After extensive algebraic manipulations elaborated in Appendix A, the coefficients in

the matrices of (4.46) are computed as follows:

𝜕𝜀𝐷 𝜕𝜌𝑙 𝜕𝜀𝑓𝑤 𝜕𝜌𝑙

𝐶𝑇𝑇 = (𝜌𝐶) + (𝜆𝐷 + 𝜆𝐸 ) (𝜌𝑙 + 𝜀𝐷 ) − 𝜆𝐸 (𝜌𝑙 + 𝜀𝑓𝑤 )
𝜕𝑇 𝜕𝑇 𝜕𝑇 𝜕𝑇
𝐶𝑇𝑃 = 0 (4.50)
𝜕𝜀𝑓𝑤
𝐶𝑇𝑉 = −𝜆𝐸 𝜌𝑙
{ 𝜕𝜌̃𝑣

𝜀𝑔 𝑃𝑔 𝜕𝜙 𝜕𝜀𝑓𝑤
𝐶𝐴𝑇 = − + 𝜌
̃𝑎 ( − )
𝑅𝑎 𝑇 2 𝜕𝑇 𝜕𝑇
𝜀𝑔
𝐶𝐴𝑃 = (4.51)
𝑅𝑎 𝑇
𝜀𝑔 𝑅𝑣 𝜕𝜀𝑓𝑤
𝐶𝐴𝑉 =− − 𝜌̃𝑎
{ 𝑅𝑎 𝜕𝜌̃𝑣

𝜕𝜙 𝜕𝜀𝑓𝑤 𝜕𝜌𝑙 𝜕𝜀𝐷

𝐶𝑀𝑇 = 𝜌̃𝑣 + (𝜌𝑙 − 𝜌̃𝑣 ) + (𝜀𝑓𝑤 − 𝜀𝐷 ) − 𝜌𝑙
𝜕𝑇 𝜕𝑇 𝜕𝑇 𝜕𝑇
𝐶𝑀𝑃 = 0 (4.52)
𝜕𝜀𝑓𝑤
𝐶𝑀𝑉 = 𝜀𝑔 + (𝜌𝑙 − 𝜌̃𝑣 )
{ 𝜕𝜌̃𝑣

70
 𝑆𝐵 𝐾𝐾𝑙 𝑇 𝜕𝜌𝑙 𝜌̃𝑣 𝑅𝑣 𝑇
𝐾𝑇𝑇 = 𝑘 − 𝜆𝐸 𝜀𝑓𝑤 𝜌𝑙 {(1 − ) 𝑅𝑣 𝜌𝑙 [(1 + ) ln ( )
𝑆 𝜇𝑙 𝜌𝑙 𝜕𝑇 𝑃𝑠𝑎𝑡
𝑇 𝜕𝑃𝑠𝑎𝑡 𝑆𝐵 𝐷𝐵 𝜕𝜀𝑓𝑤 𝜀𝑓𝑤 𝜕𝜙
+1 − ]+ ( − )}
𝑃𝑠𝑎𝑡 𝜕𝑇 𝑆𝜙 𝜕𝑇 𝜙 𝜕𝑇
(4.53)
𝑆𝐵 𝐾𝐾𝑙
𝐾𝑇𝑃 = −𝜆𝐸 𝜀𝑓𝑤 𝜌𝑙 (1 − )
𝑆 𝜇𝑙
𝑆𝐵 𝐾𝐾𝑙 𝑅𝑣 𝜌𝑙 𝑇 𝑆𝐵 𝐷𝐵 𝜕𝜀𝑓𝑤
𝐾 = −𝜆𝐸 𝜀𝑓𝑤 𝜌𝑙 [(1 − ) + ]
{ 𝑇𝑉 𝑆 𝜇𝑙 𝜌̃𝑣 𝑆𝜙 𝜕𝜌̃𝑣

𝜀𝑓𝑤 𝐷𝑎𝑣 𝜌̃𝑣 𝑃𝑔

𝐾𝐴𝑇 = −
𝜌̃𝑔 𝑅𝑎 𝑇 2
𝜌̃𝑎 𝐾𝐾𝑔 𝐷𝑎𝑣 𝜌̃𝑣
𝐾𝐴𝑃 = 𝜀𝑔 ( + ) (4.54)
𝜇𝑔 𝜌̃𝑔 𝑅𝑎 𝑇
𝜀𝑔 𝐷𝑎𝑣 𝜌̃𝑣 𝑅𝑣
𝐾𝐴𝑉 =− (𝜌̃𝑎 + )
{ 𝜌̃𝑔 𝑅𝑎

𝜀𝑔 𝐷𝑣𝑎 𝜌̃𝑣 𝑃𝑔 𝑆𝐵 𝐾𝐾𝑙 𝑇 𝜕𝜌𝑙 𝜌̃𝑣 𝑅𝑣 𝑇

K MT = 2
+ 𝜀𝑓𝑤 𝜌𝑙 {(1 − ) 𝑅𝑣 𝜌𝑙 [(1 + ) ln ( )
𝜌̃𝑔 𝑅𝑎 𝑇 𝑆 𝜇𝑙 𝜌𝑙 𝜕𝑇 𝑃𝑠𝑎𝑡
𝑇 𝜕𝑃𝑠𝑎𝑡 𝑆𝐵 𝐷𝐵 𝜕𝜀𝑓𝑤 𝜀𝑓𝑤 𝜕𝜙
+1 − ]+ ( − )}
𝑃𝑠𝑎𝑡 𝜕𝑇 𝑆𝜙 𝜕𝑇 𝜙 𝜕𝑇
(4.55)
𝐾𝐾𝑔 𝐷𝑣𝑎 𝑆𝐵 𝐾𝐾𝑙
𝐾𝑀𝑃 = 𝜀𝑔 𝜌̃𝑣 ( − ) + 𝜀𝑓𝑤 𝜌𝑙 (1 − )
𝜇𝑔 𝜌̃𝑔 𝑅𝑎 𝑇 𝑆 𝜇𝑙
𝜀𝑔 𝐷𝑣𝑎 𝜌̃𝑣 𝑅𝑣 𝑆𝐵 𝐾𝐾𝑙 𝑅𝑣 𝜌𝑙 𝑇 𝑆𝐵 𝐷𝐵 𝜕𝜀𝑓𝑤
𝐾𝑀𝑉 = (𝜌̃𝑎 + ) + 𝜀𝑓𝑤 𝜌𝑙 [(1 − ) + ]
{ 𝜌̃𝑔 𝑅𝑎 𝑆 𝜇𝑙 𝜌̃𝑣 𝑆𝜙 𝜕𝜌̃𝑣

4.4.1 Initial and Boundary Conditions

Initial and boundary conditions need to be defined so as to ensure that a unique solution
exists for the numerical model. Initial conditions specify the initial state of all primary
variables at the beginning of heating, i.e.,

x = x 0 = (𝑇 0 , 𝑃𝑔0 , 𝜌̃𝑣0 ), 𝑎𝑡 𝑡 = 0 (4.56)

71
Boundary conditions can normally be categorized into three types. The first type, also
called the Dirichlet boundary condition, is to prescribe known values of the primary
variables on the boundary:

𝑇 = 𝑇∞ 𝑜𝑛 Γ𝑇1 , 𝑃𝑔 = 𝑃𝑔 ∞ 𝑜𝑛 Γ𝑔1 , 𝜌̃𝑣 = 𝜌̃𝑣 ∞ on 𝑜𝑛 Γ𝑣1 (4.57)

The second type, also known as the Neumann boundary condition, is to prescribe fluxes
on the boundary, in terms of the normal derivative of primary variables:

𝑘∇𝑇 ∙ 𝒏 = 𝑞𝑇 𝑜𝑛 Γ𝑇2
−𝐽𝑔 ∙ 𝒏 = 𝑞𝑔 𝑜𝑛 Γ𝑔2 (4.58)
−𝐽𝑣 ∙ 𝒏 = 𝑞𝑣 𝑜𝑛 Γ𝑣2

The third type, also called the Cauchy boundary condition, could be regarded as
imposing both a Dirichlet and Neumann boundary condition. It specifies both the value
and normal derivative of the state variables on the boundary of the domain:

−𝑘∇𝑇 ∙ 𝒏 = ℎ𝑞(𝑇 − 𝑇∞ ) + ℎ𝑟(𝑇 − 𝑇∞ ) 𝑜𝑛 Γ𝑇3

𝐽𝑔 ∙ 𝒏 = 𝛽 (𝜌̃𝑔 − 𝜌̃𝑔 ∞ ) 𝑜𝑛 Γ𝑔3 (4.59)
𝐽𝑣 ∙ 𝒏 = 𝛽(𝜌̃𝑣 − 𝜌̃𝑣 ∞ ) 𝑜𝑛 Γ𝑣3

where the whole boundary Γ = Γπ1 ∪ Γ𝜋2 ∪ Γ𝜋3 , 𝒏 is the unit vector normal to the
boundary, ℎ𝑞 and ℎ𝑟 are the convection and radiation heat transfer coefficient, 𝛽 is the
coefficient of water vapour mass transfer on the boundary. 𝑇∞ , 𝑃𝑔 ∞ and 𝜌̃𝑣 ∞ are

respectively the temperature, gas pressure and vapour content of the surrounding
atmosphere.

First, the energy conservation on the boundary is analysed:

72
 ∂T 0 0
k − (𝐻
⏟ 𝑔 − 𝐻𝑔 )𝐽𝑔 ∙ 𝒏 + (𝐻
⏟ 𝑔 − 𝐻𝑔 )𝛽 (𝜌
̃𝑔 − 𝜌̃𝑔 ∞ ) + 𝜆
⏟𝐸 𝐽𝑓𝑤 ∙ 𝒏
⏟ ∂n
𝑎 𝑏 𝑐 𝑑
(4.60)
⏟𝑞𝑟 (𝑇 − 𝑇∞ ) = 0
+ℎ
𝑒

where term a is the energy transferred from inside the domain to the boundary by heat
conduction; term b is the energy accumulated in the gaseous mixture and entering the
boundary from the inside domain; term c represents the energy accumulated in the
gaseous mixture that leaves the boundary; term d represents the energy consumed for
evaporation of liquid water on the boundary; term e is the heat dissipated by convection
and radiation to the surrounding environment. 𝐻𝑔 and 𝐻𝑔0 are the gaseous enthalpies at
current and ambient conditions, ℎ𝑞𝑟 is the combined convection-radiation heat transfer
coefficient on the boundary.

From the mass conservation of gaseous mixture on the boundary, the gaseous material
transferred from inside the domain to the boundary must be equal to that dissipated into
the surrounding environment:

𝐽𝑔 ∙ 𝒏 − 𝛽 (𝜌̃𝑔 − 𝜌̃𝑔 ∞ ) = 0 (4.61)

It could be reasonably assumed that the surface of concrete exposed to the atmosphere
will be dried shortly after the fire commences, if not already dried during service life,
and the free water (both liquid and adsorbed) flux can thus be omitted, i.e., 𝐽fw = 0.
Finally, (4.60) could be reduced into the form of Cauchy boundary condition as shown
in (4.59) and further simplified as:

∂T ∂T ℎ𝑞𝑟
−k = ℎ𝑞𝑟 (𝑇 − 𝑇∞ ) ⇒ = (𝑇 − 𝑇) (4.62)
∂n ∂n 𝑘 ∞

The gas pressure gradient on the boundary will always be zero, since the gas pressure
will always equal to the atmospheric pressure as long as the surface is in contact with
the surrounding atmosphere:

73
 ∂Pg
= 0 ⇒ 𝑃𝑔 = 𝑃𝑔 ∞ (4.63)
𝜕𝑛

The mass conservation of water vapour on the boundary is reached as shown in the last
equation of (4.59) following the same manner of the gaseous mixture, where the water
vapour flux is written as:

−𝐽𝑣 = (𝐾𝑉𝑇 ∇𝑇 + 𝐾𝑉𝑃 ∇𝑃𝑔 + 𝐾𝑉𝑉 ∇𝜌̃𝑣 ) (4.64)

Substituting back into (4.59) yields:

∂T 𝜕𝑃𝑔 ∂𝜌̃𝑣
− (K VT + K VP + 𝐾𝑉𝑉 ) = 𝛽(𝜌̃𝑣 − 𝜌̃𝑣 ∞ ) (4.65)
∂n 𝜕𝑛 𝜕𝑛

Substituting (4.62) - (4.63) into (4.65) and rearranging give the water vapour gradient
across the boundary as:

∂ρ̃𝑣 𝐾𝑉𝑇 ℎ𝑞𝑟 𝛽

= (𝑇 − 𝑇∞ ) + (𝜌̃ − 𝜌̃𝑣 ) (4.66)
𝜕𝑛 𝐾𝑉𝑉 𝑘 𝐾𝑣𝑣 𝑣 ∞

4.4.2 Spatial Discretization by Finite Element

The general finite element method (Cheung, 1996) is employed to discretise the
governing equations in space, with the state variables expressed in terms of their nodal
values, i.e., x ≅ Nx̅:

𝑇 = 𝑇(𝑡) ≅ 𝑁𝑇̅(𝑡)
𝑃𝑔 = 𝑃𝑔 (𝑡) ≅ 𝑁𝑃̅𝑔 (𝑡) (4.67)
ρ̃𝑣 = ρ̃𝑣 (𝑡) ≅ 𝑁𝜌̃̅𝑣 (𝑡)

74
The variational or weak form of the mass conservation equations is obtained by the
Galerkin Method (Weighted residuals), with respect to the problem domain Ω:

𝜕x̅
∫ 𝑁 𝑇 {𝑪𝑁 − ∇ ∙ (𝑲∇𝑁x̅)} 𝑑Ω = 0 (4.68)
Ω 𝜕𝑡

Exploiting the product rule and linear property of the divergence operator:

𝑁 𝑇 ∇ ∙ (𝑲∇𝑁x̅) = ∇ ∙ (𝑁 𝑇 𝑲∇𝑁x̅) − ∇𝑁 𝑇 𝐾∇(𝑁x̅) (4.69)

(4.68) can thus be rewritten as:

𝜕x̅
∫ 𝑁 𝑇 𝑪𝑁 𝑑Ω − ∫ ∇ ∙ (𝑁 𝑇 𝑲∇𝑁x̅) 𝑑Ω + ∫ ∇𝑁 𝑇 𝑲∇(𝑁x̅) 𝑑Ω = 0 (4.70)
Ω 𝜕𝑡 Ω Ω

Applying the Divergence theorem in two-dimension, also called the Green’s theorem,
to the second term in (4.70), the weak form of governing equations is finally reached:

𝜕x̅ 𝜕𝑁
∫ 𝑁 𝑇 𝑪𝑁 𝑑Ω + ∫ ∇𝑁 𝑇 𝑲∇𝑁x̅ 𝑑Ω = ∫ 𝑁 𝑇 𝑲 x̅ 𝑑Γ (4.71)
Ω 𝜕𝑡 Ω Γ 𝜕𝑛

with the natural boundary conditions derived in (4.62) and (4.66) as:

𝜕𝑁
x̅ = 𝑅∞ − 𝑹𝒌 𝑁x̅ (4.72)
𝜕𝑛

or more explicitly,

∂T
𝐾 0 −1 ℎ𝑞𝑟 𝑇∞ 𝑘 0 −1 ℎ𝑞𝑟 0 𝑇
{ ∂n } = [ 𝑇𝑇 ] { }−[ ] [ ]{ } (4.73)
𝜕𝜌̃𝑣 ⏟𝐾𝑉𝑇 𝐾𝑉𝑉 𝛽𝜌̃𝑣 ∞ 𝐾
⏟ 𝑉𝑇 𝐾𝑉𝑉 0 𝛽 𝜌̃𝑣
𝜕𝑛 𝐿𝑜𝑎𝑑 𝑣𝑒𝑐𝑡𝑜𝑟 𝑆𝑡𝑖𝑓𝑓𝑛𝑒𝑠𝑠 𝑚𝑎𝑡𝑟𝑖𝑥

It may be noted that the boundary integral in (4.71) contains only the Neumann or
Cauchy type boundary. Because by Galerkin approximation, the arbitrary weight

75
functions are chosen on the condition that the essential (Dirichlet) boundary conditions
are always satisfied. Therefore, the term corresponding to boundary Γ1 is equal to zero.
(4.71) can then be rewritten in a more compact form as:

𝜕x̅
𝐂̂ ̂ x̅ = 𝐑
+𝐊 ̂ (4.74)
∂t

where 𝐂̂ and 𝐊
̂ are the global capacitance and stiffness matrices, and 𝐑
̂ is the global
force vector of the system:

𝐂̂ = ∑𝐂̂ 𝑒 , ̂ = ∑(𝐊
𝐊 ̂𝑒 + 𝐊
̂ 𝑒𝑅 ) , ̂ = ∑𝐑
𝐑 ̂𝑒 (4.75)

The global matrices are the sum of the element matrices:

𝐂̂ 𝑒 = ∫ 𝑁 𝑇 𝑪𝑁𝑑Ω , ̂ 𝑒 = ∫ 𝑁 𝑇 𝑲𝑅∞ 𝑑Γ
𝐑 (4.76)
Ωe Γe

̂ 𝒆 = ∫ ∇𝑁 𝑇 𝑲∇𝑁 𝑑Ω,
𝑲 ̂ 𝑒𝑅 = ∫ 𝑁 𝑇 𝑲𝑹𝒌 𝑁𝑑Γ
𝐊 (4.77)
Ωe Γe

4.4.3 Temporal Discretization by Finite Difference

The finite difference method (Özışık, 1994) is employed for temporal discretization to
solve the system in (4.74). The most commonly used time integration methods are those
of the one-parameter family or α-family of approximation, also called the generalised
trapezoidal method or the generalised midpoint rule (Reddy, 2010):

1
𝐂̂(x θ )(x n+1 − x n ) + θ𝐊
̂ (x θ )x n+1 = 𝐑
̂ (x θ ) − (1 − θ)𝐊
̂ (x θ )x n (4.78)
𝛥𝑡

where time interval 𝛥𝑡 = 𝑡 𝑛+1 − 𝑡 𝑛 , the superscript n denotes the nth time step while
x θ represents the solution computed at a dummy time step ( 𝑡 𝑛 + θΔ𝑡) , as x θ =
(1 − θ)x n + θx 𝑛+1 . 𝐂̂(x θ ) is the capacitance matrix evaluated using the solution x θ ,
̂ (x θ ) and 𝐑
the same applies to 𝐊 ̂ (x θ ). With parameter θ ranging from 0 to 1, different

76
members of the family could be identified. The several most well-known approximation
schemes are:

• θ = 0 , the forward difference (forward Euler) scheme, explicit and

conditionally stable;
1
• θ = , the Crank-Nicolson scheme, implicit and unconditionally stable;
2
2
• θ = , the Galerkin scheme, implicit and unconditionally stable;
3

• θ = 1 , the backward difference (backward Euler) scheme, implicit and

unconditionally stable

1
Though the schemes are unconditionally stable when θ ≥ , they may still suffer from
2

“noise” or temporal oscillations in the solution if the time step is not suitably controlled.
The Crank-Nicolson (CN) scheme is chosen to be applied in this study, due to its
superior second-order accuracy in both space and time. An implicit predictor-corrector
integration method that is second-order accurate in time is developed by combining the
CN scheme with an explicit prediction method. The prediction method employed here
is the Adams-Bashforth predictor, which is second-order accurate, variable step and
explicit, given by:

Δ𝑡 𝑛 Δ𝑡 𝑛 Δ𝑡 𝑛
xpn+1 n
=x + [(2 + 𝑛−1 ) ẋ − ( 𝑛−1 ) ẋ n−1 ]
n (4.79)
2 Δ𝑡 Δ𝑡

where the subscript p denotes a predicted solution for time step 𝑡 𝑛+1 , Δ𝑡 𝑛 = 𝑡 𝑛+1 − 𝑡 𝑛 ,
and similarly Δ𝑡 𝑛−1 = 𝑡 𝑛 − 𝑡 𝑛−1 . The “velocity” vectors (ẋ 𝑛 and ẋ 𝑛−1 ) are updated
using the new solutions and could be computed with:

Δ𝑡 𝑛 n+1
ẋ n+1 = (x − x n ) − ẋ n (4.80)
2

1
The compatible CN scheme (θ = ) corrector for use with (4.79) is formulated in the
2

form of the trapezoid rule:

77
 2 2
[ ̂(x θ ) + 𝐊
𝐂 ̂ (x θ )] x n+1 = 𝐂̂(x θ )x n + 𝐂̂(x θ )ẋ n + 𝐑
̂ (x θ ) (4.81)
Δ𝑡 𝑛 Δ𝑡 𝑛

Assuming all solutions and “velocity” vectors are known at the beginning of a new time
step and the new time interval has been determined, the time integration procedure to
advance the solution from time 𝑡 𝑛 to 𝑡 𝑛+1 can be summarized in the following steps:

1) A tentative solution vector, xpn+1 , is computed using (4.79);

2) Solve for the “true” solution x n+1 using (4.81). The predicted solution xpn+1 is

used to estimate x θ to evaluate the matrices;

3) Update the “velocity” vectors using the converged solution x n+1 and (4.80);
4) Determine the new time interval Δ𝑡 𝑛+1 . This could be either based on the
analysis of previous time truncation error or simply omitted if a constant time
step is adopted;
5) Return to step 1) for the next time increment until the end of heating.

It may be noted that in step 1) the predictor equation (4.79) requires two “velocity”
vectors available at the time step and that in step 2) the process involves iterative
solution of (4.81). The starting procedure to initialize a predictor-corrector scheme, as
well as the iterative method used to solve the nonlinear system will be discussed in
details in Appendix B. Also, the procedure for a dynamic selection of time step is
presented.

78
Figure 4.1 Curved interface between gas and liquid by Bažant and Jirásek (2018)

79
5 COUPLED THERMO-MECHANICAL MODEL

5.1 Introduction

The second component of the THM model, namely the thermo-mechanical (TM) model,
dealing with the material degradation resulting from both the thermal and mechanical
loading, is introduced in this chapter. A spalling criterion considering simultaneously
the effect of thermo-hygro (pore pressure) and thermo-mechanical (thermal stresses)
behaviour is subsequently developed and presented. The spalling criterion is then
employed in the coupled THM model to determine the occurrence of fire-induced
spalling. While the TH model introduced in Chapter 4 is based mainly on the Glasgow
model, the TM model introduced in this chapter is based mainly on the Padua model
(Gray and Schrefler, 2001, Lewis and Schrefler, 1998, Gawin et al., 1999, Gawin et al.,
2003), with some modifications adopted from the Glasgow model.

5.2 Momentum Conservation

The macroscopic momentum conservation equations are deducted on the basis of the
microscopic balance equations, where the moist concrete is assumed to be non-polar
material. Local thermodynamic equilibrium still applies at all points of the porous
medium. The rather lengthy deduction is given in details in the book of Lewis and
Schrefler (1998). The volume averaged linear momentum conservation equation for
phase π:

∇ ∙ (𝐭 π ) + 𝜌𝜋 𝐠 + 𝜌𝜋 [𝑒 𝜋 (𝜌𝒓̇ ) + 𝐭̂ π ] = 0 (5.1)

where phase π = s, l, g denotes the solid skeleton, liquid water and gaseous mixture
respectively. 𝐭 π is the macroscopic partial stress tensor for phase π . 𝜌𝜋 𝐭̂ π and
𝜌𝜋 𝑒 𝜋 (𝜌𝒓̇ ) are the volumetric exchange terms of linear momentum with other phases
due to mechanical interaction, and due to phase changes or chemical reactions
respectively, with the constraint applied:

80
 ∑ 𝜌𝜋 [ [𝑒 𝜋 (𝜌𝒓̇ ) + 𝐭̂ π ] = 0 (5.2)
π

Introducing the total stress tensor

𝛔 = 𝐭 𝑠 + 𝐭𝑙 + 𝐭 𝑔, (5.3)

and the averaged density of the multiphase porous medium

𝜌 = (1 − 𝜙)𝜌𝑠 + 𝜙𝑆𝜌𝑙 + 𝜙(1 − 𝑆)𝜌𝑔 , (5.4)

summing up (5.1) of all phases and assuming continuity of stresses at the fluid-solid
interfaces, the macroscopic linear momentum conservation equation for the whole
multiphase medium is written as:

∇ ∙ 𝛔 + 𝜌𝑔 = 0 (5.5)

As regards the volume averaged angular momentum, Hassanizadeh and Gray (1979b)
have shown that the partial stress tensor is symmetric for non-polar material at
macroscopic level, namely

𝐭 π = (𝐭 π )T (5.6)

Therefore, the coupling vectors of angular momentum between the phases vanish after
summing up.

5.3 Constitutive Laws

The constitutive laws are deduced from the macroscopic entropy inequalities for the
constituents in the porous medium to ensure these relationships obey the second law of
thermodynamics. For the sake of simplicity, small displacements are assumed for the

81
solid phase. In the following, stresses in tension for the solid phase and pore pressures
in compression for the fluids are taken to be positive.

5.3.1 Bishop Stress

Effective stress is defined as the stress which controls stress-strain, volume change and
strength behaviour in a porous medium. In fully saturated conditions it is expressed as:

𝛔′ = 𝛔 + Pl 𝐈 = 𝛔 + Ps 𝐈 (5.7)

where 𝛔′ is the effective stress, 𝛔 the total stress, 𝐈 the second-order unit tensor, Pl the
pore water pressure and Ps the pressure exerted on solid phase by the surrounding fluids.

The effective stresses for partially saturated porous medium was first suggested by
Bishop (1959), where Ps is a weighted sum of the pore pressures exerted by the wetting
(water) phase and non-wetting (air) phase, also called the intrinsically averaged (or
mean) pressure of the fluid phases, written as:

Ps = 𝜒𝑃𝑙 + (1 − 𝜒)𝑃𝑛 (5.8)

where 𝑃𝑙 and 𝑃𝑛 is the pressure in the wetting and non-wetting phase, χ is the Bishop’s
parameter normally expressed as a function of the degree of saturation. For the time
being, the hypothesis of incompressible grains is introduced by assuming the stress state
induced by this weighted pressure would not cause any deformation of the grains. The
deformation of the solid skeleton is a function of grain arrangement only. Substituting
(5.8) into (5.7) yields

𝛔′ = 𝛔 + [𝜒𝑃𝑙 + (1 − 𝜒)𝑃𝑛 ]𝐈 (5.9)

If the Bishop’s parameter is replaced by the degree of saturation 𝑆𝑙 as defined in (4.29),

(5.9) then reads (for the purpose of consistency, 𝑆 is used instead of 𝑆𝑙 ):

82
 𝛔′ = 𝛔 + [𝑆𝑃𝑙 + 𝑆𝑛 𝑃𝑛 ]𝐈 (5.10)

where 𝑆𝑛 is the degree of saturation with the non-wetting phase, follows immediately
that 𝑆𝑛 = 1 − 𝑆. This form of effective stresses could be obtained either by volume
averaging for bulk materials (Lewis and Schrefler, 1987) or from the entropy inequality
using the Colemann–Noll procedure (Coleman and Noll, 1963, Hassanizadeh and Gray,
1980). However, it is noteworthy that the definition of effective stresses given by (5.9)
and (5.10) are actually at odds. The Bishop’s parameter is related to the fractional area
of water in contact with the solid, whereas the degree of saturation 𝑆𝜋 depends on the
volume occupied by the phase π. Also, to avoid confusion with the mechanical effective
stress related to the damage mechanics, the effective stresses in (5.9) and (5.10) related
to the pore pressure are preferably referred to as Bishop stresses since then. A
dimensionless corrective term called the Biot’s coefficient is introduced for more
generality:

K
𝑏 =1− (5.11)
Ks

where K and K s is the bulk modulus of the skeleton and solid phase (grains)
respectively. The more general expression of Bishop stresses accounting for the
deformability of grains now reads:

𝛔′ = 𝝈 + 𝑏[𝑆𝑃𝑙 + 𝑆𝑛 𝑃𝑛 ]𝐈 = 𝝈 + 𝑏Ps 𝐈 (5.12)

In the Padua and Glasgow model, the solid phase pressure Ps is computed based on the
solid saturation point Sssp mentioned in Section 4.3, but with slightly different
expressions. In this study, the definition of the Glasgow model is adopted, assuming
the adsorbed water to be part of the solid skeleton and applies no pressure. The solid
phase pressure is then determined by the weighted liquid and gas pressure based on the
remaining pore volume as:

83
 𝑃𝑔 − 𝑃𝑔∞ 𝑓𝑜𝑟 𝑆 ≤ 𝑠𝑠𝑠𝑝
Ps = { 𝑆 − 𝑆𝑠𝑠𝑝 1−𝑆 (5.13)
( ) 𝑃𝑙 + ( ) 𝑃 − 𝑃𝑔∞ 𝑓𝑜𝑟 𝑆 > 𝑆𝑠𝑠𝑝
1 − 𝑆𝑠𝑠𝑝 1 − 𝑆𝑠𝑠𝑝 𝑔

With introduction of Bishop stresses, the macroscopic linear momentum equation (5.5)
could be rewritten as:

∇ ∙ (𝛔′ − 𝑏Ps 𝐈) + 𝜌𝑔 = 0 (5.14)

5.3.2 Decomposition of Total Strain

The total strain ε could be decomposed into the following components:

𝜀 = 𝜀 𝑒 + 𝜀 𝑝 + 𝜀 𝑓𝑡 + 𝜀 𝑐 + 𝜀 𝑡𝑚 (5.15)

where 𝜀 𝑒 is the elastic strain, 𝜀 𝑝 the plastic strain, 𝜀 𝑓𝑡 the free thermal strain, 𝜀 𝑐 the
creep strain, and 𝜀 𝑡𝑚 the thermo-mechanical strain. The elastic strain 𝜀 𝑒 is related to
the instantaneous mechanical effective stresses. The plastic strain is not considered in
this study since an isotropic elastic-damage model is adopted. The creep and thermo-
mechanical strains are often combined and referred to as the load induced thermal strain
(LITS). The LITS is irreversible and only appears in first heating. All of these strain
components could be expressed as a function of the Bishop stress and/or temperature.
In strain rate form, the breakdown of total strain gives:

𝜀̇ = 𝜀̇ 𝑒 (𝛔′ , 𝑇) + 𝜀̇ 𝑓𝑡 (𝑇) + 𝜀̇𝑙𝑖𝑡𝑠 (𝛔′ , 𝑇) (5.16)

5.3.2.1 Free Thermal Strain

The isotropic free thermal strain is a function of temperature only, having a simple,
empirically-based expression:

𝑓𝑡
𝜀̇𝑖𝑗 = 𝛼𝑓𝑡 𝑇̇𝛿𝑖𝑗 (5.17)

84
where 𝛼𝑓𝑡 is the coefficient of free thermal strain, and 𝛿𝑖𝑗 is the Kronecker delta. Since
the free thermal strain is strongly dependent on the type and amount of aggregate, the
function of coefficient 𝛼𝑓𝑡 is normally determined according to the aggregate type
(siliceous or carbonate). (5.17) could be rewritten in matrix-vector form using Voigt’s
notation:

𝜀̇ 𝑓𝑡 = 𝛼𝑓𝑡 𝒎𝑇̇ (5.18)

where 𝒎 = (1 1 1 0 0 0)𝑇 (5.19)

5.3.2.2 Load Induced Thermal Strain

The observed difference in strains between concrete specimens heated under, and prior
to, loading is defined as the load induced thermal strain (LITS). The two components
mentioned in (5.16), creep strain and thermal-mechanical strain, of the LITS are
recognized to have little experimental distinction, and therefore, the subdivision in
(5.16) is in fact theoretical only and has little practical implications. Moreover, it has
been suggested that the creep component of the LITS could be ignored for the short
duration, rapid heating. The LITS rate is then entirely represented by the thermal-
mechanical strain rate, whose general formulation reads:

𝑡𝑚
𝜀̇𝑖𝑗 = (𝛾1 𝝈′𝑘𝑘 𝛿𝑖𝑗 + 𝛾2 𝝈′𝑖𝑗 )𝑇̇ (5.20)

where 𝛔′ is the Bishop stress, 𝛾1 and 𝛾2 are experimentally determined coefficients and
have the following expressions for the uniaxial case:

𝛽𝑡𝑚 𝛽𝑡𝑚
𝛾1 = −𝜈tc , 𝛾2 = (1 + 𝜈𝑡𝑐 ) (5.21)
𝑓𝑐0 𝑓𝑐0

where 𝜈tc is the Poisson’s ratio of transient creep, 𝑓𝑐0 the initial compressive strength at
room temperature and 𝛽𝑡𝑚 the coefficient of uniaxial thermo-mechanical strain.
Substituting (5.21) into (5.20) gives:

85
 𝛽𝑡𝑚
𝑡𝑚
𝜀̇𝑖𝑗 = ′ ′ ̇ ̇
0 (−𝜈𝑡𝑐 𝝈𝑘𝑘 𝛿𝑖𝑗 + (1 + 𝜈𝑡𝑐 )𝝈𝑖𝑗 )𝑇 𝑓𝑜𝑟 𝑇 > 0 (5.22)
𝑓𝑐

In the work of Pearce et al. (2004), the positive (tensile) projection of the stress tensor
is not considered in (5.22) by setting the coefficient of tensile uniaxial thermo-
+
mechanical strain 𝛽𝑡𝑚 to zero. The reason was that all experimental tests for LITS are
done under compressive loading, and the lack of evidence on the LITS due to tensile
+
loading made the applicability of 𝛽𝑡𝑚 unclear. However, later experimental findings
showed that it is plausible to assume tensile loadings are responsible for part of the
LITS (Neville, 2011, Mindeguia et al., 2013). Therefore, a complete stress tensor,
including both the positive (tensile) and negative (compressive) projections, is instead
used in this study. Nielsen et al. (2004) proposed a bi-parabolic function which fits the
results from multiple studies, to describe the uniaxial thermo-mechanical strain:

𝝈′
𝜀 𝑡𝑚 = 𝑦 (5.23)
𝑓𝑐0

A𝜃̅ 2 + 𝐵𝜃̅ 𝑓𝑜𝑟 0 ≤ 𝜃̅ ≤ 𝜃̅ ∗ = 4.5

with y={ (5.24)
𝐶(𝜃̅ − 𝜃̅ ∗ )2 + 𝐴(2𝜃̅ − 𝜃̅ ∗ )𝜃̅ ∗ + 𝐵𝜃̅ 𝑓𝑜𝑟 𝜃̅ > 𝜃̅ ∗

where 𝜃̅ is a dimensionless normalised temperature defined by

𝑇 − 𝑇0
𝜃̅ = (5.25)
100℃

and 𝜃̅ ∗ = 4.5 is the transition dimensionless temperature between the two expressions,
equivalent to 470℃. The thermo-mechanical strain rate thus reads:

𝝈′ 𝑑𝑦 𝑑𝜃̅ 𝝈′
𝜀̇ 𝑡𝑚 = 𝑇̇ = 𝛽 𝑇̇ (5.26)
𝑓𝑐0 𝑑𝜃̅ 𝑑𝑇 𝑓𝑐0 𝑡𝑚

86
 2A𝜃̅ + 𝐵 𝑓𝑜𝑟 0 ≤ 𝜃̅ ≤ 𝜃̅ ∗
where 𝛽𝑡𝑚 = 0.01 × { (5.27)
2𝐶(𝜃̅ − 𝜃̅ ∗ ) + 2𝐴𝜃̅ ∗ + 𝐵 𝑓𝑜𝑟 𝜃̅ > 𝜃̅ ∗

Having 𝛽𝑡𝑚 defined in (5.27), the thermo-mechanical strain rate in (5.22) could now be
used to express the LITS rate in Voigt’s notation:

𝛽𝑡𝑚
𝜀̇ 𝑙𝑖𝑠𝑡 = 𝐌𝝈′ 𝑇̇ (5.28)
𝑓𝑐0

1 −νtc −νtc 0 0 0
−νtc 1 −νtc 0 0 0
−νtc −νtc 1 0 0 0
with 𝐌= (5.29)
0 0 0 2(1 + νtc ) 0 0
0 0 0 0 2(1 + νtc ) 0
[ 0 0 0 0 0 2(1 + νtc )]

It may be noted that 𝐌 is identical to the isotropic elastic compliance tensor with the
Poisson’s ration replaced by the transient creep Poisson’s ration νtc . The LITS rate in
(5.28) only depends on the temperature variation, meaning that a variation of stress (𝝈̇ ′ )
at a constant temperature would not cause the variation of LITS.

5.3.3 Damage of Concrete

The starting basis of the damage model development of concrete in this study is the
single-scalar, isotropic mechanical damage model, whose typical expression reads:

𝛔′ = (1 − 𝐷)𝐄𝟎 : 𝜺𝑒 (5.30)

in which 𝐄𝟎 is the initial elastic stiffness matrix in the form of fourth order tensor, 𝜺𝑒
the elastic strain tensor, 𝐷 the mechanical damage parameter accounting for the
degradation of elastic stiffness due to crack development as defined in (5.32), where
𝐸(𝑇) and 𝐸0 (𝑇) is the Young’s modulus at temperature T of the mechanically damaged
and undamaged material. This model assumes the damaged material to behave

87
elastically and remain isotropic at any given temperature. The effective stress, defined
as the undamaged material stress, can then be computed as:

𝝈′
̂′ = 𝐄𝟎 : 𝜺𝑒 =
𝛔 (5.31)
1−𝐷

𝐸(𝑇)
𝐷 =1− (5.32)
𝐸0 (𝑇)

with the assumption that the strain under the damaged state is equal to the strain under
the equivalent undamaged state subjected to the effective stress:

𝛆 = 𝛆̂ (5.33)

In addition to the mechanical damage, concrete at elevated temperature may also

experience complicated physical and chemical changes, resulting in the concrete
dehydration and degradation of elastic stiffness. To describe these changes, the classical
damage model is modified to include an additional damage parameter in a
multiplicative manner such that:

𝛔′ = (1 − 𝐷)(1 − 𝑉)𝐄𝟎 : 𝜺𝑒 = 𝐄𝒔𝒆𝒄 : 𝜺𝑒 (5.34)

where 𝐄𝒔𝒆𝒄 is the secant stiffness defined as 𝐄𝒔𝒆𝒄 = (1 − 𝐷)(1 − 𝑉)𝐄𝟎 , 𝑉 is called the
thermo-chemical damage parameter, accounting for degradation of material stiffness
due to thermally induced microcracks and the degradation of concrete strength
properties. The microcracks are mainly induced by both the incompatible thermal
expansion of cement paste and aggregate, and the local volume increase of dehydration
products. The degradation of material properties is mainly caused by the dehydration
process. The effective stress in (5.31) is now computed as:

′ 𝑒
𝝈′
̂ = 𝐄𝟎 : 𝜺 =
𝛔 (5.35)
(1 − 𝐷)(1 − 𝑉)

88
Similar to the mechanical damage parameter 𝐷 , the thermo-chemical damage
parameter 𝑉 could be defined in terms of the Young’s modulus. In (5.36), 𝐸0 (𝑇𝑎 ) is the
mechanically undamaged Young’s modulus at room temperature.

𝐸0 (𝑇)
𝑉 =1− (5.36)
𝐸0 (𝑇𝑎 )

With (5.16) and (5.35), the stress rate could be evaluated from (5.34) as:

𝛔̇ ′ = 𝐄𝒔𝒆𝒄 (𝜺̇ − 𝜺̇ 𝑓𝑡 − 𝜺̇ 𝑙𝑖𝑡𝑠 ) − 𝐷̇(1 − 𝑉)𝛔

̂′ − 𝑉̇ (1 − 𝐷)𝛔
̂′ (5.37)

The evolution of the mechanical and thermo-chemical damage is governed by different

functions and will be discussed separately in the following.

5.3.3.1 Mechanical Damage

The mechanical damage function is expressed as:

𝑓 𝑚𝑑 = 𝜀̅ − 𝜅 𝑚𝑑 (5.38)

with κmd = max{𝑚𝑎𝑥(𝜀̅) , 𝜅0𝑚𝑑 (𝑇)} (5.39)

𝑓𝑡 (𝑇)
and 𝜅0𝑚𝑑 (𝑇) = (5.40)
𝐸(𝑇)

where 𝜀̅ is the non-local equivalent strain measure – a scalar mapping of the tensorial
strain, 𝜅 𝑚𝑑 is the history parameter for mechanical damage, which takes the larger of
either the maximum value obtained by 𝜀̅ or the temperature dependent threshold strain
𝜅0𝑚𝑑 (𝑇). The threshold strain measure 𝜅0𝑚𝑑 (𝑇) is defined as the ratio of tensile strength
to the Young’s modulus at a given temperature 𝑇 . Therefore, the growth of the
mechanical damage is determined by:

89
 𝐷̇ ≥ 0 𝑖𝑓 𝑓 𝑚𝑑 = 0
{ (5.41)
𝐷̇ = 0 𝑖𝑓 𝑓 𝑚𝑑 < 0 𝑎𝑛𝑑 𝜅̇ 𝑚𝑑 = 0

(5.41) implies that, during damage growth 𝜀̅ = 𝜅 𝑚𝑑 and correspondingly during

unloading or reloading conditions, 𝜀̅ < 𝜅 𝑚𝑑 with 𝜅 𝑚𝑑 remaining at the maximum
history value. The equivalent strain measure could be obtained by the following three
definitions:

Energy Release Rate Definition

1
ε̅ = √ 𝜺: 𝐄𝟎 : 𝜺 (5.42)
𝐸

Mazars Definition

3
|𝑥| + 𝑥
ε̅ = √∑(⟨𝜀𝑖 ⟩+ )2 (⟨𝑥⟩+ = ) (5.43)
2
𝑖=1

where ⟨𝜀𝑖 ⟩+ represents the positive components of the principal strains.

Modified von Mises Definition

ε̅ = A𝐼1 + √𝐴2 𝐼12 + 𝐵𝐽2 (5.44)

g−1 1
with A= ,𝐵 = (5.45)
2g(1 − 2ν) 2g(1 + 𝜈)2

where g is the ratio of compressive to tensile strength, ν is the Poisson’s ratio, 𝐼1 and
𝐽2 is the first and second invariant of the strain tensor and the deviatoric strain tensor.
In this study, the equivalent strain measure is computed based on the modified von
Mises definition. The mechanical damage parameter 𝐷 then follows:

90
 𝜅0𝑚𝑑 (𝑇) 𝑚𝑑 𝑚𝑑 (𝑇))
𝐷(𝜅 𝑚𝑑 ) = 1 − 𝑚𝑑
((1 − 𝛼𝑚𝑑 ) + 𝛼𝑚𝑑 𝑒 −𝛾(𝑇)(𝜅 −𝜅0 ) (5.46)
𝜅

The mechanical damage parameter function in (5.46) is defined according to the

modified Mazars type damage evolution. The term (1 − 𝛼𝑚𝑑 ) represents the residual
stress as 𝐷 → 1, expressed as a portion of the original strength. For simplicity, the
material is assumed to have no residual strength in this study (𝛼𝑚𝑑 = 1), (5.46) can
thus be recast as:

𝜅0𝑚𝑑 (𝑇) −𝛾(𝑇)(𝜅𝑚𝑑−𝜅0𝑚𝑑(𝑇))

𝐷(𝜅 𝑚𝑑 ) = 1 − 𝑒 (5.47)
𝜅 𝑚𝑑

(1 − 𝑉)𝑓𝑡 (𝑇)
with 𝛾(𝑇) = 𝑙𝑐 (5.48)
𝐺𝑓 (𝑇)

where 𝛾(𝑇) is the ductility parameter that controls the slope of the softening curve, the
smaller the value, the more ductile the response is. 𝑙𝑐 is a proportionality factor
depending on the size of the localization zone and has the units of length. 𝐺𝑓 (𝑇) is the
fracture energy release rate expressed as:

𝐺𝑓 (𝑇̂) = 𝐺𝑓0 (1 + 0.39𝑇̂ − 0.07𝑇̂ 2 ) , 0 ≤ 𝑇̂ ≤ 4.8 (5.49)

where 𝐺𝑓0 is the fracture energy at room temperature and 𝑇̂ = max(𝜃̅ ) is the maximum
normalised temperature.

5.3.3.2 Thermo-Chemical Damage

The thermo-chemical damage function could be expressed as:

𝑓 𝑡𝑑 = 𝑇 − 𝜅 𝑡𝑑 (5.50)

with 𝜅 𝑡𝑑 = max{𝑚𝑎𝑥(𝑇) , 𝑇0 } (5.51)

91
where 𝜅 𝑡𝑑 is the history parameter for thermo-chemical damage, which takes the larger
of either the maximum value obtained by 𝑇 or the reference temperature 𝑇0 . Similarly,
the growth of the thermo-chemical damage is determined by:

𝑉̇ ≥ 0 𝑖𝑓 𝑓 𝑡𝑑 = 0
{ (5.52)
𝑉̇ = 0 𝑖𝑓 𝑓 𝑡𝑑 < 0 𝑎𝑛𝑑 𝜅̇ 𝑡𝑑 = 0

Assuming that the degradation of Young’s modulus depends on the increase of

temperature in the following form:

2
E0 (𝑇) = 𝐸0 (𝑇𝑎 )(1 − 0.1𝑇̂) (5.53)

Combining with (5.36), the thermo-chemical damage parameter could be calculated by:

𝑉(𝜃̅ ) = 0.2𝜃̅ − 0.01𝜃̅ 2 (5.54)

5.4 Numerical Solution

The primary state variable chosen for the macroscopic linear momentum conservation
equation is the displacement vector of the solid skeleton 𝑢. The initial condition is
specified for the whole problem domain Ω and its boundary Γ as:

𝑢 = 𝑢0 , 𝑎𝑡 𝑡 = 0 (5.55)

The boundary conditions are specified either as the Dirichlet type on Γ𝑢1 or as the
Neumann or Cauchy type on Γu2 :

𝑢 = 𝑢∞ 𝑜𝑛 Γ𝑢1 , 𝑜𝑟 𝛔′ ∙ 𝒏 = 𝒕̅ 𝑜𝑛 Γu2 (5.56)

where u∞ is the prescribed displacement on the boundary, 𝒏 is the unit normal vector,
𝒕̅ is the imposed tension force. Approximating the displacement vector in terms of its
nodal values:

92
 𝑢 = 𝑢(𝑡) ≅ 𝑁𝑢̅(𝑡) (5.57)

The strain tensor is then associated with the displacement vector in the following way:

𝛆 = 𝑳𝑢 ≅ 𝑳𝑁𝑢̅ = 𝑩𝑢̅ (5.58)

where 𝑩 is called the strain matrix, and 𝑳 is the differential operator defined as:

𝜕
0 0
𝜕𝑥
𝜕
0 0
𝜕𝑦
𝜕
0 0
𝜕𝑧
𝑳= 𝜕 𝜕 (5.59)
0
𝜕𝑦 𝜕𝑥
𝜕 𝜕
0
𝜕𝑧 𝜕𝑦
𝜕 𝜕
[ 𝜕𝑧 0
𝜕𝑥 ]

The procedure for obtaining the variational form of the governing equation has been
elaborated in Section 4.4.2, and therefore is not repeated here. Together with the
previously chosen variables, the final set of primary state variables can be expressed as:

x = (𝑢, 𝑇, 𝑃𝑔 , 𝜌̃𝑣 ) ≅ 𝑁x̅ = 𝑁(𝑢̅, 𝑇̅, 𝑃̅𝑔 , 𝜌̃̅𝑣 ) (5.60)

The global matrices in the discretised system of governing equations in (4.74) can thus
be extended as:

0 0 0 0 ̂ uu
𝐊 ̂ uT
𝐊 ̂ uP
𝐊 ̂ uV
𝐊
0 𝐂̂TT 𝐂̂TP 𝐂̂TV ̂ TT ̂ TP ̂ TV
𝐂̂ = , ̂= 0
𝐊
𝐊 𝐊 𝐊
(5.61)
0 𝐂̂AT 𝐂̂AP 𝐂̂AV 0 ̂ AT
𝐊 ̂ AP
𝐊 ̂ AV
𝐊
[0 𝐂̂MT 𝐂̂MP 𝐂̂MV ] [ 0 ̂ MT
𝐊 ̂ MP
𝐊 𝐊 MV ]

93
It could be noted that the bottom right submatrix of the capacitance matrix 𝐂̂ and
̂ , related to the description of heat and mass transport in the model,
stiffness matrix 𝐊
have actually remained unchanged from that in (4.74). In other words, only the four
̂ need to be evaluated. The detailed derivation of the
new entries in the first row of 𝐊
submatrices could be found in Appendix C. The final formulations are written as:

̂ 𝑢𝑢 = ∫ 𝑩𝑻 (𝑬𝒔𝒆𝒄 − (1 − 𝑉)𝐻1𝑚𝑑 𝝈
𝐊 ̂ ′ 𝒔𝑇 )𝑩𝑑Ω (5.62)
Ω

𝛽𝑡𝑚
̂ 𝑢𝑇 = − ∫ 𝑩𝑻 {[𝑬𝒔𝒆𝒄 − (1 − 𝑉)𝐻1𝑚𝑑 𝝈
𝐊 ̂′ 𝒔𝑇 ] (𝛼𝑓𝑡 𝒎 + 𝐌𝝈′ )
Ω 𝑓𝑐0
(5.63)
+ [(1 − 𝐷)𝐻 𝑡𝑑 + (1 − 𝑉)𝐻2𝑚𝑑 ]𝝈
̂′ + 𝑏𝒎𝑻 } 𝑁𝑑Ω

̂ 𝑢𝑃 = − ∫ 𝑩𝑻 𝑏𝒎𝑷 𝑁𝑑Ω
𝐊 (5.64)
Ω

̂ 𝑢𝑉 = − ∫ 𝑩𝑻 𝑏𝒎𝑽 𝑁𝑑Ω
𝐊 (5.65)
Ω

𝑚𝑑
where the terms 𝐻1/2 and 𝐻 𝑡𝑑 are associated with the mechanical and thermo-chemical
damage evolution respectively, defined as:

𝜕𝐷
𝐻1𝑚𝑑 = (5.66)
𝜕𝜅 𝑚𝑑

𝜕𝐷 𝜕𝜅0𝑚𝑑 𝜕𝐷 𝜕𝛾
𝐻2𝑚𝑑 = + (5.67)
𝜕𝜅0𝑚𝑑 𝜕𝑇 𝜕𝛾 𝜕𝑇

𝜕𝑉
𝐻 𝑡𝑑 = (5.68)
𝜕𝜅 𝑡𝑑

and the vector 𝒔 is defined as the derivative of the equivalent strain measure with
respect to the strain tensor and computed by:

94
 𝜕𝜀̅
𝒔= (5.69)
𝜕𝜺

Using the modified von Mises definition in (5.44) and (5.45), (5.69) could be recast as:

𝜕𝜀̅ 𝜕𝐼1 𝜕𝜀̅ 𝜕𝐽2

𝐬= + (5.70)
𝜕𝐼1 𝜕𝜺 𝜕𝐽2 𝜕𝜺

with 𝐽2 = 𝐼12 − 2𝐼2 (5.71)

where 𝐼1 and 𝐼2 is the first and second invariant of the strain tensor 𝜺. The derivatives
of the invariants of a second-order tensor are as follows:

𝜕𝐼1 𝜕𝐼2
= 𝟏, = 𝐼1 𝟏 − 𝜺𝑇 = 𝐼1 𝟏 − 𝜺 (5.72)
𝜕𝜺 𝜕𝜺

where 𝟏 is the second order identity tensor. Therefore,

𝜕𝐽2 𝜕𝐼12 𝜕𝐼2

= −2 = 2𝐼1 𝟏 − 2(𝐼1 𝟏 − 𝜺) = 2𝜺 (5.73)
𝜕𝜺 𝜕𝜺 𝜕𝜺

Also, with

𝜕𝜀̅ 1
= 𝐴 + 𝐴2 𝐼1 (𝐴2 𝐼12 + 𝐵𝐽2 )− 2 (5.74)
𝜕𝐼1

𝜕𝜀̅ 𝐵 2 2 1
and = (𝐴 𝐼1 + 𝐵𝐽2 )− 2 (5.75)
𝜕𝐽2 2

The final expression of vector 𝒔 is:

𝜕𝜀̅ 𝐴2 𝐼1 𝐵
𝒔= = (𝐴 + )𝟏 + 𝜺 (5.76)
𝜕𝜺 √𝐴2 𝐼12 + 𝐵𝐽2 √𝐴2 𝐼12 + 𝐵𝐽2

95
As regards the equivalent strain measure computed based on the Mazars definition,
Burgh (2016) suggested (5.77) would be a sufficient approximation:

𝜕𝜀̅ ⟨𝜀𝑖 ⟩+
𝒔= = (5.77)
𝜕𝜺 𝜀̅

The temporal discretization of the equations by finite difference method is conducted

in like manner as discussed in Section 4.4.3, therefore is not repeated here.

5.5 Spalling Criterion

Having descriptions of thermo-hygro and thermo-mechanical phenomena developed in

Chapter 4 and Chapter 5, it is desirable to obtain a quantitative measure for the
assessment of spalling. The last and final component of the THM model is therefore
the incorporation and implementation of a spalling criterion in the model. The
development of such a criterion is given in this section following first a brief review of
the spalling criteria based on different classifications.

The early proposals of the spalling criteria were mainly based on the experimental
results, often in the form of nomograms, such as those of Harmathy (1965),
Sertmehmetoglu (1977), Zhukov (1994) mentioned in Section 2.3.3.1. Due to the
restrictions from materials and specimen geometries, the application of these
experiment-based criteria hardly achieved any major practical success. Later, as
discussed previously in Section 2.3.3.2, attempts to the formulation of more
comprehensive analytical and numerical models arose, so did the spalling criteria.
Ichikawa (2000) proposed a hollow spherical model (Figure 2.12) to connect the pore
pressure and maximum tensile stress using the expression in (5.78). Spalling is
predicted when σt,max exceeds the concrete tensile strength 𝑓𝑡 :

1 + 2𝑉𝑝
σt,max = 𝑃 > 𝑓𝑡 (5.78)
2(1 − 𝑉𝑝 )

96
where 𝑉𝑝 and 𝑃 is the pore volume and pore pressure, equivalent to the porosity 𝜙 and
vapour pressure Pv in this study. It could be noted from the expression that this criterion
considers only the effect of pore pressure and ignores that of the thermal stress.

Tenchev and Purnell (2005) presented a criterion that couples the pore pressure and
evolution of damage (also the associated stress increments) based on the damage
constitutive model originally proposed by (Ortiz, 1985). The pore pressure Pg was
applied as a body force F computed from the gradient of pore pressure:

F = 𝑏∇Pg (5.79)

where 𝑏 is the Biot’s coefficient and taken as 𝑏 = 1 with the incompressible dry
skeleton assumption. Spalling was predicted when the resulting mechanical damage
variable attained the maximum value of unity.

Gawin et al. (2006c) proposed four spalling indexes for the practical evaluation of
thermal spalling. The first three indexes were introduced based respectively on the
pressure-induced shear model, buckling model considering gas pressure and fracture
mechanics, while the fourth index was a dimensionless quantity with no mechanical
basis yet showed the best agreement with the experimental tests. Later, Gawin and
Pesavento (2009) proposed a fifth index using the same form of the fourth one, only
omitting the effect of the specific fracture energy. Still, this fifth index has no
mechanical basis. Gawin et al, the authors of the publications themselves, termed the
criteria presented as heuristic and intuitive, due to the intrinsic difficulties in the a priori
determination of the charateristic parameters.

Kodur et al.(2008b, 2009, 2008a, 2010) evaluated the fire resistance of the reinforced
concrete elements with the finite element program initially developed for strength
analysis and then later coupled with thermal analysis. They proposed the following
criterion for the prediction of spalling:

nPv > 𝑓𝑡 (𝑇) (5.80)

97
where n is the porosity (equivalent to 𝜙 in this study), Pv is the vapour pressure and
𝑓𝑡 (𝑇) is the concrete tensile strength. This criterion was relatively simple, but no
rationale was given for the derivation of criterion.

Phan et al. (2011) proposed two criteria for the simulation of spalling front. The first
criterion was related to the heat and mass transfer (pore pressure build-up) and
analogous to the one in (5.80) in form:

Fcri1 = 𝑓𝑡 (𝑇) − 𝑏𝑃𝑠 (5.81)

where 𝑏 is the Biot’s coefficient and 𝑃𝑠 is the weighted sum of pore pressure exerted on
the solid skeleton. The second criterion was a buckling check conducted only for the
external layer satisfying the first criterion, and was related to the thermal-mechanical
process:

𝜋 2𝑒2
Fcri2 = (1 − ⟨𝐷⟩)𝐸0 − ⟨𝜎⟩ (5.82)
12ℎ2

where h and e is the length and depth of the spalling zone, E0 is the initial Young’s
modulus of concrete, ⟨𝐷⟩ and ⟨𝜎⟩ is the average material damage and compressive
stresses. Though these criteria managed to consider the effects of both pore pressure
and thermal stress, the determination of geometric parameters of the spalling zone
involved in the second criterion was rather unclear and challenging. The criteria are
therefore deemed impractical to predict spalling.

Davie et al. (2012) defined the equivalent strain measure using three typical
formulations to reflect the evolution of damage. He then investigated a wall and column
problem (corresponding to the one and two-dimensional spalling case) using different
equivalent strain measures and altering the parameters associated with mechanical
damage, concluded that the modified von Mises definition may be the best
representation of the strain components. Recommendations for the values of various
parameters were also given based on the sensitivity case studies. While the mechanical
damage may provide information regarding the pattern of spalling zone in both one and

98
two-dimensional cases, it can only serve as an indicator for spalling, not a quantitative
measure predicting when and where the spalling would be initiated. Furthermore, given
the fact that the evaluation of mechanical damage is based on the equivalent strain
measure, which in term is primarily related to the thermo-mechanical phenomenon, this
approach somehow ignores the influence of pore pressure.

Lu (2015) proposed a relatively more thorough criterion by refining the combined-

stresses model of Zhukov (1976). He introduced the Biot’s coefficient to the pore
pressure-induced stress and included the effect from pore pressure in all directions. A
schematic sketch of a concrete element subjected to the acting stresses is shown in
Figure 5.1. The final expression for the criterion reads:

σspalling = √(1 − 2𝜈)𝑏 2 𝜎𝑃2 − 𝜈(2𝜎𝑡 + 𝜎𝐿 )𝑏𝜎𝑝 ≥ 𝑓𝑡 (𝑇) (5.83)

where σ𝐿 , 𝜎𝑡 and 𝜎𝑝 is the external load, thermal and pore pressure-induced stress
respectively, 𝑏 and ν is the Biot’s coefficient and Poisson’s ratio. Though this
formulation has a rational mechanical basis, the calculation of the thermal stress given
by Lu is actually problematic. He used the following expression to calculate the thermal
stress:

𝐸 𝐸
σt = 𝜀𝑡 = 𝑎𝑑𝑇 (5.84)
1−𝜈 1−𝜈

which obviously considered only the free thermal strains and neglected the load-
induced component of thermal strains. This is clearly at odds with the existence of
external loading in his formulation. Moreover, this formulation was derived in the
direction perpendicular to the heated surface, restricting its application to one-
dimensional spalling only.

From the above discussion, it can be seen that the existing criteria have limitations in
either considering only one aspect of the spalling mechanisms or being impractical in
use. In this study, a two-step criterion is developed to assess the pattern and occurrence

99
of spalling for both one and two-dimensional cases. Firstly, the evaluation of
mechanical damage using the equivalent strain mapping as adopted by Davie et al.
(2012) is implemented in the model, indicating an approximated spalling zone within
the problem domain. For areas with mechanical damage exceeding certain values, the
second step is implemented to check whether spalling occurs or not. The strain energy
density of the highly damaged areas is computed with introduction of (5.12) by:

1 1
𝑊𝑑 = 𝝈′ : 𝜺 = (𝝈 + 𝑏𝑃𝑠 𝐈): 𝜺 (5.85)
2 2

where 𝝈′ and 𝜺 is the effective stress and strain tensor. The critical energy required to
initiate spalling is calculated by:

1 𝑓𝑡 (𝑇)2
𝑊𝑠𝑝 = (5.86)
2 𝐸(𝑇)

When the strain energy density exceeds the required spalling energy, i.e., 𝑊𝑑 > 𝑊𝑠𝑝 ,
spalling is assumed to occur. It can be seen from (5.85) that the effects of both the pore
pressure and thermal strains are properly taken care of. This two-step criterion is
implemented in the THM model for the prediction of spalling.

100
Figure 5.1 Concrete element subjected to combined stresses

101
6 THERMO-HYGRO-MECHANICAL MODEL ANALYSIS

6.1 Introduction

In this chapter, the THM model developed so far is validated against the experimental
data. The numerically simulated results are compared with the published data to
examine the validity of the model and spalling criterion. Last but not least, an algorithm
to predict the occurrence of spalling is proposed. This algorithm takes into account the
stochastic nature of concrete behaviour in the process of heat and mass transfer. More
precise modelling of thermo-hygro-mechanical behaviour of concrete at elevated
temperature could be achieved by integrating this algorithm into the THM model.

6.2 Experiment of Ichikawa and England (2004)

6.2.1 Model Description

The experiment of Ichikawa and England (2004) were the reproduction and extended
analyses of experiments by Khan (1990). The test problem was a concrete wall exposed
to fire at one side as shown in Figure 6.1. The left side was subjected convective-
radiative heat transfer, mass transfer and atmospheric pressure, where the heat and mass
transfer were of flux boundary conditions while the gas pressure of prescribed (constant)
boundary condition. Both sides of the wall were unsealed. The goal was to determine
the temperature, pore pressure, pore volume and moisture content profile along the
horizontal axis, simplifying the problem into one-dimensional. The one-dimensional
problem was effectively solved with a row of two-dimensional elements. The problem
was first solved with a row of eight-node quadrilateral elements, then a row of six-node
triangular elements. Identical results were obtained from the two solutions, indicating
the results are insensitive to the element type or mesh. The length of the domain under
consideration was 0.1 m. The width (side AB and CD) was arbitrary for one-
dimensional problem, in this study, the width was chosen such that the generated
elements were squares. Side CD had the same boundary condition types as side AB
while the other side of rectangle ABCD were isolated, where no conditions were

102
specified. The concrete wall was exposed to elevated temperature in accordance with
the ISO 834 standard temperature-time curve (6.1) for one hour:

𝑇∞ = 293.15 + 345 log10 (8𝑡 + 1) (6.1)

where 𝑡 is time in minutes. In the work of Ichikawa and England, the test problem was
solved using explicit finite difference scheme. The rectangular domain was discretized
by 101 nodes in space, with time step fixed at 0.002 second.

6.2.2 Model Parameters

The surrounding absolute temperature 𝑇∞ was computed from equation (6.1). The
prescribed gas pressure 𝑃𝑔 ∞ took the value of atmospheric pressure of 101325 Pa,

approximated to 0.1 MPa. The surrounding vapour content 𝜌̃𝑣 ∞ was defined as 80% of
the initial internal vapour content, where 80% is the external humidity. The initial
internal conditions for the domain were considered to be uniformly distributed,
occupying values of temperature 𝑇 0 = 20℃ , gas pressure 𝑃𝑔0 = 101325 Pa and
vapour content 𝜌̃𝑣0 calculated from the ideal gas law (4.11) corresponding to the
saturation pressure of water vapour, 𝑃𝑠𝑎𝑡 .

6.2.3 Results and Discussions

Figure 6.2 shows the comparison between internal temperature fields. The heated face
corresponds to the position at x = 0 and the rear (unheated) face corresponds to x =
0.1 m. The internal temperature fields at times of 5, 10, 20, 30 and 60 minutes after
heating are shown in the figure. Considering a particular time step, the temperature
maximized at x = 0 and then gradually decreased with increasing distance from the
heated face. The highest temperature reached after one-hour’s heating was
approximately 900 ℃ , similar for both the experimental and numerical results.
Considering a particular position, the temperature increased with longer duration of
heating and maximized after 60 minutes. The numerically simulated temperature fields

103
including the highest temperature achieved for all time steps agreed well with the
experimental data.

A good representation of pore volume can be seen from Figure 6.3. The initial pore
volume was assumed uniformly distributed before the start of heating. As heating
continued, the pore volume grew with the rise of temperature. The increased pore
volume was mainly caused by the release of gel water and chemically-bound water
added to the initial pore volume. Moreover, the thermal dilatation of concrete could
also be responsible for the increased pore volume. Turning points could be found in the
figure where the gradient at curves of 5, 10, 20 and 30 minutes markedly changed. This
was mainly due to the fact that the release rate of gel water (in the range of 20℃ to
105℃) was greater than the release rate of chemically-bound water (after 105℃).
Therefore, the corresponding temperatures at the turning points were roughly 105℃.
And the closer to the heated face, the higher the temperature and the lower the rate of
volume change.

The distribution of moisture content at various time steps is shown in Figure 6.4. The
heated (x = 0) and unheated (x = 0.1 m) face were unsealed, meaning the two faces
were directly exposed to the surrounding environment. The moisture content at x = 0
was kept constant at zero because the moisture was assumed to be instantly dried out
by the heat source. The moisture content at the rear face was kept at a small value which
was calculated using an atmospheric relative humidity (RH) of 60%. The moisture near
x = 0, driven by the pressure- and concentration-induced gradient, was released into
the environment through the heated face. The moisture content in the neighbouring
region of the heated face rapidly decreased, which then became a dry zone accordingly.
The moisture also moved inwards due to pressure gradient, therefore, a wet zone was
formed to the right of the dry zone. As moisture accumulated with rise of temperature,
the moisture content reached the maximum limit for saturated liquid water and a
saturated zone was produced. The moisture content in the saturated zone was relatively
stable and changed moderately with position. The shape of saturation zones in Figure
6.4 (a) and (b) were slightly different. This was mainly because Ichikawa and England
(2004) ignored the contribution of diffusion to the variation of moisture content, while

104
the effect of diffusion was considered in this study. In the region near x = 0.1 m, where
the unheated face was connected to the atmosphere, moisture could be released and the
moisture content again rapidly decreased. Another wet zone was then produced to the
right of the saturated zone. It can be seen from the figure that the zones moved towards
the cold face as time went by. The length of the dry and saturated zone gradually
increased while the length of the second wet zone decreased. The numerical results
successfully reproduce the formation and movement of the dry, wet and saturated zones.

Figure 6.5 shows the distribution of the pore pressure in concrete. The initial pore
pressure was assumed uniform, which corresponded to the saturation vapour pressure
at initial temperature (20℃). As heating continued, temperature and moisture content
increased, leading to the rise of pore pressure. In the wet zone mentioned above, the
pore pressure equalled to saturation vapour pressure calculated with temperature only.
In the saturated zone, the pore pressure was a function of temperature and density of
free water. The pressure of the superheated steam and compressed free water
contributed to the pore pressure within saturated zone. It can be noted from the figures
that the shape of the curves was slightly different. This could be explained by the fact
that the pore pressure was computed with another formulation of equation (Wagner and
Pruß, 2002) different from the original work. The maximum values and evolution of
the pore pressures were properly represented by the numerical analysis.

6.3 Experiment of Mindeguia et al. (2009)

6.3.1 Model Description

The second test problem was prismatic concrete specimens (0.6×0.7×0.15 m3), made
with two types of concrete (ordinary concrete B40 and high performance concrete B60),
subjected to a quasi-unidirectional thermal load. Kalifa et al. (2000) designed a special
experimental setup capable of continuously measuring the temperature and pore
pressure profiles, as well as water mass loss, of the prismatic specimen. The thermal
load was applied on the upper face of the specimen, 3 cm above which was the radiant
heater that produced the heating. Mindeguia et al. (2009) used and redesigned a similar

105
experimental setup for their tests. Lateral faces of the specimen were insulated with
rock wool blocks to simulate conditions of one-dimensional heat and mass transfer. The
details of the experimental setup are shown in Figure 6.6. Three gauges, consisting of
thermal couple and pressure tube, were located along the centre line at depths of 10, 20
and 30 mm from the heated face respectively. The fourth gauge was an additional
thermal couple measuring the temperature at 2 mm from the heated face. The locations
and arrangement of thermal couples and pressure gauges are shown in Figure 6.7. In
this study, the numerical analyses are performed with a fire duration of two hours
concerning the B60 specimens at a moderate heating rate. The problem was modelled
by a row of 150 eight-node quadrilateral elements, assuming essentially one-
dimensional behaviour. Similar to the first validation, the problem was then modelled
by a row of 300 six-node triangular elements. Identical results were again obtained,
indicating the insensitivity to element type and mesh. The upper and lower faces of the
specimen were subjected to open boundaries, which were free to transfer heat and mass.

6.3.2 Model Parameter

The surrounding absolute temperature 𝑇∞ on the heated face was modelled by equation
(6.2) in the analysis. The prescribed gas pressure 𝑃𝑔 ∞ had a constant value of

atmospheric pressure of 101325 Pa. The vapour content of surrounding environment

𝜌̃𝑣 ∞ was 0.01497 kg/m3. The initial internal conditions for the domain were considered
to be uniformly distributed. The initial and atmospheric temperature had the same value
(𝑇 0 = 20℃). The initial gas pressure 𝑃𝑔0 and vapour content 𝜌̃𝑣0 was 101325 Pa and
0.014525 kg/m3 respectively.

6.3.3 Results and Discussions

In the original work of Mindeguia et al. (2009), the measured furnace temperature was
generated by reducing the power of the gas burners, but no equation or relevant
information was given by the author regarding the control of temperature. Instead, the
evolution of furnace temperature was only shown in figure form. The furnace
temperature was modelled with the following empirical equation after calibration:

106
 𝑇∞ = 293.15 + 550 log10 (0.05𝑡 + 1) (6.2)

where 𝑡 was time in minutes. The comparison between the measured and modelled
furnace temperature is presented in Figure 6.8. As illustrated in the figure, the proposed
equation can basically model the evolution of furnace temperature, except some
irregular fluctuations. The measured internal temperature of the specimen was only
reported at depth of 10 mm from the heated surface. Simulation was conducted with
the assumed furnace temperature described by (6.2). It can be seen from Figure 6.9 that
the simulated internal temperature agrees well with the measured temperature. The
simulated surface temperature is also plotted for comparison. As expected and shown
in the figure, the closer to the heat source, the higher the temperature.

The measured and numerically simulated internal pressures at 10, 20 and 30 mm from
the heated surface are presented in Figure 6.10 (a) to (c) respectively. Spalling was
reported at 106 minutes since heating, so the accumulated pressures were suddenly
released and dropped to zero. Internal pressures at all depths remained approximately
unchanged in the first 50 minutes of testing, after which pressures started to increase.
At around 70 minutes since heating, the pressure at depth of 10 mm increased with an
obviously smaller rate than the simulated pressure, which maximized at around 90
minutes with a peak value of 2.4 MPa. The difference in the gradient could be explained
by a possible leak in pressure gauge, caused by concrete damage in the region of probe
location and/or a faulty pipe seal (Kalifa et al., 2000, Burgh, 2016). The peak pressure
at depth of 20 mm was the largest among the experimental as well as numerical data.
The measured internal pressure reached a peak value of 2.6 MPa at the time of spalling.
The peak pressure of numerical analysis occurred at around 113 minutes with a value
of 3.14 MPa, which was considered to be within a reasonable range of variation. The
simulated pressure at the depth of 30 mm showed the best agreement with the measured
pressure among the three sets of data. The simulated pressure at the time of spalling
was around 1.3 MPa, approximately equalled to the measured value of 1.34 MPa.
Overall, the internal temperature and pressure were properly simulated by the model.

107
6.4 Experiment of Ozawa et al. (2012)

6.4.1 Model Description

The last test problem was the experiment performed by Ozawa et al. (2012) to study
the mechanism of explosive spalling in high strength concrete by acoustic emission
(AE) techniques. The experiment was conducted on prismatic specimens of the same
dimension (0.4×0.4×0.1 m3) but cured under two different conditions. All specimens
were firstly wet-cured for 64 days at temperature of 20 ± 2 ℃. The set of specimens
that were put to test directly after wet-curing was denoted as wet. The other specimens,
denoted as air-dried, were subjected to further air-drying under controlled conditions
(20℃, RH 40%) for 118 days. The curing conditions primarily affected the moisture
content of the specimens, presented as percentage weight. The moisture content of the
wet specimen was uniformly distributed at around 9.5% while for the air-dried
specimens, the moisture content was decreasing from 9% at the depth of 50 mm to 4.0%
at the heated surface. One thermal couple and five steel pipes were embedded in the
specimens at a depth of 8 mm from and parallel to the heated surface. The steel pipes
were filled with hydraulic jack oil and connected to pressure transducers to measure the
internal vapour pressure. The locations and arrangement of the thermal couple and
pressure gauges are shown in Figure 6.11. The specimen was placed on an electrical
furnace with a power rating of 56 kW. The bottom face of the specimen was heated at
a rate of about 20℃/min until spalling occurred. The experimental setup of the heating
test is shown in Figure 6.12. Both the wet and air-dried specimens were modelled by a
row of 100 eight-node quadrilateral elements, assuming essentially one-dimensional
behaviour. Then, as was done in the first two validations, both specimens were
modelled again by a row of 200 six-node triangular elements and it was proven that the
results are insensitive to element type and mesh. The upper and lower faces of the
specimen were unsealed, which were free to transfer heat and mass.

108
6.4.2 Model Parameters

The surrounding temperature 𝑇∞ on the heated surface was increased linearly with a
rate of 20℃/min in the analysis. The prescribed gas pressure 𝑃𝑔 ∞ had a constant value

of atmospheric pressure of 101325 Pa. The vapour content of surrounding environment

𝜌̃𝑣 ∞ was calculated assuming 65% relative humidity (equivalent to 0.01497 kg/m3).
The initial internal conditions for the domain were considered to be uniformly
distributed. The absolute initial internal temperature 𝑇 0 was 308.15 K (equivalent to
35℃). The initial gas pressure 𝑃𝑔0 was 101325 Pa. The initial vapour content 𝜌̃𝑣0 for wet
and air-dried specimen was calculated assuming an initial degree of saturation 𝑆 of 99%
and 50% respectively.

6.4.3 Results and Discussions

The only information regarding the furnace temperature given in the original work of
Ozawa et al. (2012) was a heating rate of 20 ℃ per minute. Except for this, no
experimental data or figure was presented to illustrate the control of furnace
temperature. The convective and radiative heat transfer boundary parameters, i.e. the
heat convection coefficient and surface emissivity, would require to be lower than the
physically realistic values to reproduce the measured temperature profiles if a heating
rate of 20℃/min was used. Not having enough information on the evolution of furnace
temperature, the heating rate was reduced to 15℃/min hypothetically. The absolute
furnace temperature could then be described by the following equation:

𝑇∞ = 308.15 + 15𝑡 (6.3)

where 𝑡 is time in minutes. Numerical simulation was conducted with the above
prescribed temperature for both the wet and air-dried specimens. Figure 6.13 (a) and
Figure 6.14 (a) show the comparison between the measured and simulated temperature
profiles at a depth of 8 mm of the wet and air-dried specimens. It can be seen from the
figures that the simulated temperature profiles of both specimens presented good
correlation with the measured temperature profiles, demonstrating the heating condition

109
adopted in this study was appropriate. The temperature field of the heated surface was
added in the figures for comparison. With decreasing distance to the heated surface, the
temperature increased as expected.

The measured and simulated internal vapour pressure at the depth of 8 mm of the wet
and air-dried specimens are shown in Figure 6.13 (b) and Figure 6.14 (b). The pressure
gauges were all embedded in the same depth, theoretically, the measured vapour
pressures should be approximately the same. However, some of the pressure gauges
had especially low readings, which were treated as abnormal data and not used for
validation in this study. The reasons for the drop in measured pressures could be
attributed to the susceptibility of leaks in pressure gauges and/or cracks in the concrete.
For the wet specimen, the internal vapour pressure remained roughly unchanged in the
first 25 minutes of heating. The numerical result was in good agreement with the
experimental data until around 38 minutes since heating, when the measured pressure
suddenly dropped about 0.3 MPa. As explained above, this was very likely to be caused
by leaks and concrete damage. The maximum vapour pressure (3.2 MPa) was measured
at approximately 47 minutes since heating. The peak value of the numerical analysis
was 3.31 MPa occurring at around 44 minutes since heating. The air-dried specimen,
on the other hand, did not experience much pressure change in the first 30 minutes of
heating. Later, both the measured and simulated vapour pressure rapidly increased until
explosive spalling was generated. The maximum vapour pressure (3.4 MPa) was
measured at approximately 44 minutes since heating. The peak value of the numerical
analysis was 3.43 MPa occurring at around 43 minutes since heating. In general, the
numerically simulated pressures for both the wet and air-dried specimens were good
representations of the experimentally measured pressures. Though moisture content of
the wet and air-dried specimen was significantly altered by different curing conditions,
the maximum internal vapour pressure and occurrence time of spalling were basically
the same, indicating a threshold over which even higher moisture content would not
have substantial influence on the evolution of vapour pressure and time of spalling.

110
6.5 Algorithm to Predict Fire-Induced Spalling

The THM model developed in Chapter 4 and Chapter 5 adopts the same initial
conditions and material properties across the problem domain. In reality, however, the
actual conditions and properties of concrete are nonuniformly distributed, resulting
from the stochastic nature of concrete behaviour. Gawin et al. (2006c) pointed out the
need to consider the stochasticity for more precise heat and mass transfer modelling,
yet no progress has been achieved so far. An algorithm is proposed in this study to
account for the stochastic concrete behaviour at elevated temperature.

The stochasticity is first considered by assigning random initial conditions and

properties throughout the domain. Initial conditions and properties are generated from
an average value, by adding or subtracting a difference value within certain range. The
sum of these difference values should equal to zero so that the law of mass conservation
is obeyed. Normally distributed random numbers with zero expectation could satisfy
this requirement. Box-Muller transform (Box and Muller, 1958) is implemented to
generate pairs of independent, normally distributed random numbers using the
following equations:

y1 = 𝜇 + 𝜎√−2 ln 𝑥1 cos(2𝜋𝑥2 )
(6.4)
y2 = 𝜇 + 𝜎√−2 ln 𝑥1 sin(2𝜋𝑥2 )

where 𝑥1 and 𝑥2 are randomly chosen, independent numbers from the uniform
distribution on unit interval (0, 1), 𝜇 and 𝜎 are the expectation and standard deviation
of the generated normal distribution. Here, 𝜇 = 0 and 𝜎 = 1 are adopted.

The position and time of spalling vary with different initial conditions and material
properties. The stochasticity is then handled by applying the Monte Carlo simulation
(Guttag, 2016) to obtain the most critical position of spalling. The algorithm is
implemented in the following steps:

111
 1) Set the number of trials (M), the number of specimens to be analysed in each
trial (N), and the required precision of results
2) For each trial, the main body of heat and moisture transfer will run for N times,
so there will be N spalling positions recorded. The mean of the N positions is
calculated at the end of each trial.
3) After the running of M trials, there will be M mean positions of spalling. The
mean and standard deviation (sDev) of the M means are then calculated. If the
required precision level is not reached (sDev > precision/1.96), then double the
number of specimens in each trial (N = N*2) and repeat from step 2).
4) When the required precision is reached, the mean of the M means is the
estimated critical position for spalling.

If the proposed algorithm is successfully implemented, the uncertainty brought by the

stochastic nature of concrete could be properly accounted for. However, one important
characteristic of the Monte Carlo simulation is that a large number of trials are required
to satisfy good precision. The computational cost to achieve high precision is, without
doubt, extremely expensive. The application of the proposed algorithm is therefore
hindered by the performance of computers. Future studies on how to shorten the
computational time are needed to make the proposed algorithm practical.

112
Figure 6.1 Modelling of concrete wall

Figure 6.2 Comparison of temperature

113
Figure 6.3 Comparison of pore volume

114
(a)

(b)

Figure 6.4 Moisture content comparison: (a) experimental data from Ichikawa and
England (2004) and (b) numerical analysis

115
(a)

(b)

Figure 6.5 Pore pressure comparison: (a) experimental data from Ichikawa and
England (2004) and (b) numerical analysis

116
Figure 6.6 Experimental setup, after Mindeguia et al. (2009)

117
Figure 6.7 Locations of gauges, by Mindeguia et al. (2009)

118
Figure 6.8 Comparison of furnace temperature

Figure 6.9 Comparison of internal temperature

119
(a)

(b)

120
(c)

Figure 6.10 Comparison of pressure at depth of: (a) 10 mm, (b) 20 mm and (c) 30 mm

Figure 6.11 Locations of thermal couple and pressure gauges by Ozawa et al. (2012)

121
Figure 6.12 Experimental setup of heating test, after Ozawa et al. (2012)

122
(a)

(b)

Figure 6.13 Comparison of wet specimen: (a) temperature and (b) pressure

123
(a)

(b)

Figure 6.14 Comparison of air-dried specimen: (a) temperature and (b) pressure

124
7 EARLY RESIDUAL SPLITTING TENSILE STRENGTH
AT ELEVATED TEMPERATURE

7.1 Introduction

Attention has been paid to researches on high-temperature properties of HSC by

different researchers in recent years, yet data on the temperature-dependent tensile
strength of HSC are very limited. Most researches focus mainly on residual tensile
strength, where measurements are made after the specimens cool down to ambient
temperature (Gencel, 2012, Chowdhury, 2014, Behnood and Ghandehari, 2009, Li et
al., 2004, Chan et al., 1999, Aslani and Bastami, 2011). Khaliq and Kodur (2012)
conducted high-temperature tests and linear regression analysis to obtain the reduction
factor of splitting tensile strength, but only for high strength fly ash concrete with and
without fibres up to 90MPa. To guarantee structural fire safety of concrete structures,
better prediction of the occurrence of fire induced spalling would be needed, which
requires the knowledge of tensile strength of concrete at elevated temperature. It is
worth mentioning that residual tensile strengths are applicable to concrete cooled down
after a fire exposure, and therefore cannot represent the behaviour of concrete being
heated. The aim of the experimental study in this chapter is to evaluate the early residual
splitting tensile strength of concrete, measured when the concrete specimen is still at
high temperature. To compare the differences between the residual splitting tensile
strength of concrete measured at ambient and elevated temperatures, a review has been
conducted to collect the experimental results from relevant studies. Considering the fact
that high strength concrete (HSC) is more susceptible to fire-induced spalling due to its
high compactness (low permeability), this study mainly focuses on concrete with
compressive strengths no less than 60 MPa in standard tests. The residual splitting
tensile strength measured at ambient temperature of ranging in C60-C76 and C80-C100
are summarized in Figure 7.1 and Figure 7.2 respectively. As shown in the figures, the
splitting tensile strength generally decreases with increasing temperatures, except for a
few abnormal points abruptly increase in the range between 60℃ to 200℃. The
maximum temperature investigated in the studies varies from 200℃ to 1200℃, and the

125
temperature interval between each data point also differs from 40℃ to 600℃.
Therefore, the data sets from different studies are distributed in a scattered manner.
Relative reduction in each temperature interval are needed to make comparison between
test results from the previous studies and this study. Data generated from the
experimental tests in this study are also used for numerical simulations of the heating
and splitting test to reproduce the test results.

7.2 Experimental Program

7.2.1 Concrete Mix and Test Specimens

In this study, high strength concrete (60 MPa and 90 MPa) are investigated. Within
each grade, cylindrical specimens of three different sizes are cast. Mix proportions and
the detailed dimension of specimens are shown in Table 7.1 and Table 7.2. Group Ⅰ
specimens are used for the ambient strength tests to obtain the 28-day compressive and
tensile strength of different mixes. Group Ⅱ specimens are designed for the splitting
tensile strength tests at elevated temperature. Group Ⅲ specimens are prepared to
evaluate the moisture content of concrete mixes. Due to limited number of concrete
moulds, the C60 and C90 specimens were cast four times each, with the mix proportions
maintaining the same except a variation in the superplasticizer (SP) content. Specimens
from four different batches correspond to tests at four different temperatures (100℃,
200℃, 300℃, 400℃ respectively). The 28-day compressive strength, splitting tensile
strength and average moisture content of each batch of concrete mix are presented in
Table 7.3. Each strength is the average value of three specimens. Three thermal couples
are inserted in each specimen in Group Ⅱ and the sitting plan of thermal couples is
illustrated in Figure 7.3. A set of customized moulds are designed and manufactured to
fix the positions and avoid the shifting of thermal couples during the casting of concrete.
One thermal couple is located at the centre and the spacing between two thermal
couples is 50 mm. The specimen is placed at an electrical furnace where another thermal
couple is configurated to detect the temperature around the specimen, as depicted in
Figure 7.4. Due to the limitation of furnace chamber, the height of heated specimens is
limited to 250 mm, instead of the standard height of 300 mm.

126
7.2.2 Heating Regime

Considering the fact that explosive spalling mainly occurs in the temperature range
from 100℃ to 400℃ (Lee et al., 2012, Phan, 2002), the target peak temperatures in the
heating process are 100℃, 200℃, 300℃ and 400℃ respectively. As concretes
subjected to higher heating rate are more susceptible to explosive spalling, a relatively
low heating rate of 20℃/min is applied in this study to avoid spalling. The specimen
then goes through a peak temperature exposure of approximately 2 hours to let the
temperature of the entire section become uniform. The heating regime is shown in
Figure 7.5.

7.2.3 Instrumentation and Test Setup

RS Pro K-Type Thermocouples were used to record the temperature of specimens. The
measurable temperature range is from -100℃ to 1100℃. Probe length and diameter are
500 mm and 1 mm respectively. The length of cable connecting the probe and data
logger is 1m. The size (diameter × height) of chamber of the electrical furnace utilized
in this study is 200 × 330 mm. The furnace could reach up to 1200℃, but the
recommended highest temperature for general use is 900℃. Uniformity of the furnace
temperature will be within ±5-8℃. The temperature is controlled by a programmable
temperature controller, which is capable of setting up 8 heating stages and 2 sets of
program memory. In the proposed experiment, only one constant heating rate was
adopted in the heating process. The temperature recorder in the control panel could
record and store the furnace temperature data. The chamber and the control panel of the
furnace are shown in Figure 7.6. A trolley installed with a jig is tailor-made for the
transport of heated specimen from the furnace to the compression machine. Figure 7.7
presents the jig for splitting tensile test and the trolley. Researchers could push the jig
with heated specimen from the trolley to the compression machine through a slide rail
to prevent direct contact of the heated specimen.

127
7.2.4 Test Procedure

The experiment will be carried out in three steps: heating of specimens, transfer of
heated specimens, and splitting tensile tests. In the first step, one cylindrical specimen
is put into the electrical furnace at a time and heated according to the heating regime in
Figure 7.5 until the proposed target temperature is reached. Meanwhile, the real-time
temperature distribution of the mid-height section can be monitored through the data
collected from the thermal couples. Approximately 2 hours of peak temperature
exposure is applied, while the actual exposure time of each specimen will be determined
according to the real-time temperature distribution. The second step is to transfer the
heated specimen from the furnace to the compression machine using the trolley shown
in Figure 7.7. To minimize as much as possible the temperature loss in the course of
transportation, the time used in the process of transfer is strictly limited. It takes about
6 minutes between removing the heated specimen from the furnace and setting it in
place on the compression machine. This period will also be accounted in numerical
simulations. The last step is to conduct the splitting tensile test on the heated specimen
using the unstressed test method (Phan, 1996). Compressive load is applied until the
cylindrical specimen is split into pieces. Loading rate (in N/s) is determined based on
the construction standard CS1 (Department, 2010) and calculated from the following
equation:

π
Rate = 𝑠 × × 𝐿 × 𝑑 (𝑁/𝑠) (7.1)
2

where L is the average measured length of the specimen (in mm); d is the average
measured diameter of the specimen (in mm); 𝑠 is the increase in rate of stress,
recommended to be 0.04 MPa/s to 0.06 MPa/s. The early residual tensile strength (in
MPa) of HSC at different elevated temperatures could then be calculated by equation:

2𝐹
𝑓𝑡,𝑇 = (7.2)
𝜋×𝐿×𝑑

128
where F is the maximum load (in N); L is the average measured length (in mm); d is
the average measured diameter (in mm). In this study, the loading rate adopted is 3000
N/s, and the corresponding increase in rate of stress is 0.05MPa/s, which satisfies the
recommendation from code. Thermal couples are connected to the data logger during
the whole test until failure, so that the temperature change of specimen in the furnace,
in the process of transfer, and at the time of splitting tensile test could be monitored.

7.3 Experimental Results

7.3.1 Temperature Distribution

Temperatures of two different positions (centre and 50 mm away from centre) of each
specimen are recorded during heating. Numerical simulations are then conducted to
compare with the test data. The procedure of heat transfer simulation will be elaborated
in Section 7.4. Figure 7.8 and Figure 7.9 show the comparison between real
temperatures measured from thermal couples inserted in the specimens and the
simulation results. The numerical simulation results fit well with the real test data,
except that a plateau is observed at the temperature range of 150℃ to 180℃ for centre
position of both the C60 and C90 specimens heated to 300℃ and 400℃. This
phenomenon is very likely to be caused by the evaporation of free water and transport
of water vapor in the concrete pores between 105℃ and 160℃, as well as the release
and diffusion of chemically bound water in the concrete matrix from 160℃ to 180℃
(Phan, 2008b, Wu et al., 2019). When the temperature reaches around 105°C, free water
starts to evaporate and migrate towards the centre of the specimen due to pore pressure
gradient. As the concrete temperature continues to increase, additional volume of water
vapor is produced in the concrete pores due to the release of chemically bound water at
around 160°C. The transfer of heat in the concrete is retarded as a result of energy being
absorbed in the process of evaporation and migration, leading to a decrease in the rate
of temperature rise or a nearly constant temperature period. The plateau ends when the
concrete temperature rises to about 180°C, by which time most of the water vapor, from
free water or chemically bound water, would have escaped through the concrete surface
and stop migrating. The temperature distribution of specimens in the process of heat

129
transfer are generally reproduced by numerical method, and the resulting profile could
then be used as input data for the simulation of splitting tensile tests.

7.3.2 Measured Early Residual Splitting Tensile Strength

The recorded compressive load at which the heated specimen split in tension was used
to calculate the splitting tensile strength measured at elevated temperatures (𝑓𝑡,𝑇 ) using
equation (7.2). The values of 𝑓𝑡,𝑇 for C60 and C90 specimens from 25°C to 400°C are
presented in Table 7.4. The average 𝑓𝑡,𝑇 and their corresponding standard deviations
are presented in Table 7.5 with respect to temperature. The early residual 𝑓𝑡,𝑇 measured
at elevated temperatures decreases with increasing temperatures, but a larger strength
reduction is observed for both C60 and C90 specimens in the range of 25°C to 100°C
and 300°C to 400°C. Detailed discussions on the comparisons with numerical
simulation and previous studies are made in Section 7.4.2 and Section 7.5.1.

7.3.3 Spalling Behaviour

It is not the purpose of this study to investigate spalling because splitting tensile test is
planned to determine the tensile strength of HSC at high temperature. A low heating
rate was adopted to maintain the specimens as intact as possible. This is because
spalling, especially explosive spalling, generally happens when temperature is rapidly
increasing. Therefore, no major spalling was observed for all the specimens in the
abovementioned high-temperature tests as expected.

7.3.4 Failure Mode

The typical failure patterns of cylindrical specimens after the splitting tensile test
include: (1) specimens split into two approximately equal halves along the loading
direction; (2) a macro crack develops in the middle of the specimens along the loading
direction. The failure mode of heated specimen tested at elevated temperatures is
similar to that of the ambient specimen, as shown in Figure 7.10.

130
7.4 Numerical Simulation

7.4.1 Temperature Distribution

Numerical simulations are conducted with the commercial finite element software
ABAQUS. Thermal properties, including conductivity, specific heat, density,
convection factor and emissivity, of high strength concrete, are adopted as
recommended in Eurocode (2004). Coefficient of thermal expansion is applied
according to Kodur and Sultan (2003). These temperature-dependent properties and
thermal constants are presented in Table 7.6 and Table 7.7. The recorded furnace
temperature of each specimen is used as the convection and radiation environment
temperature in the heat transfer step. To better simulate the process of transport, an
ambient temperature of duration of 6 minutes is added at the end of each amplitude.
The temperatures before and after transportation of a specimen heated to 200℃ are
compared as an example in Figure 7.11. The left half of the figure is the temperature at
the moment when the specimen is taken out from the furnace and ready to be
transported. The right half is the temperature at the moment when the transportation is
finished. As shown in the figure, only the layers near concrete surface have a relatively
obvious temperature drop. The drop for the outer 15 mm concrete layer is 25℃, when
it comes to the 25 mm layer, the drop is only 5℃. The core temperature remains at
around 200℃ throughout transportation. Therefore, the effect on the temperature
profile due to the transportation process is not so significant. Simulation is also run at
the exact time when the heating is terminated, excluding the consideration of
transportation time. The comparison between splitting tensile strengths with or without
the transportation process is shown in Table 7.8. It can be seen that the difference
between the two strengths are not considerable, and the strength without transportation
approximates more to a real fire where the concrete would not experience cooling or
temperature drop.

131
7.4.2 Predicted Early Residual Splitting Tensile Strength

Numerical simulations are run to reproduce the splitting tensile strengths of HSC by
ABAQUS. By inputting proper material properties and incorporating the temperature
profile from heat transfer analysis, the process of splitting tensile test could be
simulated. The material model being utilized in the splitting test step is the Concrete
Damaged Plasticity (CDP). The required concrete properties can be categorized into
two types, namely thermal properties and mechanical properties. Thermal properties
mainly include conductivity, specific heat, density, convection factor, emissivity, and
thermal expansion coefficient, adopted as mentioned in the above section (Table 7.6
and Table 7.7). Mechanical properties mainly include the temperature-dependent
Young’s modulus and Poisson’s ratio (Wu et al., 2002, Bahr et al., 2013), stress-strain
relationship of HSC at elevated temperature (BSEN1992-1-2:2004, 2004, Terro, 1998),
and temperature-dependent CDP model parameters 𝑓𝑏0 /𝑓𝑐0 (Feist et al., 2009). Other
CDP parameters are the default value as recommended by ABAQUS. The mechanical
properties and parameters are presented in Table 7.9 and Table 7.10. The procedure of
simulation would be firstly to conduct the heat transfer analysis, and then use the
resulting temperature profile of specimen as predefined field in the splitting test step.
A dynamic-explicit analysis was conducted following the heat transfer to obtain the
splitting tensile strength. The results of simulation are shown and compared with the
data obtained from the experimental tests in Figure 7.12. The numerical simulation
results fit in well with the test data. The general trend of simulation also agrees with the
test, where sharp loss in strength is found in the range of 25°C to 100°C and 300°C to
400°C, and a stable strength reduction is observed between 100°C and 300°C.

7.5 Discussions

7.5.1 Measured Early Residual Splitting Tensile Strength at Elevated

Temperature

To better understand the influence of high temperature on HSC, relative residual

strengths are obtained by converting the test data from Figure 7.1 and Figure 7.2. The

132
relative residual strength is represented by the ratio of recorded splitting tensile strength
at a target temperature (𝑓𝑡,𝑇 ) to the splitting tensile strength at room temperature (𝑓𝑡 ),
as shown in

Figure 7.13 and Figure 7.14. After removing the outliers of the relative residual
strengths, the remaining values are averaged to obtain the average relative relationship
as the representative of residual tensile strength of HSC. The average strength at room
temperature is substituted into the relationship to solve for the residual 𝑓𝑡,𝑇 as a function
of temperature between 100°C to 400°C for C60 and C90 specimens respectively. The
comparison between the average early residual and residual strengths are presented in
Figure 7.15. The corresponding relative strengths are also shown in Figure 7.16. It can
be seen that HSC generally exhibits a degradation of splitting tensile strength with the
increase of temperature and a numerical gap exists between the average early residual
and residual values. While the trend is more gradual for the residual strengths of both
C60 and C90, the average early residual strengths shows a sharp loss in the temperature
range of 25°C to 100°C and 300°C to 400°C. Early residual strengths between 100°C
to 300°C are more stable and the changes are minor compared to the residual strengths.
This phenomenon indicates that there are differences between the residual and early
residual splitting tensile strength at elevated temperature, and the differences are
considerable. This trend of strength reduction could be explained by the physical and
chemical changes occurring within HSC subjected to high temperature. The compact
structure of HSC could induce thermal stresses that cause micro- and macro-cracks.
Thermal incompatibility between aggregates and cement paste could give rise to the
formation of cracks (Chan et al., 1999). Upon heated to around 100°C, free water in the
concrete evaporates and pore pressure gradually builds up. This great vapor tension in
capillaries together with the cracks contribute to the substantial tensile strength loss of
HSC up to 100°C. Minor changes in the range of 100°C to 300°C could be attributed
to a reduction in the decomposition of hydration product in the cement paste (Behnood
and Ghandehari, 2009). When the temperature exceeds 300°C, dehydration of C-S-H
and Ca(OH)2 and other ingredients arises. The process of dehydration would generate
more cracks and vapor, leading to a faster loss of tensile strength (Li et al., 2004). In
comparison to the residual strengths measured at ambient temperatures, the simulation

133
results show better agreement with the early residual splitting tensile strengths
measured at elevated temperatures in both numerical form and general trend.

7.5.2 Predicted Early Residual Splitting Tensile Strength at Arbitrary High

Temperature

Data generated from the abovementioned experiments and discussions prove the
feasibility of obtaining the early residual splitting tensile strength at high temperature
(up to 400°C in this study) by numerical method. With the compressive and tensile
strength at room temperature, curve of heating regime, thermal and mechanical
properties of concrete at elevated temperature, the early residual splitting tensile
strength of concrete at high temperatures could be obtained through the simulation of
heat transfer and splitting tensile test. This numerical method avoids the difficulty and
complexity of conducting a large number of high-temperature experiments, while
maintaining the accuracy and repeatability of the test results. The early residual tensile
strength at any temperature in the range of 25°C to 400°C could then be used as an
input parameter in the determination of whether spalling would occur, taking into
account the pore pressure of heated concrete. In the finite element model, for any
generic point if the tensile stress exceeds the early residual tensile strength it is
considered to have failed, and when a sufficient area of internal surface has failed to
allow free body motion, spalling occurs.

7.6 Conclusions

The tensile strength of HSC would be lowered at elevated temperature between 100℃
to 400℃. In general, tensile strength decreases with increasing temperature, but the
trend for residual and early residual tensile strength are different.

For the early residual splitting tensile strength, a more significant strength loss occurs
in the range of 25℃ to 100℃ and 300℃ to 400℃. The sharp strength loss at around
100℃ is assumed to be caused by the micro- and macro-cracks as well as the vapor
pressure in capillaries. The main cause for the latter sharp loss is assumed to be the

134
dehydration of C-S-H and Ca(OH)2. The degradation of strength in 100℃ to 300℃
could be attributed to the reduction in the decomposition of hydration product in the
cement paste. For the residual strength, the changes are more gradual.

The residual strength decreases for concrete specimens subjected to elevated

temperature. However, this is not the strength of HSC at a higher temperature. The early
residual tensile strength is a direct measure of the strength of HSC at elevated
temperature, which is considerably lower than the residual strength as pore pressure is
high at higher temperatures. As a result, in the determination of spalling of HSC under
fire, the early residual tensile strength might be a better indicator for the occurrence of
spalling.

The process of heat transfer and splitting tensile test could be reproduced by the
numerical method, providing additional evidence and information that HSC loses
strength at elevated temperatures. The early residual splitting tensile strength can be
simulated by incorporating the temperature profile and inputting the required concrete
material properties in ABAQUS. Relative to the residual strength, the simulation results
agree better with the experimental results in numerical form and general trend.

The experimental tests and the numerical simulation in this study prove the feasibility
of determining the early residual splitting tensile strength of HSC at elevated
temperature (up to 400℃). The early residual splitting tensile strength of HSC could
serve as a critical parameter in predicting the occurrence of spalling, which is the cause
of major damage to structural members subjected to fire load.

135
 Table 7.1 Concrete mix proportions
Fine 10 mm 20 mm
Cement SP Water
Grade aggregate aggregate aggregate w/c
(kg/m3) (kg/m3) (kg/m3)
(kg/m3) (kg/m3) (kg/m3)
C60 651.60 488.70 488.70 465.60 1.72~2.4 180.00 0.39
C90 651.60 488.70 488.70 639.30 12.68~13.88 148.00 0.23

Table 7.2 Specimen sizes of different groups

Group Ⅰ Ⅱ Ⅲ
Diameter (mm) 150 150 100
Height (mm) 300 250 200

Table 7.3 Concrete properties

28d
28d Moisture
Splitting
Batch of Compressive Standard Standard content
Grade tensile
concrete strength deviation deviation by mass
strength
(MPa) (%)
(MPa)
B1 58 0.55 3.61 0.23 4.52
B2 56 2.35 3.71 0.13 4.32
C60
B3 54 2.14 3.43 0.12 4.84
B4 56 2.18 3.45 0.13 4.59
B1 90 2.64 5.36 0.51 2.02
B2 96 3.88 6.57 0.12 2.04
C90
B3 91 5.39 5.88 0.19 3.31
B4 93 4.23 6.02 0.15 3.56

Table 7.4 Splitting tensile strengths at elevated temperatures

Splitting tensile strength 𝑓𝑡,𝑇 (MPa)
Temperature
C60 C90
25℃ 3.43 3.45 3.71 3.61 5.88 6.02 5.36 6.56
100℃ 2.30 2.73 4.07 4.40
200℃ 2.20 2.58 3.67 4.42
300℃ 1.95 2.29 3.50 3.73
400℃ 1.47 1.63 2.27 2.40

136
 Table 7.5 Average splitting tensile strengths at elevated temperatures

C60 C90
Temperature Average 𝑓𝑡,𝑇 Standard Average 𝑓𝑡,𝑇 Standard
(MPa) deviation (MPa) deviation
25℃ 3.55 0.13 5.95 0.50
100℃ 2.52 0.31 4.24 0.24
200℃ 2.39 0.27 4.05 0.53
300℃ 2.12 0.24 3.62 0.16
400℃ 1.55 0.12 2.33 0.09

Table 7.6 Temperature-dependent thermal properties

Specific heat Coefficient
Conductivity Density
(J/kg⋅K) of thermal
(× 10-3) T(℃) T(℃) (kg/mm3) T(℃) T(℃)
-6 expansion
(W/mm⋅K) MC=3% MC=1.5% (× 10 )
(× 10-3)
1.33 20 900 900 20 2.40 20 0.02 20
1.23 100 900 900 100 2.40 115 0.90 100
1.11 200 2020 1470 115 2.37 160 2.00 200
1.00 300 1540 1249 155 2.35 200 3.10 300
0.91 400 1000 1000 200 2.33 250 4.20 400
0.82 500 1050 1050 300 2.32 300 4.75 450
0.75 600 1100 1100 400 2.30 350 6.50 500
0.69 700 1100 1100 600 2.28 400 10.10 600
0.64 800 1100 1100 800 2.24 600 11.90 650
0.60 900 1100 1100 1000 2.20 800 11.90 800
0.57 1000 1100 1100 1200 2.15 1000 11.90 1000
0.55 1200 2.11 1200

Note: MC = Moisture Content

Table 7.7 Thermal Properties

Convection Stefan-Boltzmann
Absolute zero Emissivity related to
factor constant
temperature (℃) concrete surface
(W⋅m-2⋅K-1) (W⋅mm−2⋅K−4)

-273.15 25 0.7 5.67 × 10-14

137
Table 7.8 Simulated splitting tensile strengths (𝑓𝑡,𝑇 ) with or without transportation
With transportation (MPa) Without transportation (MPa)
Temperature
C60 C90 C60 C90
100℃ 2.91 4.92 2.81 4.77
200℃ 2.50 4.21 2.42 3.92
300℃ 2.16 3.49 2.08 3.30
400℃ 1.55 2.46 1.46 2.25

Table 7.9 Temperature-dependent properties of HSC

Young's modulus (MPa)
Poisson's T
C60 C90
ratio (℃)
B1 B2 B3 B4 B1 B2 B3 B4
39483 39024 38554 39024 45710 46704 45879 46213 0.20 20
39049 38595 38129 38595 45208 46191 45374 45704 0.17 100
37509 37073 36626 37073 43425 44369 43585 43902 0.14 200
30402 30048 29686 30048 35197 35962 35327 35584 0.11 300
23295 23024 22747 23024 26969 27556 27069 27265 0.09 400
16188 16000 15807 16000 18741 19149 18810 18947 0.08 500

Table 7.10 Temperature-dependent CDP parameters

Dilation angle Eccentricity fb0/fc0 K Viscosity Parameter T (℃)
30 0.1 1.16 0.667 0.0005 20
30 0.1 1.20 0.667 0.0005 100
30 0.1 1.25 0.667 0.0005 200
30 0.1 1.30 0.667 0.0005 300
30 0.1 1.39 0.667 0.0005 400
30 0.1 1.48 0.667 0.0005 500
30 0.1 1.57 0.667 0.0005 600
30 0.1 1.66 0.667 0.0005 700
30 0.1 1.70 0.667 0.0005 800
30 0.1 1.70 0.667 0.0005 1200

138
Figure 7.1 Residual tensile strength of HSC (C60-C76)

Figure 7.2 Residual tensile strength of HSC (C80-C100)

139
Figure 7.3 Arrangement of thermal couples in a specimen

Figure 7.4 Location of thermal couple inside furnace

Figure 7.5 Schematic representation of heating regime

140
Figure 7.6 Control panel and chamber of furnace

Figure 7.7 Trolley installed with a jig

141
(a)

(b)

Figure 7.8 Comparison of temperature distribution of C60 specimens: (a) Centre and
(b) 50 mm from centre

142
(a)

(b)

Figure 7.9 Comparison of temperature distribution of C90 specimens: (a) Centre and
(b) 50 mm from centre

143
(a) Ambient specimen (b) Heated specimen

Figure 7.10 Failure mode of specimens

Figure 7.11 Temperatures before and after transportation

144
Figure 7.12 Early residual tensile strengths compared with simulation results

Figure 7.13 Relative residual tensile strength of HSC (C60-C76)

145
 Figure 7.14 Relative residual tensile strength of HSC (C80-C100)

Figure 7.15 Average early residual tensile strengths compared with residual strengths

146
Note: RS = Residual strength; ERS = Early residual strength

Figure 7.16 Comparison between average relative strengths

147
8 PROTECTIVE MEASURE – BILAYER COLUMN TEST

8.1 Introduction

In practical engineering projects in Hong Kong, normal strength concrete (NSC) is

usually adopted in the construction of basement other than high strength concrete (HSC)
despite its advantages in various aspects. The hesitation to use HSC in basement and
also other locations is mainly attributed to the potential occurrence of fire-induced
spalling at elevated temperature. However, if HSC can be successfully applied in
construction of locations such as basement and structural elements such as columns,
tremendous economic and aesthetic benefits would be achieved. For instance,
application of HSC can reduce the size of elements, not only cutting the material cost
but also creating more workable space. This is especially important in places like Hong
Kong where property and housing prices are high. Protective measures have to be
adopted to make sure that HSC is properly and safely applied. As discussed in Section
2.4, there are several ways of protecting concrete from fire-induced spalling, for
example, the introduction of fibres, the use of air-entraining agent, adjusting the mix or
structural detailing of concrete elements, and adopting thermal barrier. Among the
mentioned measures, thermal barrier is by far the one measure that does not have
obvious counter effects on or compromise the properties and performance of concrete,
while being relatively effective in preventing spalling. However, current studies on the
thermal barrier are mainly conducted on slab elements, such as those by Lu and Fontana
(2015) and Klingsch et al. (2013). In this study, a bilayer system is proposed to
investigate the effect of thermal barrier on columns. The bilayer columns consist of a
high strength concrete (HSC) core and a normal strength concrete (NSC) outer layer.
The NSC outer layer is designed to act as an insulation cover to reduce the thermal
gradient and also serve as a lateral confinement to prevent the HSC core from spalling.
Compression and fire tests are conducted on the bilayer specimens to investigate their
strength and behaviour at ambient and elevated temperature. This chapter is mainly
devoted to the experimental study on the bilayer column test, prior to which a pioneer
test is conducted on the cylinders to preliminarily investigate the bilayer system and

148
prepare for the column test in larger scale. Details on the pioneer test are presented in
Appendix D and therefore not elaborated in this chapter.

8.2 Experimental Program

8.2.1 Concrete Mix and Test Specimens

Totally 20 columns are cast to investigate the feasibility of the bilayer design, of which
fifteen are put to fire test and the other five to ambient compression test. Fifteen of the
columns are bilayer columns and the other five are normal HSC columns. The outer
layer of bilayer column is cast with C30 NSC and the inner core with C90 HSC. Details
of different types of specimen are presented in Table 8.1 and Figure 8.1. Mix proportion
of the concrete are shown in Table 8.2.

From Table 8.1, it can be seen that the outer diameter of all the columns are the same
(238 mm) and two thicknesses of the NSC layer are adopted, namely 24 mm and 34
mm. The thickness is determined mainly according to the Hong Kong Code of Practice
for Fire Safety in Buildings (2011), where the minimum concrete cover to main
reinforcement for fire resistance rating (FRR) of 60 and 120 minutes are 25 mm and 35
mm respectively. Also, taking into account the conclusions drawn by Lu and Fontana
(2015) that the minimum thickness of thermal barrier subjected to two hours’ exposure
of hydrocarbon fire is 25 mm, the thicknesses are chosen to be 24 mm and 34 mm
respectively. Among the total 20 columns, five of them are plain concrete columns
through which the performance of specimens with or without reinforcement could be
compared. The effect of confinement (tie) spacing s is also studied, where the control
spacing of 100 mm is compared with spacings of 50 mm and 200 mm. It may also be
noted that the longitudinal reinforcement has been controlled to a minimal amount so
that its effect could be ignored and its main function is to provide support for the
confinement. The fire test columns are subjected to fire exposure of either 60 or 120
minutes, reasoning that the fire-induced spalling mainly occurs in the first 10 to 30
minutes of heating. The last two specimens in Table 8.1 are entirely sealed with NSC,
meaning the top and bottom surfaces of the columns are also covered with NSC.

149
Therefore, the height of the last two columns are 100 mm (50mm each top and bottom)
less than the others.

The casting of the bilayer columns is carried out in two steps. First, the HSC core is
cast and cured for one day. After the concrete has hardened, the HSC core is demoulded
and put into a larger mould, made with PVC pipes, to cast the NSC outer layer (Figure
8.2). The PVC pipes are drilled with small holes to let the thermal couples go through.
Each fire test column has two thermal couples located at the mid-height, one detecting
the temperature of the centre and the other detecting the interface between HSC core
and NSC layer. The ends of the transverse reinforcement are bent approximately 135℃
into the section to provide better anchorage with the concrete. The averaged
compressive and tensile strengths, and moisture content of the concrete are shown in
Table 8.3.

8.2.2 Test Setup and Procedures

The compression tests are conducted at the structural laboratory of the University of
Hong Kong. The loading is applied by a 5000-kN hydraulic actuator shown in Figure
8.3. The lower end of the compression machine is the hydraulic jack while the upper
end is the fixed steel frame. Four linear variable displacement transducers (LVDTs)
with a stroke of 100 mm are placed symmetrically at four sides of the column to
measure the axial deformation. Eight strain gauges are attached to each column at three
different heights to investigate the load distribution of the column. The configuration
of the LVDTs and strain gauges are drawn in Figure 8.4. The specimens are cured under
ambient conditions in the laboratory for approximately one month before undergoing
the compression test. Prior to applying the axial loading, a transfer plate made with
fresh-mixed plaster slurry is placed on the specimen to level the top surface. The
specimen is then immediately preloaded at 5 kN for 10 minutes before the plaster
hardened. Once the preloading is finished, the axial load is applied continuously using
displacement control with test rate of 1 mm/min. After the specimen fails, the test
switches to load control and release loading at a rate of 500 kN/min.

150
The fire tests are conducted at the fire laboratory of Research Engineering Development
(RED) Fire and Façade Consultants, Ltd., in Hong Kong. The internal dimensions of
the furnace used in this study is 3m × 3m × 1m. As shown in Figure 8.5, the furnace
has nine built-in thermal couples to control the furnace temperature, which is
programmed to follow the ISO 834 standard temperature-time curve (Figure 8.6) for 60
minutes and 120 minutes, and can be calculated by the equation:

T = 345 log10 (8t + 1) + T0 (8.1)

where T and T0 is the furnace and ambient temperature (℃), t is the heating time in
minutes. The temperature achieved after 60- and 120-minutes’ firing is 945℃ and
1050℃ respectively. There are four gas burners each on the left and right side of the
furnace, from which the hydrocarbon fire comes out. Temperature of the specimens
will be detected by the RS Pro K-Type Thermocouple. The measurable temperature
range is from -100 ℃ to 1100 ℃ . Probe length and diameter are 3 m and 3 mm
respectively. The columns are placed in a row at approximately uniform distance.

8.3 Compression Test Results and Discussions

The compression test results of all five specimens are shown in Table 8.4, where Pexp
and Ppre are the experimentally obtained and predicted axial load capacity, Δu is the
axial deformation corresponding to the ultimate load. The predicted ultimate load
capacity is calculated with the following equation (Wang and Su, 2014, Campione,
2012, Su and Wang, 2012, Calderón et al., 2009):

Ppre = 0.85𝐴𝑐 𝑓𝑐 + 𝐴𝑠 𝑓𝑠𝑦 (8.2)

where 𝐴𝑐 and 𝑓𝑐 are the area and cylinder compressive strength of high strength
concrete, 𝐴𝑠 and 𝑓𝑠𝑦 are the area and yield strength of longitudinal reinforcement. The
coefficient of 0.85 is adopted to consider various factors involving the loading rate,
height-to-width ratio, defects introduced during casting and compaction of the concrete.
If the cube compressive strength 𝑓𝑐𝑢 is used instead of the cylinder compressive

151
strength 𝑓𝑐 , the corresponding coefficient would be 0.67 assuming the cylinder strength
is 80% of the cube strength (𝑓𝑐 = 0.8𝑓𝑐𝑢 ).

8.3.1 Pexp /Ppre Ratios

It can be seen from Table 8.4 that the ratio of the recorded and predicted capacity falls
between 0.93 (HSC-1) and 1.43 (2L174-1), and the mean value of the five specimens
is 1.15. The overall performance of the specimens was better than expected, where all
specimens’ experimentally obtained value exceeded the predicted value except the
control specimen (HSC-1). And the ratio increased with decreasing section of high
strength concrete. This could be due to the fact that the low water/cement ratio of HSC
renders it rather viscous and to set more quickly. Since the casting was done manually
in laboratory and the specimens were of relatively large dimensions, the casting of
larger volume of HSC required more time and involved more difficulties to vibrate the
concrete and avoid it from setting. Therefore, more defects could be introduced with
larger section of HSC in the specimen. The actual capacities of the reinforced and plain
2L174 specimens were higher than the corresponding 2L190 specimens, which
indicates that the increase of HSC section could not compensate totally for the
compromising of strength due to the defects. Comparing ratios of the specimens with
or without reinforcement, it can be found that the ratios of the reinforced columns were
slightly higher than its plain concrete counterpart. This could be explained by the
confinement effect brought by the transverse reinforcement, which was ignored in the
prediction of ultimate capacity in (8.2). Hence, the predicted value of the reinforced
specimens (2L174-1 and 2L190-1) was lowered and the ratio was increased.

8.3.2 Axial Deformations

Figure 8.7 to Figure 8.9 present the load-axial deformation curves of different
specimens. The effect of the outer layer thickness on the reinforced specimens is shown
in Figure 8.7. It can be seen that the ascending segments of the control specimen and
2L190-1 are almost coincident and the deformations at ultimate capacity Δu are
approximately the same. While the curve of 2L174-1 specimen climbs with a higher

152
rate and the deformation reached at peak load is less than the other two. This
phenomenon may be explained by the technician’s maloperation of the compression
machine. All specimens are supposed to be loaded at a rate of 1 mm/min after 10
minutes’ preloading maintained at 5 kN. During the loading phase of the 2L174-1
specimen, however, the technician accidentally applied a wrong program and the
specimen was instantaneously loaded to 1500 kN with a significantly high loading rate.
The stiffness and compressive strength of concrete would be increased at a high loading
rate (Fu et al., 1991), as a result, the curve has a sharper ascending branch and a
relatively larger ultimate capacity. Though the loading was immediately stopped and
switched to the correct rate, it is plausible to assume that some microcracks have formed
in the concrete and irreversible deformation had occurred. Therefore, even when the
specimen was reloaded according to the designated program, the recorded deformation
would be smaller than it should have been. The axial deformation curves of the
specimens cast with plain concrete are shown in Figure 8.8. The shape of curve and the
deformation corresponding to ultimate capacity of the plain specimens are similar and
the differences are within reasonable range. Thus, the influence of the layer thickness
is insignificant on the deformation of plain specimens. Figure 8.9 shows the effect of
the reinforcement on the specimens. It can be clearly seen that the compared curves
have an analogous curve shape and value at peak load. The reinforced specimen
(2L190-1) has a slightly higher ultimate capacity and steeper ascending branch. This is
reasonable because the presence of reinforcement can confine and stiffer the concrete
to certain extent. The reinforcement, though negligible in amount, could also provide
extra capacity for load bearing. Overall, the reinforcement has minimal influence on
the performance of the bilayer columns except confinement effect.

8.3.3 Failure Modes and Strains of Concrete

Figure 8.10 and Figure 8.11 shows the failure modes of the plain and reinforced
specimen. After applying compressive load for the first several minutes, vertical cracks
started to appear at the top end of the specimen. As the test proceeded, the cracks
gradually propagated downwards along the direction of height and evolved into major
cracks. Upon failure of the specimen, the outer NSC layer was penetrated by some of

153
the major cracks and then partly detached from the HSC core. Comparing the two
figures, no major difference could be observed and the failure modes of plain and
reinforced specimen were actually quite similar. The top view of a failed specimen
(Figure 8.12) clearly reveals the outer layer penetrated by major cracks. Part of the NSC
outer layer fell off when the specimen was removed from the compression machine,
and the inner HSC core was exposed (Figure 8.13). It can be seen that vertical cracks
also formed in the HSC core along the direction of height and thin layers of concrete
split from the core.

To explain the phenomenon that the outer NSC layer detaches from the HCS core, the
effect of strain incompatibility should be considered. According to previous studies
(Wee et al., 1996, Carreira and Chu, 1985, Lim and Ozbakkaloglu, 2014, Yang et al.,
2014) on the stress-strain relationship covering a wide range of concrete grade, the
strain of concrete at peak stress can be represented by a function of concrete
compressive strength. This means the concrete stain at peak stress increases with an
increase of compressive strength. Before the NSC reaches its peak stress, the outer layer
and inner core deform simultaneously and the bilayer column remains intact. Once the
NSC has reached its peak stress, the outer layer fails while the HSC continues to deform
with increasing strain. At this moment, major cracks form along the longitudinal
direction and the outer layer detacher from the core to release the stress. It is the HSC
core that takes effect in load bearing by the time a bilayer specimen fails. Therefore,
the strength of NSC is ignored when predicting the ultimate capacity using (8.2). The
effect of strain incompatibility could be amplified if the inner core is confined with
reinforcement, when the concrete strain at peak stress is increased.

8.4 Fire Test Results and Discussions

The fire tests were carried out three times for different purposes. The first and second
tests were to find out the effect of duration of fire exposure. The specimens were
subjected to hydrocarbon fire for one and two hours in the first and second test
respectively. The duration of fire exposure of the third test was also one hour, where

154
the specimens were entirely sealed with NSC and compared to the counterparts in the
first test.

8.4.1 Temperature Distribution

The temperatures inside the furnace were measured by the lower six built-in thermal
couples nearest to the specimens. The mean furnace temperature of the three fire tests
are compared with the ISO standard temperature-time curve. As shown in Figure 8.14,
the measured temperatures agreed well with the standard curve. The difference between
the thermal conductivity of NSC and HSC is not indistinguishable, but the difference
is generally minimal and can be ignored. In Figure 8.15 and Figure 8.16, “C” stands for
the centre position while “L” stands for the position at layer interface. Figure 8.15
shows the temperature distributions of unsealed specimens with different layer
thickness. Since the outer diameters were identical for all specimens, the temperature
at centre of the section should be approximately the same regardless of thickness, which
is demonstrated by the figures. The thickness of NSC layer was 24 and 34 mm
respectively for the 2L190 and 2L174 specimens, so temperatures at the layer interface
of 2L190 specimens were higher than the 2L174 specimens. The outer thermal couple
of HSC specimens was placed at the same depth of the thickness of 2L174 specimens,
thus, the measured temperatures of HSC and 2L174 specimens were similar. The results
illustrate that the presence of NSC outer layer did not alter the transfer of heat and the
temperature distribution largely depended on the depth from the heated surface.

The temperature distributions of the sealed specimens are plotted and compared with
their unsealed counterparts in Figure 8.16, where “S” stands for sealed specimens.
During the third fire test, the NSC outer layer of 2L190S and 2L174S were predicted
to spall off the core after 10 to 20 minutes since heating commenced. The HSC core
was then directly exposed to fire. Consequently, there is an abrupt jump of temperature
at the layer interface around 10 and 15 minutes respectively in Figure 8.16 (a) and (b).
Later, the centre temperatures correspondingly increased due to the loss of thermal
barrier. In general, the temperature distributions of sealed and unsealed specimens are
similar, except the sudden increase upon the falling of NSC layer of sealed specimens.

155
8.4.2 Failure mode and Spalling Analysis

Figure 8.17 shows the typical failure mode of HSC, unsealed bilayer and sealed bilayer
specimens. During the first 30 minutes of the unsealed tests, sounds of concrete splitting
and popping were heard. These sounds are predicted to come either from the spalling
of HSC specimens or the cracking of the bilayer specimen. The ordinary HSC
specimens worked as the control group, where no protective measure was adopted and
spalling occurred as expected. As illustrated in the figure, only a thin layer of concrete
spalled and no reinforcement was exposed. After measurement, the maximum depth of
the spalling area was less than 30 mm. These characteristics correspond to the aggregate
spalling discussed in Section 2.2.1, whose influence is superficial. Similar crack
patterns could be observed on the unsealed specimens regardless of layer thickness. The
cracks initially formed at the two ends of the specimen and then gradually propagated
towards the middle. The NSC outer layer at the two ends slightly detached from the
core. Though the crack patterns were similar, the specimens with smaller layer
thickness generally experienced cracks with larger length and width. The water vapour
was dissipated through the cracks, preventing the occurrence of spalling. Since no
explosive spalling was observed on HSC in all specimens, the effect of confinement
spacing cannot be concluded. The HSC core of the plain specimen was broken in the
middle during hung out from the furnace, while the HSC core of all the reinforced
specimens remained intact. This phenomenon suggests that although the steel
reinforcement could expand and generate additional stresses on the concrete material
at elevated temperature, the confinement effect brought by the steel reinforcement can
still outweigh its negative effects and help improve the spalling resistance. The
behaviour of specimens heated for one and two hours were analogous, indicating the
early stage (within the first hour) of fire exposure was the most critical period for
spalling. If spalling did not occur in the early stage, subsequent heating would not
induce spalling either, only deteriorating the properties of concrete.

The schematic representation in Figure 8.18 depicts a quarter of the unsealed specimen,
where the gap between the outer layer and inner core is exaggerated for the purpose of
demonstration. The blue arrows represent the transport of water vapour. When the

156
specimen was continuously heated, the water contained in the concrete pores started to
evaporate. Although the permeability of NSC is larger than HSC, the rate of dissipation
still cannot catch up with the rate of evaporation if vapour is dissipated through the
NSC layer only. Since the specimen was unsealed at two ends, water vapour could
escape through the gap between the outer layer and inner core. The bond connecting
the layer and core was weakened by the vapour flow and eventually destroyed at the
two ends where most the vapour was released. The red arrows represent the thermal
dilation of concrete, which is predominantly determined by the rise of temperature.
Higher temperature could lead to larger thermal dilation. The temperatures at the two
ends were unavoidably higher than the middle, although the top and bottom surfaces of
every specimen were covered with insulating cotton. The thermal dilation increased
gradually from mid-section along height and maximized at the ends, further pushing
the NSC outer layer outwards in the radial direction. The combined effect of trapped
water vapour and nonuniform thermal dilation caused the cracking and separation of
outer layer at two ends. Larger layer thickness could be beneficial to restricting the
propagation of cracks.

The sealed bilayer specimens were designed to better simulate the conditions of
columns in a structure, where the two ends would not be exposed to the surrounding
atmosphere as the unsealed specimens. Though the HSC and unsealed bilayer
specimens experienced surface spalling and layer cracking respectively, they can
basically remain intact after the fire test (the loss of concrete in Figure 8.17 (b)
happened during transportation after the test). It can be seen, however, that the NSC
outer layer of the sealed specimen barely exists after the test. The sound of loud bang
was heard at about 10 and 14 minutes of the test, followed immediately by the soaring
of temperature. Based on the sound and change of temperature, spalling of the NSC
outer layer was very likely to occur at the mentioned time. The two ends were sealed
by a 50 mm layer of mortar, stopping the release of water vapour to the environment.
Not being able to dissipate through the outer layer or escaped from the interfacial gap,
the water vapour gradually accumulated. By the time the accumulated vapour pressure
exceeded the bond strength of the interface, explosive spalling occurred resulting in the
breakaway of NSC outer layer. The interfacial bond strength has to be increased in

157
order to avoid the spalling of the outer layer. One way to strengthen the interfacial bond
is to increase the thickness of outer layer. The 24- and 34-mm thickness proved to be
insufficient to endure the fast-rising vapour pressure. As previous studies (Lu and
Fontana, 2015, Klingsch et al., 2013) pointed out, thermal barrier could prevent
explosive spalling only if the protection was sufficient, otherwise the postponed
spalling may have even more severe influence. Another way to increase bond strength
is to roughen the surface of inner core prior to the casting of the outer layer, or provide
good anchorage to the outer layer, so that the two components could be more integrated
as a whole. Methods include:

1) Generate slight scratches on the surface of inner core using an electric saw or
grinder to increase the coefficient of friction between concrete layers. Such
method should be carefully applied since the saw or grinder could also bring
potential damage to the concrete.
2) Transferring part of the inner core reinforcement to the outer layer or increasing
the amount of the reinforcement by reducing the diameter of reinforcement used.
The bond strength between steel reinforcement and concrete could be as high as
25.54 MPa after 28-day’s curing (Shen et al., 2016), which obviously exceeds
the bond strength between concrete layers and could provide better anchorage
for the outer layer.
3) Use a thin layer of reinforcement mesh in between the layers. The diameter of
wires and dimension of pitches could refer to Eurocode 2 Part 1-2 (2004). The
reinforcement mesh should be properly embedded in or connected to the inner
core to provide enough anchorage such that the outer layer is firmly hold to the
inner core.

The aforementioned methods could potentially increase the bond strength between
inner core and outer layer if properly applied. However, careful considerations should
be taken during application to guarantee the quality of concrete and performance of
specimens.

158
8.4.3 Minimum Effective Thickness of Outer Layer

As discussed in Section 8.4.2, insufficient protection from the thermal barrier may lead
to exacerbated consequences once spalling occurs. Based on the experimental results
obtained from the bilayer column tests, the adopted thicknesses (24 mm and 34 mm) of
the NSC outer layer are inadequate. Though further experimental studies are desirable,
it is more practical to first estimate the minimum effective thickness of outer layer. In
this section, therefore, numerical analyses are conducted to investigate the influence of
thickness of outer layer.

The temperature fields over time at the layer interfaces and centre of bilayer column
are simulated and presented in Figure 8.19. By comparison, it can be found that the
numerical results show good agreement with the experimental results. According to the
discussions on Figure 8.16 in Section 8.4.1, the spalling of the 24- and 34-mm outer
layer was assumed to occur respectively at 10 and 15 minutes since heating. One
important fact to be spotted in Figure 8.19 is that temperatures at the predicted times of
the layer interfaces were approximately the same, with a value of 163 and 150℃
respectively at the thickness of 24 and 34 mm. This fact indicates that the interfacial
bond strength may be weakened and damaged at temperature around 150℃, if under
the same fire scenario. Based on the data obtained from experiments introduced in
Chapter 7, strength reduction of the early residual tensile strength between 100 and
200℃ is in the range of 0.71 ~ 0.68. The averaged tensile strength of the normal strength
concrete at ambient temperature (2.95 MPa at 28-day test from Table 8.3) is taken as
the initial interfacial bond strength. But considering that the strength of interfacial bond
is lower than the tensile strength of NSC (otherwise the NSC would spall prior to the
damage of interfacial bond), a lower strength reduction coefficient of 0.68
corresponding to the temperature of 200℃ is adopted. The estimated interfacial bond
strength at around 150℃ is 2.95 × 0.68 = 2.01 MPa.

The evolution of pore pressure in bilayer column with outer layer thickness of 24 and
34 mm is shown respectively in Figure 8.20 (a) and (b). The dash line in the figures
represents the position of the interface between layers. The 24-mm outer layer of the

159
2L174 column was assumed to spall at around 10 minutes. As can be seen in Figure
8.20 (a), the maximum pressure occurred within the outer layer in the first 10 minutes
was 2.02 MPa, almost the same with the estimated bond strength. The 34-mm outer
layer of the 2L190 column was assumed to spall at around 15 minutes, and as Figure
8.20 (b) shows, the maximum pressure occurred was 2.78 MPa. Although the maximum
pore pressure was slightly higher than the estimated bond strength, the thickness of
outer layer was also increased by 10 mm. Therefore, the required pore pressure to break
the outer layer was correspondingly increased. This could be interpreted as the
interfacial bond strength being reinforced. It can also be deducted that an increase of
thickness by 10 mm could approximately improve the bond strength by 0.7 MPa. The
reduced peak pressure in Figure 8.20 (b) compared with Figure 8.20 (a) could be
explained by the higher permeability of normal strength concrete. The thicker the outer
layer, the lower the peak pressure at a same time. Based on the discussions above, the
minimum effective thickness of outer layer can be determined by considering the
combined effects of temperature, pore pressure and bond strength.

Since fire-induced spalling normally occurs within the first 30 minutes of heating
(Dong et al., 2014, Bilodeau et al., 2004, Khaliq and Kodur, 2013), the temperature at
the layer interface should not exceed 200℃ at 30 minutes. The reinforced bond strength
should be higher than the maximum pore pressure occurred within the outer layer. After
simulations of different thicknesses, the minimum value that satisfies the above
requirements is 50 mm. The simulated temperature at 50 mm from the heated surface
is shown in Figure 8.19. The dash line again represents the position of layer interface.
The temperature after 30 minutes’ heating is around 190℃, the estimated bond strength
of 2.01 MPa is applicable. The evolution of pore pressure within the first 40 minutes is
shown in Figure 8.21. The maximum pore pressure of 3.35 MPa occurs at 30 minutes,
after which the peak pressure starts to decrease. The thickness is increased by 26 mm
(from 24 to 50 mm), and the reinforced bond strength is estimated to be 3.5 MPa,
slightly higher than the maximum pressure. Therefore, the thickness of 50 mm should
be theoretically sufficient for preventing spalling. Considering the defects that could be
possibly introduced during casting and for higher safety level, the minimum effective
thickness is recommended to be 100 mm, preferably with steel reinforcement. For

160
further illustration, calculation for a realistic column of 2 m diameter with a layer of
100 mm is conducted. As shown in Figure 8.22, the maximum pore pressure of
approximately 4 MPa occurs at 30 minutes since heating, and the corresponding bond
strength is estimated to be around 7 MPa according to the abovementioned method. The
layer thickness of 100mm should be adequate for preventing spalling. Besides, methods
to improve the interfacial bond strength, suggested in Section 8.4.2, can be
simultaneously adopted to further guarantee the effectiveness of outer layer.

8.5 Conclusions

The overall performance of bilayer columns was better than expected in the
compression test. The difficulties in casting HSC introduced defects that lowered its
strength, resulting in decreasing Pexp /Ppre ratio with increasing HSC section. The
Pexp /Ppre ratio of reinforced specimens were larger than the corresponding plain
specimens primarily due to confinement effect of transverse reinforcement.

The shape of the load-deformation curve and the axial deformation at peak load of
bilayer specimens were similar to the control specimens, indicating that the layer
thickness did not have a significant effect on the axial deformation. The influence of
reinforcement on the load-deformation behaviour of bilayer specimens was also
minimal.

When the specimens were under compression, cracks developed from the top end and
propagated along height. By the time specimens failed, several major cracks penetrated
the outer layer which detached from the inner core at the end. This phenomenon was
mainly caused by the strain incompatibility of NSC and HSC at peak stress. The
observed failure mode applied to both the plain and reinforced bilayer specimens.

The temperature distributions of the unsealed and sealed specimens were compared
against the control specimen. The presence of NSC outer layer did not alter the transfer
of heat within the bilayer specimens and the temperature was mainly determined by the
depth of measured position, as long as no spalling occurred.

161
Aggregate spalling occurred on the control specimens, where the maximum depth of
spalling area was less than 30 mm. The cracks on the NSC outer layer of unsealed
specimens helped release the vapour pressure and prevent spalling. The most critical
period of heating was within the first hour, after which the duration of fire exposure
would not have large impact on spalling. Mix proportions of HSC should be adjusted
to induce explosive spalling so that the effect of confinement spacing could be
investigated.

Similar crack patterns were observed on the unsealed specimens regardless of layer
thickness, where cracks first formed at the two ends, developed along height and
penetrated the depth of layer. The combined effect of trapped water vapour and
nonuniform thermal dilation caused the cracking and separation of outer layer at two
ends. Larger layer thickness helped restrict the propagation of cracks.

The accumulated vapour pressure caused the breakaway of NSC layer of sealed
specimens. Experimental results suggested that insufficient protection may exacerbate
the influence of postponed spalling. Measures such as increasing the thickness of
thermal barrier and roughen the surface of inner core should be adopted to prevent
explosive spalling.

Based on the experimental results and numerical analyses of bilayer columns, a

minimum effective thickness of the outer layer of 100 mm is recommended for further
study. Proper steel reinforcement is preferable to guarantee higher safety level. The
recommended thickness is estimated to be efficient for preventing the breakaway of
outer layer under ISO fire scenario.

162
 Table 8.1 Details of specimens
Longitudinal FRR
Specimen D (mm) d (mm) t (mm) H (mm) Ties s (mm)
rebar (min)
HSC-1 238 238 0 750 6T6 R6 100 0
2L190-1 238 190 24 750 6T6 R6 100 0
2L174-1 238 169 34 750 6T6 R6 100 0
2L190P-1 238 190 24 750 N/A N/A N/A 0
2L174P-1 238 169 34 750 N/A N/A N/A 0
HSC-2 238 238 0 750 6T6 R6 100 60
2L190-2 238 190 24 750 6T6 R6 100 60
2L190-4 238 190 24 750 6T6 R6 200 60
2L174-2 238 169 34 750 6T6 R6 100 60
2L174P-2 238 190 24 750 N/A N/A N/A 60
2L190P-2 238 169 34 750 N/A N/A N/A 60
HSC-3 238 238 0 750 6T6 R6 100 120
HSCP 238 238 0 750 N/A N/A N/A 120
2L190-3 238 190 24 750 6T6 R6 100 120
2L190-5 238 190 24 750 6T6 R6 50 120
2L174-3 238 169 34 750 6T6 R6 100 120
2L174P-3 238 169 34 750 N/A N/A N/A 120
HSC-4 238 238 0 750 6T6 R6 100 60
2L190S 238 190 24 650 6T6 R6 100 60
2L174S 238 169 34 650 6T6 R6 100 60

Note: D = Outer diameter; d = Inner HSC core diameter; t = NSC layer thickness; H =
Height of column; s = Spacing of confinement;

163
 Table 8.2 Concrete mix proportions
Target Fine 10mm Silica
Cement Water
strength aggregate aggregate fume SP (g/m3) w/c
(kg/m3) (kg/m3)
(MPa) (kg/m3) (kg/m3) (kg/m3)
30 1684 NA NA 329 Adjust to 214 0.7
140 mm
90 633 950 20 634 slump 163 0.26

Table 8.3 Averaged properties of concrete

Compressive Tensile Moisture
Grade
(MPa) (MPa) content (%)
C30 33.6 2.95 3.36
C90 94.3 5.24 3.54

Table 8.4 Results of compression tests

Specimen Pexp (kN) Ppre (kN) Δu (mm) Pexp /Ppre
2L174P-1 1816.69 1433.86 4.14 1.27
2L174-1 2165.53 1512.06 31.33 1.43
2L190P-1 1755.97 1713.28 5.21 1.02
2L190-1 1936.47 1791.48 7.83 1.08
HSC-1 2551.53 2751.82 5.77 0.93
Mean 1.15

164
 Figure 8.1 Types of specimen

Figure 8.2 Before casting the outer layer of bilayer and sealed bilayer column

165
Figure 8.3 Compression test setup

166
Figure 8.4 Configuration of LVDTs and strain gauges

167
 Figure 8.5 Fire test setup

Figure 8.6 ISO 834 temperature-time curve

168
Figure 8.7 Effect of outer layer thickness of reinforced specimens

Figure 8.8 Effect of outer layer thickness of plain specimens

169
Figure 8.9 Effect of the presence of reinforcement

170
Figure 8.10 Failure mode of plain specimen

171
Figure 8.11 Failure mode of reinforced specimen

172
Figure 8.12 Top view of a failed specimen

Figure 8.13 HSC core of a failed specimen

173
Figure 8.14 Measured furnace temperatures versus standard curve

174
 (a)

(b)

Figure 8.15 Temperature distributions of unsealed specimens: (a) One-hour exposure

and (b) two-hours exposure

175
 (a)

(b)

Figure 8.16 Temperature distribution of sealed specimens: (a) 2L174 specimens and
(b) 2L190 specimens

176
(a) HSC (b) unsealed bilayer (c) sealed bilayer
Figure 8.17 Failure modes of specimens

Figure 8.18 Schematic representation of outer layer cracking

177
Figure 8.19 Temperatures at interface of bilayer columns

178
(a)

(b)

Figure 8.20 Pressure in bilayer column: (a) 2L190 and (b) 2L174

179
Figure 8.21 Pressure in bilayer column of 50-mm thickness

Figure 8.22 Pressure in bilayer column of 100-mm thickness

180
9 CONCLUSIONS AND RECOMMENDATIONS

9.1 Conclusions

Fire-induced spalling has been a primary concern in the application of concrete at

elevated temperature. The susceptibility of fire-induced spalling in concrete could be
assessed by fire testing, prescriptive method as well as performance-based method.
Considering the economic cost and accuracy of the assessment, performance-based
method is deemed more effective than the other two types of methods. In this study,
therefore, efforts have been spent on the development of a comprehensive thermo-
hygro-mechanical model to capture the behaviour of concrete at elevated temperature.
One important parameter required in the modelling of heat and mass transfer is the
tensile strength of concrete at high temperature. A special setup is designed and
manufactured to carry out the experimental study, which aims to investigate the early
residual tensile strength of concrete. Different from the residual tensile strength
measured when the specimen has cooled down, the early residual tensile strength is
measured when concrete is still at high temperature. Apart from the modelling of
thermo-hygro-mechanical process, this study also focuses on putting forward a
protective measure to prevent fire-induced spalling. A bilayer system, consisting of
HSC inner core and NSC outer layer, is proposed. Bilayer column tests following
preliminary cylinder tests are conducted to examine the feasibility of the bilayer system.

9.1.1 Modelling of Concrete at Elevated Temperature

With the purpose of modelling the thermo-hygro-mechanical process, the temperature-

dependent properties of concrete at elevated temperature have to be evaluated prior to
the development of THM model. Thermal properties, including thermal conductivity,
density, specific heat capacity, porosity, permeability and relative permeability are first
evaluated in Chapter 3. Following are the high temperature mechanical properties of
concrete, such as compressive and tensile strength, elastic modulus and thermal
expansion. Important material data related to water species and air employed in the
model are also discussed in Chapter 3.

181
Building upon the Glasgow and Padua model introduced in Chapter 2, the formulation
of the thermo-hygro model, dealing with the heat and mass transfer process within
concrete, is presented in Chapter 4. The macroscopic governing equations are
established on the basis of mass and enthalpy conservation. Relevant constitutive laws
and state equations needed for solving the system of governing equations are discussed.
The detailed derivation of the governing equations and the finite element solution
procedure is given respectively in Appendix A and Appendix B. The second component,
i.e. the thermo-mechanical component, of the numerical model is developed in a similar
procedure. The macroscopic governing equation is first discussed in Chapter 5 based
on the momentum conservation, then follow the constitutive laws and numerical
solution. Extended details for derivation are presented in Appendix C. A two-step
spalling criterion is developed using a combination of mechanical damage and strain
energy density to predict the pattern and occurrence of spalling.

Having established the HTM model, numerical validations are conducted against the
experimental work of Ichikawa and England (2004), Mindeguia et al. (2009) and
Ozawa et al. (2012), where both moderate (15℃/min) and rapid (ISO standard fire curve)
heating rates are involved. Each validation example presented in Chapter 6 starts with
a description of the problem and parameters applied in the model, then follows the
discussions on the results of numerical analyses. It is illustrated that the described
model is capable of reproducing the experimental behaviour of concrete at elevated
temperature. Last but not least, an algorithm is proposed at the end of Chapter 6 to
account for the stochastic nature of concrete behaviour with steps of implementation to
incorporate the algorithm into the model. However, the application of the proposed
algorithm is restricted at present due to the expensive computational cost. The
efficiency of each simulation must be improved before the proposed algorithm could
be applied under the current computational capability.

9.1.2 Experiment of Early Residual Splitting Tensile Strength

The early residual splitting tensile strength of concrete at elevated temperature was
investigated by conducting the splitting tensile test while the specimen was still at hot.

182
Key outcomes of the experimental study described in Chapter 7 could be concluded as
follows:

1) The tensile strength of HSC would be lowered at elevated temperature between 100℃
to 400℃. In general, tensile strength decreases with increasing temperature, but the
trend for residual and early residual tensile strength are different.
2) For the early residual splitting tensile strength, a more significant strength loss
occurs in the range of 25℃ to 100℃ and 300℃ to 400℃. The sharp strength loss
at around 100℃ is assumed to be caused by the micro- and macro-cracks as well as
the vapor pressure in capillaries. The main cause for the latter sharp loss is assumed
to be the dehydration of C-S-H and Ca(OH)2. The degradation of strength in 100℃
to 300℃ could be attributed to the reduction in the decomposition of hydration
product in the cement paste. For the residual strength, the changes are more gradual.
3) The residual strength decreases for concrete specimens subjected to elevated
temperature. However, this is not the strength of HSC at a higher temperature. The
early residual tensile strength is a direct measure of the strength of HSC at elevated
temperature, which is considerably lower than the residual strength as pore pressure
is high at higher temperatures. As a result, in the determination of spalling of HSC
under fire, the early residual tensile strength might be a better indicator for the
occurrence of spalling.
4) The process of heat transfer and splitting tensile test could be reproduced by the
numerical method, providing additional evidence and information that HSC loses
strength at elevated temperatures. The early residual splitting tensile strength can
be simulated by incorporating the temperature profile and inputting the required
concrete material properties in ABAQUS. Relative to the residual strength, the
simulation results agree better with the experimental results in numerical form and
general trend.
5) The experimental tests and the numerical simulation in this study prove the
feasibility of determining the early residual splitting tensile strength of HSC at
elevated temperature (up to 400℃). The early residual splitting tensile strength of
HSC could serve as a critical parameter in predicting the occurrence of spalling,
which is the cause of major damage to structural members subjected to fire load.

183
9.1.3 Experiment of Bilayer Column

The bilayer specimens were put to compression and fire test at ambient and elevated
temperature to investigate their mechanical and fire performance. In summary, the main
conclusions of the experimental study presented in Chapter 8 could be drawn as follows:

1) The overall performance of bilayer columns was better than expected in the
compression test. The difficulties in casting HSC introduced defects that lowered
its strength, resulting in decreasing Pexp /Ppre ratio with increasing HSC section.
The Pexp /Ppre ratio of reinforced specimens were larger than the corresponding
plain specimens primarily due to confinement effect of transverse reinforcement.
2) The shape of the load-deformation curve and the axial deformation at peak load of
bilayer specimens were similar to the control specimens, indicating that the layer
thickness did not have a significant effect on the axial deformation. The influence
of reinforcement on the load-deformation behaviour of bilayer specimens was also
minimal.
3) When the specimens were under compression, cracks developed from the top end
and propagated along height. By the time specimens failed, several major cracks
penetrated the outer layer which detached from the inner core at the end. This
phenomenon was mainly caused by the strain incompatibility of NSC and HSC at
peak stress. The observed failure mode applied to both the plain and reinforced
bilayer specimens.
4) The temperature distributions of the unsealed and sealed specimens were compared
against the control specimen. The presence of NSC outer layer did not alter the
transfer of heat within the bilayer specimens and the temperature was mainly
determined by the depth of measured position, as long as no spalling occurred.
5) Aggregate spalling occurred on the control specimens, where the maximum depth
of spalling area was less than 30 mm. The cracks on the NSC outer layer of unsealed
specimens helped release the vapour pressure and prevent spalling. The most
critical period of heating was within the first hour, after which the duration of fire
exposure would not have large impact on spalling. Mix proportions of HSC should

184
 be adjusted to induce explosive spalling so that the effect of confinement spacing
could be investigated.
6) Similar crack patterns were observed on the unsealed specimens regardless of layer
thickness, where cracks first formed at the two ends, developed along height and
penetrated the depth of layer. The combined effect of trapped water vapour and
nonuniform thermal dilation caused the cracking and separation of outer layer at
two ends. Larger layer thickness helped restrict the propagation of cracks.
7) The accumulated vapour pressure caused the breakaway of NSC layer of sealed
specimens. Experimental results suggested that insufficient protection may
exacerbate the influence of postponed spalling. Measures such as increasing the
thickness of thermal barrier and roughen the surface of inner core should be adopted
to prevent explosive spalling.
8) Based on the experimental results and numerical analyses of bilayer columns, a
minimum effective thickness of the outer layer of 100 mm is recommended for
further study. Proper steel reinforcement is preferable to guarantee higher safety
level. The recommended thickness is estimated to be efficient for preventing the
breakaway of outer layer under ISO fire scenario.

9.2 Recommendations for Further Study

While this study has made progress in researching the behaviour of concrete at elevated
temperature, further study is still required towards better and more reliable assessment.
Following are the main recommendations regarding future numerical and experimental
studies on concrete subjected to elevated temperature:

1) Treatment of progressive spalling: The current numerical model focuses only on the
one-time spalling. To consider progressive spalling, the model needs to be extended
to remove spalled concrete and continue with subsequent analysis. Boundary
conditions should be applied to the new spalling front. Spalling locations (depth and
area) are required for the determination of spalled concrete.
2) Incorporation of reinforcement: The current numerical model developed in this
study deals with plain concrete specimens and elements. Though reinforcement has

185
 little influence on the heat and mass transfer in concrete, transverse reinforcement
has shown to be effective in preventing spalling by providing confinement to
concrete. The incorporation of reinforcement should therefore be considered in
future study for more accurate assessment.
3) Improving efficiency of computation: The algorithm proposed to account for the
stochastic nature of concrete requires a substantially large number of trials to be
conducted. The computational time of each trial must be reduced to make the
proposed algorithm applicable. Areas of potential improvement include the
evaluation of element matrices, the solution procedures of governing equations and
solution techniques for the system of nonlinear equations, etc.
4) Measurement of vapour pressure: The experimental data of vapour pressure in
concrete has been restricted to one-dimensional scenarios due to difficulty in
measurement. No successful measurement has been made in two-dimensional
scenarios so far, though special experimental setup has been designed by various
researchers. If accurate measurements of vapour pressure could be made in 2D
heating scenarios, the validity of numerical model could be further proven.
5) Increasing thickness of the NSC outer layer: In the described experiment of bilayer
columns, the thicknesses (24 mm and 34 mm) of the NSC outer layer have shown
to be inadequate for resisting the vapour pressure. In future experimental study, the
thickness of outer layer should be increased to investigate the minimum applicable
thickness of bilayer column.
6) Increasing bond strength of the interface: Bond strength of the interface between
HSC inner core and NSC outer layer is critical to the behaviour of bilayer column
under compressive and thermal tests. Apart from increasing the thickness of outer
layer, the bond strength could also be increased by roughening the surface of the
inner core or providing better anchorage to the outer layer.
7) Improvement in casting the bilayer column: The current procedure for casting the
bilayer column is rather difficult and time-consuming. To ensure this new bilayer
design achievable in practical construction projects, efforts should be focused on
the investigation and simplification of casting procedure (especially the casting of
NSC outer layer) in further experimental study.

186
Appendix A Derivation of Thermo-Hygro Equations

Water vapour flux:

𝐾𝐾𝑔 𝜌̃𝑣
−𝐽𝑣 = 𝜀𝑔 𝜌̃𝑣 ∇𝑃𝑔 + 𝜀𝑔 𝜌̃𝑔 𝐷𝑣𝑎 ∇ ( ) (A.1)
𝜇𝑔 𝜌̃𝑔

𝜌̃𝑣 (𝜌̃𝑎 + 𝜌̃𝑣 )∇𝜌̃𝑣 − 𝜌̃𝑣 ∇(𝜌̃𝑎 + 𝜌̃𝑣 ) 𝜌̃𝑎 ∇𝜌̃𝑣 − 𝜌̃𝑣 ∇𝜌̃𝑎
∇( )= = (A.2)
𝜌̃𝑔 𝜌̃𝑔2 𝜌̃𝑔2

𝑃𝑔 − 𝑃𝑣 𝑃𝑔 𝑅𝑣 𝑇∇𝑃𝑔 − 𝑃𝑔 ∇𝑇 𝑅𝑣
∇𝜌̃𝑎 = ∇ ( ) = ∇( − 𝜌̃𝑣 ) = − ∇𝜌̃
𝑅𝑎 𝑇 𝑅𝑎 𝑇 𝑅𝑎 𝑅𝑎 𝑇 2 𝑅𝑎 𝑣
(A.3)
𝑃𝑔 1 𝑅𝑣
= (− ) ∇𝑇 + ( ) ∇𝑃𝑔 + (− ) ∇𝜌̃𝑣
𝑅𝑎 𝑇 2 𝑅𝑎 𝑇 𝑅𝑎

𝜌̃𝑣 𝜌̃𝑣 𝑃𝑔 𝜌̃𝑣 𝜌̃𝑎 𝜌̃𝑣 𝑅𝑣

𝜌̃𝑔 ∇ ( )=( ) ∇𝑇 + (− ) ∇𝑃𝑔 + ( + ) ∇𝜌̃𝑣 (A.4)
𝜌̃𝑔 𝜌̃𝑔 𝑅𝑎 𝑇 2 𝜌̃𝑔 𝑅𝑎 𝑇 𝜌̃𝑔 𝜌̃𝑔 𝑅𝑎

𝐷𝑣𝑎 𝜀𝑔 𝜌̃𝑣 𝑃𝑔 𝐾𝐾𝑔 𝐷𝑣𝑎

−𝐽𝑣 = ( 2
) ∇𝑇 + 𝜀𝑔 𝜌̃𝑣 ( − ) ∇𝑃𝑔
𝜌̃𝑔 𝑅𝑎 𝑇 𝜇𝑔 𝜌̃𝑔 𝑅𝑎 𝑇
𝐷𝑣𝑎 𝜀𝑔 𝑅𝑣 (A.5)
+ (𝜌̃𝑎 + 𝜌̃ ) ∇𝜌̃𝑣
𝜌̃𝑔 𝑅𝑎 𝑣
−𝐽𝑣 = 𝐾𝑉𝑇 ∇𝑇 + 𝐾𝑉𝑃 ∇Pg + 𝐾𝑉𝑉 ∇𝜌̃𝑣

Air flux:

𝐾𝐾𝑔 𝜌̃𝑎
−𝐽𝑎 = 𝜀𝑔 𝜌̃𝑎 ∇𝑃𝑔 + 𝜀𝑔 𝜌̃𝑔 𝐷𝑎𝑣 ∇ ( ) (A.6)
𝜇𝑔 𝜌̃𝑔

𝜌̃𝑎 𝜌̃𝑔 − 𝜌̃𝑣 𝜌̃𝑣

∇( ) = ∇( ) = −∇ ( ) , Using (A. 4): (A.7)
𝜌̃𝑔 𝜌̃𝑔 𝜌̃𝑔

187
 𝐷𝑎𝑣 𝜀𝑔 𝜌̃𝑣 𝑃𝑔 𝜀𝑔 𝜌̃𝑎 𝐾𝐾𝑔 𝐷𝑎𝑣 𝜀𝑔 𝜌̃𝑣
−𝐽𝑎 = (− ) ∇𝑇 + ( + ) ∇𝑃𝑔
𝜌̃𝑔 𝑅𝑎 𝑇 2 𝜇𝑔 𝜌̃𝑔 𝑅𝑎 𝑇
𝑅𝑣 𝐷𝑎𝑣 𝜀𝑔 (A.8)
+ (−𝜌̃𝑎 − 𝜌̃𝑣 ) ∇𝜌̃v
𝑅𝑎 𝜌̃𝑔
−𝐽𝑎 = 𝐾𝐴𝑇 ∇𝑇 + 𝐾𝐴𝑃 ∇𝑃𝑔 + 𝐾𝐴𝑉 ∇𝜌̃𝑣

Free water flux:

𝑆𝐵 𝐾𝐾𝑙 𝑆𝐵
−𝐽𝑓𝑤 = (1 − ) 𝜀𝑓𝑤 𝜌𝑙 ∇𝑃𝑙 + ( ) 𝜀𝑓𝑤 𝜌𝑙 𝐷𝐵 ∇𝑆𝐵 (A.9)
𝑆 𝜇𝑙 𝑆

𝜕𝑃𝑐 𝜕𝑃𝑐
∇𝑃𝑙 = ∇(𝑃𝑔 − 𝑃𝑐 ) = ∇𝑃𝑔 − ∇𝑇 − ∇𝜌̃ (A.10)
𝜕𝑇 𝜕𝜌̃𝑣 𝑣

𝑃𝑣 𝑅𝑣 𝜌̃𝑣 𝑇
Pc = −𝑅𝑣 𝑇𝜌𝑙 ln ( ) = −𝑅𝑣 𝑇𝜌𝑙 ln ( ) (A.11)
𝑃𝑠𝑎𝑡 𝑃𝑠𝑎𝑡

𝑅𝑣 𝜌̃𝑣 𝑇
𝜕 (𝑇𝑙𝑛 ( ))
∂Pc 𝜕𝜌𝑙 𝑅𝑣 𝜌̃𝑣 𝑇 𝑃𝑠𝑎𝑡
= −𝑅𝑣 𝑇𝑙𝑛 ( ) + 𝜌𝑙 (A.12)
𝜕𝑇 𝜕𝑇 𝑃𝑠𝑎𝑡 𝜕𝑇
[ ]

𝑅𝑣 𝜌̃𝑣 𝑇 𝑅𝑣 𝜌̃𝑣 𝑇
𝜕 (𝑇𝑙𝑛 ( )) 𝜕 (𝑙𝑛 ( ))
𝑃𝑠𝑎𝑡 𝑅𝑣 𝜌̃𝑣 𝑇 𝑃𝑠𝑎𝑡 (A.13)
= 𝑙𝑛 ( )+𝑇
𝜕𝑇 𝑃𝑠𝑎𝑡 𝜕𝑇

𝑅𝑣 𝜌̃𝑣 𝑇 𝑅𝑣 𝜌̃𝑣 𝑇 𝜕𝑃𝑠𝑎𝑡

𝜕 (𝑙𝑛 ( ))
𝑃𝑠𝑎𝑡 𝜕 ( 𝑃𝑠𝑎𝑡 ) 𝑃𝑠𝑎𝑡 𝑇 𝑃𝑠𝑎𝑡 − 𝑇 𝜕𝑇
′
𝑃𝑠𝑎𝑡
= = 2
𝜕𝑇 𝑅𝑣 𝜌̃𝑣 𝑇 𝜕𝑇 𝑇 𝑃𝑠𝑎𝑡
(A.14)
𝜕𝑃𝑠𝑎𝑡
1 𝑃𝑠𝑎𝑡 − 𝑇 𝜕𝑇
=
𝑇 𝑃𝑠𝑎𝑡

∂𝑃𝑐 𝑇 𝜕𝑃𝑠𝑎𝑡 𝑇 𝜕𝜌𝑙 𝑅𝑣 𝜌̃𝑣 𝑇

= 𝑅𝑣 𝜌𝑙 [ − (1 + ) ln ( ) − 1] (A.15)
𝜕𝑇 𝑃𝑠𝑎𝑡 𝜕𝑇 𝜌𝑙 𝜕𝑇 𝑃𝑠𝑎𝑡

188
 ∂𝑃𝑐 𝑅𝑣 𝑇𝜌𝑙
=− (A.16)
𝜕𝜌̃𝑣 𝜌̃𝑣

𝑇 𝜕𝑃𝑠𝑎𝑡 𝑇 𝜕𝜌𝑙 𝑅𝑣 𝜌̃𝑣 𝑇

∇𝑃𝑙 = ∇𝑃𝑔 − 𝑅𝑣 𝜌𝑙 [ − (1 + ) ln ( ) − 1] ∇𝑇
𝑃𝑠𝑎𝑡 𝜕𝑇 𝜌𝑙 𝜕𝑇 𝑃𝑠𝑎𝑡
(A.17)
𝑅𝑣 𝑇𝜌𝑙
+ ∇𝜌̃𝑣
𝜌̃𝑣

𝜀𝑓𝑤
∇𝑆 = ∇ ( ) 𝑓𝑜𝑟 𝑆 ≤ 𝑆𝑠𝑠𝑝
∇SB = { 𝜙 (A.18)
∇𝑆𝑠𝑠𝑝 =0 𝑓𝑜𝑟 𝑆 > 𝑆𝑠𝑠𝑝

∵ ε𝑓𝑤 = 𝜀𝑓𝑤 (𝑃𝑣 (𝜌̃𝑣 , 𝑇), 𝑃𝑠𝑎𝑡 (𝑇), 𝜌𝑙 (𝑇), 𝑚(𝑇)), 𝜙 = 𝜙(𝑇) (A.19)

𝜕𝜀𝑓𝑤 𝜕𝜀𝑓𝑤 𝜕𝜙
∴ ∇𝜀𝑓𝑤 = ∇𝜌̃𝑣 + ∇𝑇, ∇𝜙 = 𝛻𝑇 (A.20)
∇𝜌̃𝑣 𝜕𝑇 𝜕𝑇

𝜕𝜀𝑓𝑤 𝜕𝜀𝑓𝑤 𝜕𝜙
( ∇𝜌̃𝑣 + ∇𝑇) 𝜙 − 𝜀𝑓𝑤 𝛻𝑇
𝜕𝜌̃𝑣 𝜕𝑇 𝜕𝑇
∇SB =
𝜙2 𝑓𝑜𝑟 𝑆 ≤ 𝑆𝑠𝑠𝑝 (A.21)
1 𝜕𝜀𝑓𝑤 𝜀𝑓𝑤 𝜕𝜙 1 𝜀𝑓𝑤
= ( − ) ∇𝑇 + ∇𝜌̃
𝜙 𝜕𝑇 𝜙 𝜕𝑇 𝜙 𝜕𝜌̃𝑣 𝑣

−𝐽𝑓𝑤 = 𝐾𝐿𝑇 ∇𝑇 + 𝐾𝐿𝑃 ∇Pg + 𝐾𝐿𝑉 ∇𝜌̃𝑣 (A.22)

𝑆𝐵 𝐾𝐾𝑙
where K LP = (1 − ) 𝜀𝑓𝑤 𝜌𝑙 (A.23)
𝑆 𝜇𝑙

For 𝑆 > 𝑆𝑠𝑠𝑝 (∇SB = 0):

𝑆𝐵 𝐾𝐾𝑙 𝑅𝑣 𝜌𝑙 𝑇 𝜕𝑃𝑠𝑎𝑡
𝐾𝐿𝑇 = 𝜀𝑓𝑤 𝜌𝑙 {− (1 − ) [
𝑆 𝜇𝑙 𝑃𝑠𝑎𝑡 𝜕𝑇
(A.24)
𝑇 𝜕𝜌𝑙 𝑅𝑣 𝜌̃𝑣 𝑇
− (1 + ) ln ( ) − 1]}
𝜌𝑙 𝜕𝑇 𝑃𝑠𝑎𝑡

189
 𝑆𝐵 𝐾𝐾𝑙 𝑅𝑣 𝑇𝜌𝑙
K LV = 𝜀𝑓𝑤 𝜌𝑙 (1 − ) (A.25)
𝑆 𝜇𝑙 𝜌̃𝑣

For 𝑆 ≤ 𝑆𝑠𝑠𝑝 [∇SB from (A. 21)]:

𝑆𝐵 𝐷𝐵 𝜕𝜀𝑓𝑤 𝜀𝑓𝑤 𝜕𝜙
𝐾𝐿𝑇 = 𝜀𝑓𝑤 𝜌𝑙 { ( − )
𝑆𝜙 𝜕𝑇 𝜙 𝜕𝑇
𝑆𝐵 𝐾𝐾𝑙 𝑅𝑣 𝜌𝑙 𝑇 𝜕𝑃𝑠𝑎𝑡
− (1 − ) [ (A.26)
𝑆 𝜇𝑙 𝑃𝑠𝑎𝑡 𝜕𝑇
𝑇 𝜕𝜌𝑙 𝑅𝑣 𝜌̃𝑣 𝑇
− (1 + ) ln ( ) − 1]}
𝜌𝑙 𝜕𝑇 𝑃𝑠𝑎𝑡

𝑆𝐵 𝐾𝐾𝑙 𝑅𝑣 𝑇𝜌𝑙 𝑆𝐵 𝐷𝐵 𝜕𝜀𝑓𝑤

K LV = 𝜀𝑓𝑤 𝜌𝑙 [(1 − ) + ] (A.27)
𝑆 𝜇𝑙 𝜌̃𝑣 𝑆𝜙 𝜕𝜌̃𝑣

Moisture flux (water vapour + free water):

−(𝐽𝑓𝑤 + 𝐽𝑣 ) = (𝐾𝑉𝑇 + 𝐾𝐿𝑇 )∇𝑇 + (𝐾𝑉𝑃 + 𝐾𝐿𝑃 )∇𝑃𝑔 + (𝐾𝑉𝑉 + 𝐾𝐿𝑉 )∇𝜌̃𝑣
(A.28)
= 𝐾𝑀𝑇 ∇𝑇 + 𝐾𝑀𝑃 ∇𝑃𝑔 + 𝐾𝑀𝑉 ∇𝜌̃𝑣

̃𝒂 )
𝛛(𝛆𝐠 𝝆
Time derivative:
𝝏𝒕

∂(εg 𝜌̃𝑎 ) 𝜕𝜌̃𝑎 𝜕𝜀𝑔

= 𝜀𝑔 + 𝜌̃𝑎 (A.29)
𝜕𝑡 𝜕𝑡 𝜕𝑡

𝑃𝑎 𝑃𝑔 − 𝑃𝑣 𝑃𝑔 − 𝜌̃𝑣 𝑅𝑣 𝑇 1 𝑃𝑔 𝑅𝑣
𝜌̃𝑎 = = = = − 𝜌̃ (A.30)
𝑅𝑎 𝑇 𝑅𝑎 𝑇 𝑅𝑎 𝑇 𝑅𝑎 𝑇 𝑅𝑎 𝑣

𝜕𝑃𝑔 𝜕𝑇
𝜕𝜌̃𝑎 1 𝜕𝑡 𝑇 − 𝑃𝑔 𝜕𝑡 𝑅𝑣 𝜕𝜌̃𝑣
= −
𝜕𝑡 𝑅𝑎 𝑇2 𝑅𝑎 𝜕𝑡 (A.31)
𝑃𝑔 𝜕𝑇 1 𝜕𝑃𝑔 𝑅𝑣 𝜕𝜌̃𝑣
= (− 2
) +( ) + (− )
𝑅𝑎 𝑇 𝜕𝑡 𝑅𝑎 𝑇 𝜕𝑡 𝑅𝑎 𝜕𝑡

190
 𝜀𝑔 = 𝜙 − 𝜀𝑓𝑤 (𝑃𝑣 (𝜌̃𝑣 , 𝑇), 𝑃𝑠𝑎𝑡 (𝑇), 𝜌𝑙 (𝑇), 𝑚(𝑇)) (A.32)

∂𝜀𝑔 𝜕(𝜙 − 𝜀𝑓𝑤 ) 𝜕𝜙 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣

= = −( + )
𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑇 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡
(A.33)
𝜕𝜙 𝜕𝜀𝑓𝑤 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣
=( − ) + (− )
𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡

∂(εg 𝜌̃𝑎 ) 𝜀𝑔 𝑃𝑔 𝜕𝜙 𝜕𝜀𝑓𝑤 𝜕𝑇 𝜀𝑔 𝜕𝑃𝑔

= [− 2
+ 𝜌̃𝑎 ( − )] +( )
𝜕𝑡 𝑅𝑎 𝑇 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝑅𝑎 𝑇 𝜕𝑡
𝜀𝑔 𝑅𝑣 𝜀𝑓𝑤 𝜕𝜌̃𝑣
+ (− − 𝜌̃𝑎 ) (A.34)
𝑅𝑎 𝜕𝜌̃𝑣 𝜕𝑡
𝜕𝑇 𝜕𝑃𝑔 𝜕𝜌̃𝑣
= 𝐶𝐴𝑇 + 𝐶𝐴𝑃 + 𝐶𝐴𝑉
𝜕𝑡 𝜕𝑡 𝜕𝑡

̃𝒗 )
𝛛(𝛆𝐠 𝝆
Time derivative:
𝝏𝒕

∂(εg 𝜌̃𝒗 ) 𝜕𝜌̃𝑣 𝜕𝜙 𝜕𝜀𝑓𝑤 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣

= 𝜀𝑔 + 𝜌̃𝑣 [( − ) + (− ) ]
𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡
(A.35)
𝜕𝜙 𝜕𝜀𝑓𝑤 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣
= [𝜌̃𝑣 ( − )] + (𝜀𝑔 − 𝜌̃𝑣 )
𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡

𝛛(𝛆𝐟𝐰 𝝆𝒍 )
Time derivative:
𝝏𝒕

∂(εfw 𝜌𝑙 ) 𝜕𝜀𝑓𝑤 𝜕𝜌𝑙

= 𝜌𝑙 + 𝜀𝑓𝑤
𝜕𝑡 𝜕𝑡 𝜕𝑡
𝜕𝜀𝑓𝑤 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣 𝜕𝜌𝑙 𝜕𝑇
= 𝜌𝑙 ( + ) + 𝜀𝑓𝑤 (A.36)
𝜕𝑇 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡 𝜕𝑇 𝜕𝑡
𝜕𝜀𝑓𝑤 𝜕𝜌𝑙 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣
= (𝜌𝑙 + 𝜀𝑓𝑤 ) + 𝜌𝑙
𝜕𝑇 𝜕𝑇 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡

𝛛(𝛆𝐃 𝝆𝒍 )
Time derivative:
𝝏𝒕

191
 ∂(εD 𝜌𝑙 ) 𝜕𝜀𝐷 𝜕𝜌𝑙 𝜕𝜀𝐷 𝜕𝑇 𝜕𝜌𝑙 𝜕𝑇
= 𝜌𝑙 + 𝜀𝐷 = 𝜌𝑙 + 𝜀𝐷
𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑇 𝜕𝑡 𝜕𝑇 𝜕𝑡
(A.37)
𝜕𝜀𝐷 𝜕𝜌𝑙 𝜕𝑇
= (𝜌𝑙 + 𝜀𝐷 )
𝜕𝑇 𝜕𝑇 𝜕𝑡

Left hand side of (4.8):

∂(εg 𝜌̃𝒗 ) ∂(εfw 𝜌𝑙 ) ∂(εD 𝜌𝑙 ) 𝜕𝑇 𝜕𝑃𝑔 𝜕𝜌̃𝑣

+ − = 𝐶𝑀𝑇 + 𝐶𝑀𝑃 + 𝐶𝑀𝑉
𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑡
𝜕𝜙 𝜕𝜀𝑓𝑤 𝜕𝜌𝑙
= [𝜌̃𝑣 + (𝜌𝑙 − 𝜌̃𝑣 ) + (𝜀𝑓𝑤 − 𝜀𝐷 ) (A.38)
𝜕𝑇 𝜕𝑇 𝜕𝑇
𝜕𝜀𝐷 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣
− 𝜌𝑙 ] + [𝜀𝑔 + (𝜌𝑙 − 𝜌̃𝑣 ) ]
𝜕𝑇 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡

Left hand side of (4.9):

𝜕𝑇 𝜕(𝜀𝑓𝑤 𝜌𝑙 ) 𝜕(𝜀𝐷 𝜌𝑙 )
(𝜌𝐶) − 𝜆𝐸 + (𝜆𝐷 +𝜆𝐸 )
𝜕𝑡 𝜕𝑡 𝜕𝑡
𝜕𝜀𝐷 𝜕𝜌𝑙
= [(𝜌𝐶) + (𝜆𝐷 + 𝜆𝐸 ) (𝜌𝑙 + 𝜀𝐷 )
𝜕𝑇 𝜕𝑇
(A.39)
𝜕𝜀𝑓𝑤 𝜕𝜌𝑙 𝜕𝑇 𝜕𝜀𝑓𝑤 𝜕𝜌̃𝑣
− 𝜆𝐸 (𝜌𝑙 + 𝜀𝑓𝑤 ) + (−𝜆𝐸 𝜌𝑙 )
𝜕𝑇 𝜕𝑇 𝜕𝑡 𝜕𝜌̃𝑣 𝜕𝑡
𝜕𝑇 𝜕𝑃𝑔 𝜕𝜌̃𝑣
= 𝐶𝑇𝑇 + 𝐶𝑇𝑃 + 𝐶𝑇𝑉
𝜕𝑡 𝜕𝑡 𝜕𝑡

192
Appendix B Finite Element Solution

Initialization of Solution

To start the prediction-correction process, two velocity vectors are needed in the
predictor formula (4.79). However, in the very beginning of the solution, these vectors
are obviously not available and thus a special initializing procedure should be employed
instead. The initialization approach is to use the dissipative backward Euler method for
the first few steps, and switch to the standard predictor-corrector method. The
advantage of this initializing approach is that the non-physical features of the numerical
model could be damped quickly by the backward Euler method. The approach is
implemented in the following steps:

1) The initial condition of state variables x 0 are taken as the predictor x1p for the
first time step (n = 0). This is equivalent to using the backward Euler method
alone. The corresponding corrector formula therefore becomes:

1 1
[ 𝐂̂(xpn+1 ) + 𝐊
̂ (xpn+1 )] x n+1 = 𝐂̂(xpn+1 )x n + 𝐑
̂ (xpn+1 ) (B.1)
𝛥𝑡 Δ𝑡

2) The second step (n = 1) utilizes a forward Euler predictor in the form of:

xpn+1 = x n + Δ𝑡[𝐂̂ −1 (x n )𝐑
̂ (x n ) − 𝐂̂ −1 (x n )𝐊
̂ (x n )x n ] (B.2)

or alternatively an extrapolation predictor in the form of:

3 1
xpn+1 = x n − x n−1 (B.3)
2 2

together with the backward Euler corrector (B.1). It should be noted that in the
cases of strong nonlinearities, the forward Euler predictor would be more
favourable but it would require the inversion of capacitance matrix 𝐂̂.

193
 3) Starting from the third step, the standard second-order predictor-corrector
scheme is implemented with the Adams-Bashforth predictor (4.79) and Crank-
Nicolson corrector (4.81). The automatic time selection also commences from
here if the time step control, rather than a constant time step, is adopted.

Time Step Control

The a priori selection of a reasonable time interval may be difficult in some cases. One
of the benefits of the predictor-corrector scheme outlined above is the provision of a
rational basis for the automatic time selection procedure. The following content
presents a dynamic time estimation formula based on the time truncation errors of the
explicit predictor and implicit corrector with a user-defined error tolerance. The
detailed derivation of this formula could be found in Gresho et al. (1978, 2008). The
estimation formula reads:

𝑚
𝑛+1 𝑛
𝜖𝑡
Δ𝑡 = Δ𝑡 (𝑏 ∙ 𝑛+1 ) (B.4)
𝑑

1 Δ𝑡 𝑛−1
where m = , b = 3 (1 + ), 𝜖 𝑡 is the user-defined error tolerance, with a typical
3 Δ𝑡 𝑛

default value as 0.001. And 𝑑𝑛+1 is the norm on the integration error, defined as the
difference between the predicted and corrected solution and given by:

1
𝑁 2
1 2 (B.5)
𝑑𝑛+1 = [∑(x𝑖n+1 − x𝑖𝑝
n+1
) ]
x𝑚𝑎𝑥 √𝑁 𝑖=1

where N is the total number of nodes in the problem domain, x𝑚𝑎𝑥 is a user-defined
n+1
constant scale for the problem, x𝑖𝑝 and x𝑖n+1 are the predicted and corrected solutions
at the nodes. This norm is obtained separately with respect to different state variables.
The time step is then computed using each of these norms and the resulting smallest
one is used as the new time interval, allowing the dominant equation to take control of
the integration algorithm.

194
Iteration Procedure

The incremental iterative solution of (4.81) is implemented via Picard’s method in this
study. Suppose all the capacitance and stiffness matrices 𝐂̂(x θ ) and 𝐊
̂ (x θ ), and the
̂ (x θ ) defined in (4.81) have been evaluated using x θ and known, where
force vectors 𝐑
the solution x θ is now computed with:

x θ = (1 − θ)x n + θxkn+1 (B.6)

where the subscript k is the iteration index representing the number of iterations. A new
variable 𝑑x is defined to represent the difference between the previous and current
estimate of solution x θ for the (n + 1)th time step:

n+1
𝑑x = xk+1 − xkn+1 (B.7)

1
Using the notation of (B.7) with CN scheme (θ = ) and rearranging the equation by
2

moving the term containing xkn+1 to the RHS, (4.81) can be rewritten as:

[𝐊]𝑑x = {𝐑} (B.8)

where

2
[𝐊] = 𝐂̂(x θ ) + 𝐊
̂ (x θ ) (B.9)
Δ𝑡 𝑛

2
{𝐑} = 𝐂̂(x θ )x n + 𝐂̂(x θ )ẋ n + 𝐑
̂ (x θ ) − [𝐊]xkn+1 (B.10)
Δ𝑡 𝑛

Matrices 𝐂̂, 𝐊
̂ and 𝐑
̂ are updated using the “new” xkn+1 in each iteration, so are the [𝐊]
and {𝐑}. The system of linear equations in (B.8) is then solved for the increment 𝑑x.
An absolute and relative tolerance is set respectively for the norm of increment 𝑑x and
residual force {𝐑}. When both norms are less than their corresponding tolerance, the
solution is considered to be converged and the procedure is then advanced to the next

195
time step. It may be noted that the Picard’s method adopted here has the advantage of
not computing the Jacobian matrix involved in a monolithic Newton-Raphson type
method. Though having a faster rate of convergence, the evaluation of the derivatives
with respect to the state variables associated with the Jacobian matrix substantially
increases the computational cost of a Newton-Raphson iteration. Moreover, the use of
the Adam-Bashforth predictor together with the second-order implicit corrector in the
Picard’s method can compensate for the convergence rate to some extent. Consequently,
the Picard’s method is chosen as the iteration algorithm in this study.

196
Appendix C Derivation of Thermo-Mechanical Equation

Writing (5.37) in incremental form of the Bishop stress with (5.35):

𝑑𝛔′ = 𝐄𝒔𝒆𝒄 (𝑑𝜺 − 𝑑𝜺 𝑓𝑡 − 𝑑𝜺𝑙𝑖𝑡𝑠 ) − (1 − 𝑉)𝛔

̂′ 𝑑𝐷 − (1 − 𝐷)𝛔 ̂′ 𝑑𝑉
𝛽𝑡𝑚 (C.1)
= 𝐄𝒔𝒆𝒄 : 𝑑𝜺 − 𝐄𝒔𝒆𝒄 (𝛼𝑓𝑡 𝒎 + 0 𝐌𝝈′ ) 𝑑𝑇 − (1 − 𝐷)𝛔 ̂′ 𝑑𝑉 − (1 − 𝑉)𝛔
̂′ 𝑑𝐷
𝑓𝑐

The increment of mechanical damage 𝐷(𝜅 𝑚𝑑 , 𝜅0𝑚𝑑 , 𝛾) is computed when 𝜅 𝑚𝑑 = 𝜀̅:

𝜕𝐷 𝜕𝜅 𝑚𝑑 𝜕𝐷 𝜕𝜅0𝑚𝑑 𝜕𝐷 𝜕𝛾
𝑑𝐷 = 𝑚𝑑 𝑑𝜀̅ + 𝑚𝑑 𝑑𝑇 + 𝑑𝑇
𝜕𝜅 𝜕𝜀̅ 𝜕𝜅0 𝜕𝑇 𝜕𝛾 𝜕𝑇
𝜕𝐷 𝜕𝐷 𝜕𝜅0𝑚𝑑 𝜕𝐷 𝜕𝛾 (C.2)
= 𝑚𝑑 𝑑𝜀̅ + ( 𝑚𝑑 + ) 𝑑𝑇
𝜕𝜅 𝜕𝜅0 𝜕𝑇 𝜕𝛾 𝜕𝑇
= 𝐻1𝑚𝑑 𝑑𝜀̅ + 𝐻2𝑚𝑑 𝑑𝑇

The increment of thermal-chemical damage 𝑉(𝜅 𝑡𝑑 ) is computed when 𝜅 𝑡𝑑 = 𝑇:

𝜕𝑉 𝜕𝜅 𝑡𝑑 𝜕𝑉
𝑑𝑉 = 𝑑𝑉(𝜅 𝑡𝑑 ) = 𝑡𝑑
𝑑𝑇 = 𝑡𝑑 𝑑𝑇 = 𝐻 𝑡𝑑 𝑑𝑇 (C.3)
𝜕𝜅 𝜕𝑇 𝜕𝜅

The increment of equivalent strain measure:

𝜕𝜀̅ 𝜕𝜺𝒆 𝜕𝜺𝒆 𝜕𝜺𝒇𝒕 𝜕𝜺𝒍𝒊𝒕𝒔

𝑑𝜀̅ = ( 𝑑𝜺 + 𝑑𝑇) = 𝑠 [𝑑𝜺 − ( + ) 𝑑𝑇]
𝜕𝜺𝒆 𝜕𝜺 𝜕𝑇 𝜕𝑇 𝜕𝑇
(C.4)
𝛽𝑡𝑚
= 𝒔 [𝑑𝜺 − (𝛼𝑓𝑡 𝒎 + 0 𝐌𝝈′ ) 𝑑𝑇]
𝑓𝑐

The increment of total stress:

𝑑𝛔 = 𝑑𝛔′ − 𝑏𝐈𝑑𝑃𝑠 = 𝑑𝛔′ − 𝑏𝑑𝑷𝒔 (C.5)

197
 𝑑 (𝑃𝑔 − 𝑃𝑔 ∞ ) 𝑓𝑜𝑟 𝑆 ≤ 𝑆𝑠𝑠𝑝
𝑑𝑃𝑠 = { 𝑆 − 𝑆𝑠𝑠𝑝 1−𝑆 (C.6)
𝑑 [( ) 𝑃𝑙 + ( ) 𝑃 − 𝑃𝑔∞ ] 𝑓𝑜𝑟 𝑆 > 𝑆𝑠𝑠𝑝
1 − 𝑆𝑠𝑠𝑝 1 − 𝑆𝑠𝑠𝑝 𝑔

𝑃𝑣
𝑃𝑙 = 𝑃𝑔 − 𝑃𝑐 , 𝑃𝑐 = −𝑅𝑣 𝑇𝜌𝑙 (𝑇) ln ( ) = 𝑃𝑐 (𝑇, 𝜌̃𝑣 ) (C.7)
𝑃𝑠𝑎𝑡

𝑑𝑃𝑔 𝑓𝑜𝑟 𝑆 ≤ 𝑆𝑠𝑠𝑝

𝑑𝑃𝑠 = { 𝑆 − 𝑆𝑠𝑠𝑝 𝜕𝑃𝑐 𝑆 − 𝑆𝑠𝑠𝑝 𝜕𝑃𝑐 (C.8)
( ) 𝑑𝑇 + 𝑑𝑃𝑔 + ( ) 𝑑𝜌̃ 𝑓𝑜𝑟 𝑆 > 𝑆𝑠𝑠𝑝
1 − 𝑆𝑠𝑠𝑝 𝜕𝑇 1 − 𝑆𝑠𝑠𝑝 𝜕𝜌̃𝑣 𝑣

Adopting Voigt’s notation:

𝑑𝑷𝒔 = 𝒎𝑻 𝑑𝑇 + 𝒎𝑷 𝑑𝑃𝑔 + 𝒎𝑽 𝑑𝜌̃𝑣 (C.9)

𝑆 − 𝑆𝑠𝑠𝑝 𝜕𝑃𝑐
𝒎𝑻 = ( ) 𝒎
𝒎𝑻 = 0 1 − 𝑆𝑠𝑠𝑝 𝜕𝑇
where S ≤ Sssp {𝒎𝑷 = 𝒎 , 𝑆 > 𝑆𝑠𝑠𝑝 𝒎𝑷 = 𝒎 (C.10)
𝒎𝑽 = 0 𝑆 − 𝑆𝑠𝑠𝑝 𝜕𝑃𝑐
𝒎𝑽 = ( ) 𝒎
{ 1 − 𝑆𝑠𝑠𝑝 𝜕𝜌̃𝑣

and 𝒎 = (1 1 1 0 0 0 )𝑇 (C.11)

The incremental form of total stress reads:

𝑑𝛔 = 𝑑𝛔′ − 𝑏𝑑𝑷𝒔
𝛽𝑡𝑚 ′
= 𝐄𝒔𝒆𝒄 : 𝑑𝜺 − 𝐄𝒔𝒆𝒄 (𝛼𝑓𝑡 𝒎 + 0 𝐌𝝈 ) 𝑑𝑇 − ̂′ 𝐻 𝑡𝑑 𝑑𝑇
(1 − 𝐷)𝛔 (C.12)
𝑓𝑐
̂′ (𝐻1𝑚𝑑 𝑑𝜀̅ + 𝐻2𝑚𝑑 𝑑𝑇) − 𝑏𝒎𝑻 𝑑𝑇 − 𝑏𝒎𝑷 𝑑𝑃𝑔 − 𝑏𝒎𝑽 𝑑𝜌̃𝑣
−(1 − 𝑉)𝛔

Rearranging (C.12) gives:

198
 𝑑𝛔 = [𝐄𝒔𝒆𝒄 − (1 − 𝑉)𝐻1𝑚𝑑 𝛔
̂′ 𝒔𝑇 ]𝑑𝜺 − {[𝐄𝒔𝒆𝒄 − (1 − 𝑉)𝐻1𝑚𝑑 𝛔
̂′ 𝒔𝑇 ]
𝛽𝑡𝑚
× (𝛼𝑓𝑡 𝒎 + 0 𝐌𝝈′ ) + [(1 − 𝐷)𝐻 𝑡𝑑 + (1 − 𝑉)𝐻2𝑚𝑑 ]𝛔 ̂′ (C.13)
𝑓𝑐
+𝑏𝒎𝑻 }𝑑𝑇 − 𝑏𝒎𝑷 𝑑𝑃𝑔 − 𝑏𝒎𝑽 𝑑𝜌̃𝑣

External (body) force:

̂ 𝑒𝑢 = − ∫ 𝑁 𝑇 𝜌𝑔𝑑Ω
𝑹
Ω
(C.14)
= − ∫ 𝑁 𝑇 [(1 − 𝜙)𝜌𝑠 + 𝜙𝑆𝜌𝑙 + 𝜙(1 − 𝑆)𝜌𝑔 ]𝑔𝑑Ω
Ω

199
Appendix D Pioneer Bilayer Cylinder Test

D1) Test Specimens and Test Setup

Two identical groups, one for compression tests and the other for fire tests, of totally
12 cylindrical specimens were designed. Three HSC and three bilayer specimens were
prepared within each group. Thermal couples were located at mid-height of all
specimens and placed in a perpendicular way as shown in Figure D.1. The detailed
dimensions of specimens can be found in Table D.2. The ratio between outer layer
thickness and the outer diameter was 0.1. As shown in Table D.1, the compressive
strength of C65 concrete is only 50.7 MPa, which was much lower than the design
strength. Since the concrete used in this test was mixed and cast manually, the low
compressive strength was probably due to the poor workmanship such as insufficient
mixing and inaccurate material weighing. Loading rate of 265 kN/min was adopted as
recommended in British Standard. The electric furnace (Figure D.2) could be
programmed to follow the ISO 834 standard temperature-time curve (Figure 8.6),
calculated by equation (8.1). The highest temperature achieved after 4-hours’ heating
is approximately 1150℃. Type K Mineral Insulated Thermocouple Sensor with probe
length and diameter of 2500mm and 1.6mm respectively, was adopted in this study.
The cable length is 5000mm and the maximum temperature that could be detected is
1300℃. A steel cage was placed outside each specimen to protect the furnace in case
that vigorous spalling happened, shown in Figure D.3. Specimens were located at the
two ends of the furnace. Thermal couples were connected to the data logger through
the openings on two sides of the furnace.

D2) Test Procedures

To obtain the compressive strength of all specimens at ambient temperature and

investigate the performance of the bond between HSC core and NSC outer layer of the
bilayer specimen, compression tests were conducted. Dental stone was first applied to
smooth the top surface of specimens. Since the specimens were only 300 mm high, a

200
300 mm steel cube had to be placed on the compression machine to raise the specimens
(Figure D.4). Compressive loading was then applied to the specimen until failure.

The top and bottom surface of each specimen was insulated with insulating pad made
of asbestos wrapped in insulating cloth. The heating time was mainly determined by
the core temperature of the specimen represented by the measured at the centre. The
heating criterion was that, when the core temperature reached a certain value, the
furnace would be shut down immediately. Since most of the explosive spalling occurs
in the temperature range of 200℃ to 400℃ (Kanéma et al., 2011, Fu et al., 2005, Fu
and Li, 2011, Cheng et al., 2004, Kanema et al., 2011), 400℃ was chosen as the
threshold value to guarantee the whole specimen would be out of the suspected range
of spalling, because the peripheral temperature must be higher than 400℃ when the
core temperature reaches 400 ℃ . However, considering that spalling only occurs
somewhat 50mm from the surface of the specimen at most, it is unnecessary to heat the
whole specimen over 400℃. The threshold value for the core temperature of 250 mm
and 300 mm specimens were therefore changed to 300℃ to shorten the fire exposure.
By this time, the temperature of the outermost 50mm layer would already be far beyond
400℃. The sounds during heating were recorded to indicate whether spalling occurred.

D3) Results and Discussions

The compression test results are presented in Table D.2. Specimens were named by
“specimen type + specimen outer diameter”. For instance, 2L200 stands for the bilayer
specimen with outer diameter of 200 mm. The predicted capacity was the theoretical
values of failure load calculated from the strengths of HSC and NSC. The experimental
capacity represents the failure loads obtained from the compression tests. The final
column is the ratio between the predicted and experimentally obtained values.
Comparing the Pexp /Ppre ratios of bilayer and HSC specimens, it could be found that
the bilayer specimens generally had a lower ratio than HSC specimens, which may
indicate a strength reduction for bilayer specimens due to the interface between HSC
core and NSC outer layer. The value of Pexp /Ppre ratio should normally fall in the range
of 0.7~1, however, the value of 2L250 specimen was larger than 1, which was assumed

201
to be caused by the varying strength and quality of different batches of concrete mix.
Another thing to note is that no NSC outer layer fell off throughout the loading process,
so it is reasonable to assume that the bond between HSC core and NSC outer layer
should be strong enough to withstand the loading and avoid the separation of two layers.
Specimens were too brittle to perform residual strength test after several hours’ heating.
The gas temperatures inside the furnace of all three sets of fire tests were compared
with the ISO834 standard curve (Figure D.5), showing good agreement with the target
temperature.

Figure D.6 to Figure D.11 present the comparisons between measured temperatures and
simulated results of each specimen. Experimental curves are named by “Specimen type
+ outer diameter + thermal couple position (distance to the centre of specimen)”. Due
to manufacture or heating reasons, some of the thermal couples were broken and data
was missing for some of the positions. The simulated curves are named by “thermal
couple position + sim”. A plateau was observed for all specimens at the temperature
between 100℃ and 150℃, and the nearer to the centre, the longer the plateau continued.
This phenomenon was very likely to be caused by the evaporation and diffusion of free
water in the concrete pores (Phan, 2008a, Lie and Celikkol, 1991). When the
temperature reached around 105°C, free water started to evaporate and migrate towards
the centre of the specimen due to the pore pressure gradient. The transfer of heat in the
concrete was retarded as a result of energy being absorbed in the process of evaporation
and migration, leading to a decrease in the rate of temperature rise or a nearly constant
temperature period in the early stage of fire tests. The plateau ended when the concrete
temperature rises to about 150°C, by which time most of the water vapor would have
escaped through the concrete surface and stop migrating. Regarding the HSC and
bilayer specimens with the same outer diameter, the temperatures at corresponding
positions had similar values and trends, which may indicate the interface between the
HSC core and NSC outer layer did not affect the heat transfer in concrete. The NSC
outer layer took effect in preventing the heat from attacking the HSC core, improving
the performance of HSC under fire to some extent.

202
No explosive spalling was observed for all specimens. The poor workmanship when
casting concrete, leading to the low strength and high permeability of HSC, could be
the main reason that prevented explosive spalling. Water vapor inside the concrete was
released through the pores under high temperature, pore pressure was not able to build
up and exceed the tensile strength of concrete. However, sounds of cracking and
popping were continuously heard during heating, which may suggest the occurrence of
minor spalling in the HSC specimens. NSC outer layer of bilayer specimens remained
intact after heating, except that the outer layer of 200 mm specimen fell off when
hanging out from furnace. The outer layer of the 200 mm bilayer specimen was only
20 mm, causing certain difficulties to the casting and manual vibration. The minimum
thickness of NSC outer layer should be no less than 20 mm, reasoning that once the
thickness is less than 20mm it would be difficult to cast and vibrate the outer layer.
Visible cracks developed on the surface of all specimens.

D4) Post-Cooling Behaviour

Post cooling spalling occurred when the 300 mm HSC specimen was taken out from
furnace and cooled down in the ambient temperature. As mentioned in Section 2.2.1,
thermal chemical spalling consists of sloughing-off spalling at extremely high
temperature and post cooling spalling after exposing to elevated temperature (Xing et
al., 2011, Annerel, 2009). The main cause of thermal-chemical spalling is the break-
down of aggregate cement bond, such as calcium silicate hydroxide and calcium
hydroxide (Schneider, 1988a). The threshold temperature of thermal-chemical spalling
is relatively high at around 750℃. The 300 mm specimens were heated for the longest
time at around 3.5 hours, and the core temperature would continue to rise for a period
of time even the furnace was shut down. Therefore, the highest core temperature and
gas temperature recorded was 713℃ and 1213℃ respectively, meaning most part of the
specimen had been heated to over 700℃. Thus, the occurrence of post cooling spalling
can be deemed reasonable.

203
 Table D.1 Concrete mix proportions
Compressive
Cement Water Sand Stone SP
Grade Strength
(kg/m3) (kg/m3) (kg/m3) (kg/m3) (kg/m3)
(MPa)
C65 535 162 676 930 0.47 50.7
C35 350 168 720 1070 / 32.9

Table D.2 Compression Test Results

NSC Predicted
Total D Core d Experimental Pexp /Ppre
Specimen layer t capacity
(mm) (mm) capacity (kN) Ratio
(mm) (kN)
HSC200 200 0 200 1592.30 1554 0.98

2L200 200 20 160 1390.91 1051 0.76

HSC250 250 0 250 2487.96 1934 0.78

2L250 250 25 200 2173.30 2233 1.03

HSC300 300 0 300 3582.67 3009 0.84

2L300 300 30 240 3129.55 2304 0.74

204
Figure D.1 Cylindrical specimens for experimental tests

205
 Figure D.2 Electric furnace

Figure D.3 Steel cage protecting the specimen

206
 Figure D.4 Specimen on a steel cube base

Figure D.5 Furnace temperatures versus standard curve

207
Figure D.6 Comparison of HSC200 specimen

Figure D.7 Comparison of 2L200 specimen

208
Figure D.8 Comparison of HSC250 specimen

Figure D.9 Comparison of 2L250 specimen

209
Figure D.10 Comparison of HSC300 specimen

Figure D.11 Comparison of 2L300 specimen

210
Reference

ACI 1989. ACI 216R-89: Guide for determining the fire endurance of concrete
elements. ACI Farmington Hills, MI, USA.

ALI, F., NADJAI, A., SILCOCK, G. & ABU-TAIR, A. 2004. Outcomes of a major
research on fire resistance of concrete columns. Fire safety journal, 39, 433-
445.

ANNEREL, E. E. M. E. A. U. B. 2009. Revealing the temperature history in concrete

after fire exposure by microscopic analysis. Cement and Concrete Research, 39.

ASLANI, F. & BASTAMI, M. 2011. Constitutive relationships for normal-and high-

strength concrete at elevated temperatures. Materials Journal, 108, 355-364.

ATKINSON, A. & NICKERSON, A. 1984. The diffusion of ions through water-

saturated cement. Journal of materials science, 19, 3068-3078.

BAHR, O., SCHAUMANN, P., BOLLEN, B. & BRACKE, J. 2013. Young's modulus
and Poisson's ratio of concrete at high temperatures: Experimental
investigations. Mater. Des., 45, 421-429.

BAROGHEL-BOUNY, V., MAINGUY, M., LASSABATERE, T. & COUSSY, O.

1999. Characterization and identification of equilibrium and transfer moisture
properties for ordinary and high-performance cementitious materials. Cement
concrete research, 29, 1225-1238.

BARY, B. 1996. Etude du couplage hydraulique-mécanique dans le béton endommagé.

Cachan, Ecole normale supérieure.

BAŽANT, P. Z. & THONGUTHAI, W. 1979. Pore pressure in heated concrete walls:

theoretical prediction. Magazine of Concrete Research, 31, 67-76.

211
BAŽANT, Z. & CUSATIS, G. Concrete creep at high temperature and its interaction
with fracture: recent progress. Creep, shrinkage and durability of concrete and
concrete structures, 2005.

BAŽANT, Z. P. & CHERN, J.-C. 1987. Stress-induced thermal and shrinkage strains
in concrete. Journal of engineering mechanics, 113, 1493-1511.

BAŽANT, Z. P. & JIRÁSEK, M. 2018. Creep and hygrothermal effects in concrete

structures, Springer.

BAZANT, Z. P. & THONGUTHAI, W. 1978. Pore pressure and drying of concrete at

high temperature. ASCE J Eng Mech Div, 104, 1059-1079.

BAŽANT, Z. P. & ZI, G. 2003. Decontamination of radionuclides from concrete by

microwave heating. I: theory. Journal of engineering mechanics, 129, 777-784.

BEAR, J. 2013. Dynamics of fluids in porous media, New York: Dover, Courier
Corporation.

BEHNOOD, A. & GHANDEHARI, M. 2009. Comparison of compressive and splitting

tensile strength of high-strength concrete with and without polypropylene fibers
heated to high temperatures. Fire Safety Journal, 44, 1015-1022.

BILODEAU, A., KODUR, V. & HOFF, G. 2004. Optimization of the type and amount
of polypropylene fibres for preventing the spalling of lightweight concrete
subjected to hydrocarbon fire. Cement and Concrete Composites, 26, 163-174.

BISHOP, A. W. 1959. The principle of effective stress. Teknisk ukeblad, 39, 859-863.

BOX, G. E. P. & MULLER, M. E. 1958. A Note on the Generation of Random Normal

Deviates. Ann. Math. Statist., 29, 610-611.

212
BSEN1992-1-2:2004 2004. Eurocode 2: Design of concrete structures-Part 1-2:
General rules-Structural fire design. European Standards, London.

BUCKINGHAM, E. 1907. Studies on the movement of soil moisture. US Dept. Agic.

Bur. Soils Bull., 38.

BUILDINGS DEPARTMENT, H. K. 2013. Code of Practice for Structural Use of

Concrete. Buildings Department, Hong Kong.

BURGH, J. M. D. 2016. Hygro-thermo-mechanical study of concrete elements subject

to elevated temperatures: assessment of spalling risk and moisture interactions.
University of New South Wales.

CALDERÓN, P. A., ADAM, J. M., IVORRA, S., PALLARÉS, F. J. & GIMÉNEZ, E.

2009. Design strength of axially loaded RC columns strengthened by steel
caging. Materials Design, 30, 4069-4080.

CAMPIONE, G. 2012. Load carrying capacity of RC compressed columns

strengthened with steel angles and strips. Engineering Structures, 40, 457-465.

CARREIRA, D. J. & CHU, K.-H. Stress-strain relationship for plain concrete in

compression. Journal Proceedings, 1985. 797-804.

CASTILLO, C. 1987. Effect of transient high temperature on high-strength concrete.

In: DURRANI, A. J. (ed.).

ÇENGEL, Y. A. 2003. Heat transfer : a practical approach, Boston, McGraw-Hill.

CHAN, Y. N., PENG, G. F. & ANSON, M. 1999. Residual strength and pore structure
of high-strength concrete and normal strength concrete after exposure to high
temperatures. Cement and Concrete Composites, 21, 23-27.

213
CHEN, B. & LIU, J. 2004. Residual strength of hybrid-fiber-reinforced high-strength
concrete after exposure to high temperatures. Cement Concrete Research, 34,
1065-1069.

CHENG, F.-P., KODUR, V. & WANG, T.-C. 2004. Stress-strain curves for high
strength concrete at elevated temperatures. Journal of Materials in Civil
Engineering, 16, 84-90.

CHEUNG, Y. K. 1996. Finite element implementation, Oxford, [England], Blackwell

Science.

CHOWDHURY, S. 2014. Effect of elevated temperature on mechanical properties of

high strength concrete.

CHUNG, J. H. 2003. Numerical simulation of hydro-thermo-mechanical behavior of

concrete structures exposed to elevated temperatures.

CHUNG, J. H., CONSOLAZIO, G. R., MCVAY, M. C. J. C. & STRUCTURES 2006.

Finite element stress analysis of a reinforced high-strength concrete column in
severe fires. 84, 1338-1352.

CHUNG, J. H., CONSOLAZIO, G. R. J. C. & RESEARCH, C. 2005. Numerical

modeling of transport phenomena in reinforced concrete exposed to elevated
temperatures. 35, 597-608.

COLEMAN, B. D. & NOLL, W. 1963. The thermodynamics of elastic materials with

heat conduction and viscosity. Archive for Rational Mechanics and Analysis,
13, 167-178.

CONNOLLY, R. J. 1995. The spalling of concrete in fires. Ph.D. thesis, The University
of Aston in Birmingham.

214
DAL PONT, S., MEFTAH, F. & SCHREFLER, B. 2011. Modeling concrete under
severe conditions as a multiphase material. Nuclear engineering design, 241,
562-572.

DAVIE, C., ZHANG, H. & GIBSON, A. 2012. Investigation of a continuum damage

model as an indicator for the prediction of spalling in fire exposed concrete.
Computers structures, 94, 54-69.

DAVIE, C. T., PEARCE, C. J. & BIĆANIĆ, N. 2006. Coupled Heat and Moisture
Transport in Concrete at Elevated Temperatures—Effects of Capillary Pressure
and Adsorbed Water. Numerical Heat Transfer, Part A: Applications, 49, 733-
763.

DAVIE, C. T., PEARCE, C. J. & BIĆANIĆ, N. 2010. A fully generalised, coupled,

multi-phase, hygro-thermo-mechanical model for concrete. Materials
structures, 43, 13-33.

DEPARTMENT, C. E. A. D. 2010. Construction Standard CS1:2010 Testing Concrete

Volume 2 of 2. The Government of the Hong Kong Special Administrative
Region.

DEPARTMENT, T. B. 2011. Code of Practice for Fire Safety in Buildings. The

Government of the Hong Kong Special Administrative Region.

DIEDERICHS, U., JUMPPANEN, U.-M. & PENTTALA, V. 1988. Material properties

of high strength concrete at elevated temperatures. 13.

DIEDERICHS, U., JUMPPANEN, U. & SCHNEIDER, U. High temperature properties

and spalling behaviour of high strength concrete, 4th Int. Workshop on High
Performance Concrete—Characteristics, Material Properties and Structural
Performance, Weimar, Germany, 1995.

215
DONG, H., CAO, W., BIAN, J. & ZHANG, J. 2014. The fire resistance performance
of recycled aggregate concrete columns with different concrete compressive
strengths. Materials, 7, 7843-7860.

DOUGILL, J. 1972. Modes of failure of concrete panels exposed to high temperatures.

Magazine of Concrete Research, 24, 71-76.

DWAIKAT, M. & KODUR, V. 2008. A numerical approach for modeling the fire
induced restraint effects in reinforced concrete beams. Fire Safety Journal, 43,
291-307.

DWAIKAT, M. & KODUR, V. 2010. Fire induced spalling in high strength concrete
beams. Fire technology, 46, 251.

DWAIKAT, M. B. & KODUR, V. K. R. 2009. Hydrothermal model for predicting fire-

induced spalling in concrete structural systems. Fire Safety Journal, 44, 425-
434.

ELMOHANDES, F. 2013. Advanced three-dimensional nonlinear analysis of

reinforced concrete structures subjected to fire and extreme loads.

ENV1992-1-2:1996 1996. Eurocode 2: Design of concrete structures—Part 1.2:

General rules—Structural fire design. European Committee for Standardization
(CEN), 63.

FEIST, C., ASCHABER, M. & HOFSTETTER, G. 2009. Numerical simulation of the

load-carrying behavior of RC tunnel structures exposed to fire. Finite Elements
in Analysis & Design, 45, 958-965.

FELICETTI, R., GAMBAROVA, P. G., ROSATI, G., CORSI, F. & GIANNUZZI, G.

Residual mechanical properties of high-strength concretes subjected to high-
temperature cycles. 4th International Symposium on Utilization of High-
Strength/High-Prformance Concrete, 1996.

216
FERAILLE-FRESNET, A., TAMAGNY, P., EHRLACHER, A. & SERCOMBE, J.
2003. Thermo-hydro-chemical modelling of a porous medium submitted to high
temperature: an application to an axisymmetrical structure. Mathematical
computer modelling, 37, 641-650.

FORSYTH, P. A. & SIMPSON, R. J. I. J. F. N. M. I. F. 1991. A two‐phase, two‐

component model for natural convection in a porous medium. 12, 655-682.

FU, H., ERKI, M. & SECKIN, M. 1991. Review of effects of loading rate on concrete
in compression. Journal of structural engineering, 117, 3645-3659.

FU, Y. & LI, L. 2011. Study on mechanism of thermal spalling in concrete exposed to
elevated temperatures. Materials and structures, 44, 361-376.

FU, Y., WONG, Y., POON, C. & TANG, C. 2005. Stress-strain behaviour of high-
strength concrete at elevated temperatures. Magazine of Concrete Research, 57,
535-544.

FURUMURA, F., ABE, T. & SHINOHARA, Y. Mechanical properties of high strength

concrete at high temperatures. Proceedings of the Fourth Weimar Workshop on
High Performance Concrete: Material Properties and Design, 1995. Germany:
Weimar.

GAWIN, D., ALONSO, C., ANDRADE, C., MAJORANA, C. & PESAVENTO, F.

2005a. Effect of damage on permeability and hygro-thermal behaviour of HPCs
at elevated temperatures: Part 1. Experimental results. Computers Concrete, 2,
189-202.

GAWIN, D., MAJORANA, C., PESAVENTO, F. & SCHRELFER, B. 2005b. Effect

of damage on permeability and hygro-thermal behaviour of HPCs at elevated
temperatures: Part 2. Numerical analysis. Computers Concrete, 2, 203-214.

217
GAWIN, D., MAJORANA, C. & SCHREFLER, B. 1999. Numerical analysis of
hygro ‐ thermal behaviour and damage of concrete at high temperature.

Mechanics of Cohesive‐frictional Materials, 4, 37-74.

GAWIN, D. & PESAVENTO, F. Prediction of the thermal spalling risk of concrete

structures exposed to high temperatures. Conference Proceedings of the 6th
International Conference “Analytical Models and New Concepts in Concrete
and Masonry Structures-AMCM, 2009.

GAWIN, D. & PESAVENTO, F. 2012. An overview of modeling cement based

materials at elevated temperatures with mechanics of multi-phase porous media.
Fire technology, 48, 753-793.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. 2002. Simulation of damage–

permeability coupling in hygro‐thermo‐mechanical analysis of concrete at high

temperature. Communications in numerical methods in engineering, 18, 113-

119.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. 2003. Modelling of hygro-thermal

behaviour of concrete at high temperature with thermo-chemical and
mechanical material degradation. Computer methods in applied mechanics and
engineering, 192, 1731-1771.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. A. 2006a. Hygro‐thermo‐chemo‐

mechanical modelling of concrete at early ages and beyond. Part I: hydration

and hygro‐thermal phenomena. International Journal for Numerical Methods

in Engineering, 67, 299-331.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. A. 2006b. Hygro‐thermo‐chemo‐

mechanical modelling of concrete at early ages and beyond. Part II: shrinkage

218
 and creep of concrete. International Journal for Numerical Methods in
Engineering, 67, 332-363.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. A. 2006c. Towards prediction of

the thermal spalling risk through a multi-phase porous media model of concrete.
Computer Methods in Applied Mechanics and Engineering, 195, 5707-5729.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. A. 2011a. What physical

phenomena can be neglected when modelling concrete at high temperature? A
comparative study. Part 1: Physical phenomena and mathematical model.
International journal of solids structures, 48, 1927-1944.

GAWIN, D., PESAVENTO, F. & SCHREFLER, B. A. 2011b. What physical

phenomena can be neglected when modelling concrete at high temperature? A
comparative study. Part 2: Comparison between models. International journal
of solids structures, 48, 1945-1961.

GENCEL, O. 2012. Effect of elevated temperatures on mechanical properties of high‐

strength concrete containing varying proportions of hematite. Fire and

Materials, 36, 217-230.

GRAY, W. G. & SCHREFLER, B. A. 2001. Thermodynamic approach to effective

stress in partially saturated porous media. J European Journal of Mechanics-
A/Solids, 20, 521-538.

GRESHO, P. M., GRIFFITHS, D. F. & SILVESTER, D. J. J. S. J. O. S. C. 2008.

Adaptive time-stepping for incompressible flow part I: Scalar advection-
diffusion. 30, 2018-2054.

GRESHO, P. M., LEE, R. L., SANI, R. L. & STULLICH, T. 1978. Time-dependent

FEM solution of the incompressible Navier--Stokes equations in two-and three-
dimensions. California Univ.

219
GUTTAG, J. 2016. Introduction to computation and programming using Python: With
application to understanding data, MIT Press.

HAMMER, T. A. 1995. High Strength Concrete Phase 3: Compressive strength and

E-modulus at elevated temperatures, SP6 Fire Resistance, Report 6.1, SINTEF
Structures and Concrete.

HANSEN, P. & JENSEN, J. 1995. High strength concrete phase-3, fire resistance and
spalling behaviour of LWA beams. SP-6-Fire Resistance, Report, 6.

HARMATHY, T. 1965. Effect of moisture on the fire endurance of building elements.

Moisture in materials in relation to fire tests. ASTM International.

HASSANIZADEH, M. & GRAY, W. G. 1979a. General conservation equations for

multi-phase systems: 1. Averaging procedure. Advances in water resources, 2,
131-144.

HASSANIZADEH, M. & GRAY, W. G. 1979b. General conservation equations for

multi-phase systems: 2. Mass, momenta, energy, and entropy equations.
Advances in water resources, 2, 191-203.

HASSANIZADEH, M. & GRAY, W. G. 1980. General conservation equations for

multi-phase systems: 3. Constitutive theory for porous media flow. Advances in
water resources, 3, 25-40.

HERTZ, K. 1984. Heat-induced explosion of dense concretes. Technical University of

Denmark, Institute of Building Design, No. 166.

HERTZ, K. D. 1992. Danish investigations on silica fume concretes at elevated

temperatures. ACI Materials Journal, 89, 345-347.

HERTZ, K. D. 2003. Limits of spalling of fire-exposed concrete. Fire Safety Journal,

38, 103-116.

220
HERTZ, K. D. 2005. Concrete strength for fire safety design. Magazine of Concrete
Research, 57, 445-453.

HIMMELWRIGHT, A. L. A. 1906. The San Francisco Earthquake and Fire: A Brief

History of the Disaster, Roebling Construction Company.

ICHIKAWA, Y. 2000. Prediction of pore pressures, heat and moisture transfer leading
to spalling of concrete during fire.

ICHIKAWA, Y. & ENGLAND, G. L. 2004. Prediction of moisture migration and pore

pressure build-up in concrete at high temperatures. Nuclear Engineering and
Design, 228, 245-259.

INSTITUTION, B. S. 1985. BS 8110 Part 2, Structural Use of Concrete. Code of

Practice for Design and Construction. BSI, London, UK.

JANSSON, R. 2013. Fire spalling of concrete: theoretical and experimental studies.

KTH Royal Institute of Technology.

JANSSON, R. & BOSTRÖM, L. Fire spalling–the moisture effect. 1st International

workshop on concrete fire spalling due to fire exposure, 2009.

JANSSON, R. & BOSTRÖM, L. Determination of fire spalling of concrete–relevance

of different test methods. 7th Int Conf on Structures in Fire, 2012.

JEONGWON, K., DONGWOO, R. & TAKAFUMI, N. 2011. The spalling mechanism

of high-strength concrete under fire. Magazine of Concrete Research, 63, 357-
370.

KALIFA, P., MENNETEAU, F.-D. & QUENARD, D. 2000. Spalling and pore
pressure in HPC at high temperatures. Cement and Concrete Research, 30,
1915-1927.

221
KANÉMA, M., PLIYA, P., NOUMOWÉ, A. & GALLIAS, J. 2011. Spalling, thermal,
and hydrous behavior of ordinary and high-strength concrete subjected to
elevated temperature. Journal of Materials in Civil Engineering, 23, 921-930.

KANEMA, M., PLIYA, P., NOUMOWE, A. & GALLIAS, J. L. 2011. Spalling,

thermal, and hydrous behavior of ordinary and high-strength concrete subjected
to elevated temperature. Journal of Materials in Civil Engineering, 23, 921.

KHALIQ, W. & KODUR, V. 2012. High Temperature Mechanical Properties of High-

Strength Fly Ash Concrete with and without Fibers. ACI Materials Journal, 109,
665-665.

KHALIQ, W. & KODUR, V. 2013. Behavior of high strength fly ash concrete columns
under fire conditions. Mater Struct, 46, 857-867.

KHAN, S. A. 1990. Pore pressure and moisture migration in concrete at high and non
uniform temperatures. University of London.

KHOURY, G., ANDERBERG, Y., BOTH, K., FELLINGER, J., HØJ, N. &
MAJORANA, C. 2007. Fire design of concrete structures—materials, structures
and modelling, state-of-the art report. Fib bulletin 38.

KHOURY, G., MAJORANA, C., PESAVENTO, F. & SCHREFLER, B. 2002.

Modelling of heated concrete. Magazine of concrete research, 54, 77-101.

KHOURY, G. A. 2000. Effect of fire on concrete and concrete structures. Progress in

Structural Engineering Materials, 2, 429-447.

KHOURY, G. A. 2008. Passive fire protection of concrete structures. Proceedings of

the Institution of Civil Engineers-Structures Buildings, 161, 135-145.

KHOURY, G. A. & ANDERBERG, Y. 2000. Concrete spalling review. Fire Safety

Design, 60, 5-12.

222
KLINGSCH, E., FRANGI, A. & FONTANA, M. 2013. Explosive spalling of concrete
in fire, Test report. IBK Tests report No. 352. ETH Zurich, Switzerland: Institute
of Structural Engineering (IBK).

KLINGSCH, E. W. 2014. Explosive spalling of concrete in fire. 356.

KODUR, V. 2000. Spalling in high strength concrete exposed to fire: concerns, causes,
critical parameters and cures. Advanced Technology in Structural Engineering.

KODUR, V. 2014. Properties of Concrete at Elevated Temperatures. ISRN Civil

Engineering.

KODUR, V. & DWAIKAT, M. 2008a. Effect of fire induced spalling on the response
of reinforced concrete beams. International journal of concrete structures
materials, 2, 71-81.

KODUR, V. & DWAIKAT, M. 2008b. High-Temperature Properties of Concrete for

Fire Resistance Modeling of Structures. ACI Materials Journal, 105, 517-527.

KODUR, V. & DWAIKAT, M. 2008c. A numerical model for predicting the fire
resistance of reinforced concrete beams. Cement and Concrete Composites, 30,
431-443.

KODUR, V., DWAIKAT, M. & RAUT, N. 2009. Macroscopic FE model for tracing
the fire response of reinforced concrete structures. Engineering Structures, 31,
2368-2379.

KODUR, V. & KHALIQ, W. 2011. Effect of Temperature on Thermal Properties of

Different Types of High-Strength Concrete. Journal of Materials in Civil
Engineering, 23, 793-801.

KODUR, V. & MCGRATH, R. 2003. Fire Endurance of High Strength Concrete

Columns. Fire Technology, 39, 73-87.

223
KODUR, V., MCGRATH, R., LEROUX, P. & LATOUR, J. J. N. R. C. C., INTERNAL
REPORT 2005. Experimental studies for evaluating the fire endurance of high-
strength concrete columns.

KODUR, V. & PHAN, L. 2007. Critical factors governing the fire performance of high
strength concrete systems. Fire Safety Journal, 42, 482-488.

KODUR, V., WANG, T. & CHENG, F. 2004. Predicting the fire resistance behaviour
of high strength concrete columns. Cement and Concrete Composites, 26, 141-
153.

KODUR, V. K. & MCGRATH, R. 2006. Effect of silica fume and lateral confinement
on fire endurance of high strength concrete columns. Canadian journal of civil
engineering, 33, 93-102.

KODUR, V. K. R. & SULTAN, M. A. 2003. Effect of Temperature on Thermal

Properties of High-Strength Concrete. Journal of Materials in Civil Engineering,
15, 101-107.

LEE, J.-H., SOHN, Y.-S. & LEE, S.-H. 2012. Fire resistance of hybrid fibre-reinforced,
ultra-high-strength concrete columns with compressive strength from 120 to
200 MPa. Magazine of concrete research, 64, 539-550.

LEWIS, R., MAJORANA, C. & SCHREFLER, B. 1986. A coupled finite element

model for the consolidation of nonisothermal elastoplastic porous media.
Transport in porous media, 1, 155-178.

LEWIS, R. W. & SCHREFLER, B. A. 1987. The finite element method in the

deformation and consolidation of porous media, Chichester, Wiley.

LEWIS, R. W. & SCHREFLER, B. A. 1998. The finite element method in the static
and dynamic deformation and consolidation of porous media, John Wiley.

224
LI, L.-Y. & PURKISS, J. 2005. Stress–strain constitutive equations of concrete material
at elevated temperatures. Fire Safety Journal, 40, 669-686.

LI, M., QIAN, C. & SUN, W. 2004. Mechanical properties of high-strength concrete
after fire. Cement and Concrete Research, 34, 1001-1005.

LI, W. & GUO, Z. H. 1993. Experimental investigation on strength and deformation of

concrete under high temperature (in Chinese). Journal ofBuilding Structures,
14, 8-16.

LI, Y., PIMIENTA, P., PINOTEAU, N. & TAN, K. H. 2019. Effect of aggregate size
and inclusion of polypropylene and steel fibers on explosive spalling and pore
pressure in ultra-high-performance concrete (UHPC) at elevated temperature.
Cement Concrete Composites, 99, 62-71.

LIE, T. 1992. Structural Fire Protection, New York, NY, American Society of Civil
Engineers (ASCE).

LIE, T. & KODUR, V. 1996. Thermal and mechanical properties of steel-fibre-

reinforced concrete at elevated temperatures. Canadian Journal of Civil
Engineering, 23, 511-517.

LIE, T. T. & CELIKKOL, B. 1991. Method to calculate the fire resistance of circular
reinforced concrete columns. ACI Materials Journal, 88, 84-91.

LIM, J. C. & OZBAKKALOGLU, T. 2014. Stress–strain model for normal-and light-

weight concretes under uniaxial and triaxial compression. Construction
Building Materials, 71, 492-509.

LIU, J.-C., TAN, K. H., YAO, Y. J. C. & MATERIALS, B. 2018a. A new perspective
on nature of fire-induced spalling in concrete. 184, 581-590.

225
LIU, J.-C., TAN, K. H. J. C. & COMPOSITES, C. 2018b. Fire resistance of ultra-high
performance strain hardening cementitious composite: residual mechanical
properties and spalling resistance. 89, 62-75.

LIU, J.-C., TAN, K. H. J. C. & COMPOSITES, C. 2018c. Mechanism of PVA fibers

in mitigating explosive spalling of engineered cementitious composite at
elevated temperature. 93, 235-245.

LIU, J.-C. & ZHANG, Y. 2019. A simplified model to predict thermo-hygral behaviour
and explosive spalling of concrete. Journal of Advanced Concrete Technology,
17, 419-433.

LOTTMAN, B. B. G. 2017. The spalling mechanism of fire exposed concrete.

LU, F. 2015. On the prediction of concrete spalling under fire. ETH Zurich.

LU, F. & FONTANA, M. A thermo-hydro model for predicting spalling and evaluating
the protective methods. 4th International Workshop on Concrete Spalling due
to Fire Exposure, 2015. MFPA Leipzig, 2015, 385-395.

LU, Z.-D. 1989. A research on fire response of reinforced concrete beams (in Chinese).
PhD, Tongji University.

LUCKNER, L., VAN GENUCHTEN, M. T. & NIELSEN, D. 1989. A consistent set of

parametric models for the two‐phase flow of immiscible fluids in the subsurface.

Water Resources Research, 25, 2187-2193.

LUO, X., SUN, W. & CHAN, S. Y. N. 2000. Effect of heating and cooling regimes on
residual strength and microstructure of normal strength and high-performance
concrete. Cement Concrete Research, 30, 379-383.

226
MAJORANA, C. E., SALOMONI, V. A., MAZZUCCO, G. & KHOURY, G. 2010.
An approach for modelling concrete spalling in finite strains. Mathematics
Computers in Simulation, 80, 1694-1712.

MEYER-OTTENS, C. J. T. U. O. B. 1972. The question of spalling of concrete

structural elements of standard concrete under fire loading.

MINDEGUIA, J.-C., HAGER, I., PIMIENTA, P., CARRÉ, H. & LA BORDERIE, C.

2013. Parametrical study of transient thermal strain of ordinary and high
performance concrete. Cement Concrete Research, 48, 40-52.

MINDEGUIA, J.-C., PIMIENTA, P., CARRÉ, H. & LA BORDERIE, C. Experimental

study on the contribution of pore vapour pressure to the thermal instability risk
of concrete. 1st International Workshop on Concrete Spalling due to Fire
Exposure, 2009.

MINDEGUIA, J.-C., PIMIENTA, P., NOUMOWÉ, A. & KANEMA, M. 2010.

Temperature, pore pressure and mass variation of concrete subjected to high
temperature — Experimental and numerical discussion on spalling risk. Cement
and Concrete Research, 40, 477-487.

NEVILLE, A. M. 2011. Properties of concrete, Harlow, England and Hong Kong,

Pearson.

NIELSEN, C. V., PEARCE, C. J. & BICANIC, N. 2004. Improved phenomenological

modelling of transient thermal strains for concrete at high temperatures.
Computers and Concrete, 1, 189-209.

NOUMOWE, A. N., SIDDIQUE, R. & DEBICKI, G. 2009. Permeability of high-

performance concrete subjected to elevated temperature (600 °C). Construction
and Building Materials, 23, 1855-1861.

227
ORTIZ, M. 1985. A constitutive theory for the inelastic behavior of concrete.
Mechanics of materials, 4, 67-93.

OZAWA, M., UCHIDA, S., KAMADA, T. & MORIMOTO, H. 2012. Study of

mechanisms of explosive spalling in high-strength concrete at high
temperatures using acoustic emission. Construction and Building Materials, 37,
621-628.

ÖZıŞıK, M. N. 1994. Finite difference methods in heat transfer, Boca Raton, CRC
Press.

PEARCE, C. J., NIELSEN, C. V., BIĆANIĆ, N. J. I. J. F. N. & GEOMECHANICS, A.

M. I. 2004. Gradient enhanced thermo‐mechanical damage model for concrete

at high temperatures including transient thermal creep. 28, 715-735.

PENG, G.-F., YANG, W.-W., ZHAO, J., LIU, Y.-F., BIAN, S.-H. & ZHAO, L.-H.
2006. Explosive spalling and residual mechanical properties of fiber-toughened
high-performance concrete subjected to high temperatures. Cement and
Concrete Research, 36, 723-727.

PESAVENTO, F., SCHREFLER, B., GAWIN, D. & PRINCIPE, J. Prediction of the

thermal spalling risk of concrete structures during fire by means of a finite
element model. 6th International Conference on Engineering Computational
Technology, ECT 2008, 2008.

PHAN, L. 2008a. Pore pressure and explosive spalling in concrete. Materials and
Structures, 41, 1623-1632.

PHAN, L. T. 1996. Fire performance of high-strength concrete : a report of the state-

of-the art. U.S. Dept. of Commerce, Technology Administration, National
Institute of Standards & Technology.

228
PHAN, L. T. 2002. High-strength concrete at high temperature-an overview.
Proceedings of 6th International Symposiumon Utilization of High
Strength/High Performance Concrete, Leipzig, Germany, 501-518.

PHAN, L. T. 2008b. Pore pressure and explosive spalling in concrete. Materials and
structures, 41, 1623-1632.

PHAN, M., MEFTAH, F., RIGOBERT, S., AUTUORI, P., LENGLET, C. & DAL
PONT, S. 2011. A finite element modeling of thermo-hydro-mechanical
behavior and numerical simulations of progressing spalling front. Procedia
Engineering, 10, 3128-3133.

PURKISS, J. A. & LI, L.-Y. 2013. Fire safety engineering design of structures, CRC
press.

REDDY, J. N. 2010. The finite element method in heat transfer and fluid dynamics,
Boca Raton, FL, CRC Press.

SAITO, H. 1966. Explosive spalling of prestressed concrete in fire. Bulletin of Japan

Association for Fire Science Engineering, 15, 23-30.

SANJAYAN, G. & STOCKS, L. J. 1993. Spalling of high-strength silica fume concrete

in fire. ACI Materials Journal, 90, 170-173.

SCHNEIDER, U. 1988a. Concrete at high temperatures—a general review. Fire safety

journal, 13, 55-68.

SCHNEIDER, U. 1988b. Concrete at high temperatures — A general review. Fire

Safety Journal, 13, 55-68.

SCHNEIDER, U. & HORVATH, J. 2003. Behaviour of ordinary concrete at high

temperatures, Techn. Univ. Wien, Inst. Baustofflehre, Bauphysik und
Brandschutz.

229
SERTMEHMETOGLU, Y. 1977. On a mechanism of spalling of concrete under fire
conditions. University of London.

SHAH, A. H., SHARMA, U. J. C. & MATERIALS, B. 2017. Fire resistance and

spalling performance of confined concrete columns. 156, 161-174.

SHEN, D., SHI, X., ZHANG, H., DUAN, X. & JIANG, G. 2016. Experimental study
of early-age bond behavior between high strength concrete and steel bars using
a pull-out test. Construction

Building materials, 113, 653-663.

SHIN, K.-Y., KIM, S.-B., KIM, J.-H., CHUNG, M. & JUNG, P.-S. 2002. Thermo-
physical properties and transient heat transfer of concrete at elevated
temperatures. Nuclear Engineering and Design, 212, 233-241.

SHIRLEY, S. T., BURG, R. G. & FIORATO, A. E. 1988. Fire endurance of high-

strength concrete slabs. Materials Journal, 85, 102-108.

SHORTER, G. & HARMATHY, T. J. P. O. T. I. O. C. E. 1961. Discussion on the fire

resistance of prestressed concrete beams. 20, 313-315.

SU, R. & WANG, L. 2012. Axial strengthening of preloaded rectangular concrete

columns by precambered steel plates. Engineering structures, 38, 42-52.

SULLIVAN, P. & SHARSHAR, R. 1992. The performance of concrete at elevated

temperatures (as measured by the reduction in compressive strength). Fire
Technol, 28, 240-250.

TENCHEV, R. & PURNELL, P. 2005. An application of a damage constitutive model

to concrete at high temperature and prediction of spalling. International Journal
of Solids Structures, 42, 6550-6565.

230
TENCHEV, R. T., LI, L. & PURKISS, J. 2001. Finite element analysis of coupled heat
and moisture transfer in concrete subjected to fire. Numerical Heat ransfer: Part
A: Applications, 39, 685-710.

TERRO, M. 1998. Numerical modeling of the behavior of concrete structures in fire.

ACI Struct. J., 95, 183-193.

TOUMI, B. & RESHEIDAT, M. 2010. Influence of high temperatures on surface

cracking of concrete studied by image scanning technique. Jordan Journal of
Civil Engineering, 4, 155-163.

VAN DER MERWE, J. E. 2019. Constitutive Models Towards the Assessment of

Concrete Spalling in Fire. ETH Zurich.

WAGNER, W. & PRUß, A. 2002. The IAPWS formulation 1995 for the
thermodynamic properties of ordinary water substance for general and scientific
use. Journal of physical chemical reference data, 31, 387-535.

WANG, L. & SU, R. K.-L. 2014. Repair of fire-exposed preloaded rectangular concrete
columns by postcompressed steel plates. Journal of Structural Engineering, 140,
04013083.

WEE, T., CHIN, M. & MANSUR, M. 1996. Stress-strain relationship of high-strength

concrete in compression. Journal of Materials in Civil Engineering, 8, 70-76.

WITEK, A., GAWIN, D., PESAVENTO, F. & SCHREFLER, B. 2007. Finite element
analysis of various methods for protection of concrete structures against spalling
during fire. Comput Mech, 39, 271-292.

WOOLSON, I. H. 1905. Investigation of the effect of heat upon the crushing strength
and elastic properties of concrete.

231
WU, B., SU, X.-P., LI, H. & YUAN, J. 2002. Effect of high temperature on residual
mechanical properties of confined and unconfined high-strength concrete.
Materials Journal, 99, 399-407.

WU, Z., LO, S. H., TAN, K. H. & SU, K. L. 2019. High Strength Concrete Tests under
Elevated Temperature. Athens Journal of Technology and Engineering, 6, 141-
162.

XING, Z., BEAUCOUR, A.-L., HEBERT, R., NOUMOWE, A. & LEDESERT, B.

2011. Influence of the nature of aggregates on the behaviour of concrete
subjected to elevated temperature. Cement and concrete research, 41, 392-402.

YANG, K.-H., MUN, J.-H., CHO, M.-S. & KANG, T. H.-K. 2014. Stress-Strain Model
for Various Unconfined Concretes in Compression. ACI Structural Journal, 111.

ZEIML, M. 2008. Concrete subjected to fire loading-From experimental investigation

of spalling and mass-transport properties to structural safety assessment of
tunnel linings under fire.

ZHANG, Y., ZEIML, M., MAIER, M., YUAN, Y. & LACKNER, R. 2017. Fast
assessing spalling risk of tunnel linings under RABT fire: From a coupled
thermo-hydro-chemo-mechanical model towards an estimation method.
Engineering structures, 142, 1-19.

ZHANG, Y., ZEIML, M., PICHLER, C. & LACKNER, R. 2014. Model-based risk
assessment of concrete spalling in tunnel linings under fire loading. Engineering
structures, 77, 207-215.

ZHAO, J. 2012. Fire-induced spalling modeling of high-performance concrete. Delft

University of Technology.

ZHUKOV, V. 1976. Reasons of explosive spalling of concrete by fire. Beton I

zhelezobeton (Concrete and Reinforcement Concrete), 3.

232
ZHUKOV, V. 1994. Forecast of a brittle failure of concrete by fire. Scientific Research
Institute for Concrete and Reinforced Concrete, Moscow, 235.

ZHUKOV, V. V. 1975. Explosive failure of concrete during a fire. Joint Fire Research
Organisation, Borehamwood, Translation No. DT 2124.

233