Are you sure?
This action might not be possible to undo. Are you sure you want to continue?

Based Materials under EarlyAge Drying and Shrinkage Conditions
by
Mehdi Bakhshi
A Dissertation Presented in Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy
Approved July 2011 by the
Graduate Supervisory Committee:
Barzin Mobasher, Chair
Subramaniam D. Rajan
Claudia E. Zapata
ARIZONA STATE UNIVERSITY
August 2011
i
ABSTRACT
Earlyage cracks in fresh concrete occur mainly due to high rate of surface
evaporation and restraint offered by the contracting solid phase. Available test
methods that simulate severe drying conditions, however, were not originally
designed to focus on evaporation and transport characteristics of the liquidgas
phases in a hydrating cementitious microstructure. Therefore, these tests lack
accurate measurement of the drying rate and data interpretation based on the
principles of transport properties is limited.
A vacuumbased test method capable of simulating earlyage cracks in 2
D cement paste is developed which continuously monitors the weight loss and
changes to the surface characteristics. 2D crack evolution is documented using
timelapse photography. Effects of sample size, w/c ratio, initial curing and fiber
content are studied. In the subsequent analysis, the cement paste phase is
considered as a porous medium and moisture transport is described based on
surface mass transfer and internal moisture transport characteristics. Results
indicate that drying occurs in two stages: constant drying rate period (stage I),
followed by a falling drying rate period (stage II). Vapor diffusion in stage I and
unsaturated flow within porous medium in stage II determine the overall rate of
evaporation. The mass loss results are analyzed using diffusionbased models.
Results show that moisture diffusivity in stage I is higher than its value in stage II
by more than one order of magnitude.
The drying model is used in conjunction with a shrinkage model to predict
the development of capillary pressures. Similar approach is implemented in
ii
drying restrained ring specimens to predict 1D crack width development. An
analytical approach relates diffusion, shrinkage, creep, tensile and fracture
properties to interpret the experimental data.
Evaporation potential is introduced based on the boundary layer concept,
mass transfer, and a driving force consisting of the concentration gradient. Effect
of wind velocity is reflected on Reynolds number which affects the boundary
layer on sample surface. This parameter along with Schmidt and Sherwood
numbers are used for prediction of mass transfer coefficient. Concentration
gradient is shown to be a strong function of temperature and relative humidity and
used to predict the evaporation potential. Results of modeling efforts are
compared with a variety of test results reported in the literature.
Diffusivity data and results of 1D and 2D image analyses indicate
significant effects of fibers on controlling earlyage cracks. Presented models are
capable of predicting evaporation rates and moisture flow through hydrating
cementbased materials during earlyage drying and shrinkage conditions.
iii
To my family for their support, patience and understanding during the past four
years
iv
ACKNOWLEDGMENTS
I would like to start by expressing my gratitude to my advisor Professor
Barzin Mobasher for giving me the opportunity to work with him and study at
Arizona State University. His guidance and support are the reasons I was able to
accomplish this work. Thank you to Professor Subramaniam D. Rajan and Dr.
Claudia E. Zapata for serving on my defense committee. Thank you to Dr. M.
Shekarchi and late Professor Ghalibafian for starting me on this trek of concrete at
University of Tehran.
I would like to gratefully acknowledge financial support for this project
received from the Arizona Department of Transportation under project number
SR633. This project was conducted in the Structural Mechanics and Materials
Testing Laboratories at Arizona State University and as such I would like to
acknowledge the support that has made these experiments possible. I would also
like to acknowledge the assistance of Mr. Peter Goguen, the laboratories manager,
Mr. Danny Clevenger and Mr. Jeff Long, for assistance in performance and
preparation of experimental testing.
In addition, I want to acknowledge the assistance of my colleagues at
Arizona State University including, Dr. Amir Bonakdar, Dr. Deju Zhu, Dr. Chote
Soranakom, Dr. Flavio Silva, Geoffrey Minor, Christopher Barsby, Sean Krauss,
and Nathan Rodriguez.
Thank you to all my friends who helped keep me sane during the process
including Jonathan Fein, Erblina Vokshi, Vikram Dey, Aditya Vaidya, Kapil
Krishnan, Pedram Shafieian, Shahrzad Badvipour and Kaveh Behbahani.
v
I would finally like to express my deepest appreciation and gratitude to my
family. I would like to thank my parents for their love and support through my
path of selfdiscovery. And finally I would like to thank my brothers for being a
constant reminder to worry about the important things in life.
vi
TABLE OF CONTENTS
Page
LIST OF TABLES…………………………………………………………...…....x
LIST OF FIGURES…………………………………………………........……....xi
NOMNECLATURE ………………………………………………........…….....xx
CHAPTER
1. INTRODUCTION ...............................................................................................1
1.1. Motivation .....................................................................................................1
1.2. Earlyage Shrinkage Cracking Mechanisms .................................................5
1.3. Background in Shrinkage Testing and Modeling .........................................6
1.3.1. Free Shrinkage Testing of Cementbased Materials ..............................6
1.3.2. Restrained Shrinkage Testing of Cementbased Materials ....................8
1.3.3. Earlyage (Plastic) Shrinkage Testing of Cementbased Materials .....10
1.3.4. Modeling the Drying of Cementbased Materials................................11
1.4. Objectives of the Dissertation .....................................................................13
1.5. Overview of Thesis .....................................................................................14
2. RESTRAINED SHRINKAGE CRACKING .....................................................17
2.1. Introduction .................................................................................................17
2.2. Review of Drying Shrinkage Testing Methods ..........................................17
2.3. Restrained Shrinkage Cracking ...................................................................20
2.4. Restrained Drying Shrinkage Test Methodology .......................................21
2.4.1. Mixture Properties,Casting and Specimen Preparation .......................22
2.4.2. Results of Strain Gauges Attached to Steel Ring .................................25
vii
CHAPTER Page
2.4.3. Capturing Shrinkage Cracks on the Ring Samples ..............................29
2.4.4. Image Analysis of Shrinkage Cracks ...................................................32
2.5. Theoretical Modeling of Restrained Shrinkage ..........................................35
2.5.1. Moisture Diffusion and Free Shrinkage ...............................................36
2.5.2. Creep of Concrete ................................................................................38
2.5.3. Age Dependent Concrete Strength .......................................................39
2.5.4. Restraining Effect ................................................................................43
2.5.5. Stress Strain Development ...................................................................44
2.5.6. Algorithm for Strain History in Steel Ring and Crack Width at
Concrete Surface ............................................................................................47
2.6. Comparing Experimental Data and Simulations .........................................48
2.7. Conclusion ..................................................................................................50
3. EXPERIMENTAL OBSERVATIONS OF EARLYAGE DRYING ...............51
3.1. Introduction .................................................................................................51
3.2. Testing methodology ..................................................................................54
3.3. Analysis of typical evaporation data under low pressure test condition .....56
3.3.1. Cumulative moisture loss and evaporation rate versus time ................56
3.3.2. Calculation of Moisture Diffusivity .....................................................62
3.4. Experimental program ................................................................................68
3.4.1. Scope of Test Program .........................................................................68
3.4.2. Materials, Mixing, Placing and Curing procedures .............................70
3.5. Parameter Estimation ..................................................................................71
viii
CHAPTER Page
3.5.1. Effect of sample size ............................................................................72
3.5.2. Effect of w/c ratio ................................................................................75
3.5.3. Effect of duration of initial curing .......................................................77
3.5.4. Effect of fiber content and cracking .....................................................79
3.6. Conclusion ..................................................................................................84
4. THEORY AND MODELING OF EARLYAGE DRYING .............................87
4.1. Introduction .................................................................................................87
4.2. Physics of water evaporation ......................................................................89
4.3. Theory of evaporation from cementitious materials ...................................93
4.3.1. Stage I Drying ......................................................................................98
4.3.2. Stage II Drying ...................................................................................100
4.4. TwoStage Modeling of the Drying of Cementitious Materials ...............101
4.4.1. Governing Equations and Geometry of the Problem .........................101
4.4.2. Simulation and Parametric Studies ....................................................106
4.4.3. BackCalculation Procedures for Modeling of Stage II Drying ........113
4.5. Comparison with Experimental Data and Finite Element (FE) Analysis .115
4.6. Conclusion ................................................................................................122
5. MODELING RATE OF EVAPORATION POTENTIAL ..............................126
5.1. Introduction ...............................................................................................126
5.2. Available Method to Estimate Initial Rate of Evaporation .......................129
5.3. Evaporation as a Mass Transfer Problem .................................................131
5.3.1. Mass Transfer at Low Rates ..............................................................133
ix
CHAPTER Page
5.3.2. Mass Transfer at High Rates ..............................................................142
5.3.3 Algorithm for determination of evaporation rate ................................147
5.4. Parametric Studies ....................................................................................148
5.5. Model Comparison with Experimental Results of Concrete Evaporation 150
5.6. Conclusion ................................................................................................157
APPENDIX
A Fee Shrinkage Test Results ..........................................................................158
B Vacuum Evaporation Test Method and Data Transfer Description .............164
C Finite Element Analysis of Evaporation Test Data ......................................175
D Drying Tests on Clays: An Application of the Developed Test Method .....204
E Image Analysis on Earlyage Cracks: Quantification and Orientation ........213
F Modeling Earlyage Shrinkage of Concrete Due to Evaporation………….227
REFERENCES ....................................................................................................248
x
LIST OF TABLES
Table Page
2.4.1.1. Mixture proportions of the Control and GRC samples (kg/m
3
) ................23
2.4.4.1. Mean Crack width and standard deviation of samples (mm) ...................34
2.5.3.1. Backcalculated parameters of Control and GRC3 samples .....................42
3.4.1.1. Scope of the test program .........................................................................70
3.4.2.1. Mix proportions of the test series ..............................................................71
3.5.1. Initial evaporation rates, transition time, evaporation rate at 24 h,
cumulative moisture loss, and diffusivities of tested samples .......................72
5.5.1. Results of the evaporation rate in comparison with the experiments ........154
5.5.2. Parameters used for calculation of evaporation rates ................................155
A.1. Mixture proportions of the ARGs and Control samples (lb/ft
3
) ...................159
C.1. Analogy and corresponding terms between two BVPs ................................187
C.2. Results of FE analysis with ABAQUS using Q4 elements comparing to the
experimental results .....................................................................................191
C.3. Results of FE analysis with ABAQUS using Q8 elements comparing to the
experimental results .....................................................................................192
D.1. Mix proportions of the test series (kg/m
3
) ...................................................205
D.2. Initial evaporation rates, transition time, evaporation rate at 24 h, cumulative
moisture loss, and diffusivities of tested samples ........................................209
E.1. Results of crack analysis on a plain and FRC samples ................................217
E.2. Results of crack analysis and quantifications ...............................................225
F.1. Properties of PVA fiber used in this study………………………………...241
xi
LIST OF FIGURES
Figure Page
1.1.1. Greenhouse gases trap some of the heat causing global warming ................4
1.1.2. Sustainable development principle ...............................................................4
1.2.1. Equilibrium between liquid water and humid air at the interface of
liquid/gas phase inside a pore and different pressures applied on solid
skeleton ............................................................................................................6
1.3.1.1. Molds, samples and digital comparator dial used in free shrinkage tests ...8
1.3.1.2. The configuration and geometry of restrained shrinkage ring specimen ..10
2.4.1.1. Experimental setup of shrinkage test and recording strain gauge results .24
2.4.2.1. Typical result of an attached strain gauge for a plain concrete sample ....26
2.4.2.2. (a) Results of two strain gauges mounted on a plain concrete sample, (b)
Location of the crack and strain gauges .........................................................27
2.4.2.3. Smoothing typical results of a strain gauge attached to steel ring ............27
2.4.2.4. Effect of glass fibers on results of strain gauges attached to the rings .....28
2.4.3.1. Imaging with the digital camera for crack investigations .........................30
2.4.3.2. Transverse cracks due to restrained drying shrinkage; a) control sample,
b) GRC3 sample after 14 days of drying in the shrinkage chamber ..............30
2.4.3.3. Reconstructed shrinkage crack images of control and GRC3 samples .....31
2.4.4.1. Image analysis of a single image taken from a Control sample ...............33
2.4.4.2. Mean and SD of shrinkage crack widths at different times of drying ......35
2.5.1.1. Fick's law of diffusion is used to simulate humidity profile through the
thickness of the concrete section using an error function ..............................37
xii
Figure Page
2.5.1.2. Fitted free shrinkage results by the modified ACI 209R92 model ..........38
2.5.2.1. Creep model for Control and GRC mixes.................................................40
2.5.3.1. Flexural response of concrete samples under threepoint bending tests ..40
2.5.3.2. Tensile stressstrain and crack width model .............................................41
2.5.3.3. Backcalculated tensile stress strain model at different ages ....................43
2.5.5.1. Schematic drawing for the proposed drying shrinkage model; (a) ring
specimen; (b) strain components ...................................................................45
2.6.1. Comparison of experimental results and model simulation; (a) history of
strain in steel ring; and (b) crack width dimension history at the outer
concrete surface .............................................................................................49
3.2.1b. Plan view of the mold (numbers in mm) ...................................................56
3.3.1.1. Typical cumulative moisture loss and evaporation rate of a cement paste
sample versus time, (a) in linear scale, (b) in log scale .................................58
3.3.1.2. Schematics of moisture transport during drying process of porous media
representing solid (S), liquid (L) and vapor phases (V): (a) initial condition ,
(b) capillary saturation, (c) hygroscopic state (continuous vapor phase) ......59
3.3.1.3. Evaporation test results from water surface comparing to cement paste ..60
3.3.1.4. 2D Crack development during drying of a fresh plain cement paste ......61
3.3.2.1. (a) Deriving aln by fitting a straight line to the curve of
( )
max
ln 1 /
t
M M −Δ Δ vs. time in stage II, (b) Simulation of cumulative
moisture loss vs. experimental data for a cement paste sample .....................69
xiii
Figure Page
3.5.1.1. Effects of surface area and thickness on results of evaporation tests on
plain cement paste ..........................................................................................74
3.5.2.1. Effects of w/c ratio on evaporation results of plain cement pastes ...........76
3.5.3.1. Effect of curing duration on results of evaporation tests ..........................78
3.5.4.1. Results of drying tests on Portland cement paste with different content of
ARglass fibers ...............................................................................................81
3.5.4.2. Crack pattern of cement paste specimens with and without ARglass
fibers after 24 hours of drying under lowpressure test condition .................82
4.2.1. Schematics of the interfacial boundary layer during evaporation from free
water surface ..................................................................................................91
4.2.2. Variation of the diffusion coefficient and saturated vapor pressure versus
temperature ....................................................................................................92
4.2.3. Parametric study on the effects of temperature, boundary layer thickness
and relative humidity on evaporation rate from water surface ......................94
4.3.1. Two stages of drying during evaporation of capillary porous materials .....97
4.3.1.1. (a) Fully saturation state of porous material, (b) capillary saturation of
porous material, (c) variation of hydraulic potential and vapor pressure from
the interior of the material through ambient in stage I drying. ......................99
4.3.2.1. (a) Low saturation state of porous material, (b) variation of hydraulic
potential and vapor pressure using simplified model (model A), (c) the
variations using convection model (model B) .............................................101
xiv
Figure Page
4.3.2.2. Geometry, governing equations and boundary conditions during: (a) stage
I drying, (b) stage II drying assuming convective moisture flow ................106
4.4.2.1. Parametric study on the effect of moisture diffusivity at stage I on profiles
of moisture concentration drawn at 1 h intervals during drying ..................109
4.4.2.2. Parametric study of moisture diffusivity at stage II drying and convective
moisture transfer coefficient on cumulative moisture loss and evaporation
rates of cement paste samples ......................................................................110
4.4.2.3. Parametric study on the effect of initial evaporation rate, F0, and
transition time, ttrans, on cumulative moisture loss and evaporation rates
during drying of cement paste samples ........................................................112
4.4.2.4. Parametric study on the effect of moisture transfer coefficient, k, on
cumulative moisture loss and evaporation rates during drying of cement
pastes ............................................................................................................113
4.4.3.1. (a) Prediction of cumulative moisture loss and evaporation rate as a
function of time for the experimental data of an evaporation test, (b) Back
calculated D and k for the best fit of data ....................................................116
4.5.1. Backcalculated moisture diffusivity of an earlyage noncured cement
paste as a function of normalized concentration expressed by CEBFIP
models ..........................................................................................................118
4.5.2. Distribution of moisture flux (kg/(m2h)) at different drying times. ..........121
xv
Figure Page
4.5.3. (a) Prediction of cumulative moisture loss and evaporation rate as a
function of time for the experiment performed by Šelih (1996), (b) Back
calculated D’ and k for the best fit of data ...................................................123
4.5.4. Results of analysis on drying data by Šelih (1996) ....................................124
5.1.1. Development of 2D plastic shrinkage cracks during drying of a fresh plain
cement paste .................................................................................................127
5.1.2.Two stages of drying during evaporation of capillary porous materials ....128
5.2.1. ACI nomograph for estimating rate of evaporation of surface moisture from
concrete (ACI 305.106) ..............................................................................130
5.3.1.1. Schematic representation of the concentration distribution near an
interface ........................................................................................................134
5.3.1.2. Diffusion boundary layer thickness in a mixed laminar and turbulent
regions for free stream flow over a smooth isothermal flat plate ................141
5.3.2.1. The mass concentration boundary layer in the highrate mass transfer
problem ........................................................................................................143
5.4.1. Effect of temperature, relative humidity, wind velocity and characteristic
length on the rate of evaporation from water surfaces. ................................151
5.5.1. Prediction of experimental data of Azenha et al. (2007a,b) by a method
based on a boundary layer mass transfer concept ........................................156
5.5.2. The comparison between the experimental data and the modeling values of
evaporation potential based on a boundary layer mass transfer theory .......156
xvi
Figure Page
A.1. Free shrinkage results of Control and ARG samples and their comparison 161
A.2. Modeling free shrinkage data of Control sample using ACI 209R92 model162
A.3. Fitting experimental free shrinkage data of ARG samples with modified ACI
209R92 model .............................................................................................163
B.1. 2D shrinkage cracking test setup based on vacuum technology ..................165
C.1. Governing differential equations and boundary conditions of the problem .176
C.2. Input diffusivity and moisture transfer coefficient as a function of time .....177
C.3. Finding the critical time step (152<Δtcrit<153) by trial and error for FE
analysis with 1 element ................................................................................182
C.4. Finding the critical time step (37<Δtcrit<38) by trial and error for FE analysis
with 2 elements ............................................................................................182
C.5. Finding the critical time step (16<Δtcrit<17) by trial and error for FE analysis
with 3 elements ............................................................................................183
C.6. Finding the critical time step (9<Δtcrit<10) by trial and error for FE analysis
with 4 elements ............................................................................................183
C.7. Tracking the values of cumulative moisture loss at end of the test by
increasing number of elements ....................................................................184
C.8. (a) Results of FE analysis with different elements and time steps. (b)
Comparing the results with experimental data .............................................185
C.9. Comparing results of 1D FE analysis by the MATLAB code with the
analytical results ...........................................................................................185
xvii
Figure Page
C.10. Cement paste slab subjected to different boundary conditions during stag I
and II drying .................................................................................................186
C.11. A representative thin slice cut perpendicular to top surface for modeling 1
D moisture transport ....................................................................................187
C.12. Finite element domain ................................................................................188
C.13. Input values of diffusivity as a function of moisture concentration ...........188
C.14. The boundary condition of stage I drying defined as a load ......................189
C.15. Different mesh sizes used for convergence studies ....................................190
C.16. Convergence studies on the results of evaporation tests with Q4 elements191
C.17. Convergence studies on the results of evaporation tests with Q8 elements192
C.18. Final cumulative moisture loss and initial evaporation rates versus number
of nodes ........................................................................................................193
C.19. Distribution of moisture concentration (kg/m3) at: a) t=2 h, b) t=10 h, c)
t=14 h, d) t=20 h, e) t=40 h ..........................................................................194
C.20. Distribution of moisture flux (kg/(m2h)) at: a) t=2 h, b) t=10 h, c) t=14 h, d)
t=20 h, e) t=40 h ...........................................................................................195
C.21. Comparing results of FE analysis by ABAQUS with the analytical results196
C.22. Comparing results of analysis by ABAQUS with the experimental results
and results of 1D analysis by the MATLAB code .......................................197
D.1. Results of evaporation tests on clay and their comparison with cement paste
sample: (a) cumulative moisture loss curves; (b) drying rate curves ...........206
xviii
Figure Page
D.2. Boundary conditions and Fick’s 2nd law of diffusion for the presented drying
problem ........................................................................................................207
D.3. Results of diffusion analysis on wet clay and cement paste samples .........209
D.4. (a) Prediction of cumulative moisture loss and evaporation rate as a function
of time for the experimental data of an evaporation test, (b) Backcalculated
D and k for the best fit of data .....................................................................211
D.5. 2D Crack development during drying of a clay sample .............................211
D.6. Comparing cracking area of a clay sample vs. time with a Portland cement
paste .............................................................................................................212
D.7. Image of a 2D crack pattern in a clay sample after vacuum drying for 48 h212
E.1. The process of image analysis on a cracked paste sample ...........................215
E.2. Normal probability and cumulative density curves of crack width and crack
length ............................................................................................................216
E.3. Analysis by counting number of intersections of cracks network with the
oriented secants (parallel equidistant lines) .................................................218
F.1. Geometry, governing equations and boundary conditions during………...231
F.2. Prediction of cumulative moisture loss and evaporation rate as a function of
time (Selih 1996) ………………………………………………………....233
F.3. Results of analysis on drying data by Šelih (1996)………………………..234
F. 4. Equilibrium between liquid water and humid air at the interface of liquid/gas
phase ……………………………………………………………...………237
F.5. Effective stress on solid skeleton based on unit cell modeling……………239
xix
Figure Page
F.6. Simulation of degree of hydration in comparison with experimental data..239
F.7. Modeling development of modulus of elasticity in early ages (0150 h)….240
F.8. Unrestrained earlyage shrinkage test setup (Wongtanakitcharoen and
Naaman 2007)…………………………………………………………..…241
F.9. Prediction of cumulative moisture loss and evaporation rate
(Wongtanakitcharoen and Naaman 2007)…………………………..…….242
F.10. Simulation of moisture concentration relative humidity profiles ……….243
F.11. Simulation of capillary pores pressure and macroscopic pressure on solid
skeleton profiles…………………………………………………………..244
F.12. Change in Capillary porosity during aerlyage………………………….245
F.13. Introduced elastic modulus for plain and PVA 0.2% concrete …………245
F.14. Simulated earlyage shrinkage strain profiles during drying …………...246
F.15. Simulation of earlyage shrinkage strain averaged over the thickness of the
sample in comparison with the experimental data reported by
Wongtanakitcharoen and Naaman (2007) ………………………………247
xx
NOMENCLATURE
B
m
: mass transfer driving force for water vapor [kg.m
3
]
BF: blowing factor
c: mass concentration of water vapor [kg.m
3
]
c
crit
: critical water vapor concentration for transition from stage I to II [kg.m
3
]
c
e
: mass concentration of water vapor at the air stream [kg.m
3
]
c
s
: mass concentration of water vapor at the interface [kg.m
3
]
D: diffusivity of the water vapor in the air [m
2
.s
1
]
E: Rate of evaporation based on Menzel’s equation [kg.m
2
.h
1
]
h
c
: heat transfer coefficient by convection [W.m
2
.
o
C
1
]
h
m
: mass transfer coefficient for water vapor [m.s
1
]
m
h : average lowrate mass transfer coefficient for water vapor [m.s
1
]
J
s
: rate of diffusion of water vapor per unit area of the surface [kg.m
2
.s
1
]
K
H
: drying coefficient per unit humidity difference [kg.s
1
.m
2
]
l
a,v
: characteristic length of the mixture of air and water vapor, i.e. 3.176x10
10
[m]
L: total characteristic length in the direction of air flow [m]
M: molar mass of the water [kg.mol
1
]
M
a
: molar mass of the air [kg.mol
1
]
m
s
: mass fraction of water vapor at the surface
m′′ ` : total mass flux of water vapor [kg.m
2
.s
1
]
a
m′′ ` : total mass flux of all components in the humid air [kg.m
2
.s
1
]
p
atm
: atmospheric pressure [kPa]
p: water vapor pressure [kPa]
xxi
p
sat
: saturated water vapor pressure [kPa]
p
v
: partial pressure of the water vapor in the air far from the surface [kPa]
Re
crit
: critical Reynolds number
Re
L
: Reynolds number for the characteristic length in the direction of flow
Re
x
: local Reynolds number
Sc: Schmidt number
Sh
x
: local Sherwood number
L Sh : average Sherwood number
T: absolute temperature [K]
T
a
: temperature of the air far from the water surface [
o
C]
T
c
: temperature of the concrete surface [
o
C]
T
film
: temperature of the boundary layer (film)
T
wb
: wetbulb temperature of water [
o
C]
t: transport time for the mass diffusion [s]
t
trn
: time of transition from stage I to stage II drying [h]
u
∞
: freestream velocity [m.s
1
]
v
s
: velocity of water vapor at the interface [m.s
1
]
v
a
: mass average velocity of all components in the humid air at interface [m.s
1
]
V: wind velocity [km.h
1
]
x: the distance from the leading edge [m]
x
crit
: distance from leading edge at which critical Reynolds number is reached [m]
y: distance in the direction of diffusion [m]
d
δ : diffusion boundary layer thickness [m]
xxii
λ: latent heat of evaporation at T
wb
[kJ.kg
1
]
μ : dynamic viscosity of the air [kg.m
1
.s
1
]
ν: kinematic viscosity of the air [m
2
.s
1
]
ρ: density of the air [kg.m
3
]
ρ
a
: total density of all components of the humid air at the interface [kg.m
3
]
ρ
s
: density of the water vapor at the interface [kg.m
3
]
φ: relative humidity of the air far from the surface
D
Ω : dimensionless collision integral for diffusion
1
1. Introduction
1.1. Motivation
The construction industry has to support continuing population growth and
social and economic development. It is one of the largest industries in any country
of the world with a vast number of participants. Only in the United States, there
were 709,590 establishments with 6.57 million paid employees in 2000. The value
of new construction put in place was $842.5 billion in 2001, or 8.4% of the U.S.
gross domestic product (GDP) that year. Construction services contributed 4.7%
to the GDP, i.e., over three times more than agriculture, forestry, and fishing, and
almost four times more than motor vehicles and equipment manufacturing [1].
These numbers show that construction industry’s contribution to the GDP in 2000
was larger than the GDP of 212 countries of the world [2]. Construction involves
numerous manufacturing sectors such as construction materials, building systems,
and service sectors. Among these sectors construction materials is very important
as for almost a century, the construction industry uses the most materials by
weight comparing to other industries [1]. Among all construction materials,
crushed rock, gravel, sand, and cement are the ones which are used the most by
volume. These data indicate that the materials that make up the bulk of Portland
cement concrete and asphalt concrete are used in the largest volume and a special
attention has to be given to concrete and cement industry.
Heavy civil construction is increasingly faced with considerable public
opposition due to construction activities’ environmental impacts. Although
residential construction is still growing in the industrialized countries primarily
2
owing to renovations and retrofits, suburban sprawl, and automobility [1], the
growth cannot be maintained without tackling global economical and
environmental issues namely the rising cost and demand of natural resources and
energy and increasing environmental concerns due to industrial waste products.
Focusing on building material sector, concrete and cement industry is considered
as a large polluter of environment due to production of CO
2
. It is now wellknown
that CO
2
as one of the main greenhouse gases traps some of the sun’s heat energy
causing a significant climate challenge called as ‘global warming’. CO
2
production due to calcination of limestone is one way of CO
2
production (CaCO
3
Æ CaO+ CO
2
) which leads to CO
2
emission of 525 kg per 1 ton of clinker [3]. On
the other hand, the combustion of fossil fuels to heat up the clinker is another way
of CO
2
production. Fossil fuels are hydrocarbons and burning process is the
oxidation of carbon that leads to CO
2
emission (C + O
2
Æ CO
2
+ thermal energy)
by as much as 165 kg CO
2
per 1 ton of clinker [3]. Therefore, production of 1 ton
clinker emits 0.69 ton CO
2
. Altogether, the cement production contributes
approximately 5% of the earth’s CO
2
emissions which is a very significant
number.
Different methods to reduce CO
2
emissions include but not limited to:
improving energy efficiency, finding alternative fuels and energy, reformulation,
capturing CO
2
, and establishing tough regulations against CO
2
emissions. While
these solutions are directly related to the production of cement, extending service
life of cementbased materials and enhancing performance of cement systems also
result in reduction of CO
2
emissions as well as reducing materials demand and
3
economical benefits. This is a sustainable solution since it leads to environmental
and social responsibility as well as economical performance and therefore meets
the needs of the present without compromising the ability of future generations to
meet their own needs. Using sustainable construction materials will eventually
lead to structures which are more durable and economical, while meeting the ever
increasing demand of the public. The main thrust of this effort is to use novel
characterization and simulation techniques in the development of a range of
materials for applications in various infrastructure systems. This research program
will focus on addressing sustainability metrics through prolonging the service life
of products currently used. This task can be achieved through a fundamental
scientific understanding of degradation mechanisms in cement systems. The focus
is on increasing durability of cementbased materials by controlling earlyage
shrinkage cracks. The main topic is the drying which is a major cause of earlyage
shrinkage.
Cementitious materials are weak against tensile stresses, especially during
early hours after casting, and therefore they crack when restrained against
shrinkage strains. Earlyage shrinkage cracks in concrete reduce load carrying
capacity, and accelerate deterioration, resulting in increased maintenance costs
and reduced service life [4, 5]. These cracks are the main routes through which
4
Figure 1.1.1. Greenhouse gases trap some of the heat causing global warming [6]
Figure 1.1.2. Sustainable development principle
aggressive agents such as chloride ions penetrate into the concrete mass and
threaten the longterm durability of structures [7]. Plastic shrinkage in presence of
restraints leads to tensile stresses which may easily exceed the low tensile strength
of fresh concrete and result in cracking. Although attributed to several driving
forces such as differential settlement, thermal dilation, and autogenous
5
deformation [8], plastic shrinkage cracking in concrete occurs principally due to a
high rate of water evaporation from the concrete surface [9, 10, 11]. The drying as
the main cause of shrinkage is studied thoroughly.
1.2. Earlyage Shrinkage Cracking Mechanisms
Earlyage cracking is a threat to structural integrity of concrete
structures and if not inhibited would lead to service life reduction and
unsightliness. High rates of evaporation which depends on air and concrete
temperature, wind speed and relative humidity increase the risk of early age
cracking [11]. Due to evaporation, the solid particles at the surface of fresh
concrete are no longer covered by the water and water menisci are formed in the
interparticle spaces because of adhesive forces and surface tension [ 12]. A
negative pressure in the capillary water is built up due to the curvature of the
water surface [13, 14] and continues to rise as the evaporation proceeds. The
developed capillary pressure in drying fresh concrete can reach 50 kPa in less
than few hours [10, 12, 15] and exceed 1MPa [16, 17]. This pressure acts on the
solid particles resulting in the contraction of the still plastic material [18] ranging
from strain levels of 14x10
3
[12, 14, 19, 20]. If concrete is restrained, the
developed shrinkage strain may easily exceed low tensile strength of the fresh
concrete [21] and cracking occurs.
6
Figure 1.2.1. Equilibrium between liquid water and humid air at the interface of
liquid/gas phase inside a pore and different pressures applied on solid skeleton
[22]
1.3. Background in Shrinkage Testing and Modeling
1.3.1. Free Shrinkage Testing of Cementbased Materials
The free shrinkage test is performed according to ASTM standard C 157
[23]. This test method covers the determination of the length changes that are
produced by causes other than externally applied forces and temperature changes
in hardened hydrauliccement mortar and concrete specimens made in the
laboratory and exposed to controlled conditions of temperature and moisture.
Measurement of length change permits assessment of the potential for volumetric
contraction (shrinkage) of concrete due to drying. This test method is particularly
useful for comparative evaluation of shrinkage potential in different hydraulic
cement mortar or concrete mixtures. In this test, specimens of 1” [25mm] square
crosssection and approximately 111⁄4” [285 mm] in length are casted for each
mixture. Assuming the length of the specimen is much larger than the cross
7
sectional dimensions, then shrinkage takes place only in the length direction. The
measurement of change in length with time can then provide a measure of one
dimensional shrinkage of the material. In this method, the test specimens are
cured in the molds covered with a plastic sheet for 24 hours while water dripping
from specimen is protected. Upon removal of the specimens from the molds, the
initial length reading is taken using a digital comparator dial. After the initial
comparator reading, the specimens are stored in the drying room, so that the
specimens have a clearance of at least 1” [25 mm] on all sides. Comparator
readings of each specimen are taken every one or two days after initial reading.
These readings are done in a room maintained at a relative humidity of 30% while
the specimens are at a temperature of 73°F [23°C]. The difference between the
average of readings on at least two specimens and the initial length of specimens
is used to calculate shrinkage strain. Since the specimens are not restrained in this
test method, this test cannot be an indicator of cracking performance of cement
systems against shrinkage and does not differentiate the contribution of different
materials such as fibers in controlling shrinkage cracks.
8
Figure 1.3.1.1. Molds, samples and digital comparator dial used in free shrinkage
tests
1.3.2. Restrained Shrinkage Testing of Cementbased Materials
If the prismatic specimen is restrained on the length direction, uniaxial
tensile stresses are produced which is similar to a uniaxial tensile test. The linear
specimens have the advantage of the relatively straightforward data
interpretation, however, it is difficult to provide sufficient restraint to produce
cracking with linear specimens, especially when cross sectional dimensions are
large [24]. It is also difficult to restrain linear specimens, just as it is difficult to
conduct a uniaxial tensile test for concrete [24]. Paillere, Buil, and Serrano [25]
used long specimens (1.5 m) with flared ends that were restrained, and used small
crosssectional dimensions (70 x 100 mm) to produce shrinkage cracking. Other
researchers have used similar methods and linear specimens to assess the
shrinkage cracking potential of concrete mixtures [26, 27, 28, 29]. Unfortunately
(b)
(a) (c)
9
due to difficulties associated with providing sufficient end restraint, these test
methods are generally not used for quality control procedures [28, 30]. Other
investigators have used different types of specimens to simulate cracking due to
restrained shrinkage. Platetype specimens have been used by Kraai [31], Shaeles
and Hover [32], Opsahl and Kvam [33], and Padron and Zollo [34]. When
restraint to shrinkage is provided in two directions, a biaxial state of stress is
produced. Consequently, the results obtained from platetype specimens may
depend on specimen geometry in addition to the material properties [24].
A restrained shrinkage test using a steel ring was done as early as 1939 to
1942 by Carlson and Reading [35]. They used a 25 mm thick and 38 mm wide
concrete ring cast around a 25 mm thick steel ring with an external diameter of
175 mm. As a result of drying, a concrete ring would tend to shrink, but the steel
ring would prevent this and cracking occurs. More recently, to better quantify
earlyage cracking tendency of cementitious material, instrumented rings have
been used by researchers to measure the magnitude of tensile stresses that develop
inside the material [24, 36, 37, 38, 39, 40]. Due to its simplicity and economy, the
ring test has been developed into both AASHTO [41] and ASTM [42] standards.
The main difference between these standards is the relative ratio of the concrete to
steel ring thickness which influences the degree of restraint provided to the
concrete. Although this test method can be used to study cracking tendency of
different materials, the test is not applicable to fresh concrete when plastic
cracking is the main concern.
10
Cardboard
form
Groove cut in
plywood
Steel tube
Threaded rod
and a bolt
267 mm
289 mm
423 mm
Figure 1.3.1.2. The configuration and geometry of restrained shrinkage ring
specimen
1.3.3. Earlyage (Plastic) Shrinkage Testing of Cementbased Materials
Several test methods have been implemented using fans [43, 44, 45], fans
and heaters [46, 47], or heat lamps [48] to simulate severe evaporation conditions
causing plastic shrinkage cracks in fresh concrete. Wind tunnel [49] and vacuum
drying [50] have also been used to expedite the drying process. However, these
tests were not designed to focus on evaporation characteristics of the drying
material as the principal driving force of plastic shrinkage. Also, results of test
methods in which the drying of fresh concrete have been studied lack accurate
measurement of the evaporation rate and data interpretation based on transport
characteristics. While Wongtanakitcharoen and Naaman [20] and Wang et al. [51]
studied the effect of fiber content, w/c ratio, and fly ash on normalized cumulative
moisture loss using fan, heater, and methanol treatment respectively, parameters
addressing rates of evaporation were not utilized in their work. Samman et al. [52]
used an electric fan to investigate the effect of w/c ratio, admixtures, and silica
fume on the normalized cumulative moisture loss and maximum rate of
11
evaporation. However, maximum rate of evaporation is not sufficient to
characterize different materials characteristics. Berhane [53] evaluated the effect
of w/c ratio and ambient relative humidity using evaporation rates obtained from
moisture loss curves. However, the scale of data requires the evaporation rates to
be interpreted based on mass transport parameters. Hall and Hoff [54] analyzed
the drying of clay brick ceramics using two distinct stages: a constant drying rate
period and a falling drying rate period. They then extended the results from
porous media to stones and concrete. Analysis of the evaporation rates of cement
based materials with direct attention to stages of drying process has not been
thoroughly investigated.
1.3.4. Modeling the Drying of Cementbased Materials
In existing literature, the moisture movement during drying of concrete is
often described as a diffusion process where the diffusivity depends highly on the
moisture content. Torrenti et al [55], West and Holmes [56], Garbalińska [57],
Huldén and Hansen [58], Kim and Lee [59] modeled the moisture movement by
the Fick’s second law of diffusion without explanation of the physics of the
problem. Most of these works refer to Bazant and Najjar [60, 61] who considered
the drying of concrete as a nonlinear diffusion problem. Using an approach based
on soil science, Kodikara and Chakrabarti [62] expressed the moisture movement
during drying as a two separate stage mechanism [63], however, they modeled the
phenomena as a onestage diffusion process. Chen and Mahadevan [64] referred
to Walton et al [ 65] in using diffusion as the dominant moisture transport
12
mechanism in concrete. Shimomura and Maekawa [50] used mass conservation of
the vapor and liquid water to derive a diffusiontype equation. Šelih and Bremner
[66] concluded from the experimental results that the diffusiondriven moisture
transport approach is appropriate only for the late stage of drying when the
material is predominantly unsaturated, and the moisture movement in the form of
vapor flux is dominant. Although Shimomura and Maekawa [50] discussed the
movement of liquid water as governing mechanism regarding moisture movement
near the saturated state, referring to Young [67], the concept was not implemented
in their model. Šelih and Bremner [66] expressed the drying process as two
different types of mechanisms of movement which can be described by a Darcy
type equation near saturation, succeeding by a diffusiontype equation.
Nonetheless, their work was only limited to experimental observations.
The evaporation through the surface as a boundary condition is applied
differently in available models. While ACI 305R99 [ 68 ] and Uno [ 69 ]
recommended a constant flux equal to evaporation rate of water surface for early
age hot weather concreting, a convective boundary condition proportional to the
difference of internal and ambient moisture concentration was most often
employed [50, 56, 58, 59]. Torrenti et al [55] and Shimomura and Maekawa [50]
also used evaporative boundary conditions similar to convection. In another
approach, Bazant and Najjar [60], and Garbalinska [57] applied a constant
moisture concentration equal to ambient moisture concentration as the boundary
condition in their models. However, the assumed boundary conditions are not
supported by the physics and driving forces of surface evaporation process.
13
1.4. Objectives of the Dissertation
The objectives of this research are as follows:
(1) Build and calibrate an evaporation tests system to investigate
evaporation characteristics of the drying cementbased material as the
principal driving force of plastic shrinkage.
(2) Accurate measurement of the evaporation rates and data interpretation
based on mass transport parameters.
(3) Investigate two distinct stages of drying which are generally proposed
for drying of various capillary porous media.
(4) Equip the tests system with restraining methods and imaging device to
impose and track 2D shrinkage cracks from the early initiation stage
to crack propagation and stabilization stages using timelapse
photography.
(5) Develop an image analysis technique to perform crack surface
morphology and distinguish various surface crack patterns.
(6) Propose a simplified nonlinear diffusion equation, which is accounted
for all drying processes namely, Darcian transport of liquid water and
diffusional transport of vapor water.
(7) Propose appropriate boundary conditions for different stages of drying
based on the physics of the problem which is useful when modeling
the drying process.
14
(8) Model earlyage shrinkage from the drying model using the
intermediate tools namely, desorption isotherms and capillary
pressurebased models.
(9) Predict cracking tendency of cementbased materials from material
models by comparing predicted developed shrinkage with strain
capacity of the material.
1.5. Overview of Thesis
Chapter 2 describes testing methodology of a ringtype restrained drying
shrinkage test. Materials and mixture properties of different concrete samples
including plain and fiber reinforced concrete samples are investigated
experimentally. Results of the tests including strain history in restraining steel
captured by strain gauges attached to the midheight of the steel ring is monitored
as concrete is exposed to an arid environment. Also a systematic image analysis
approach is followed to measure crack width growth in concrete ring specimen.
Along with the experimental results, an analytical approach that relates key
influential parameters of modeling including diffusion, shrinkage, creep, aging
material properties, and restraining effect is presented.
Chapter 3 presents a test method for characterizing evaporation parameters
and simulating the sequential formation of shrinkage cracks in twodimensional
cement paste samples under lowpressure condition. Effects of sample size, w/c
ratio, initial curing and fiber content on weight loss and evaporation rates are
explained. Two major stages of drying including a constant drying rate period
15
(stage I) and a falling drying rate period (stage II) is investigated through
experimental results. The results are interpreted based on mass transfer
parameters using continuity of liquid and vapor phases.
Chapter 4 summarizes physics of water evaporation followed by a
description of theory of evaporation from cementitious matrix that is represented
as a capillary porous media. A dualstage methodology for modeling the drying is
introduced based on surface moisture transfer and internal moisture transport
characteristics. After studying the effect of different parameters on the model, the
methodology is applied to the results of drying experiments under low and normal
pressure conditions. A 1D Finite Element (FE) analysis code is also presented in
order to verify the results of proposed analytical method. The FE method is
further extended to 2D analysis using ABAQUS v.6.101 [ 70] utilizing the
analogy between the heat transfer and the moisture diffusion.
Chapter 5 suggests a new approach in the absence of other theoretical
methods for calculation of initial rate of evaporation from cementbased
materials. In this chapter theory of water evaporation is described based on the
boundary layer theory, mass transfer, diffusion, and fluid mechanics. A driving
force of concentration gradient is introduced as the main cause of mass transfer
during the drying process. Mass transfer at low and high rates is discussed
separately. The Schmidt number, Sc, which is a dimensionless number defined as
the ratio of momentum diffusivity (ν) and mass diffusivity (D), physically relates
the relative thickness of the hydrodynamic (velocity) and masstransfer
(concentration) boundary layers. On the other hand, the local Sherwood number,
16
Sh
x
is a dimensionless number representing the effectiveness of mass convection
at the surface. Using these parameters and Reynolds number, Re, which gives a
measure of the ratio of inertial forces to viscous forces, mass transfer coefficient
can be derived. An algorithm for determination of evaporation rate is suggested.
Considering similarity between rate of evaporation from water surface and
concrete surface during initial stage of drying, predicted evaporation rates are
verified by the recent experiments. The results show that given the environmental
parameters as inputs, the model is capable of predicting the evaporation rates with
a good accuracy assuming surface cooling occurrence.
17
2. Restrained Shrinkage Cracking
2.1. Introduction
This chapter deals with the role of fibers in extending the cracking
resistance of concrete subjected to drying shrinkage. In hot and low humidity
environments, concrete shrinks due to loss of moisture from capillary and gel pore
microstructure. When concrete is restrained from free shrinkage, tensile stresses
develop and may result in cracking if the stress exceeds materials’ low tensile
strength. This is more dominant when at an early age the tensile strength is quite
low and rate of moisture evaporation is high. The main objective of this chapter is
to study effect of fibers on responses of a wellknown restrained shrinkage test.
The methodologies of restrained drying shrinkage tests are described.
Materials and mixture properties of different concrete samples including plain and
fiber reinforced concrete samples are investigated experimentally. The
experiments are performed according to the ringtype restrained shrinkage test
method and strain history in the steel ring is monitored. A systematic
methodology based on image analysis approach is used to measure crack width
growth in concrete ring specimen. An analytical approach that relates key
influential parameters of modeling including diffusion, shrinkage, creep, aging
material properties, and restraining effect is presented.
2.2. Review of Drying Shrinkage Testing Methods
Understanding the mechanism of shrinkage cracking is essential to design
of durable structures. Plastic shrinkage occurs during the early age period when
18
the strength of the paste is quite low and drying volume changes occur due to
depletion of water due to evaporation, chemical reactions, in addition to chemical
shrinkage. While the fibers may not affect the evaporation rate, their addition
increases the strength and strain capacity sufficiently during the early ages so that
the potential for tensile cracking is minimized. Fiber addition to concrete
therefore reduces cracking potential due to restrained shrinkage. As the concrete
hardens, high stiffness fibers such as steel and macro synthetic fibers serve to
increase the strength, crack growth resistance, and strain carrying capacity,
providing a mechanism for additional restraint to distribute volumetric shrinkage
and maintain a low crack width.
To measure free drying shrinkage, ASTM C 157 [23] recommends a
prismatic specimen of 25, 75 or 100 mm square cross section and 285 mm in
length. If it is assumed that the length of the specimen is much larger than the
cross sectional dimensions, then shrinkage takes place only in the length
direction. The measurement of change in length with time can then provide a
measure of onedimensional shrinkage of the material. In this test method no
tensile stress and consequently no crack develops in the specimen. Therefore, this
test method does not differentiate the contribution of fibers which reveals in the
resistance of concrete to cracking.
If the prismatic specimen is restrained on the length direction, uniaxial
tensile stresses are produced which is similar to a uniaxial tensile test. The linear
specimens have the advantage of the relatively straightforward data
19
interpretation, however, it is difficult to provide sufficient restraint to produce
cracking with linear specimens, especially when cross sectional dimensions are
large [24]. It is also difficult to restrain linear specimens, just as it is difficult to
conduct a uniaxial tensile test for concrete [24]. Paillere et al. [25] used long
specimens (1.5 m) with flared ends that were restrained, and used small cross
sectional dimensions (70 x 100 mm) to produce shrinkage cracking. Other
researchers have used similar methods and linear specimens to assess the
shrinkage cracking potential of concrete mixtures [26, 27, 28, 29]. Unfortunately
due to difficulties associated with providing sufficient end restraint, these test
methods are generally not used for quality control procedures [28, 30].
Other investigators have used different types of specimens to simulate
cracking due to restrained shrinkage. Platetype specimens have been used by
Kraai [31], Shaeles and Hover [32], Opsahl and Kvam [33], and Padron and Zollo
[34]. When restraint to shrinkage is provided in two directions, a biaxial state of
stress is produced. Consequently, the results obtained from platetype specimens
may depend on specimen geometry in addition to the material properties [24].
A restrained shrinkage test using a steel ring was done as early as 1939 to
1942 by Carlson and Reading [35]. They used a 25 mm thick and 38 mm wide
concrete ring cast around a 25 mm thick steel ring with an external diameter of
175 mm. As a result of drying, a concrete ring would tend to shrink, but the steel
ring would prevent this and cracking occurs. More recently, to better quantify
earlyage cracking tendency of cementitious material, instrumented rings have
20
been used by researchers to measure the magnitude of tensile stresses that develop
inside the material [24, 36, 37, 38, 39, 40]. Due to its simplicity and economy, the
ring test has been developed into both AASHTO [41] and ASTM [42] standards.
The main difference between these standards is the relative ratio of the concrete to
steel ring thickness which influences the degree of restraint provided to the
concrete. A ringtype restrained shrinkage testing method similar to AASHTO PP
3499 is used in this chapter to study effects of fibers on shrinkage cracking.
2.3. Restrained Shrinkage Cracking
Uniaxial restrained shrinkage tests and tensile tests conducted on large
scale steel FRC specimens with fiber contents ranging from 0100 kg/m
3
indicate
that multiple cracking affects the overall response of the steel FRC in the
hardened state [71]. The Ring test however is unable to capture the effect of
multiple cracking simply because as a single crack forms the boundary conditions
on the wall change. Several approaches to evaluate the influence of ring
geometry and drying direction on the behavior of the restrain shrinkage test using
the steel ring specimen have been proposed [ 72 , 73 , 39]. These studies
demonstrate use of the steel ring in measuring the residual stress development as
well as the stress relaxation once cracking occurs. Mane el al. [74] developed an
experimental and analytical simulation algorithm to study the restrained shrinkage
cracking in plain and FRC. A constant humidity chamber holding the restrained
shrinkage specimens was utilized with a fan providing constant flow of air around
the specimens. The strain in the restraining steel and the crack width in the
21
concrete samples were monitored continuously. The results are correlated with the
specimen geometry, humidity and temperature conditions, stiffness of the steel
ring, and concrete stiffness, ductility, shrinkage, and creep characteristics.
Shah and Weiss [75] demonstrated that prior to cracking, the stresses that
develop in a plain and a FRC are very similar. They also developed an analytical
procedure for stress development in the steel ring, the stress transfer across the
crack, and crack size [75]. Acoustic emission measurements indicated similar
energy release in plain and fiber reinforced specimens; however, fibers appear to
delay the development of a localized, visible crack [76]. In order to eliminate the
influence of test conditions, an analytical approach is needed to incorporate
influential parameters of shrinkage, creep, aging, and microcracking in the stress
analysis of a restrained concrete section. Using the theoretical models, it is
possible to calibrate and interpret the experimental test results.
2.4. Restrained Drying Shrinkage Test Methodology
Restrained shrinkage due to drying environment may cause cracking in
concrete and cementitious materials. Shrinkage cracking can be a critical problem
in concrete construction. One possible method to reduce the adverse effect of
cracking is the addition of randomly distributed short fibers to the concrete
mixture. An experimental study is conducted to evaluate effects of adding low
contents of ARglass fibers (1.5, 3, 4.5 and 6 kg/m3) on shrinkage cracking [77].
An instrumented ring specimen similar to AASHTO PP3499 [41] has
been used to quantify the restrained shrinkage and tensile creep behavior of
22
concrete. The specimen consists of a 66.6 mm thick annulus of concrete cast
around a rigid steel ring 11.2 mm in thickness with outer diameter of 289.4 mm
and a height of 133.3 mm. A schematic configuration and geometry of shrinkage
ring specimen is shown in Figure 1.3.1.2. In order to cast this sample, a cardboard
form is used to hold the concrete from the outer sides. Since the height of
specimen is two times to its thickness, it is assumed that uniform shrinkage takes
place along the height of specimen. Two strain gauges are mounted on interior
surface of steel ring at the mid height level and 90 degrees apart to measure
developed strains in steel due to shrinkage of concrete. The width of the shrinkage
cracks are measured at different ages using a highresolution digital camera and
implementing a systematic procedure of image analysis.
2.4.1. Mixture Properties, Casting and Specimen Preparation
To understand the drying shrinkage and cracking properties of glassfiber
reinforced concrete (GRC), four concrete mix designs were developed with water
cement ratio of 0.55 and slump of 65 90 mm. The cement content was 680 Kg/m
3
and only fine aggregates were used to cast concrete. Four mixtures of Control,
GRC1.5, GRC3 and GRC4.5 are designed with 1.5, 3 and 4.5 kg/m
3
of glass
fibers (24 mm long) respectively. These dosages are equivalent to 0.06%, 0.11%,
0.17% and 0% volumetric fractions, respectively. The AlkaliResistant (AR)
Glass fibers were provided by St. Gobain Vetrotex America Inc. as multifiber
strand of 100 round filaments bonded together, with filament diameter of 14
micron. The length, elastic modulus and specific gravity of fibers are 24 mm, 72
23
GPa and 2.68, respectively. The AR glass fibers contain zirconium dioxide
(ZrO2) which makes the glass composition resistant against alkaline environment,
and thus stable in concrete [78, 79, 80, 81, 82]. The mixture proportions of all
mixes are shown in Table 2.4.1.1 and include Type I/II Portland cement as the
primary binding agent.
Table 2.4.1.1.
Mixture proportions of the Control and GRC samples (kg/m
3
)
Mix ID Control GRC1.5 GRC3 GRC4.5
Portland cement 680 680 680 680
Fine aggregates 1360 1360 1360 1360
Water 374 374 374 374
AR glass fibers 0 1.5 3 4.5
w/c 0.55 0.55 0.55 0.55
Sand/cement 2.0 2.0 2.0 2.0
The following mixing procedure is adopted in this study. First half of dry
materials including cement and fine aggregates are introduced into the mixer and
blended for 2 minutes. Then, water is added to the mixer thoroughly and blended
for 5 minutes before adding the rest of dry materials and mixing for an extra 5
minutes. Finally, glass fibers are added gradually to the GRC mixer and blended
for 3 minutes. The dosage of AR glass fibers in GRC mixes are limited to the
weight fractions of 4.5 kg/m
3
which is equivalent to very low volume fraction of
0.17%. All molds are filled in three layers with proper compaction by a thin rod in
24
between the layers. Due to high fluidity of all mixes, no special technique such as
vibration is used to help with the compaction of the fresh mixture in molds.
Four replicate specimens are made from each batch. The samples are kept
in a moist curing room, covered with sealed plastic sheets for 24 hours. After 24
hours, the specimens are demolded and top surface of ring samples are sealed by a
silicon rubber adhesive sealant to allow drying to occur only from the outer
circumferential surface of the concrete specimen. Samples are placed in the
shrinkage chamber shown in Figure 2.4.1.1.
Figure 2.4.1.1. Experimental setup of shrinkage test and recording strain gauge
results
The rings are placed in a constant humidity chamber at the temperature of
40º C. The strain gauges attached to the inner surface of the steel ring are
connected to the interface module and a dummy temperature compensating gauge
in the form a halfbridge. The response from the strain gauges is collected using
25
SCXI1321(16channel, Half Bridge) module, SCXI1000 carrier and NI data
acquisition device and transferred into a LABVIEW programmed computer. The
program records the strain gauges readings at specified intervals.
2.4.2. Results of Strain Gauges Attached to Steel Ring
The results of a typical strain gauge attached to steel ring for a plain
concrete sample is shown in Figure 2.4.2.1. As shown in the figure, three stages
can be seen in the response of strain gauge vs. time. The first stage which lasts 3
hours to less than 24 hours shows the effect of rising temperature and expansion
of steel ring due to placement in chamber. After steel ring reaches the same
constant temperature as the chamber, the strain gauge response reflects effect of
shrinkage in concrete at stage 2. In this stage the response of concrete cast around
the steel ring is linear elastic. When shrinkage strain in concrete exceeds ultimate
tensile strength, concrete cracks and its effect on steel ring in plain concrete is a
drop in the strain values. This drop in strain gauge results shows time of cracking
due to very low residual postcrack tensile strength in plain concrete. In this stage
(stage 3), strain in steel drop approximately to zero in plain concrete samples and
visible crack width increases by the time. Strain gauge vs. time response can be
used to detect time of cracking in the specimen, e.g. in the Figure 2.4.2.1, 6 days
after drying in the chamber.
Two strain gauges mounted on each sample give different responses
especially after cracking. Different responses before cracking can be related to
different distances of strain gauges relative to heater location or source of
26
temperature. In the postcrack response, the difference can be related to different
distances of strain gauges from crack location in the sample. Results of two strain
gauges mounted on a steel ring along with location of crack and strain gauges are
shown in Figure 2.4.2.2 representing the strain responses of a plain concrete
sample. A simple code is used to reduce and smooth the data. The raw and
smoothened results of a strain gauge attached to steel ring are shown in Figure
2.4.2.3. The smoothened results of strain gauges were used to compare results of
different mixtures.
0 2 4 6 8 10
Drying Time, days
100
75
50
25
0
25
50
S
t
r
a
i
n
i
n
S
t
e
e
l
R
i
n
g
,
m
i
c
r
o
s
t
r
a
i
n
s
Stage 3
Stage 2
Stage 1
Prepeak region
(elastic loading)
PostPeak
Cracking
Corresponding to
ultimate tensile strength
Expansion
equilibrium
Figure 2.4.2.1. Typical result of an attached strain gauge for a plain concrete
sample
27
strain gauge 2
strain gauge 1
0 2 4 6 8 10 12 14
Drying Time, days
100
75
50
25
0
25
50
S
t
r
a
i
n
i
n
S
t
e
e
l
R
i
n
g
,
m
i
c
r
o
s
t
r
a
i
n
s
Strain Gage 1
Strain Gage 2
w/c=0.55,
curing time = 1 day
(a) (b)
Figure 2.4.2.2. (a) Results of two strain gauges mounted on a plain concrete
sample, (b) Location of the crack and strain gauges
0 2 4 6 8 10 12 14
Drying Time, days
100
75
50
25
0
25
50
S
t
r
a
i
n
i
n
S
t
e
e
l
R
i
n
g
,
m
i
c
r
o
s
t
r
a
i
n
s
Raw Strain Gage Data
Smoothed Strain Gage Data
w/c=0.55
curing time = 1 day
Figure 2.4.2.3. Smoothing typical results of a strain gauge attached to steel ring
28
From two strain gauges attached on each sample’s steel ring, the results of
the one showing more stability and not so close to the crack location was used to
be compared with other samples. The results of strain gauges of different mixtures
are shown in Figure 2.4.2.4. The figure clearly indicates that using ARglass
fibers delayed cracking in concrete samples for nearly 1 or 2 days. According to
the figure, only Control sample and somehow GRC1.5, which are the samples
with no or very low amount of fibers show abrupt drop in the strain gauge data
due to cracking. Cracking happens between 68 days for all different samples, and
the postcrack response of GRC samples with higher fiber dosages are
significantly more than Control sample.
0 2 4 6 8 10 12 14
Drying Time, days
100
80
60
40
20
0
S
t
r
a
i
n
i
n
S
t
e
e
l
R
i
n
g
,
m
i
c
r
o
s
t
r
a
i
n
s
Control
GRC1.5
GRC3
GRC4.5
Figure 2.4.2.4. Effect of glass fibers on results of strain gauges attached to the
rings
29
2.4.3. Capturing Shrinkage Cracks on the Ring Samples
In order to measure crack width, capturing digital images followed by
image analysis approach is adopted in this study. Capturing is performed for all
specimens after 14, 21 and 28 days of drying in the chamber. Since the average
cracking time is 7 days, 14, 21 and 28 days are chosen as convenient ages that
provide sufficient time for shrinkage cracks to extend along the height of samples.
The samples were mounted on a traveling round plate and a 10x digital camera is
used to take pictures. Using this setup, the entire surface of the specimen can be
scanned and examined for new cracks and the measurements of the widths of
already existing cracks. A view of the digital camera when taking picture of
cracks of a concrete sample is shown in Figure 2.4.3.1. Images taken by this
camera are grayscale images with RGB color format. In Figure 2.4.3.2, wide
images of transversal shrinkage cracks in control and GRC3 samples are shown.
As one can see, the difference between crack widths of different samples is
obvious, but in order to quantitatively measure them, images are acquired along
the crack length using a series of 812 images. To do reconstruction of whole
image, overlapping portion of images is cut off and images are attached along the
crack length. Figures 2.4.3.3 show that crack width in GRC3 sample is almost 3
times smaller than in control.
30
Figure 2.4.3.1. Imaging with the digital camera for crack investigations
(a) (b)
Figure 2.4.3.2. Transverse cracks due to restrained drying shrinkage; a) control
sample, b) GRC3 sample after 14 days of drying in the shrinkage chamber
31
Control (Sample 1) Control (Sample 2) GRC3 (Sample 1) GRC3 (Sample 2)
Figure 2.4.3.3. Reconstructed shrinkage crack images of control and GRC3
samples
32
2.4.4. Image Analysis of Shrinkage Cracks
Image analysis technique is used to measure crack width of concrete
samples precisely. First all images are converted from RGB color to 8bit format.
Then a code is developed to read the 8bit images, ask the user of scale factor and
crack boundaries and then makes a binary image of crack boundaries. In the next
step, crack width is automatically measured by pixel counting at 64 different lines
perpendicular to crack length and implementing the scale factor. Finally code
calculates mean crack width, standard deviation, and demonstrates crack width
histogram and normal probability density. Figure 2.4.4.1 shows image analysis
process and output statistical results of a single crack image. To report crack
width of a sample, average crack width and Standard Deviation (SD) obtained
from 812 discrete images along the crack length are calculated. These values for
four replicates of Control, GRC1.5, GRC3 and GRC4.5 mixes at 14, 21 and 28
days are shown in Table 2.4.4.1.
Crack width for each mixture is reported as an average crack width of four
replicate samples. These results are shown in Table 2.4.4.1. Results indicate that
in all shrinkage samples, crack width dimension increases by increasing drying
time. Effect of fiber addition is so significant on reducing crack width dimension.
By adding 1.5, 3 and 4.5 kg/m3 ARglass fiber to the plain concrete (Control mix)
crack width dimension at 14 days is reduced by 51%, 72% and 82%, respectively.
Approximately the same trend is observed after 21 and 28 days of drying in
shrinkage chamber.
33
(a) Input grey scale image
(b) Binary image of crack region (c) Crack width measurement along the
crack length
(d) Crack width histogram (e) Crack width normal
probability
Figure 2.4.4.1. Image analysis of a single image taken from a Control sample
34
Table 2.4.4.1.
Mean Crack width and standard deviation of samples (mm)
* Crack width dimension of these samples were not measured at 28ays.
Sample
Code
Crack Width at 14
Days
Crack Width at 21
Days
Crack Width at
28 Days
Average SD Average SD Average SD
Control 1* 1.168 0.094 1.461 0.079  
Control 2* 1.146 0.066 1.427 0.147  
Control 3 1.280 0.064 1.534 0.084 1.742 0.069
Control 4 1.017 0.053 1.196 0.046 1.364 0.061
Average 1.153 0.069 1.405 0.089 1.553 0.065
GRC1.5 1 0.340 0.104 0.470 0.168 0.528 0.104
GRC1.5 2 0.551 0.010 0.704 0.061 0.823 0.046
GRC1.5 3 0.343 0.041 0.391 0.023 0.488 0.048
GRC1.5 4 0.993 0.048 1.189 0.041 1.227 0.048
Average 0.557 0.051 0.688 0.073 0.766 0.062
GRC3 1* 0.363 0.046 0.445 0.020  
GRC3 2* 0.323 0.046 0.404 0.046  
GRC3 3 0.333 0.051 0.427 0.074 0.528 0.041
GRC3 4 0.244 0.038 0.348 0.043 0.480 0.046
Average 0.316 0.045 0.406 0.046 0.504 0.043
GRC4.5 1 0.229 0.033 0.307 0.033 0.386 0.038
GRC4.5 2 0.216 0.048 0.300 0.020 0.351 0.020
GRC4.5 3 0.157 0.028 0.218 0.015 0.246 0.025
GRC4.5 4 0.114 0.025 0.173 0.043 0.229 0.020
Average 0.179 0.034 0.250 0.028 0.303 0.026
35
0 5 10 15 20 25 30
Time, days
0
0.5
1
1.5
2
C
r
a
c
k
W
i
d
t
h
,
m
m
Control
GRC1.5
GRC3
GRC4.5
Figure 2.4.4.2. Mean and SD of shrinkage crack widths at different times of
drying
According to Figure 2.4.4.2 and Table 2.4.4.2, standard deviation of the
crack widths for GRC1.5 samples is higher than that of Control and other GRC
mixtures. The resistance of GRC1.5 samples against crack extension can be
similar to both plain concrete (plain sample) and other GRC samples with higher
fiber dosage. It is concluded that glass fiber dosage of 1.5 kg/m
3
is not enough to
ensure of crack controlling. However, addition of 3 kg/m
3
glass fibers or more to
the plain mixture results in much lower deviation in the results and reduction of
crack width dimension by 70%.
2.5. Theoretical Modeling of Restrained Shrinkage
Various theoretical approaches for modeling the restrained shrinkage
cracking of concrete have been developed [40, 83, 84]. These models try to
36
address the interaction between materials properties and the shrinkage
characteristics of concrete. Meanwhile, there is no model to properly predict post
crack behavior of plain concrete and GRC in the restrained shrinkage test based
on the experimentally obtained tensile stressstrain response of the material. The
simulation of crack width opening which is of great importance from the aspects
of durability and serviceability is also lacking.
A research attempt to develop an analytical approach that relates key
influential parameters of modeling including diffusion, shrinkage, creep, aging
material properties and restraining effect is presented in this section [74]. The
theoretical model is used to calibrate and interpret the experimental test results,
which provides better understanding on how drying environment generates tensile
stress and cracks in a restrained concrete specimen. This model is capable of
showing effects of adding fiber on the restrained shrinkage and cracking of
concrete. Finally a summary of results concluding significant effect of ARglass
fibers on minimizing crack width dimension supported by both experimental and
analytical data are presented.
2.5.1. Moisture Diffusion and Free Shrinkage
As concrete loses its moisture to the environment, free shrinkage is taken
place and the humidity profile h(z) through the thickness of the concrete section is
simplified to follow the Fick's law of diffusion:
( ) ( ) ( )
s s i
h z h h h erf z = − − (2.5.1.1)
37
where z is the distance measured from the outside surface inward the specimen, h
s
and h
i
represent the humidity (fraction) at the outside surface and interior section
and erf(z) represents the error function [85].
steel
concrete
evaporation
humidity
d
e
p
t
h
(
z
)
( ) h z
s
h
i
h
Figure 2.5.1.1. Fick's law of diffusion is used to simulate humidity profile
through the thickness of the concrete section using an error function
A cubic function is used to relate the free shrinkage strain as a function of
the humidity profile throughout the thickness.
3
( )
( ) (1 ( ) )
sh sh t
z h z ε ε = − (2.5.1.2)
where ε
sh(t)
is the free drying shrinkage strain at any time. Free drying shrinkage
can be obtained by experiments or from the empirical relationships proposed by
BazantPanula [86] for drying shrinkage at infinite time (ε
sh∞
)
and timedependent
shrinkage formulation suggested by ACI 209 [87] as;
( ) sh t sh
t
f t
ξ
ξ
ε ε
∞
=
+
(2.5.1.3)
where f and ξ are constants and t is time in days. ACI 209R92 recommends an
average value for f of 35 for 7 days of moist curing, while an average value of 1.0
38
is suggested for ξ. In this study, free drying shrinkage is obtained by experiments
performed according to ASTM standard C 15704. Meanwhile, the experimental
data of free shrinkage is fitted by the ACI equation as shown in Figure 2.5.1.2,
and the proper values of f and ξ are selected as 28 days and 1.5, respectively. As
shown in the figure, the best fitted values for drying shrinkage at infinite time
(ε
sh∞
) for Control, GRC1.5, GRC3 and GRC 4.5 samples are 2160, 2052, 1944
and 1836 micro strains, respectively.
0 7 14 21 28
Time ,days
0
500
1000
1500
2000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
,
m
i
c
r
o
s
t
r
a
i
n
Control
GRC1.5
GRC3
GRC4.5
Figure 2.5.1.2. Fitted free shrinkage results by the modified ACI 209R92 model
2.5.2. Creep of Concrete
As concrete is subjected to tensile stress, it produces creep strain with time
and offset the strain due to free shrinkage, resulting in relaxation of the elastic
stresses in the material. Due to lacking of early age tensile creep model, it is
39
assumed that the creep coefficient of concrete in compression ν(t) used in ACI
209 report [87] is applicable to the present tensile mode of loading.
( )
u
t
v t v
d t
ψ
ψ
=
+
(2.5.2.1)
where ν
u
is the ultimate creep coefficient, d and ψ are constants and t is time in
days. In this study ν
u
is assumed as 12 for all samples. ACI209R92 recommends
an average value of 10 and 0.6 for d and ψ, respectively. However, since the
recommendation is for the plain concrete under compression tests, the values of d
and ψ are modified to 6 and 1 respectively for the tensile creep behavior of
Control mixture. For the case of GRC1.5, 3 and 4.5, values of d are chosen as 4, 3
and 1.5 and values of ψ are chosen as 1.1, 1.3 and 0.5, respectively. Note that the
presented creep model needs further studies to be verified by experimental data.
2.5.3. Age Dependent Concrete Strength
To obtain tensile strength properties of the mixtures along the time,
monotonic threepoint bending tests are performed on the Control and GRC3
samples at 1, 3, 7 and 28 days. Figure 2.5.3.1 shows the flexural responses of
these two mixtures at 1 and 7 days.
40
0 7 14 21 28
Time ,days
0
4
8
12
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
,
m
i
c
r
o
s
t
r
a
i
n
Control
GRC1.5
GRC3
GRC4.5
Figure 2.5.2.1. Creep model for Control and GRC mixes
0 0.25 0.5 0.75 1 1.25
CMOD (Crack Opening), mm
0
1000
2000
3000
4000
F
l
e
x
u
r
a
l
L
o
a
d
,
N
Control
GRC3
Age: 1 day
0 0.25 0.5 0.75 1 1.25
CMOD (Crack Opening), mm
0
1000
2000
3000
4000
5000
F
l
e
x
u
r
a
l
L
o
a
d
,
N Control
GRC3
Age: 7 days
Figure 2.5.3.1. Flexural response of concrete samples under threepoint bending
tests
41
To calculate tensile stressstrain parameters of the samples from three
point bending tests, a model shown in below figure, which is properly applicable
to both plain and GRC samples is hired.
Figure 2.5.3.2. Tensile stressstrain and crack width model
The parameters of stressstrain backcalculations model of Control and GRC3
samples at 1, 3, 7 and 28 days are summarized in the following Table 2.5.3.1.
Similar to ACI timedependent compressive strength development
function, back calculated tensile strength, ft, residual strength, st3, fracture
energy, Gf and young’s modulus, Ec are assumed to follow these functions:
28
( ) ( )
t t
t
f t f
a bt
=
+
,
3 3 28
( ) ( )
t t
t
s t s
a bt
−
=
+
28
( ) ( )
f f
t
G t G
a bt
=
+
,
28
( ) ( )
c c
t
E t E
a bt
=
+
(2.5.3.1)
42
Table 2.5.3.1.
Backcalculated parameters of Control and GRC3 samples at different ages
Age
(days)
ε
t1
ε
t2
w
t1
(mm)
w
t2
(mm)
f
t
(MPa)
S
t3
(MPa)
G
f
(MPa.
mm)
E
c
(GPa)
Mix ID
Control
1 0.00012 0.00020 0.02794 0.1524 1.758 0.246 0.041 11.264
3 0.00012 0.00019 0.0254 0.1524 2.034 0.270 0.043 13.141
7 0.00011 0.00017 0.02286 0.1143 2.068 0.310 0.046 15.019
28 0.00011 0.00017 0.02286 0.10922 2.103 0.315 0.048 15.770
GRC3
1 0.00017 0.00026 0.04064 1.0668 1.758 0.475 0.255 8.448
3 0.00012 0.00019 0.03048 0.889 2.034 0.529 0.269 13.141
7 0.00011 0.00017 0.0254 0.8128 2.068 0.538 0.275 15.019
28 0.00011 0.00017 0.0254 0.762 2.103 0.547 0.289 15.770
where, f
t28
, s
t328
, G
f28
and E
c28
are tensile strength, residual strength, fracture energy
and Young’s modulus of concrete at 28 days, respectively. t is the time in days, a
and b are material parameters. ε
t1
and ε
t2
, are the tensile strains at limit of
proportionality and peak stress respectively, and w
t1
and w
t2
, are the transition and
ultimate crack widths, are assumed to follow these functions:
t1
(t) = a + c
bt
e ε ⋅ ,
t2
(t) = a + c
bt
e ε ⋅
t1
(t) = a + c
bt
w e ⋅ ,
t2
(t) = a + c
bt
w e ⋅ (2.5.3.2)
In the above mentioned functions, t is time in days, a, b and c are constant
material parameters. The best values for the coefficients a, b, and c after fitting
equation (2.5.3.1) and equation (2.5.3.2) with the values of Table 2.5.3.1 are
being used in this modeling approach.
43
0 400 800 1200
Tensile Strain, microstrain
0
1
2
0
1
2
3
Control
GRC3
T
e
n
s
i
l
e
S
t
r
e
s
s
,
M
P
a
A
g
e
:
2
8
,
1
4
,
7
,
3
,
1
d
a
y
s
A
g
e
:
2
8
,
1
4
,
7
,
3
,
1
d
a
y
s
Figure 2.5.3.3. Backcalculated tensile stress strain model at different ages
2.5.4. Restraining Effect
In this model, tensile strain is positive while compressive and shrinkage
strains are negative. Other related terms such as forces and stresses follow the
same sign as strains. Symbol Δ is used for the incremental change of quantities
between previous time step t
j1
and the current time step t
j
and index i is used for
the layer number at particular location z
i
. The restraining effect provided by steel
ring that prohibits concrete to shrink freely can be determined by equilibrium of
force between the tension force in concrete ΔF
c
(positive) and compression force
in steel (negative) ΔF
st
.
0
c st
F F Δ + Δ = (2.5.4.1)
44
In the equilibrium, only the elastic tensile strain component of the concrete Δε
el
produces stress and it is balanced with the compressive stress in steel, which has
elastic compressive strain Δε
st
. Using modulus of materials, the equilibrium of
force in (2.5.4.1) can be written as
*
1
( ) 0
el c j c st s s
E t A E A ε ε
−
Δ + Δ = (2.5.4.2)
where,
*
1
( )
c j
E t
−
is the secant modulus of the concrete at previous time step
averaged from all discretized concrete layers. This is only an approximation since
the modulus at the current time step is not yet known. E
s
is the young modulus of
steel, A
c
and A
s
are area of concrete and steel, respectively. Rearrange the terms in
equation (2.5.4.2), the incremental compressive steel strain Δε
st
is obtained:
*
1
( )
c j c
st el
s s
E t A
E A
ε ε
−
Δ = − Δ (2.5.4.3)
2.5.5. Stress Strain Development
Figure 2.5.5.1(a) shows a concrete ring specimen subjected to moisture
loss until it cracks. An arch segment of the ring as shown in a dash box can be
approximated to 1D problem depicted in Figure 2.5.5.1(b); the thickness is
assumed to be flat. The figure represents a schematic drawing of strain
components developed in the concrete ring specimen. The free shrinkage strain
Δε
sh
(negative) will be restrained by steel ring creating compressive strain in the
steel Δε
st
(negative) and the steel compressive force must be balanced with the
tension force in concrete, which immediately generates elastic strain in concrete
45
Δε
el
(positive). For time interval Δt, the concrete tensile stress will generate creep
strain Δε
cp
(positive). In addition, the previous stress history before time t
j1
also
contributes to the creep strain increment between time t
j1
and t
j
,
( ) prev
cp
σ
ε Δ , which
can be calculated by using Bolzmann’s superposition principle:
1
( ) 1
1 1
28 2
1
( ) ( )
2
j
el el
prev k k
cp j k cp j
c k
v t t t
E
σ
σ σ
ε ε
−
−
− −
=
⎧ ⎫ −
⎪ ⎪
Δ = − −
⎨ ⎬
⎪ ⎪
⎩ ⎭
∑
(2.5.4.4)
Figure 2.5.5.1. Schematic drawing for the proposed drying shrinkage model; (a)
ring specimen; (b) strain components
46
Therefore, the equilibrium of incremental strains developed between time step t
j1
and t
j
can be written as:
( ) prev
sh el cp cp st
σ
ε ε ε ε ε −Δ = Δ + Δ + Δ − Δ (2.5.4.5)
Substitute the restraining steel strain Δε
st
from equation (2.5.4.3) into equation
(2.5.4.5), yields
*
1
( )
( )
c j c
prev
sh el el cp cp
s s
E t A
E A
σ
ε ε ε ε ε
−
−Δ = Δ + Δ + Δ + Δ (2.5.4.6)
Use the relationship between tensile stress and creep coefficient Δε
cp
=
Δσ
el
C(dt)/E
c28
to determine creep strain, and substitute in equation (2.5.4.6)
results in:
*
1
( )
28
( )
( )
c j c s s
prev
sh el el cp
s s c
E t A E A
v dt
E A E
σ
ε ε σ ε
−
+
−Δ = Δ + Δ + Δ (2.5.4.7)
To obtain stress and strain distribution, the thickness of concrete is discretized
into N
c
sub layers while steel has only one layer. The incremental elastic stress in
layer i located at z
i
can be expressed as:
*
1 1
*
1
( ) [ ( , ) ( , )] ( , )
( ) ( , )
el i el i j el i j c i j
el i c i j
z z t z t E z t
z E z t
σ ε ε
ε
− −
−
Δ = −
= Δ
(2.5.4.8)
Substitute incremental stress Δσ
el
(z
i
) defined in equation (2.5.4.8) into equation
(2.5.4.8) and rearrange the terms, the incremental concrete elastic strain at each
sub layer can be expressed as:
47
( )
*
1
28
( ) ( )
( )
( , )
( )
prev
sh i cp i
el i
c i j
c
z z
z
E z t
Q v dt
E
σ
ε ε
ε
−
−Δ − Δ
Δ =
+
;
*
1
( )
c j c s s
s s
E t A E A
Q
E A
−
+
=
(2.5.4.9)
It should be noted that the concrete is considered as a whole thickness with an
average secant modulus
*
1
( )
c j
E t
−
in accounting for the restraining effect from
steel while concrete is treated individually with its secant modulus at each layer
*
1
( , )
c i j
E z t
−
for the effect of free shrinkage and creep.
2.5.6. Algorithm for Strain History in Steel Ring and Crack Width at Concrete
Surface
1) Calculate free shrinkage strain distribution at each layer i from equation
(2.5.1.2) and equation (2.5.1.3), and its increment by:
1
( ) ( , ) ( , )
sh i sh i j sh i j
z z t z t ε ε ε
−
Δ = − (2.5.6.1)
2) Calculate incremental elastic tensile strain at each layer due to free
shrinkage, restraining effect and creep from equation (2.5.4.9).
3) Update total elastic strain, stress and secant modulus using concrete
model described in Figure 2.5.3.2.
1
( , ) ( , ) ( )
t i j t i j el i
z t z t z ε ε ε
−
= + Δ (2.5.6.2)
( , ) [ ( , )] 2.5.3.2
t i j t i j
z t function z t from Figure σ ε =
(2.5.6.3)
*
( , )
( , )
( , )
t i j
c i j
t i j
z t
E z t
z t
σ
ε
= (2.5.6.4)
48
4) If the updated strain ε
t
(z
i
,t
j
) exceeds the strain at peak stress ε
t2
(t
j
), the
crack width at each layer of concrete is calculated by:
_
( , ) ( , ) ( , ) ( )
t i j t i j t unload i j i
w z t z t z t D z ε ε π
⎡ ⎤
= −
⎣ ⎦
(2.5.6.5)
where ε
t_unload
(z
i
,t
j
) is the unloading strain in the prepeak stress strain
curve corresponding to the same stress level as the strain in post peak
response ε
t
(z
i
,t
j
) and D(z
i
) is the diameter of the concrete ring at location z
i
.
5) Sum concrete force at each concrete layer A
ci
to obtain the total tensile
force, which the magnitude equals to compressive the force in steel.
1
( ) ( , )
N
c
c j ci t i j
i
F t A z t σ
=
=
∑
and ( ) ( )
s j c j
F t F t = − (2.5.6.6)
6) Calculate nominal stress and strain in steel, defined by:
( )
( )
s j
ns j
s
F t
t
A
σ = ;
( )
( )
ns j
ns j
s
t
t
E
σ
ε = (2.5.6.7)
7) A complete strain history at steel ring and crack width at concrete
surface can be obtained by repeating steps 16 until the time t
j
reaches the
specified age.
2.6. Comparing Experimental Data and Simulations
Results of simulation by this model and comparison with the experimental
data are shown in Figure 2.6.1. The results show a good correlation between the
experimental data including strain in steel ring and crack width opening histories
and simulation data. While the best correlation of the simulation with
experimental data is for Control sample, the model overestimates the crack widths
49
of GRC samples after 14 days. However, the model is capable of capturing
significant effects of adding glassfibers to control shrinkage cracking in concrete.
0 2 4 6 8 10 12 14
Drying Time, days
100
80
60
40
20
0
S
t
r
a
i
n
i
n
S
t
e
e
l
R
i
n
g
,
m
i
c
r
o
s
t
r
a
i
n
s
Control
ARG2.5
ARG5.0
ARG7.5
0 5 10 15 20 25 30
Time, days
0
0.5
1
1.5
2
C
r
a
c
k
W
i
d
t
h
,
m
m
Control
GRC1.5
GRC3
GRC4.5
Figure 2.6.1. Comparison of experimental results and model simulation; (a)
history of strain in steel ring; and (b) crack width dimension history at the outer
concrete surface
50
2.7. Conclusion
Reducing crack width dimension by 3 times by using low content of AR
glassfibers is the most significant effect of adding fibers to the mixture in
controlling shrinkage cracking. This effect in addition to delaying crack
occurrence by 1 or 2 days represent the low dosage glassfiber concrete as an
appropriate construction material to be used in dry and hot regions with the risk of
high drying shrinkage. The analytical model for predicting steel strain history and
concrete crack width dimension for restrained shrinkage test shows a relatively
good correlation with the experimental data. This model is capable of capturing
significant effects of fiber addition in order to control cracking of shrinkage.
Collectively speaking, ringtype restrained shrinkage test method is a good
method to capture the behavior of concrete materials against restrained drying
shrinkage.
51
3. Experimental Observations of EarlyAge Drying
3.1. Introduction
Plastic shrinkage cracks in concrete reduce load carrying capacity, and
accelerate deterioration, resulting in increased maintenance costs and reduced
service life [4, 5]. These cracks are the main routes through which aggressive
agents such as chloride ions penetrate into the concrete mass and threaten the
longterm durability of structures [7]. Plastic shrinkage in presence of restraints
leads to tensile stresses which may easily exceed the low tensile strength of fresh
concrete and result in cracking. Although attributed to several driving forces such
as differential settlement, thermal dilation, and autogenous deformation [8],
plastic shrinkage cracking in concrete occurs principally due to a high rate of
water evaporation from the concrete surface [9, 10, 11].
Several test methods have been implemented using fans [43, 44, 45], fans
and heaters [46, 47], or heat lamps [48] to simulate severe evaporation conditions
causing plastic shrinkage cracks in fresh concrete. Wind tunnel [49] and vacuum
drying [50] have also been used to expedite the drying process. However, these
tests were not designed to focus on evaporation characteristics of the drying
material as the principal driving force of plastic shrinkage. Also, results of test
methods in which the drying of fresh concrete have been studied lack accurate
measurement of the evaporation rate and data interpretation based on transport
characteristics. While Wongtanakitcharoen and Naaman [20] and Wang et al. [51]
studied the effect of fiber content, w/c ratio, and fly ash on normalized cumulative
moisture loss using fan, heater, and methanol treatment respectively, parameters
52
addressing rates of evaporation were not utilized in their work. Samman et al. [52]
used an electric fan to investigate the effect of w/c ratio, admixtures, and silica
fume on the normalized cumulative moisture loss and maximum rate of
evaporation. However, maximum rate of evaporation is not sufficient to
characterize different materials characteristics. Berhane [53] evaluated the effect
of w/c ratio and ambient relative humidity using evaporation rates obtained from
moisture loss curves. However, the scale of data requires the evaporation rates to
be interpreted based on mass transport parameters. Hall and Hoff [54] analyzed
the drying of clay brick ceramics using two distinct stages: a constant drying rate
period and a falling drying rate period. They then extended the results from
porous media to stones and concrete. Analysis of the evaporation rates of cement
based materials with direct attention to stages of drying process has not been
thoroughly investigated.
Evaporation is viewed as the diffusion of water molecules through the
boundary layer at the surface to maintain the equilibrium of liquidgas phase
transition [ 88, 89, 90, 91, 92, 93]. During the drying phase, moisture flows
through the porous material toward the evaporative surface to supply the
necessary surface flux. The relationship between unsaturated flow within concrete
pores and evaporation rate on the surface can be used to measure moisture
diffusivity [94, 95]. By analyzing moisture loss data during the drying of well
hydrated cement paste samples, Garbalińska used desorptive measurement
techniques to determine moisture diffusivity [57]. A Similar analytical method is
proposed to measure moisture diffusivity in drying fresh cement pastes.
53
Plastic shrinkage cracks due to a high rate of evaporation are two
dimensional [96]. However, a majority of recent publications have focused on
onedimensional cracking of the concrete by utilizing special restraints [8, 9, 10,
47, 97 ]. Characterization of images taken from a drying surface helps in
understanding crack initiation, propagation, and stabilization during the drying
process.
A test method to monitor the drying of fresh cement paste and the
formation of plastic twodimensional cracks is presented in this work. This test
method allows measuring material properties that characterize mass transfer
during the early stages of fresh paste drying. A drying technique based on a low
pressure condition was employed to cause a rapid rate of evaporation while
weight loss was measured. This drying method imposes a onedimensional
moisture flow through the thickness and facilitates parameter estimation from the
mass transfer data. The test method was applied to a variety of samples in order to
evaluate the impact of sample thickness, surface area, w/c ratio, duration of initial
curing and fiber content on the evaporation characteristics being tested. An
analytical method to determine moisture diffusivity using experimental data is
presented and the effects of different testing parameters on diffusion coefficients
are investigated. The development of crack patterns during drying is documented
using timelapse photography. The images have been used to determine the ability
of the fibers in controlling shrinkage cracking in drying cement pastes.
54
3.2. Testing methodology
Apparatus of the developed lowpressure drying test method is shown in
Figure 3.2.1a. A prismatic sample is filled with fresh paste such that its face is
exposed and all other sides are sealed. The mold consists of interlocking pieces
made of polycarbonate as shown in Figure 3.2.1b and uses anchor hooks to
connect the fresh paste with the mold, providing shrinkage restraint in two
directions. The sample is placed on a load cell which serves as a digital scale, and
the entire assembly is placed inside a glass desiccator. The weight of the sample is
continuously monitored throughout the drying cycle. Using a vacuum pump and a
pressure regulator the air pressure inside the desiccator is lowered to absolute
1700 Pa (0.5 inHg) and maintained at this pressure throughout the test. This level
of air pressure was selected as a system level parameter after extensive pressure
calibration procedures. If the pressure is too low, evaporation occurs at normal
rates, and if it is too high, microstructural damage due to cavitation occurs. A
controlled test condition therefore requires evaporation rates which simulate the
range of severe atmospheric conditions. A condensing system including a D
Drying apparatus is used similar to Copeland and Hayes [98] to remove the water
vapor from the desiccator. Similar preliminary results from comparison of the
weight of the condensed water in the flask with the total moisture loss at the end
of the test show the efficiency of the condensing system. The weight loss
measured was recorded using a computer interface unit. The specimen surface
was photographed at 15 min intervals using a digital camera mounted 10 cm
above the sample. To avoid excessive moisture removal at the beginning of the
55
test, a calibration procedure was developed using a cement paste exposed to
several vacuum pressures such that the rate of evaporation was maintained at a
constant rate. Based on this procedure, a test procedure consisting of 30 minutes
at 34 kPa (10 inHg), followed by a decrease in pressure to 1700 Pa (0.5 inHg) for
the remainder of the test was selected. The test procedure was subjected to
additional calibration using liquid water as the evaporating material. This
indicated the free water evaporation rate of the test setup.
load cell
sample
camera
PC
data
acquisition
system
computer
unit
interface
vacuum
vacuum pump
dry ice
alcohol
vacuum vessel
(25 cm height  25 cm Diameter)
strain
gage
transducer
amplifier
Tbore
stopcock
pressure
gage
pressure
regulator
condenser
Figure 3.2.1a. Schematic of vacuum drying test setup
56
Figure 3.2.1b. Plan view of the mold (numbers in mm)
3.3. Analysis of typical evaporation data under low pressure test condition
3.3.1. Cumulative moisture loss and evaporation rate versus time
The cumulative moisture loss curve was obtained for periods of up to 24
hours or more, and adjusted by means of a baseline calibration curve. Cumulative
moisture loss data were subjected to a numerical differentiation procedure in
order to measure the evaporation rate. Results were expressed with respect to the
exposed surface area in accordance to equation (3.3.1.1).
1 M dM
J
A t A dt
Δ
= =
Δ
(3.3.1.1)
57
where J is the evaporation rate (kg/m
2
s), ΔM is the mass change at specified time
steps (kg), Δt is the time step (s) and A is the surface area of the original sample
(m
2
).
Cumulative moisture loss and evaporation rate curves for a typical cement
paste sample are shown in Figure 3.3.1.1a. Under constant drying condition, the
drying rate of cement paste at the beginning is roughly constant with an
evaporation rate of 0.42 kg/(m
2
hr) during first 10 hours, and gradually reduces to
0.05 kg/(m
2
hr) afterwards. This change in the evaporation rate is also presented in
log format. Figure 3.3.1.1b shows the evaporation rate versus log of drying time
which is in accordance with results obtained by Hall et al. [99], Cooling [100] for
clay brick ceramics, and Kowalski [101] for wet capillaryporous materials such
as paper, leather, ceramics and clay. Similar to their observations, drying of
cement pastes under constant external conditions occurs in two distinct stages of a
constant drying rate period (stage I), and a falling drying rate period (stage II).
58
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40 50
Drying Time, h
Raw Data
Smoothened Data
0
0.2
0.4
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
(A) (B)
(A) Stage I Drying
(constant drying rate period)
(B) Stage II Drying
(falling drying rate period)
Figure 3.3.1.1a
0.1 1 10 100
Log of Drying Time, h
0
0.1
0.2
0.3
0.4
0.5
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
Stage II Drying Stage I Drying
transition time from
stage I to stage II: 10 h
Figure 3.3.1.1b
Figure 3.3.1.1. Typical cumulative moisture loss and evaporation rate of a cement
paste sample versus time, (a) in linear scale, (b) in log scale
59
Different stages of drying can be described based on the degree of
continuity between liquid and vapor phases. Schematics of state phases and
moisture transport during drying process of porous media according to Scherer
[102] and Plumb [103] are presented in Figures 3.3.1.2ac. During stage I drying,
the main phase transitions occur at the boundary surface and vapor phase
diffusion into the air is determining rate of evaporation. Hall and Hoff [54]
referred to several experimental data showing that stage I behavior is independent
of capillary processes inside the material. Drying rate at early time is constant and
about the same as rate of evaporation from water surface exposed to the same
conditions [8, 99]. This was verified by the evaporation tests on water samples.
Results of water surface evaporation compared with a plain cement paste sample
are shown in Figure 3.3.1.3.
(a) (b) (c)
Figure 3.3.1.2. Schematics of moisture transport during drying process of porous
media representing solid (S), liquid (L) and vapor phases (V): (a) initial condition
(full saturation), (b) capillary saturation (continuous liquid phase), (c)
hygroscopic state (continuous vapor phase)
60
0 10 20 30 40 50
Drying Time, h
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
water
cement paste (w/c=0.45)
Figure 3.3.1.3a
0.1 1 10 100
Drying Time, h
0
0.2
0.4
0.6
0.8
D
r
y
i
n
g
R
a
t
e
,
k
g
/
(
m
2
.
h
)
water
cement paste (w/c=0.45)
Figure 3.3.1.3b
Figure 3.3.1.3. Evaporation test results from water surface comparing to cement
paste
61
Both samples show evaporation rates in the range of 0.45 kg/(m
2
hr). It is
therefore postulated that within stage I drying, a continuous liquid phase is
present in the pore structure and external evaporative flux away from the surface
causes a gradient in capillary pressure in the pores which is the main mechanism
of liquid migration from inside toward the boundary [104]. Note that visible
plastic cracks were seen as early as 3 h as shown in Figure 3.3.1.3, and fully
developed up to 4 h, a time period well within the stage I drying. Since in this
stage, drying takes place under external mass transfer control, potential cracking
could have no significant effect in increasing evaporation rate which was the same
before and after cracking.
Drying time: 15 min, 3h, 3h 15min
Drying time: 3h 30min, 4h, 24h
Figure 3.3.1.4. 2D Crack development during drying of a fresh plain cement
paste
62
As the liquid phase water moves to the surface and evaporates with a
constant rate during stage I drying, the moisture content drops continuously. The
moisture content at the surface reaches a critical value at and below which the
material is unable to support the necessary flux to the surface to satisfy the
potential evaporation [105]. At this time, stage II drying begins and rate of
evaporation starts to fall. Referring to the results presented in Figure 3.3.1.1b, this
transition in drying stage is gradual. An empirical definition of transition time is
defined by the intersection point of two asymptotic linear approximations to the
evaporation rate data at stage I and stage II. As shown in Figure 3.3.1.1b, this
transition time for a plain cement paste sample is determined as 9.7 h. Figure
3.3.1.2c shows that in stage II drying, the liquid phase becomes discontinuous
while vapor phase becomes continuous within the matrix. Therefore, phase
transitions take place within capillary pores and mass transfer begins to occur
through the pore vapor phase. Thus, unsaturated capillary flow determines the rate
of evaporation and drying happens under internal mass transfer control [105]. In
stage II drying, the removal of moisture is mainly controlled by diffusion which
dominates the capillary forces [106]. During this period, the moisture content
decreases slowly until it reaches the equilibrium value below which the material
cannot be dried.
3.3.2. Calculation of Moisture Diffusivity
Evaporation from water surface can be treated as a boundary layer
problem, in which the water vapor diffuses from a saturated state on the surface to
63
the ambient moisture concentration [8893]. Rate of evaporation is controlled by
two parameters of binary airwater vapor diffusivity and the concentration
gradient. Evaporation from a fresh cement paste at stage I is quite similar to the
case of clay, ceramics, and other capillary porous media [107]. It consists of a
boundary layer identical to water evaporation boundary, and an unsaturated flow
for transfer of liquid water from within the material to supply the flux. The focus
of this work is on calculation of diffusivity of water within the cement paste rather
than airwater vapor diffusivity at the boundary layer. The calculated diffusivities
at stage I and II show the potential for moisture conductivity through the fresh
cement paste.
Flow within the porous medium can be described by the extended Darcy
equation for unsaturated flow [99, 104] which in the case of a onedimensional
expression is written as:
Κ
t x x
θ ψ ∂ ∂ ∂ ⎛ ⎞
=
⎜ ⎟
∂ ∂ ∂
⎝ ⎠
(3.3.2.1)
where θ is the volume fraction moisture content (), ψ is the hydraulic or
capillary potential (m), Κ is moisture conductivity (m/s), x is the pass of moisture
transfer (m), and t is the time (s).
Definition of capillary diffusivity (m
2
/s) as ( ) D Κ ψ θ = ∂ ∂ leads us to Fick’s
second law of diffusion which governs a majority of moisture transfer problems in
porous media [50, 58, 108].
2
2
C C
D
t x
∂ ∂
=
∂ ∂
(3.3.2.2)
64
where C is the moisture concentration (kg/m
3
), D is the moisture (liquid and
vapor) diffusion coefficient (m
2
/s) and t is the time (s).
In order to simulate the drying test results using a dualstage drying
concept, two different boundary conditions at the top surface are required. For
the first stage, boundary condition of constant flux at the surface can be applied
which is obtained directly from experimental data. Boundary condition for the
second stage which is a constant concentration at the surface is set to ambient
moisture concentration. At ambient pressure of 1 atm, the water vapor pressure is
the product of relative humidity and the saturated water vapor pressure. However,
in this low pressure test condition, the sum of the partial pressure of all the species
in the air including dry air and water vapor is 1700 Pa. Assuming the percentage
of water vapor remains the same after running the vacuum pump, water vapor
pressure can be estimated as 2% of barometric pressure which is 34 Pa. The
corresponding moisture content defined as ω = M
water
/M
Dry Air
= 0.622 x P
water
/(1700P
water
) is equal to 0.0127 kg of Water/ Kg of Air. Therefore, moisture
concentration can be assumed to be zero as a reasonable approximation. Using the
experimental evaporation rate during stage I drying, and zero moisture
concentration at stage II, the two required boundary conditions are:
0
,
Stage I drying: at ,
Stage II drying: at ,
s w low pressure
x L J F
x L C C C
∗
= =
= = =
(3.3.2.3)
where, x is the pass length of diffusion, measured from the bottom face (m), L is
the thickness of sample (m), F
0
is the constant flux in stage I drying (kg/(m
2
s)),
and C
*
is the ambient moisture concentration (kg/m
3
).
65
Considering C
i
as the initial moisture concentration (kg/m
3
), the initial boundary
condition is
0, 0
i
t C C x L = = ≤ ≤ (3.3.2.4)
Since the bottom surface is impermeable, the boundary condition at the bottom is
0 at 0
C
J D x
x
∂
= − = =
∂
(3.3.2.5)
The analytical solution of equation (3.3.2.2), for the initial and boundary
conditions of stage I drying, assuming a constant diffusivity is given by equation
(3.3.2.6) [109].
( )
2 2 2 2
0
2 2 2 2 2
1
3 2 ( 1)
, exp( ) cos
6
n
I I
i
I
n
F L D t x L D n t n x
C t x C
D L
L L n L
π π
π
∞
=
⎫
− − − ⎪ ⎧
= + + −
⎨ ⎬
⎩
⎪
⎭
∑
(3.3.2.6)
where, D
I
is the diffusivity in stage I drying.
Using equation (3.3.2.6), moisture concentration at the top surface is obtained
providing constant F
0
and D
I
values. Since stage I drying ends when the surface
reaches equilibrium moisture concentration, moisture diffusivity at stage I drying
is calculated as long as the time of transition from stage I to stage II has not
lapsed. Applying this method on a typical result of a plain cement paste sample
results in calculating a diffusivity value of 5.15x10
7
m
2
/s at stage I drying. Unlike
the stage I drying, diffusion coefficients in Stage II drying depend on moisture
concentration [ 110]. For simplicity, diffusivity in stage II is taken constant
resulting in calculation of an average diffusivity, which is averaged over space
and time. Analytical solution for the boundary conditions of stage II drying
subjected to constant concentration, C
*
, at the surface is [109]:
66
2 2 2
0
4 ( 1) (2 1)
1 exp{ (2 1) / 4 }cos
2 1 2
n
i
II
n
i
C C n x
D n t l
C C n L
π
π
π
∞
=
∗
− − +
= − − +
− +
∑
(3.3.2.7)
where, D
II
is the diffusivity in stage II drying.
By integrating equation (3.3.2.7) over the thickness of sample, total amount of
diffusing moisture in stage II drying, which has left the sample at time t, M
t
(kg),
can be obtained and related to its corresponding quantity after infinite time,
M
∞
(kg). Therefore, moisture loss during the advanced phase of drying is given
by
2 2 2
2 2
0
8
1 exp{ (2 1) / 4 }
(2 1)
t
II
n
M
D n t L
M n
π
π
∞
=
∞
= − − +
+
∑
(3.3.2.8)
Results of analyses performed by Garbalińska [57] on cement based
materials show that in the expansion of equation (3.3.2.8), the higher terms with
n>0 can be neglected after releasing 40% of the total mass change. As discussed
in the following sections, for almost all samples tested in this study, stage II
drying begins when more than 60% of the moisture is lost. By considering only
first term of the series, moisture diffusivity, D
II
(m
2
/s), can be derived by
rearranging equation (3.3.2.8) as
2 2
2
8
ln 1 ln
4
t
II
M
M
D
L t
π π
∞
⎛ ⎞
− −
⎜ ⎟
⎝ ⎠
⋅ =
−
(3.3.2.9)
Assuming the logarithmic terms on the right hand side of the equation are
following a linear relationship with the time, slope of the curve,
ln
a , can be used
to determine moisture diffusivity during stage II drying.
67
2 ln
8
ln 1 ln
t
M
a t
M π
∞
⎛ ⎞
− − =
⎜ ⎟
⎝ ⎠
(3.3.2.10)
2
2 ln
4
II
L
D a
π
−
= ⋅ (3.3.2.11)
Procedures for obtaining the slope,
ln
a for a plain cement paste sample is
shown in Figure 3.3.2.1a. This process results in calculation of diffusivity as
3.33x10
9
m
2
/s at stage II drying. Using calculated values of diffusivity from stage
I and stage II drying, total amount of moisture loss at any time can be predicted.
Comparison of the simulation and experimental data for a drying cement paste
sample is shown in Figure 3.3.2.1b. This figure shows that dualstage drying
model and corresponding moisture diffusivity values can be used to predict drying
characteristics of cementbased materials in this low pressure condition. Since
two different boundary conditions are applied in stage I and stage II, the
continuity of the slope is not completely achieved as shown by the slight bump in
the simulation data. Further studies on variable diffusion coefficients and
applying different boundary conditions such a convectivediffusive boundary
condition at stage II are required in order to better fit the results.
Note that the evaporation rate in the normal room pressure is different than
that in the low pressure desiccator, where rate of water vapor diffusion in the
desiccator controls the rate of vaporization and drying. The reason is that
diffusion rates of water vapor in low pressure are much higher than that under
normal room pressure, because the diffusion coefficient of gasses is roughly
inversely proportional to the gas pressure at constant temperature [ 111 ].
68
Therefore, the calculated diffusion coefficients for stage I and II drying are only
representatives for this low pressure drying condition. However, the analysis is
applicable to drying at normal room pressure providing evaporation rates and
ambient moisture concentration of that test condition.
3.4. Experimental program
3.4.1. Scope of Test Program
Effects of different variables including sample size, w/c ratio, duration of
initial curing, and fiber content were studied. The scope of the test program is
shown in Table 3.4.1.1 and includes a test matrix consisting of twelve sets of
evaporation tests conducted on different portland cement pastes. Variables of the
study included sample thickness, surface area, w/c ratio, duration of initial curing
and fiber content. Two sample thicknesses of 11 and 21 mm, and two surface
dimensions of 127 mm x 127 mm and 100 mm x 100 mm were used. w/c ratios
of 0.45, 0.5, 0.55 and 0.6 were chosen as material variables. Effect of fibers was
investigated by adding 1.5, 3, 4.5 and 6 kg/m
3
Alkali Resistant (AR) glass fibers
(St. Gobain Vetrotext) to the plain cement paste, equivalent to 0.06%, 0.11%,
69
0 10 20 30 40
Drying Time, h
8
6
4
2
0
l
n
(
1

Δ
m
t
/
Δ
m
m
a
x
)
Experimental Data
Straight Line Fitting
a
ln
Evaporation Data of a
Cement Paste Sample
Stage II Drying
1
Figure 3.3.2.1a
0 10 20 30 40
Drying Time, hr
0
20
40
60
80
M
o
i
s
t
u
r
e
L
o
s
s
,
g
Experimental Data
Simulation
D
stage I
= 5.15 × 10
7
m
2
/s
D
stage II
= 3.33 × 10
9
m
2
/s
Transition Point
Figure 3.3.2.1b
Figure 3.3.2.1. (a) Deriving aln by fitting a straight line to the curve of
( )
max
ln 1 /
t
M M −Δ Δ vs. time in stage II, (b) Simulation of cumulative moisture
loss vs. experimental data for a cement paste sample
70
Table 3.4.1.1.
Scope of the test program
Test series w/c
Initial
curing
(h)
Fiber
content
(kg/m
3
)
Thickness
(mm)
Surface
area
(m
2
)
P0.45IC0F0 (control) 0.45 0 0 11 0.01
P0.45IC0F0Th21 0.45 0 0 21 0.01
P0.45IC0F0SA0.016 0.45 0 0 11 0.016
P0.50IC0F0 0.50 0 0 11 0.01
P0.55IC0F0 0.55 0 0 11 0.01
P0.60IC0F0 0.60 0 0 11 0.01
P0.45IC3F0 0.45 3 0 11 0.01
P0.45IC24F0 0.45 24 0 11 0.01
P0.45IC0F1.5 0.45 0 1.5 11 0.01
P0.45IC0F3 0.45 0 3 11 0.01
P0.45IC0F4.5 0.45 0 4.5 11 0.01
P0.45IC0F6 0.45 0 6 11 0.01
0.17% and 0.23% volumetric fractions. Cement paste samples with initial curing
of 0, 3 and 24 hours were also tested to evaluate effect of curing duration.
3.4.2. Materials, Mixing, Placing and Curing procedures
The test program consisted of variations in the physical parameters of
testing such as sample area, depth, and materials mixture variations. The mixture
proportions for cement pastes are provided in Table 3.4.2.1 and include Type I/II
portland cement as the primary binding agent. In fiber reinforced cement pastes,
AlkaliResistant (AR) glass fibers of 24 mm in length and aspect ratio of 150
were used. The fibers are multifiber strand of 100 round filaments bonded
together. The filament diameter is 14 microns and fiber has elastic modulus of 72
GPa, specific gravity of 2.68 and tensile strength of 1,700 MPa. After mixing in
71
accordance with ASTM C 1116 [ 112 ], samples were cast in the molds
immediately and subjected to test within 15 minutes. No curing was applied
except for samples of two test series which were covered with a plastic sheet for 3
and 24 h at the room temperature before running tests.
Table 3.4.2.1.
Mix proportions of the test series
Test series
Portland
cement
Water
ARglass
fiber
w/c
P0.45ICxxF0* 1450 650 0 0.45
P0.50IC0F0 1400 700 0 0.50
P0.55IC0F0 1355 745 0 0.55
P0.60IC0F0 1310 790 0 0.60
P0.45IC0F1.5 1450 650 1.5 0.45
P0.45IC0F3 1450 650 3 0.45
P0.45IC0F4.5 1450 650 4.5 0.45
P0.45IC0F6 1450 650 6 0.45
* This mix proportion was used for P0.45IC0F0, P0.45IC3F0, P0.45IC24F0,
P0.45IC0F0Th21 and P0.45IC0F0SA0.016 samples
3.5. Parameter Estimation
Parameters obtained from evaporation tests includes aspects of initial rate
of evaporation, time of transition from stage I to stage II drying, evaporation rate
at 24 h, cumulative moisture loss at 24 h and moisture diffusivities. Results are
compiled in Table 3.5.1 and are discussed in the following sections.
72
Table 3.5.1.
Initial evaporation rates, transition time, evaporation rate at 24 h, cumulative
moisture loss, and diffusivities of tested samples
Test
series
Initial
evaporation
rate
(kg/(m
2
.h))
Transition
time of
drying
stages
(h)
Evaporation
rate at 24 h
(kg/(m
2
.h))
Cumulative
moisture
loss at 24 h
(g)
Moisture
diffusivity
at stage I
(m
2
/s)
Moisture
diffusivity
at stage II
(m
2
/s)
P0.45
IC0F0
0.42 9.7 0.024 65 5.15E07 3.33E09
P0.45
IC0F0
0.44 18.9 0.300 127 5.27E07 3.61E09
P0.45
IC0F0
0.37 12.8 0.079 108 4.91E07 3.25E09
P0.50
IC0F0
0.44 9.8 0.031 68 5.27E07 3.50E09
P0.55
IC0F0
0.45 11.0 0.039 74 5.33E07 4.03E09
P0.60
IC0F0
0.46 12.3 0.033 78 5.27E07 4.16E09
P0.45
IC3F0
0.39 7.3 0.017 49 1.94E07 1.25E09
P0.45
IC24F0
0.34 3.0 0.019 26 2.78E08 2.36E10
P0.45
IC0
0.42 10.0 0.043 64 5.05E07 3.27E09
P0.45
IC0F3
0.41 11.2 0.036 65 5.19E07 2.64E09
P0.45
IC0
0.40 11.5 0.028 65 5.15E07 2.22E09
P0.45
IC0F6
0.38 11.3 0.087 66 5.11E07 2.08E09
3.5.1. Effect of sample size
Evaporation characteristics of samples with the two different thicknesses
and surface areas were studied. Two sample thicknesses of 11 mm and 21 mm in
73
addition to two exposed surface dimensions of 100 mm x 100 mm and 127 mm x
127 mm were used. Cumulative moisture loss and evaporation rates of these test
series are shown in Figure 3.5.1.1. As expected, increasing sample thickness and
surface area, increased the cumulative moisture loss at the end of the test. Results
indicate a 95% increase in the total moisture loss due to doubling of the thickness.
A 66% increase in the total moisture loss was observed as the surface area was
increased by 62%. Initial evaporation rates changed slightly from 0.42 kg/(m
2
.hr)
to 0.44 and 0.37 kg/(m
2
.hr) for thicker sample and sample with larger surface
area, respectively. Thicker sample has a slightly higher initial evaporation rate
(~4%) and if this is attributed to within sample variation, then the initial
evaporation is primarily a surface phenomenon. Larger sample however has a
lower evaporation rate by 12% during stage I. The difference may be attributed to
the side and edge effects, as in the vacuum evaporation the mass flux is only a
function of total pressure gradient and the water vapor diffusion coefficient is not
related to the size of evaporative surface [113, 114, 115]. Since control and
thicker samples have approximately the same initial evaporation rates as shown in
Table 3.5.1, critical moisture concentration reached later for thicker sample due to
higher initial moisture content. Therefore the transition time changes from 9.7 h
for control to 18.9 h for the thicker sample. This indicates that the critical
moisture concentration was reached after
74
0
40
80
120
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40 50
Drying Time, h
Surface Dimension
10 mm x 10 mm
12.7 mm x 12.7 mm
0
0.2
0.4
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
P0.45IC0F0
Thickness: 10 mm
Figure 3.5.1.1a
0
30
60
90
120
150
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40 50
Drying Time, h
Sample Thickness
12.7 mm
25.4 mm
0
0.2
0.4
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
P0.45IC0F0
Surface Dimension:
100 mm x 100 mm
Figure 3.5.1.1b
Figure 3.5.1.1. Effects of surface area and thickness on results of evaporation
tests on plain cement paste
75
losing 63%65% of the initial moisture content. Similarly, the transition time for
larger sample was obtained as 12.8 h which refers to the time when 68% of initial
moisture is lost. The values of diffusivity at stage I and stage II drying determined
for these three samples are shown in Table 3.5.1. Values of moisture diffusivities
at stage I drying is within range of 4.91x10
7
to 5.27x10
7
m
2
/s. Moisture
diffusivity at stage II is calculated as 3.33x10
9
m
2
/s for the control sample (P045
IC0F0), while for thicker (P045IC0F0Th21) and larger (P045IC0F0
SA0.016) is determined as 3.6x10
9
and 3.25x10
9
m
2
/s, respectively. Such similar
values for moisture diffusivity of Portland cement paste mixture validate the
range of accuracy of the analytical method.
3.5.2. Effect of w/c ratio
Figure 3.5.2.1a shows the cumulative moisture losstime curves for water
cement ratios of 0.4, 0.45, 0.5, and 0.6 after a 24 h drying period, which range
from 0 to 65, 68, 74 and 78 g respectively. As expected, the initial rate of
evaporation is rather the same for all samples, however higher w/c ratio
resulted in a higher cumulative moisture loss. An increase in w/c from 0.45 to
0.6 led to 20% increase in total moisture loss. On the other hand, initial drying
rate as shown in Figure 3.5.2.1a was slightly affected by increasing w/c ratio.
Average rates of evaporation at first drying stage were 0.42, 0.44, 0.45 and 0.46
kg/(m
2
.hr) for samples with w/c ratio of 0.45, 0.50, 0.55 and 0.60, respectively.
These magnitudes are sufficiently close to the rate of evaporation from water
surface (i.e. 0.45 kg/(m
2
.hr)) as the upper limit, hence increasing w/c ratio does
76
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 5 10 15 20 25
Drying Time, h
w/c
0.45
0.50
0.55
0.60
0
0.2
0.4
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
PxxxIC0F0
Figure 3.5.2.1a
4.4E007
5.2E007
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0.4 0.45 0.5 0.55 0.6 0.65
w/c
3.2E009
3.6E009
4E009
4.4E009
Stage II Drying
Stage I Drying
Figure 3.5.2.1b
Figure 3.5.2.1. Effects of w/c ratio on evaporation results of plain cement pastes
77
not significantly increase the initial evaporation rate. Rates of evaporation at
the second drying stage indicate that samples with higher w/c ratios have
consistently higher evaporation rates. After 24 h of drying, evaporation rates
dropped to 0.03 kg/(m
2
.hr) for all samples. But the transition time changes from
9.7 h for the series with w/c ratio of 0.45 to 9.8 h, 11 h, and 12.3 h for series with
w/c ratios of 0.5, 0.55 and 0.6, respectively. Diffusivity values determined for
different w/c ratios are shown in Figure 3.5.2.1b and indicate that moisture
diffusivities of different samples at stage I drying are very similar. However, at
stage II drying moisture diffusivities increased by 5%, 21% and 25% when w/c
ratio varied from 0.45 to 0.5, 0.55 and 0.6.
3.5.3. Effect of duration of initial curing
Effect of duration of curing prior to initiation of drying was studied by test
series P0.45IC0F0, P0.45IC3F0 and P0.45IC24F0. These samples had the
same mix proportions and differed only in duration of initial curing which ranged
from 0, to 3, and to 24 h. Figure 3.5.3.1 shows the results of the cumulative
moisture loss and evaporation rates versus time. The cumulative moisture loss
after 24 h of drying was substantially
reduced by increasing the duration of initial curing. This parameter fell from 65 g
for the noncured cement paste to 49 g and 26 g for cement pastes with initial
curing of 3 and 24 h, respectively translating into 24% and 60% reduction in
cumulative moisture loss of cement paste. This transition can only be justified by
the change of microstructure, pore size distributions and physical properties of the
78
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 5 10 15 20 25
Drying Time, h
Duration of Curing
0 h
3 h
24 h
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
P0.45ICxxF0
Figure 3.5.3.1a
0
2E007
4E007
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 5 10 15 20 25
Duration of Initial Curing, h
0
1E009
2E009
3E009
4E009
Stage II Drying
Stage I Drying
Figure 3.5.3.1b
Figure 3.5.3.1. Effect of curing duration on results of evaporation tests on plain
cement pastes
79
transition layer through which evaporation occurs [114]. Results of stage I drying
indicate considerable reduction in initial evaporation rates from a control of 0.42
to 0.39 and 0.34 kg/(m
2
.hr) for 3 and 24 h cured samples. However, the
reduction of drying rates in second stage of drying is more significant, i.e. after
12 h drying when all three test series were on stage II drying, evaporation rates
dropped from 0.3 kg/(m
2
hr) for control sample to 0.2 kg/(m
2
.hr) and 0.07
kg/(m
2
.hr) for 3 hcured and 24 hcured samples, respectively. The transition time
from stage I to stage II drying decreased from 9.7 h to 7.3 h for the 3 hcured
specimens and down to 3 h for the 24 hcured sample. The reduction is clearly
attributed to the reduction of moisture diffusivities at stage I by increasing curing
duration. As shown in Table 3.5.1, moisture diffusivity of control sample is 2.7
and 18.5 times than diffusion coefficients of 3h and 24hcured samples,
respectively. Increasing curing duration led to decreasing stage II diffusion
coefficient of 3 hcured and 24 hcured samples comparing to the control sample
by as much as 62% and 93%, respectively.
3.5.4. Effect of fiber content and cracking
Effect of fiber addition and its volume fraction was studied by utilizing
ARglass fiber in the mixture. The effect of fiber reinforcement is compared with
the plain cement paste in Figure 3.5.4.1. Fiber addition to the cement paste did not
result in any significant reduction in cumulative moisture loss at the end of the
test. This is in agreement with the result of Naaman et al. [47] which shows fibers
result in the reduction of the cumulative moisture loss by only 5% or less.
80
However, ARglass fibers caused slight reduction of evaporation rates at the first
drying stage. The initial evaporation rate dropped from 0.42 kg/(m
2
hr) for plain
sample to 0.42, 0.41, 0.40 and 0.38 kg/(m
2
.hr) associated with samples with 1.5,
3, 4.5 and 6 kg/m
3
ARglass fiber additions, respectively. A more significant
effect of fiber addition was observed in the manner of transition from the first
stage of drying to the second. Addition of fibers resulted in a gradual transition of
the modes of drying which could be attributed to the effect of fibers in controlling
earlyage cracks. As shown in the micrographs of Figure 3.5.4.2, fiber addition
reduced the maximum crack width and cracking area of the cement paste. Since
earlyage cracks cause the evaporative surface to be more exposed to the low
pressure condition in the desiccator, they may contribute to higher evaporation
rates than uncracked surfaces. The transition time from stage I to stage II drying
increased by the addition of fiber for all fiber reinforced paste samples to ranges
of 10 to 11.5 h compared to 9.7 h for the control sample.
According to Figure 3.5.4.1b, diffusivity results at stage I drying do not
vary significantly by the change of fiber dosage. Since fiber controls plastic
shrinkage cracking during early hours (see Figure 3.5.4.2), results support the
minimal effect of cracks on drying rate in stage I. However, results show a
decreasing trend for the diffusivities at stage II by increasing fiber dosage as
shown in Figure 3.5.4.1b. The diffusivities of fiber reinforced cement paste
samples with fiber content of 1.5, 3, 4.5 and 6 kg/m
3
are 2%, 20%, 33% and
38% less than corresponding value for plain control sample, respectively. This
81
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 5 10 15 20 25
Drying Time, h
0
1.5
3
4.5
6
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
P0.45IC0Fxx
Fiber
Content
(kg/m
3
)
Figure 3.5.4.1a
4.4E007
5.2E007
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 2 4 6
Fiber Content, kg/m
3
1.8E009
2.4E009
3E009
3.6E009
Stage II Drying
Stage I Drying
Figure 3.5.4.1b
Figure 3.5.4.1. Results of drying tests on cement paste with different content of
ARglass fibers
82
P0.45IC0F0
P0.45IC0F3 P0.45IC0F6
Figure 3.5.4.2. Crack pattern of cement paste specimens with and without AR
glass fibers after 24 hours of drying under lowpressure test condition
effect can be explained by the higher degree of surface cracking in the presence of
fibers which reduces the mean path length of moisture transfer to reach the
surface. The result can thus be interpreted as the effect of cracking on the drying
83
rate in stage II when unlike to stage I, the restraint of cracks by fibers, the
increased number of microcracks, and reduced crack widths are effective on
lowering moisture diffusivity. This explanation can be supported by the fact that
in stage II, drying happens under internal mass transfer control and the
microstructure plays a significant role on the drying process. As shown in Figure
3.5.4.2, 22% and 61% reduction in areal fraction of cracks were observed by
adding 3 and 6kg/m
3
fibers to cement paste, respectively. Also, maximum crack
widths of samples associated with 3 and 6kg/m
3
fibers were 47% and 71% less
than corresponding value for the control sample, respectively. The length of
microcracks however was increased by addition of fibers. Higher moisture
diffusivities at stage II in plain pastes with wider cracks compared to fiber
reinforced pastes with narrower but more plentiful cracks are in accordance with
the studies performed by Bažant and Raftshol [116] and Aldea et al. [117]. While
they found the moisture diffusivity and the water permeability are proportional to
the crack width cubed, results of another experimental study [118] indicated that
diffusivity of concrete increased 2.25 times by shrinkage cracks with a width of
0.1 mm and a spacing of 70 mm. Also, the recent experimental data reported by
Vejmelková et al. [ 119] show that moisture diffusivity of high performance
concrete and cement pastes increased by one order of magnitude due to the
cracking. In addition, Torrijos et al. [ 120 ] reported the increase of water
permeability by one order of magnitude due to increase of crack density from 0.25
to 0.45 cm/cm
2
. Considering all these results, it can be concluded that fibers
reduce moisture diffusivities of cracked samples at stage II by reducing crack
84
width and crack density which results in the durability improvement of cement
based materials.
3.6. Conclusion
A test method capable of characterizing evaporation parameters and
simulating sequential formation of shrinkage cracks in twodimensional samples
under lowpressure condition was developed. The cumulative moisture loss and
evaporation rates were calculated. Based on the results of experiments and
analyses in this study, following conclusions may be drawn:
(1) Drying of cement pastes can be segmented into two distinct stages:
constant drying rate period (stage I) and falling drying rate period
(stage II). During stage I drying, evaporation rate is constant, main
phase transitions take place at the surface and vapor phase diffusion
process is ratedetermining. At the transition time, moisture content at
the surface reaches a critical value. During stage II drying, phase
transitions take place within capillary pores and unsaturated capillary
flow is ratedetermining. During this stage, moisture content decreases
slowly until reaching the equilibrium value below which the material
cannot be dried.
(2) The cumulative moisture loss and evaporation rate curves can be used
to differentiate evaporation characteristics of different cement pastes.
Parameters including thickness and surface area of samples, w/c ratio,
85
and duration of initial curing have significant effects on evaporation
results.
(3) A dualstage drying model based on two different boundary conditions
at the top surface and a diffusional moisture transfer within the pores
can be used to simulate experimental results. The model can be used to
determine moisture diffusivity of the samples during drying process.
Moisture diffusivity is higher in stage I by more than one order of
magnitude than its values in stage II.
(4) Fibers contribute to the control of plastic shrinkage cracks. However,
presence of fibers and cracking control do not affect the drying rate in
stage I drying. In stage II drying, the presence of fibers reduces the
diffusivity values. This could be due to the restraint of cracks by fibers
which is effective in stage II when internal mass transfer controls the
drying. Results show that 22% and 61% reduction in areal fraction of
cracks were observed by adding 3 and 6kg/m
3
fibers to cement paste,
respectively. Also, maximum crack widths of samples associated with
3 and 6kg/m
3
fibers were 47% and 71% less than corresponding value
for the control sample, respectively. Since there is a cubic relationship
between permeability rate and the crack width, effect of fibers in
controlling the plastic crack width results in significant durability
enhancement.
86
(5) Among several different parameters investigated in this study, duration
of curing has the most significant effect on the reduction of drying rate
and moisture diffusivities.
87
4. Theory and Modeling of EarlyAge Drying
4.1. Introduction
Drying of cementitious materials is a preventable, but often inevitable
phenomenon during early age. With regard to longterm durability, shrinkage
cracking is the most anticipated consequence of drying due to low earlyage
strength of cement based materials. Therefore, understanding the physics of
drying and modeling the process is of great importance.
In existing literature, the moisture movement during drying of concrete is
often described as a diffusion process where the diffusivity depends highly on the
moisture content. Torrenti et al [55], West and Holmes [56], Garbalińska [57],
Huldén and Hansen [58], Kim and Lee [59] modeled the moisture movement by
the Fick’s second law of diffusion without explanation of the physics of the
problem. Most of these works refer to Bazant and Najjar [60, 61] who considered
the drying of concrete as a nonlinear diffusion problem. Using an approach based
on soil science, Kodikara and Chakrabarti [62] expressed the moisture movement
during drying as a two separate stage mechanism [63], however, they modeled the
phenomena as a onestage diffusion process. Chen and Mahadevan [64] referred
to Walton et al [65] in using diffusion as the dominant moisture transport
mechanism in concrete. Shimomura and Maekawa [50] used mass conservation of
the vapor and liquid water to derive a diffusiontype equation. Šelih and Bremner
[66] concluded from the experimental results that the diffusiondriven moisture
transport approach is appropriate only for the late stage of drying when the
material is predominantly unsaturated, and the moisture movement in the form of
88
vapor flux is dominant. Although Shimomura and Maekawa [50] discussed the
movement of liquid water as governing mechanism regarding moisture movement
near the saturated state, referring to Young [67], the concept was not implemented
in their model. Šelih and Bremner [66] expressed the drying process as two
different types of mechanisms of movement which can be described by a Darcy
type equation near saturation, succeeding by a diffusiontype equation.
Nonetheless, their work was only limited to experimental observations.
The evaporation through the surface as a boundary condition is applied
differently in available models. While ACI 305R99 [68] and Uno [69]
recommended a constant flux equal to evaporation rate of water surface for early
age hot weather concreting, a convective boundary condition proportional to the
difference of internal and ambient moisture concentration was most often
employed [50, 56, 58, 59]. Torrenti et al [55] and Shimomura and Maekawa [50]
also used evaporative boundary conditions similar to convection. In another
approach, Bazant and Najjar [60], and Garbalinska [57] applied a constant
moisture concentration equal to ambient moisture concentration as the boundary
condition in their models. However, the assumed boundary conditions are not
supported by the physics and driving forces of surface evaporation process.
In this chapter, physics of water evaporation is presented followed by a
description of theory of evaporation from cementitious matrix that is represented
as a capillary porous media. A dualstage methodology for modeling the drying is
introduced based on surface moisture transfer and internal moisture transport
characteristics. After studying the effect of different parameters on the model, the
89
methodology is applied to the results of drying experiments under low and normal
pressure conditions and is further verified by a Finite Element (FE) analysis.
4.2. Physics of water evaporation
Water evaporation is a phase transition process by which molecules are
converted from the liquid state into a vapor state. It usually occurs at the water
surface exposed to the atmosphere as the vapor is carried away by the air flow
[121, 91]. This process in the atmosphere arises from the action of molecular
diffusivity, since water vapor concentration at the water surface is higher than its
concentration away from the surface [88, 89, 90, 92, 93]. This concentration
gradient drives water vapor from higher concentration to lower. Vapor is
transferred through molecular exchange in the same manner as heat and
momentum are transferred, as most of transfer takes place within a few molecular
free path lengths of the surface [89]. According to Fick’s first law, the net flux of
a material (e.g. vapor) in any direction is proportional to its concentration gradient
in that direction and therefore the rate of evaporation at a horizontal surface is
given by [122]
w
w
dC
J D
dx
= − (4.2.1)
where J is the mass flux (kg/(m
2
s)) or drying rate in evaporation problems, D
w
is
the molecular diffusivity of water vapor in air (m
2
/s), C
w
is the water vapor
concentration (kg/m
3
) and x is diffusion distance (m). Assuming water vapor as an
ideal gas ( /
w w
p C RT M = ), evaporation rate can be described by:
90
w w
D M dp
J
RT dx
= − (4.2.2)
where p
w
is the water vapor pressure (Pa), M is the molar mass of water (Kg/mol),
R is the universal gas constant (m
3
Pa/mol.K) and T is the absolute temperature
(
o
K).
At the water surface of an open dish, the partial pressure of water vapor, p
w
, is
fixed by the saturated vapor pressure of water, p
w0
, since the air in direct contact
with the water is saturated [54]. Based on definition of fractional relative
humidity as H = p
w
/ p
w0
, local relative humidity at water surface is 1 (100%). If
relative humidity of the air well away from the surface is below 100%, water
vapor moves upwards from the surface and evaporation rate can be obtained as
0 w w
p D M dH
J
RT dx
= − ⋅ (4.2.3)
The approach based on equation (4.2.3) has been suggested in studying the
evaporation processes and determining diffusion coefficients of volatile liquids
diffused into air [123, 124, 125, 126].
In mass transfer problems, the air near the surface may be regarded as a
boundary layer, a concept set forth by Prandtl [127] for the momentum transport
in the neighborhood of a solid wall. Bringing the concept to the evaporation
problem, as shown in Figure 4.2.1, a thin interfacial layer covering the water
surface is considered as the boundary layer through which the vapor molecules
transfer to the surrounding air. The thickness of the boundary layer is of the order
of 1mm [128] and the movement through this boundary layer is presumed to be
governed by the molecular diffusivity [129].
91
Figure 4.2.1. Schematics of the interfacial boundary layer during evaporation
from free water surface
For simplicity we assume linear variation of vapor pressure across the boundary
layer which is of thickness δ, so that from equation (4.2.2) we have
0 w w w
D M p p
J
RT δ
∗
−
= (4.2.4)
where
0 w
p is the saturation vapor pressure and
* w
p is the ambient vapor pressure
and J is the mass flux or rate of evaporation (kg/m
2
s). Since H= p
w
/p
w0
, moisture
flux can be expressed as
0
(1 )
w w
D M p H
J
RT δ
−
= (4.2.5)
This equation shows that rate of evaporation is a function of relative humidity, H,
saturated water vapor pressure, P
w0
, and boundary layer thickness, δ. Since
saturated water vapor pressure is a function of temperature itself [ 130] and
boundary layer thickness is affected by the air flow, this equation reflects the
effects of relative humidity, temperature and wind velocity on the evaporation
rate. Following equation proposed by Wiederhold [130] relate saturation water
vapor pressure to the temperature.
92
17.502( 273.15)
8
32.18
0
611.21(1.0007 3.46 10 )
−
−
−
= + ×
T
T
w
p p e (4.2.6)
where p
w0
is saturated vapor pressure of water (Pa), p is the ambient pressure (Pa),
and T is the absolute temperature (K). The relationship between saturated water
vapor pressure and temperature is shown in Figure 4.2.2. Also, diffusion
coefficient of water vapor in air is a function of temperature which can be
obtained by a regression curve fit to data from Bolz and Tuve [131, 132].
6 8 10 2
2.775 10 (4.479 10 ) (1.656 10 )
w
D T T
− − −
= − × + × + × (4.2.7)
where D
w
is the diffusion coefficient of water vapor in the air (m
2
/s) and T is the
absolute temperature (
o
K). This relationship is shown in Figure 4.2.2 as well.
280 300 320 340 360 380
Temperature (K)
1.6
2.4
3.2
4
B
i
n
a
r
y
D
i
f
f
u
s
i
o
n
C
o
e
f
f
i
c
i
e
n
t
,
x
1
0

5
m
2
/
s
0
25
50
75
100
S
a
t
u
r
a
t
e
d
V
a
p
o
r
P
r
e
s
s
u
r
e
,
k
P
a
AirWater Vapor
(By Bolz and Tuve, 1976)
0 20 40 60 80 100
Temperature (
o
C)
Water
Figure 4.2.2. Variation of diffusion coefficient and saturated vapor pressure vs.
temperature
93
Having molar mass of water vapor as M=18.016 x 10
3
kg/mol and the universal
gas constant as R= 8.314472 m
3
Pa/mol.K and using equations (4.2.4) to (4.2.7),
effects of temperature, relative humidity and boundary layer thickness on rate of
evaporation is studied and shown in Figures 4.2.3a and 4.2.3b. As shown in the
figures, increasing ambient temperature leads to increase in the rate of
evaporation, which more significantly increases by lowering ambient relative
humidity and boundary layer thickness. Increasing boundary layer thickness
however results in decreasing the evaporation rate, e.g., in an ambient temperature
and relative humidity of 40
o
C and 50%, an increase in the boundary layer
thickness from 0.2 mm to 3 mm translate into a reduction of evaporation rate from
13.4 to 0.9 kg/(m
2
h). Similar to the effect of the boundary layer thickness,
increasing the ambient relative humidity cause a drastic reduction in the rate of
moisture evaporation. For instance assuming a boundary layer thickness of 0.7
mm, the evaporation rate at 40
o
C decreases from 6.8 to 0.7 kg/(m
2
h) when the
relative humidity increases from 10% to 90%.
4.3. Theory of evaporation from cementitious materials
Moisture evaporation of Portland cement paste and concrete can be
explained by the concept of moisture transfer through porous media using an
isothermal drying model. Since drying of porous media includes complex
moisture transfer in both liquid and vapor states, the driving force of vapor
pressure gradient which is the case in water vapor evaporation cannot be used
directly here. Instead, a general concept of hydraulic potential ψ, applicable to
94
0 10 20 30 40 50
Temperature (
o
C)
0
4
8
12
16
20
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
(
m
2
.
h
)
RH=50%
3mm
0.7mm
δ = 0.2mm
0.4mm
1mm
(a)
0 10 20 30 40 50
Temperature (
o
C)
0
2
4
6
8
10
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
(
m
2
.
h
)
δ = 0.7mm
RH = 10%
20%
30%
40%
50%
60%
70%
80%
90%
(b)
Figure 4.2.3. Parametric study on the effects of temperature, boundary layer
thickness and relative humidity on evaporation rate from water surface
95
both vapor phase transfer and water liquid transfer through unsaturated porous
medium is used. The concept comes from unsaturated flow which is governed by
extended Darcy law, where unsaturated flow rate, u (m/s) is related to the
hydraulic potential ψ (m) by a liquid conductivity coefficient K (m/s) [133, 134,
135].
= − ∇ u Κ ψ (4.3.1)
Hydraulic potential is indeed the energy required to transfer unit weight of liquid
from the porous materials to a reservoir of the same liquid at the same
temperature and elevation, and hence it can be called capillary potential as well.
In this approach u is the massaveraged velocity which stands for the water in
both liquid and vapor states. The drying of porous media has been recently
analyzed with multiphase approaches considering the moisture transfer through
porous space partially saturated by liquid water, water vapor and dry air [136,
137 , 138 , 139 ]. However, it has been shown that by introducing certain
assumptions above formulation in terms of one single driving force can be
retrieved [ 140, 141, 142]. The validity of this hypothesis was discussed by
Mainguy et al. [139]. Equation (4.3.1) together with continuity equation leads to
fundamental Richards equation as [135, 143, 144]
( ) Κ
t
θ
ψ
∂
= ∇⋅ ∇
∂
(4.3.2)
where θ is the volume fraction moisture content (), and t is the time (s).
Note that in this approach of drying, ψ depends on only one state variable,
namely θ, as well as K which also depends explicitly on θ . Equation (4.3.2)
96
shows that gradient of hydraulic or capillary potential is the main cause of
moisture migration through capillary media [104]. Kelvin equation relates
capillary potential to capillary pressure or vapor pressure by:
0
ln =
ws
w
P RT
M P
ψ (4.3.3)
This relationship shows that unsaturated flow in porous media can be
expressed either by hydraulic potential gradient or by vapor pressure gradient.
Accordingly, Coussot [145] discussed the cause of moisture motion through the
porous network by the difference in capillary pressure between the top and the
bottom of the sample. In this study, variation of both parameters as the driving
force from the interior of material through ambient with no discontinuity is used
to explain the drying process. Nonetheless, the focus of the work is on moisture
loss analysis and therefore to calculate moisture content, ψ is replaced by θ in
equation (4.3.2) using the definition of capillary diffusivity as ( ) = D Κ d d ψ θ ,
and we have
( ) D
t
θ
θ
∂
= ∇⋅ ∇
∂
(4.3.4)
where D is the moisture (liquid and vapor) diffusion coefficient (m
2
/s). Equation
(4.3.4) can also be derived from the mass conservation and transport laws for
vapor and liquid water in concrete [50, 141]. This equation turns out to be a non
linear diffusion equation, with D depending on θ.
Studies of drying behavior of clay brick ceramics by Hall et al. [99],
Platten [146], Cooling [100], and different wet capillaryporous materials by
97
Kowalski [147] as shown in Figure 4.3.1 indicate that evaporative flux at the
boundary is roughly constant for a period of time at the beginning of the drying
referred to as stage I and then falls significantly by the time in the second period
or stage II. The same drying behavior was observed for the Portland cement
pastes in recent experimental works by authors [148], as well as in mortar [149]
and lightweight concrete [66]. These two stages of drying based on variation of
driving forces and degree of continuity between liquid and vapor phases are
discussed individually in the following sections.
Figure 4.3.1. Two stages of drying during evaporation of capillary porous
materials
98
4.3.1. Stage I Drying
Mass loss experiments of cement paste subjected to low pressure has been
recently conducted [148]. Results show that the external evaporation potential is
independent of internal capillary microstructure, and this aspects controls the
moisture transport within the material in stage I drying. The phenomenon can be
described by the variation of hydraulic potential or vapor pressure and the degree
of continuity between liquid and vapor phases. As shown in Figure 4.3.1.1a at the
beginning of drying, the porous media (i.e. cementitious materials) are fully
saturated and therefore, the vapor pressure inside the open cell microstructure is
equal to the saturated vapor pressure. There is no vapor pressure or hydraulic
potential gradient to drive the moisture inside the porous body, but the gradient of
vapor pressure at the boundary layer, which varies from saturated vapor pressure
just on top of the material to the ambient vapor pressure, drives the moisture to
the ambient. In analogy to vapor pressure, the hydraulic potential inside the
porous body is also uniform and surface hydraulic potential is equal to initial
value, =
s i
ψ ψ , which for a saturated material is 0 = ψ , referring to equation
(4.3.3) and varies to
*
ψ as the ambient hydraulic potential. Variation of hydraulic
potential and vapor pressure from the interior of material through ambient is
depicted in Figure 4.3.1.1c.
A drying begins, the moisture evaporates at the surface and the surface
vapor pressure falls slightly from the initial value. Although the vapor pressure is
still approximately equal to the saturated vapor pressure, ≈
s i
p p , but the
moisture concentration and hydraulic diffusivity are sufficiently high that an
99
adequate capillary flow to the surface can be generated by small value of vapor
pressure gradient or water content gradient [54]. Continuation of drying leads to
continuous decrease of moisture concentration deep inside the material. The
material is not fully saturated as shown in Figure 4.3.1.1b but the liquid phase is
still continuous, and therefore, the porous material remains capillary saturated.
The drying is in stage I while the material is capillary saturated, vapor pressure is
approximately equal to saturated vapor pressure, and the profile of internal water
concentration remains flat [54, 105]. This drying scenario is referred to as the
funicular drying regime [150].
(a) (b) (c)
Figure 4.3.1.1. (a) Fully saturation state of porous material as the initial
condition, (b) capillary saturation of porous material as a later stage of drying, (c)
variation of hydraulic potential and vapor pressure from the interior of the
material through ambient in stage I drying.
100
4.3.2. Stage II Drying
As drying continues, the moisture concentration in liquid phase decreases
and reaches the insular saturation which is associated with the transition time
from stage I to stage II, when the liquid phase becomes discontinuous [105]. As
shown in Figure 4.3.2.1a, at this stage of drying, the diffusion of water vapor
through continuous vapor space becomes the dominant mechanism for moisture
transport [105]. Since moisture transfer is limited only to the diffusion of water
vapor generated by local evaporation of bound liquid water, and not liquid water
movement, the evaporation rate falls and drying happens under internal mass
transfer control [106]. From the driving force aspect, at this stage, the saturation
level is lower than capillary saturation and therefore, the vapor pressure falls from
saturated vapor pressure to a lower value. During this period, the evaporation
continues until vapor pressure reaches ambient vapor pressure or hydraulic
potential balance created which can be referred to as hygroscopic equilibrium. At
this stage, two different theories can be suggested for the variation of hydraulic
potential or vapor pressure at the surface. As shown in Figure 4.3.2.1b, based on
first theory designated by stage II(A), water vapor pressure can fall almost to the
ambient vapor pressure instantaneously to reach hygral equilibrium with the
environment. On the other hand, the second theory suggests a gradual decrease for
the water vapor pressure from the saturated pressure to the ambient vapor pressure
as shown in Figure 4.3.2.1c and designated by stage II (B) modeling. The former
condition in terms of moisture concentration can be expressed as equal moisture
concentration between surface and ambient levels [148]. The latter condition is
101
tantamount to imposing a convectivediffusive transfer of vapor at the surface for
modeling stage II drying.
4.4. TwoStage Modeling of the Drying of Cementitious Materials
4.4.1. Governing Equations and Geometry of the Problem
A model is developed to address the moisture transport properties of
cement pastes during drying tests under low pressure condition [148]. The
moisture transport through the sample ( 0 ≤ ≤ x L ) is assumed to be one
directional (1D) because of high surface areatodepth ratio, and exposing only
top surface to the drying condition.
(a) (b) (c)
Figure 4.3.2.1. (a) Low saturation state of porous material as the advanced stage
of drying, (b) variation of hydraulic potential and vapor pressure from the interior
of the material through ambient in stage II using simplified model (model A), (c)
variation of hydraulic potential and vapor pressure from the interior of the
material through ambient in stage II using convection model (model B)
102
By assuming an isothermal condition, effects of change in temperature
during drying process is negligible. For modeling the internal moisture transport
in both stage I and stage II drying, 1D form of equation (4.3.4) is used with the
replacement of dimensionless moisture concentration variable, θ (), with the
dimensional variable C (kg/m
3
) [55, 61, 63, 110, 151].
2
2
2
( )
C C dD C C
D C D
t x x dC x x
∂ ∂ ∂ ∂ ∂ ⎡ ⎤ ⎛ ⎞
= = +
⎜ ⎟
⎢ ⎥
∂ ∂ ∂ ∂ ∂
⎣ ⎦ ⎝ ⎠
(4.4.1.1)
The only difference in applying this formula in stage I and II, is the assumption of
constant diffusivity at stage I, while diffusivity can be variable at stage II drying.
The assumption is based on Garrabrants and Kosson’s experimental observations
[105] which support relatively constant moisture diffusivity in the early stage of
drying, followed by variable moisture diffusivity at later stages of drying. While
the first term in the right hand side of equation (4.4.1.1) is zero in the stage I
drying, it is nonzero at stage II [141], although it looks very small considering
higher order terms. For simplification, / dD dC can be assumed to be zero during
stage II as well. However, since
D dD C
t dC t
∂ ∂
=
∂ ∂
(4.4.1.2)
the assumption would lead to 0
D
t
∂
=
∂
, which contradicts the experimental
observations. To take care of this problem, the time of drying was divided into
several time steps at stage II and constant diffusivity was applied at each time step
with respect to the time and the location. Therefore, the timevarying diffusivity
values at stage II are only averaged over the space. This can be acceptable
103
considering high evaporative surfacetodepth ratio of the samples. Consequently,
equation (4.4.1.1) can be simplified to equation (4.4.1.3) for stage I as well as
each time step for stage II.
2
2
C C
D
t x
∂ ∂
=
∂ ∂
(4.4.1.3)
Hydration of cement and its implications on transport properties need to
be taken into account during earlyage drying of cementitious materials. With
time, liquid water will be chemically bounded into hydration products as well as
physically bound water to increasing internal solid surfaces. Both chemically and
physically bound water may be assumed to be in local equilibrium with free water
due to rapid rate of hydration at the early age. To simplify the concept, the
concentration of bound water, S, is directly taken proportional to the
concentration of the free water, C, the substance free to diffuse.
S = RC (4.4.1.4)
During earlyage drying when the moisture diffusion is accompanied by
immobilization of bound water, equation (4.4.1.3) is modified as:
2
2
C C S
D
t x t
∂ ∂ ∂
= −
∂ ∂ ∂
(4.4.1.5)
Equation (4.4.1.5) can be presented in a new form by substituting for S from
equation (4.4.1.4).
2 2
2 2
1
C D C C
D
t R x x
∂ ∂ ∂
′ = =
∂ + ∂ ∂
(4.4.1.6)
Replacing the term D/(1+R) with D’ as the effective moisture diffusion
coefficient, the equation is seen in the usual form of diffusion equations.
104
Therefore, D’ presents the potential for conductivity of free water through the
fresh cementitious materials.
The cement paste is assumed to be homogeneous at initial conditions with
an impermeable bottom surface. Therefore, the initial and boundary conditions are
0, 0
i
t C C x L = = ≤ ≤ (4.4.1.7)
0 at 0
C
J D x
x
∂
′ = − = =
∂
(4.4.1.8)
where C
i
is the initial moisture concentration (kg/m
3
), and J is the moisture flux
(kg/m
2
).
Moisture transport during stage I drying is considered as an external mass transfer
control process with a constant drying rate. Thus, the boundary condition at the
top surface is
0
at J F x L = = (4.4.1.9)
where, F
0
is the constant flux in stage I drying (kg/m
2
s). Solution to moisture
concentration values as a function of time and location during stage I drying
based on equation (4.4.1.64.4.1.9) is [109]
( )
2 2 2 2
0
2 2 2 2 2
1
3 2 ( 1)
, exp( ) cos
6
n
i
n
F L D t x L D n t n x
C t x C
D L
L L n L
π π
π
∞
=
⎫
′ ′ − − − ⎪ ⎧
= + + −
⎨ ⎬
′
⎩
⎪
⎭
∑
(4.4.1.10)
In this model transition time from stage I to stage II, referring to the loss
of continuity in the liquid phase, is defined by the user which may be obtained
from experimental data. Prediction of the transition time based on physical
grounds requires further studies. For modeling stage II drying based on the
105
concept of convectivediffusive transfer of vapor at the surface, the exchanged
flux of moisture, J, can be characterized by the convection equation [109, 152].
*
( )
s
C
J D k C C
x
∂
′ = − = −
∂
(4.4.1.11)
where k (m/h) is convective moisture transfer coefficient or surface factor, C
s
is
the surface moisture concentration and C
*
is the ambient moisture concentration
(kg/m
3
). k parameter takes into account the wind velocity, the surface
temperature, and the relative humidity for drying to the atmosphere or in the case
of lowpressure drying encompasses vacuum condition and temperature. The
analytical solution for equations (4.4.1.64.4.1.9) during each time step at stage II
drying with the convective boundary condition as of equation (4.4.1.11) is given
by [153]
( ) ( )
2
2
1
2sin( )
, cos( )
sin( )cos( )
D t
n
n L
i n
n n n
n
x
C t x C C C e
L
λ
λ
λ
λ λ λ
⎛ ⎞ ′
−
∞ ⎜ ⎟
⎝ ⎠
∗ ∗
=
= + − +
+
∑
(4.4.1.12)
Where,
tan( )
n n
kL
D
λ λ =
′
(4.4.1.13)
Geometry, governing equations and boundary conditions during stages I and II
drying are presented in Figure (4.3.2.2).
106
Figure 4.3.2.2a Figure 4.3.2.2b
Figure 4.3.2.2. Geometry, governing equations and boundary conditions during:
(a) stage I drying, (b) stage II drying assuming convective moisture flow on the
top surface
4.4.2. Simulation and Parametric Studies
Modeling of the drying process is presented based on dualstage moisture
movement through the porous media and imposing a constant flux and convective
boundary conditions at stage I and II, respectively. Model simulates the
evaporation process using equation (4.4.1.10) as the solution for stage I drying,
and equation (4.4.1.12) as the solution for each time step at stage II drying. In the
code, an extra function is developed to find sufficient number of solutions of λ
n
in
equation (4.4.1.13) and return them to the main code. The code calculates
moisture concentration profile over the volume. The difference between
integration of moisture concentration and initial total moisture yields the
107
cumulative moisture loss, a variable that is measured during a drying test. The
code calculates flux at the surface according to equation (4.4.1.11).
Parametric studies are performed to verify capability of the model in
capturing effects of major moisture transport parameters, including effective
diffusivity, D’, and surface factor, k. Parametric studies also include effects of
initial external rate of evaporation, F
0
, transition time, t
trans
, and sample thickness,
L, on the cumulative moisture loss and evaporation rates. Throughout the
parametric study, a cement paste with w/c ratio of 0.45 is considered as the drying
material. A drying surface area of 100 mm x 100 mm and thickness of 10 mm are
used as the size of specimen, except for study of the depth in which the thickness
is variable.
As expected, at stage I drying, diffusivity does not control cumulative
moisture loss and evaporation rates due to the imposed boundary condition at this
stage, but it controls moisture concentration profiles. Effect of diffusivity at stage
I, D’
I
, on moisture concentration profiles are studied at 1 h intervals with three
orders of magnitude in diffusion coefficients of 5x10
9
, 5x10
8
and 5x10
7
m
2
/s as
shown in Figure 4.4.2.1, assuming an initial evaporation rate of 0.4 kg/(m
2
.h) and
a transition time of 10 h. As shown in Figure 4.4.2.1a and 4.4.2.1b, moisture
concentrations reach negative values or large moisture concentration gradients
show up imposing 0.4 kg/(m
2
.h) as the initial evaporation rate. It can be
concluded that assumed diffusivities are too low to balance the flux on the top
surface. Although in gravimetrical methods, moisture concentration profiles are
not obtained, but funicular flow, the natural internal moisture flow in stage I,
108
leads to values of initial diffusivity of the order of 5x10
7
m
2
/s which satisfy the
flux with small values of internal moisture concentration gradients. In Figure
4.4.2.2a effect of diffusivity at stage II drying, D’
II
, on cumulative moisture loss
and evaporation rates is studied with five different diffusivity values increasing
from 5x10
10
m
2
/s to 5x10
8
m
2
/s, assuming D’
I
as 5x10
7
m
2
/s, k as 0.001 m/h, F
0
as 0.4 kg/(m
2
.h) and t
trans
as 5 h. These results reflect increases in the cumulative
moisture loss from 44 g to 63 g translating into a 28% increase in total moisture
loss after 40 h of drying, and also significant increase in evaporation rates after 10
h of drying as shown in Figure 4.4.2.2a. The results of simulation are in
agreement with the concept of diffusivity which is moisture transport conductivity
through capillary medium under driving force of a concentration gradient. As
shown in the figure, an increase in moisture diffusivity translates into higher
evaporation rate and higher cumulative moisture loss at stage II drying.
Effect of convective moisture transfer coefficient, k, is studied with
increasing this parameter from 0.0005 m/h to 0.002 m/h assuming D’
I
as 5x10
7
m
2
/s, D’
II
as 1.5x10
8
m
2
/s, F
0
as 0.5 kg/(m
2
.h) and t
trans
as 5 h. As shown in
Figure 4.4.2.2b, an increase in cumulative moisture loss from 55 g to 64 g,
equivalent to an increase by 13% in total moisture loss, is simulated after 40 h of
drying. A significant increase in evaporation rates at early hours of stage II drying
is also simulated by increasing this parameter. A much smoother fall in
evaporation rates are observed at the beginning of stage II drying by assuming
109
1200 800 400 0 400 800
Moisture Concentration, kg/m
3
0
0.2
0.4
0.6
0.8
1
t
/
h
D
I
= 5x10
9
m
2
/s
t=1 2
t = 10 h
t
h
h = 10 mm
0 200 400 600 800
Moisture Concentration, kg/m
3
0
0.2
0.4
0.6
0.8
1
t
/
h
D
I
=
5x10
8
m
2
/s
t=1 2
t = 10 h
t
h
h = 10 mm
Figure 4.4.2.1a Figure 4.4.2.1b
0 200 400 600 800
Moisture Concentration, kg/m
3
0
0.2
0.4
0.6
0.8
1
t
/
h D
I
=
5x10
7
m
2
/s
h = 10 mm
t=1 2 t = 10 h
t
h
Figure 4.4.2.1c
Figure 4.4.2.1. Parametric study on the effect of moisture diffusivity at stage I on
profiles of moisture concentration drawn at 1 h intervals during drying of cement
paste samples
110
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
5.0x10
8
1.5x10
8
5.0x10
9
1.5x10
9
5.0x10
10
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
F
0
= 0.4 kg/(m
2
.h)
t
trans
= 5 h
D
I
= 5x10
7
m
2
/s
k = 0.001 m/h
D
II
(m
2
/s)
Figure 4.4.2.2a
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
0.002
0.0015
0.001
0.0005
0
0.2
0.4
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
t
trans
= 5 h
D
I
= 5.0x10
7
m
2
/s
D
II
= 1.5x10
8
m
2
/s
F
0
= 0.5 kg/(m
2
.h)
k
(m/h)
Figure 4.4.2.2b
Figure 4.4.2.2. Parametric study of moisture diffusivity at stage II drying and
convective moisture transfer coefficient on cumulative moisture loss and
evaporation rates of cement paste samples
111
higher values of k such as 0.002 m/h, while low k values such as 0.0005 m/h
results in a marked drop in evaporation rates at the beginning of stage II.
Figure 4.4.2.3a and Figure 4.4.2.3b show parametric studies on initial
external evaporation rate, F
0
, and time of transition from stage I to stage II, t
trans
,
respectively. In both simulations, D’
I
, D’
II
, and k are given as 5x10
7
m
2
/s, 1.5x10

8
m
2
/s, and 0.001 m/h, respectively. As expected, increasing F
0
from 0.3 to 0.5
kg/(m
2
.h) reflects in higher cumulative moisture loss and evaporation rates at
early hours of drying, but has minimal effects on cumulative moisture loss after
24 h. However, at stage II drying, lower evaporation rates along with a large fall
from initial values are simulated by increasing F
0
, which is due to lower surface
moisture concentration.
Effect of transition time is studied by five different values from 6 to 14 h.
As shown in Figure 4.4.2.3b, higher values result in longer period of stage I and
more significant fall in evaporation rate between stage I and II, which is expected
due to lower surface moisture concentrations.
The specimen thickness is the last parameter studied by increasing from 5
to 25 mm in five intervals, assuming D’
I
, D’
II
, k, F
0
and t
trans
are given as 5x10
7
m
2
/s, 1.5x10
8
m
2
/s, 0.001 m/h, 0.5 kg/(m
2
.h) and 5 h, respectively. As shown in
Figure 4.4.2.4, cumulative moisture loss after 40 h drying increases from 32 to 92
g and also evaporation rates at the beginning of stage II drying increase by
increasing sample thickness, make drying stage transition smoother.
112
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
0.50
0.45
0.40
0.35
0.30
0
0.2
0.4
0.6
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
F
0
(kg/(m
2
.h))
D
I
= 5x10
7
m
2
/s
D
II
= 1x10
8
m
2
/s
k = 0.001 m/h
t
trans
= 10 h
Figure 4.4.2.3a
0
25
50
75
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
14
12
10
8
6
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
F
0
= 0.5 kg/(m
2
.h)
D
I
= 5.0x10
7
m
2
/s
D
II
= 1.5x10
8
m
2
/s
k
= 0.001 m/h
t
trans
, h
Figure 4.4.2.3b
Figure 4.4.2.3. Parametric study on the effect of initial evaporation rate, F0, and
transition time, t
trans
, on cumulative moisture loss and evaporation rates during
drying of cement paste samples
113
0
25
50
75
100
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
2.5
2.0
1.5
1.0
0.5
0
0.25
0.5
0.75
1
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
F
0
= 0.5 kg/(m
2
.h)
t
trans
= 5 h
D
I
= 5.0x10
7
m
2
/s
D
II
= 1.5x10
8
m
2
/s
k
= 0.001 m/h
L (cm)
Figure 4.4.2.4. Parametric study on the effect of moisture transfer coefficient, k,
on cumulative moisture loss and evaporation rates during drying of cement pastes
4.4.3. BackCalculation Procedures for Modeling of Stage II Drying
A back calculation model is used to predict diffusivity and moisture
transfer coefficient at any time step based on given cumulative moisture loss and
evaporation rate curves. Optimization is implemented to find two best fit
parameters, D’(t) and k(t) to predict the experimental results. The objective
function for minimization is the weighted sum of two errors of cumulative
moisture loss and flux. In optimization algorithm, normalized design variables (D’
and k) are used to avoid numerical problems caused by mixing big and small
numbers in numerical operations. Thus, diffusion coefficient at the current time
step is normalized with its value at previous time step, D
norm
=D’
i
/D’
i1
, so does
the mass transfer coefficient k
norm
= k
i
/k
i1
. Once the optimal solution D
norm
and
114
k
norm
are found, they are reverted back to D’ and k. Inequality constraints are
imposed to the design variables D’ and k in terms of lower and upper bounds.
Initial values for D’ are obtained from analysis of stage I drying, while initial
values for k are estimated from equation (4.4.3.1) which is derived by rearranging
equation (4.4.1.11).
(exp)
( )
*
( )
est
trans
crit
J t t
k
C C
=
=
−
(4.4.3.1)
where, k
(est)
is the estimated surface factor (m/h) and
(exp)
( ) =
trans
J t t is the
experimental rate of evaporation at the transition time which is roughly equal to
initial evaporation rate. C
crit
is the surface moisture concentration at end of stage I
or critical moisture concentration (kg/m
3
) and C
*
is the ambient moisture
concentration. For other time steps, initial D’ values for optimization process,
( ) est
D are estimated by the assumption of proportional decrease of diffusivity
with the flux, while initial k values, k
(est)
are estimated by a convection transport
formula.
(exp) ( )
( )
1
(exp) ( 1)
( )
( )
−
−
=
i
est
i
i
J t
D D
J t
(4.4.3.3)
(exp) ( )
( )
( 1)
*
( )
( , )
i
est
i
J t
k
C t x L C
−
=
= −
(4.4.3.3)
In equation (4.4.3.1) and equation (4.4.3.3),
( ) i
t is the current time step, while
( 1) − i
t is the previous time step. At each time step, these two initially estimated
parameters, D
(est)
and k
(est)
are passed to the optimizer and the optimizer calls the
function evaluation to calculate the error and update these two parameters several
115
times until optimal solution for D’ and k are found. Results are discussed in the
following section and shown in Figures 4.4.3.1a and 4.4.3.1b.
4.5. Comparison with Experimental Data and Finite Element (FE) Analysis
An experimental procedure for drying under lowpressure conditions was
developed by authors that is capable of characterizing evaporation parameters
using measurement of mass change during drying process [148]. Experimental
data including cumulative moisture loss and evaporation rates of a drying
Portland cement paste with w/c ratio of 0.45 are used in the backcalculation
procedures. Results of prediction of experimental data and the best solutions for
D’ and k parameters are shown in Figure 4.4.3.1a and 4.4.3.1b. The model fits
cumulative moisture loss and evaporation rates accurately throughout the drying
period, while diffusion and moisture transfer coefficients are constant through the
initial stage of drying and variable during stage II. As shown in Figure 4.4.3.1b,
diffusivity fell almost at the beginning of stage II drying, but the moisture transfer
coefficient remains constant even in the first 7 h of stage II before it gradually
reduced from 0.0017 m/h to 0.0011 m/h. The reduction of moisture diffusivity at
stage II drying is much more significant when this parameter reduced from initial
value of 5.15x10
7
m
2
/s to 1.18x10
8
m
2
/s. This is in agreement with the literature
where diffusivity after being relatively constant during early stages of drying, falls
by decreasing moisture content below midrange saturation [105].
116
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
Experiment
Analytical data
FE Analysis
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
w/c = 0.45
No curing
T = 24
o
C
Figure 4.4.3.1a
2x10
7
4x10
7
6x10
7
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 10 20 30 40
Drying Time, h
K
D'
0.001
0.0015
0.002
M
o
i
s
t
u
r
e
T
r
a
n
s
f
e
r
C
o
e
f
f
i
c
i
e
n
t
,
m
/
h
6x10
9
Figure 4.4.3.1b
Figure 4.4.3.1. (a) Prediction of cumulative moisture loss and evaporation rate as
a function of time for the experimental data of an evaporation test, (b) Back
calculated D and k for the best fit of data
117
From the physical point of view, time is not the natural parameter for
characterization of moisture diffusivity, hence it was related to the average
moisture concentration to obtain an intrinsic characterization. This relationship is
presented in Figure 4.5.1 which can be fitted by a model similar to CEBFIP
[154]. While in CEBFIP model, the moisture diffusivity is related to the relative
humidity [56, 155], in this work the moisture diffusivity is modeled versus
average normalized moisture concentration.
1
1
( )
1
1
1
norm n
norm
c
D C D
C
C
α
α
⎛ ⎞
⎜ ⎟
⎜ ⎟
−
= +
⎜ ⎟
⎛ ⎞
⎜ ⎟
−
+
⎜ ⎟
⎜ ⎟
⎜ ⎟ ⎜ ⎟
−
⎝ ⎠ ⎝ ⎠
(4.5.1)
where, D
1
represents D(C
norm
) when the samples is initially fully saturated and
normalized moisture concentration is 1, D
0
is the minimum D(C
norm
) and α is the
parameter that represents the ratio D
0
/D
1
. In this equation, C
c
is the normalized
moisture concentration at D(C
norm
) = 0.5D
1
characterizing the location of the drop
in the curve D(C
norm
). According to Bazant and Najjar [60], Baluch et al [156] and
Kim and Lee [155], calculated moisture diffusivity for wellhydrated and cured
concrete samples is within range of 3x10
10
to 5.4x10
8
m
2
/s. Comparing the
values with the results of this analysis which ranges from 1.86x10
8
to 5.15x10
7
m
2
/s, it may be concluded that very earlyage moisture diffusivity of a poorly
cured cement paste is more than tenfold of the moisture diffusivity of a cured
concrete. However, the results are well within the range of the data recently
reported by Vejmelková et al. [119] which range from 1.78x10
8
to 2.53x10
8
m
2
/s
118
for noncracked and 5.46x10
7
to 2.00x10
7
m
2
/s for cracked cement paste
samples. On the other hand, the moisture diffusivity ratio, α, was obtained as
0.036 which is in accordance with the results of Bazant and Najjar [60], who
reported values in the range of 0.025 to 0.10 for different types of concrete.
0 0.2 0.4 0.6 0.8 1
Normalized Moisture Concentration
1.0E008
1.1E007
2.1E007
3.1E007
4.1E007
5.1E007
6.1E007
7.1E007
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
CEBFIP: n = 10
CEBFIP: n = 15
CEBFIP: n = 20
CEBFIP: n = 25
BackCalculated Diffusivities
D
1
= 5.15 x 10
7
m
2
/s
D
0
= 1.86 x 10
8
m
2
/s
α = D
0
/D
1
= 0.03
C
c
= 0.2
Figure 4.5.1. Backcalculated moisture diffusivity of an earlyage noncured
cement paste as a function of normalized moisture concentration expressed by
CEBFIP models
As shown in Figure 4.5.1, among different values for variable n, n = 15 gives the
best fit to the calculated diffusivity results. On the other hand, the calculated
values of parameter k, the convective moisture transfer coefficient, are within the
range of the data reported in the literature [156, 157] varying from 0.0003 to
0.003 m/h.
A Finite Element (FE) analysis was carried out using ABAQUS v.6.101
[70] to verify the results of proposed analytical method including the flux at the
119
top surface and the integration of the surface flux or cumulative moisture loss.
The technique utilizes the analogy between the heat transfer and the moisture
diffusion. In equivalent transient thermal problem in 2D space, the corresponding
equations are
2 2
2 2
T k T T
t c x y ρ
⎡ ⎤ ⎛ ⎞ ∂ ∂ ∂
= +
⎜ ⎟ ⎢ ⎥
∂ ∂ ∂
⎝ ⎠ ⎣ ⎦
(4.5.2)
*
( )
fT s
T
h T T
n
∂
= −
∂
(4.5.3)
whrere, T(x,y,t) is the temperature varying in the domain with the time. k is the
isotropic thermal conductivity which may be a function of T, and ρ and c are the
density and the specific heat, respectively.
T
n
∂
∂
is the thermal gradient with a unit
normal “n”,
fT
h is the convective transfer coefficient for thermal problem, T
s
is
the temperature at the boundary, and T
*
is the ambient temperature. While
equation (4.5.3) is for the heat convection at the boundary, analogous to moisture
convection during stage II at the boundary, boundary conditions for stage I can be
introduced as a constant heat flux. Comparing equations (4.5.2 and 4.5.3) with
equations (4.4.1.6) and (4.4.1.11) leads to the conclusion of onetoone analogy
between these two boundary value problems where C, D, and k in the diffusion
problem are corresponding to T, k/ρc and h
fT
in the heat transfer problem. In this
analysis, the boundary condition of stage I drying is defined as a load at surface
with the magnitude of initial evaporation rate. The other boundary conditions
which are impermeable surfaces on the sides and bottom of the sample are
ignored, since no load corresponds to zero flux or impermeability. Also an
120
interaction was created for the purpose of imposing a convective boundary
condition during stage II drying. For the sake of comparison with the results of
analytical method, the same number of layers (100 layers) is used along y
direction which is along the direction of external flux. Since there is no heat
transfer corresponding to moisture transfer along the xdirection which is
perpendicular to the direction of external flux, the domain is divided by only 20
layers. The type of analysis was an implicit heat transfer analysis, and since there
was no interest on the edge effects, a uniform mesh with 4node linear heat
transfer quadrilateral elements (DC2D4) was used. Results of FE analysis for the
flux at different drying times are shown in Figure 4.5.2. The comparison between
results of the surface flux by FE analysis and the analytical and experimental
results are shown in Figure 4.4.3.1a. Cumulative moisture loss can also be
obtained as the product of the integration of the flux obtained from FE analysis
and the surface area. As shown in Figure 4.4.3.1a, the FE results show a good
agreement with analytical and experimental data.
In order to further confirm the soundness of this analytical method, the
method can be applied to different sets of drying data with available experimental
moisture profiles. The experimental data of Šelih et al. [66, 158, 159] were chosen
for this purpose which involve onedimensional drying of concrete cylinders
under normal room pressure, temperature of 22
o
C and relative humidity of 50%.
Among different drying tests
121
(a) t = 2 h
(b) t = 10 h
(c) t = 14 h
(d) t = 20 h
(e) t = 40 h
Figure 4.5.2. Distribution of moisture flux (kg/(m2h)) at: a) t = 2 h, b) t = 10 h, c)
t = 14 h, d) t = 20 h, e) t = 40 h.
122
performed by Šelih et al., results of drying tests on noncured normal aggregate
concrete with w/c ratio of 0.40 was chosen [159] as a representative data set for
drying of earlyage cementitious materials. The amounts of Portland cement,
water, coarse and fine aggregate in this mixture were 489.5, 195.8, 912.3 and
804.9 kg/m
3
, respectively. The ambient moisture concentration, C
*
, can be
obtained from the desorption isotherm curves which relate relative humidity to the
water content. Using the BSB model [110], C
*
was obtained 49 kg/m
3
. As shown
in Figure 4.5.3a, the evaporation rates were computed from cumulative moisture
loss data using a numerical differentiation procedure. The drying results were
used to backcalculate diffusion and convective moisture transfer coefficients as
shown in Figure 4.5.3b, and expressed as a function of average moisture
concentration in accordance to the CEBFIP model as shown in Figure 4.5.4a.
The distribution of moisture concentration is compared with the experimental data
as shown in Figure 4.5.4b, indicating a good agreement between the simulation
and the experimental results.
4.6. Conclusion
Results of modeling moisture evaporation from the water surface and
cementbased materials based on the physics of drying. Evaporation from free
water surface is modeled implementing Prandtl’s interfacial boundary layer
concept through which the water vapor molecules diffuse to the surrounding air
due to gradient of moisture concentration. The model shows direct relationship of
water evaporation rate to relative humidity, along with saturated vapor pressure,
123
0
20
40
60
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30
Drying Time, d
Selih (1996)
Simulation
0
0.05
0.1
0.15
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, d
w/c = 0.40
T = 22
o
C
RH = 50%
No Initial Curing
Figure 4.5.3a
6x10
9
1x10
7
2x10
7
3x10
7
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 10 20 30
Drying Time, d
K
D'
0
0.0005
0.001
0.0015
M
o
i
s
t
u
r
e
T
r
a
n
s
f
e
r
C
o
e
f
f
i
c
i
e
n
t
,
m
/
h
Figure 4.5.3b
Figure 4.5.3. (a) Prediction of cumulative moisture loss and evaporation rate as a
function of time for the experiment performed by Šelih (1996), (b) Back
calculated D’ and k for the best fit of data
124
0.4 0.6 0.8 1
Normalized Moisture Concentration
7.00E009
1.07E007
2.07E007
3.07E007
4.07E007
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
BackCalculated Diffusivities
CEBFIP: n = 10
CEBFIP: n = 15
CEBFIP: n = 20
D
1
= 2.78 x 10
7
m
2
/s
D
0
= 6.94 x 10
9
m
2
/s
α = D
0
/D
1
= 0.025
C
c
= 0.7
Analysis on
Experiments by
Selih (1996)
Figure 4.5.4a
0 40 80 120 160 200
Moisture Concentration, kg/m
3
0
0.2
0.4
0.6
0.8
1
t
/
h
Exp
Sim
Exp
Sim
Exp
Sim
Exp
Sim
1 d
3 d
7 d
28 d
Time
h = 0.1 m
t
h
Figure 4.5.4b
Figure 4.5.4. Results of analysis on drying data by Šelih (1996): (a) Moisture
diffusivity as a function of normalized moisture concentration expressed by CEB
FIP models, (b) moisture concentration distributions through the thickness of the
sample in comparison with experimental results
125
and boundary layer thickness, which are functions of temperature and wind
velocity themselves. Water evaporation is used as a boundary condition for the
initial stages of drying. The moisture movement inside is modeled based on
moisture transfer theories of capillary porous media. The initial constant drying
rate period (stage I) is modeled as a funicular liquid water transport stage and
falling drying rate period (stage II) as a vapor diffusion transport stage. While
Fick’s second law of diffusion is used for modeling moisture movement in both
stages, it can be concluded that the moisture diffusivity indicates liquid water
conductivity at stage I, rather than vapor diffusion at stage II. Physics of drying
allows applying a constant flux boundary condition and a convective boundary
condition at stage I and II, respectively. Results of applying the methodology to
the drying experiments of cementbased materials under low and normal room
pressure shows that calculated moisture diffusivities reduce drastically from stage
I to stage II. The agreement between the results of the proposed analytical method
and an FE analysis and also experimental profiles of moisture distribution verifies
the soundness of this analytical method. Results of the analysis show that
proposed analytical method is capable of predicting experimental data during
drying process of cementbased materials and presenting the variation of
introduced internal and surface moisture transfer parameters by the time.
126
5. Modeling Rate of Evaporation Potential during Early Stage of Drying
5.1. Introduction
Earlyage cracking is a threat to structural integrity of concrete structures
and if not inhibited would lead to service life reduction and unsightliness. High
rates of evaporation which depends on air and concrete temperature, wind speed
and relative humidity increase the risk of early age cracking [11]. Due to
evaporation, water menisci are formed in the interparticle spaces [12] and a
negative pressure in the capillary water is built up [13, 14] and continues to rise as
the evaporation proceeds. The developed capillary pressure in drying fresh
concrete reaching 50 kPa in few hours [10, 12, 15] and exceeding 1MPa in less
than 8 h [16, 17] acts on the solid particles and results in the contraction of the
still plastic material [18] ranging from strain levels of 1 to 3x10
3
[12, 14, 19]. If
concrete is restrained, the developed shrinkage strain may easily exceed low
tensile strength of the fresh concrete [21] and cracking occurs (see Figure 5.1.1).
The initial rate of evaporation is very significant from the capillary
pressure point of view as comparably small changes in the evaporation rate
noticeably influence the capillary pressure versus time curve [12]. The differences
in crack tendency are most likely a consequence of the initial rate of evaporation
[15]. As shown in Figure 5.1.2, results of drying experiments on fresh cement
based materials reveals that the initial evaporation rate remains roughly constant
during a period, before beginning of a succeeding falling rate period [19, 66, 148,
149]. Visible plastic cracks are observed few hours after placing, a time period
well within the initial stage of drying [148].
127
Drying time: 15 min, 3h, 3h 15min
Drying time: 3h 30min, 4h, 24h
Figure 5.1.1. Development of 2D plastic shrinkage cracks during drying of a
fresh plain cement paste
It was found that the initial rate of evaporation from concrete surface is the
same as rate of evaporation from water surface exposed to the same condition [8,
10, 148]. Moisture transport during initial stage of drying when concrete is still
plastic is controlled by the external evaporation potential. Therefore, the rate of
water evaporation is required for modeling the moisture transport at this initial
stage [148, 160] which can be found by either experiments [161] or using a mass
transfer approach.
128
Figure 5.1.2.Two stages of drying during evaporation of capillary porous
materials
In the absence of a theoretical method to estimate initial rate of
evaporation from cementbased materials, a nomograph based on an empirical
equation is used [68] which was calibrated by the data from Lake Hefner. In this
paper theory of water evaporation based on the boundary layer theory, mass
transfer, diffusion, and fluid mechanics is described. A driving force of
concentration gradient is introduced as the main cause of mass transfer during the
drying process. Mass transfer at low and high rates is discussed separately along
with other related parameters such as mass transfer coefficient. An algorithm for
determination of evaporation rate is suggested. Considering similarity between
rate of evaporation from water surface and concrete surface during initial stage of
drying, predicted evaporation rates are verified by the recent experimental data
from concrete surfaces.
129
5.2. Available Method to Estimate Initial Rate of Evaporation
From the mechanical point of view, rate of evaporation can be related to
plastic cracking only when there is a model to predict capillary pressure and
pressure on solid skeleton due to moisture loss and the earlyage tensile properties
of the material is available. In the absence of such a model, available standards
and specifications such as ACI 305.1 [68], USACE TM 5822 [162], Canadian
CSA Standard A23.1 [163] and Australian Technical Bulletin 95/1 [164] set
requirements on evaporation rates below 1, 0.75, 0.75 and 0.5 kg/(m
2
h),
respectively to prevent plastic shrinkage cracking. As shown in Figure 5.2.1, the
evaporation rate in the atmospheric pressure is commonly estimated using a
nomograph in ACI 305 [68] given inputs for temperature, relative humidity, and
wind velocity. According to Hover [165] the underlying relationship for Menzel’s
equation [166] which is the source of ACI nomograph is 200 years old Dalton’s
law [167].
E = 0.313(p
sat
– p
v
) (0.253+0.06V) (5.2.1)
where E is the rate of evaporation [kg.m
2
.h
1
], p
sat
is the saturated water vapor
pressure [kPa] at the surface temperature of concrete T
c
, p
v
is the partial pressure
of the water vapor in the air [kPa] and V is the wind velocity [km.h
1
].
Uno [69] used the relationship between temperature and saturation vapor pressure
to replace saturation vapor pressure in Menzel’s equation by the temperature.
p
sat
= 0.61exp(17.3 T
c
/(237.3+ T
c
)) (5.2.2)
where T
c
is the temperature of the concrete surface [
o
C].
130
Figure 5.2.1. ACI nomograph for estimating rate of evaporation of surface
moisture from concrete (ACI 305.106)
Using equation (5.2.1) and (5.2.2), Uno [69] developed a single equation which
captures effect of temperature, relative humidity and wind velocity.
[ ] [ ]
( )
( ) 18 18
2.5 2.5
6
c a
E = 5 T  . T V+4 10 ϕ + + × (5.2.3)
131
where E is the rate of evaporation [kg.m
2
.h
1
], T
c
is the (surface) temperature of
concrete [
o
C], T
a
is the air temperature [
o
C], φ is the relative humidity and V is the
wind velocity [km.h
1
].
However, ACI 308R01 [168] reported that the accuracy of the prediction
is within ±25% for actual evaporation rates up to 1 kg/(m
2
h) and consistently
overestimates at higher evaporation rates by 50% or more. A Similar conclusion
was reached by AlFadhala and Hover [169] who found that the nomograph is
accurate for evaporation rates up to 0.5 kg/(m
2
h). This inaccuracy can be due to
the source of evaporation data used for deriving Menzel’s equation which is Lake
Hefner in Oklahoma in 1950 to 1951 [165]. Different lakes such as Lake Meade,
AZ., Fort Collins, CO., and Lake Kickapoo, TX. have different evaporation
equations [ 170, 171, 172, 173] which may be due to wind speed gradient,
cloudiness, sunlight intensity and water quality [165]. On the other hand, the rate
of evaporation which is the diffusion of water vapor through the boundary layer
covering the water surface is directly related to the characteristic length of the
surface [122, 128]. Menzel’s equation completely ignores this effect and therefore
the predicted evaporation rate can be used only when evaporation surface is about
the same size as of BPI (Bureau of Plant Industry) evaporation pan, i.e. 6’ in
diameter and 2’ deep [170].
5.3. Evaporation as a Mass Transfer Problem
Mass transfer is mass in transit as the result of a species concentration
difference in a mixture. Evaporation is not a mass transport by bulk fluid motion,
132
but rather a transport of one component, i.e. water vapor within a mixture of
components, i.e. humid air that occurs as result of a concentration gradient.
Humid air is a mixture of dry air containing a mixture of oxygen, nitrogen and
other inert gases and the water vapor which is saturated very close to the water
surface. Further from the surface, the humid air has lower water vapor
concentration and the nonuniformity of water vapor concentration near the
surface causes a mass transfer due to the effect of diffusion. The total rate of mass
transfer of component i (i.e. water vapor) is:
, , i i s i s
m v ρ ′′ = ` (5.3.1)
where
i
m′′ `
[kg.m
2
.s
1
] is the total mass flux, ρ
i,s
is the density of the component i
at the interface [kg.m
3
], and v
i,s
is the velocity component i at the interface [m.s
1
].
Mass flux due to diffusion,
, i s
J [kg.m
2
.s
1
] is related to the variation in the
distance between molecules of the component i due to change in concentration of
the component with time and can be expressed as:
, , ,
( )
i s i s i s s
J v v ρ = − (5.3.2)
where v
s
is the mass average velocity of all components [m.s
1
]. Combining
equation (5.3.1) and equation (5.3.2) leads to a general equation for the total mass
flux as:
¸
, , , , ,
diffusion
convection
( )
i i s i s s i s s i s i s s
m v v v J v ρ ρ ρ ′′ = − + = + `
¸_¸
(5.3.3)
133
The first term on the righthand side of equation (5.3.3), J
i,s
is the mass diffusional
flux at the interface and the second term represents the transfer of component i by
the accompanying flow due to diffusional flux at the interface, which is called the
convective mass flux. Considering these two terms, the mass transfer due to the
evaporation can be studied separately in regard to the flow rate.
5.3.1. Mass Transfer at Low Rates
All mass transfer processes including evaporation generate flow at the
interface. Nonetheless, when the mass transfer rates are sufficiently low, the
velocities caused by mass transfer are negligible. Thus, a stationary medium will
remain at rest and a flowing fluid will have the same velocity field as if there were
no mass transfer. More generally, when concentration of the diffusing component
is low and diffusional mass flux is low enough that it does not affect the imposed
velocity field, the total mass flux is principally carried only by diffusion as given
in the following equation:
, , ,
( )
i i s i s s i s
m v v J ρ ′′ = − = ` (5.3.1.1)
According to Fick’s law of diffusion, diffusional flux can be represented by:
i
i i
c
J D
y
∂
= −
∂
(5.3.1.2)
where, c
i
is the concentration of component i [kg.m
3
], D
i
is the diffusivity [m
2
.s

1
], J
i
is the rate of diffusion of component i per unit area of the surface [kg.m
2
.s

1
], and y is the distance in the direction of diffusion [m].
Therefore in the low rate mass transfer condition, the mass flux is:
134
,
0
i
i i s i
y
c
m J D
y
=
∂
′′ = =−
∂
` (5.3.1.3)
Figure 5.3.1.1 shows a schematic representation of the concentration distribution
near an interface for the case of evaporation from a flat plate.
Figure 5.3.1.1. Schematic representation of the concentration distribution near an
interface
Although the variation in the concentration near the interface is very
sharp, it becomes more gradual in the region slightly above the interface and the
concentration slowly approaches that in bulk fluid. The gradient of concentration
near the interface is directly related to
d
δ , the diffusion boundary layer thickness
[m], which is also shown in Figure 5.3.1.1 for the case of evaporation due to
laminar stream flow over a flat plate. This gradient can be written as:
, ,
0
( )
i e i s
i
d y
c c
c
y δ
=
−
∂
≈
∂
(5.3.1.4)
135
where, c
i,e
and c
i,s
are the concentration of component i at the air stream and
interface [kg.m
3
], respectively. This equation is not exact because the gradient is
not constant throughout the boundary layer. However, it does correctly reflect the
fact that the concentration of component i (i.e. water vapor) will change from c
i,s
to c
i,e
in the region between the surface of the plate and the edge of the diffusion
boundary layer. Therefore, using equation (5.3.1.3) and equation (5.3.1.4), the
rate of mass transfer can be taken to be proportional to the concentration
difference between the interface and the bulk fluid.
, ,
( )
i s i e
i i
d
c c
J D
δ
−
≈ (5.3.1.5)
In mass transfer problems, the diffusional mass flux from a surface, J
i,s
,
can be expressed as the product of a mass transfer coefficient and the
concentration difference between the surface and the fluid bulk (air flow). Hence,
the mass transfer coefficient for component i, h
m,i
[m.s
1
] can be defined as:
, ,
, , ,
( )
( )
i s i e
i i m i i s i e i
d
c c
m J h c c D
δ
−
′′ = = − ≈ ` (5.3.1.6)
Rearranging this equation leads to:
,
i
m i
d
D
h
δ
≈ (5.3.1.7)
To find the mass transfer coefficient for water vapor in the case of
evaporation, we need to solve an “external forced convection mass transfer”
problem. Conceptually, the free stream behaves much like a semiinfinite body
that experiences a stepchange in its surface vapor concentration at the instant that
136
fluid encounters the leading edge of the plate. The disturbance associated with the
change in the surface vapor concentration diffuses as a mass wave into the free
stream. This diffusion process takes time and, for the external flow problem, the
fluid motion transports the wave downstream from the leading edge. Therefore, as
shown in Figure 5.3.1.1, at x = x
2
, the mass wave has propagated further into the
free stream than it had at x = x
1
. The motion of mass diffusion wave can be
characterized by a mass diffusion boundary layer thickness given approximately
by
2
d vapor
D t δ ≈ (5.3.1.8)
Where D
vapor
is the diffusivity of the water vapor [m
2
.s
1
], and t is the time relative
to the disturbance at the surface [s]. The transport time for the mass diffusion is
approximately related to the distance from the leading edge, x [m], and the free
stream velocity, u
∞
[m.s
1
], according to:
x
t
u
∞
≈ (5.3.1.9)
Combining equation (5.3.1.9) and equation (5.3.1.8) gives:
2
vapor
d
D x
u
δ
∞
≈ (5.3.1.10)
To use dimensionless quantities to reduce number of independent variables,
equation (5.3.1.10) can be rearranged as:
137
2 2
2 2
vapor vapor vapor
d
vapor
D x D D
x x u u x u x u x
D
δ ν
ν ν
ν
∞ ∞ ∞ ∞
≈ = = = (5.3.1.11)
where ν is the kinematic viscosity of the water vapor [m
2
.s
1
]. Substituting the
definition of kinematic viscosity into equation (5.3.1.11) leads to:
2
d
vapor
x u x
D
δ
ρ ν
μ
∞
≈ (5.3.1.12)
where μ is the dynamic viscosity [kg.m
1
.s
1
], and ρ is the density of water vapor
[kg.m
3
].
Since Reynolds number is defined generally as
char char
u L
Re
ρ
μ
=
, for the flow over
a flat plate, the characteristic length is the distance from the leading edge, x [m],
and the characteristic velocity is the freestream velocity, u
∞
[m.s
1
]. Therefore,
the Reynolds number for the flow over a flat plate is defined according to:
x
u x
Re
ρ
μ
∞
= (5.3.1.13)
Also the Schmidt number, Sc, which provides a measure of the relative ability of
the fluid to transport momentum and mass of component, is defined as:
vapor
Sc
D
ν
= (5.3.1.14)
Inserting equation (5.3.1.13) and equation (5.3.1.14) into equation (5.3.1.12) leads
to:
138
2
d
x
x Re Sc
δ
≈ (5.3.1.15)
Combining equation (5.3.1.7) and equation (5.3.1.15), an approximate equation is
provided for the local mass transfer coefficient associated with the laminar flow
over a flat plate as:
,
2
vapor vapor
m vapor x
d
D D
h Re Sc
x δ
≈ = (5.3.1.16)
The mass transfer is typically made dimensionless using the Sherwood number,
Sh. The Sherwood number is defined in general according to:
, m i char
vapor
h L
Sh
D
= (5.3.1.17)
where L
char
is the characteristic dimension of the problem [m]. For stream flow
over a flat plate, the characteristic dimension is the distance from the leading
edge, x [m]. Therefore, the Sherwood number for flow over a flat plate is defined
according to:
, m vapor
x
vapor
h x
Sh
D
= (5.3.1.18)
Substituting equation (5.3.1.16) which is derived from conceptual model of the
mass diffusion boundary layer thickness, into equation (5.3.1.18) provides an
approximate equation for the local Sherwood number associated with the laminar
flow over a flat plate as:
139
,
0.5
m vapor
x x
vapor
h x
Sh Re Sc
D
= ≈ (5.3.1.19)
Although above equation is only an approximate solution for a laminar flow, the
exact analytical solution of local Sherwood number for laminar and turbulent
flows over a smooth isothermal flat plate which are correlated based on
experimental data can be found elsewhere [132, 174, 175].
1 2 1 3
1 4
2 3
0.3387
for
0.0468
1
x
x x crit
Re Sc
Sh Re Re
Sc
= <
⎡ ⎤
⎛ ⎞
+
⎢ ⎥
⎜ ⎟
⎝ ⎠
⎢ ⎥
⎣ ⎦
(5.3.1.20)
4 5 1 3
0.0296 for
x x x crit
Sh Re Sc Re Re = > (5.3.1.21)
where
crit
Re is the critical Reynolds number for flow passing a flat plate of water.
The critical Reynolds number for the transition from laminar to turbulent flow
varies with experimental conditions. The reported data [176, 177, 178, 179, 180]
are in the range of 10
4
10
6
. In this study, 5x10
5
was used for critical Reynolds
number.
The values of Sherwood numbers in equation (5.3.1.20) and equation
(5.3.1.21) are local, while an average Sherwood number associated with average
mass transfer coefficient is more useful to specify the boundary conditions. For a
stream flow over a flat plate, the average mass transfer coefficient is given by:
, ,
0
1
L
m vapor m vapor
h h dx
L
=
∫
(5.3.1.22)
and average Sherwood number is defined as:
140
, m vapor
L
vapor
h L
Sh
D
= (5.3.1.23)
Inserting equation (5.3.1.22) into equation (5.3.1.23) and using equation
(5.3.1.18) leads to:
,
0 0
0
L L
x vapor
m vapor L
x
L
vapor vapor
Sh D
h dx dx
x Sh
Sh dx
D D x
= = =
∫ ∫
∫
(5.3.1.24)
Since the Sherwood number correlations are expressed in terms of the
Reynolds number, it is convenient to change the coordinates of integration in
equation (5.3.1.24) from x to Re
x
by substituting equation (5.3.1.13) and its
derivative in equation (5.3.1.24):
0
L
Re
x
L
x
x
Sh
Sh dRe
Re
=
∫
(5.3.1.25)
Equation (5.3.1.25) can be integrated using any correlation or set of correlations
for the local Sherwood number [132]. In the laminar region, equation (5.3.1.20) is
substituted into equation (5.3.1.25) in order to obtain:
1 2 1 3
1 4
2 3
0.6774
for
0.0468
1
L
L
L crit
Re Sc
Sh Re Re
Sc
= <
⎡ ⎤
⎛ ⎞
+
⎢ ⎥
⎜ ⎟
⎝ ⎠
⎢ ⎥
⎣ ⎦
(5.3.1.26)
The average Sherwood number for the combined laminar and turbulent regions as
shown in Figure 5.3.1.2 is obtained by integrating equation (5.3.1.25) in two
141
parts; equation (5.3.1.20) is used in the laminar region, from
x crit
Re Re < , and
equation (5.3.1.21) is used in the turbulent region, for
crit x L
Re Re Re < < to obtain:
1 2 1 3
1 3 0.8 0.8
1 4
2 3
0.6774
0.037 ( ) for
0.0468
1
crit
L
L crit L crit
Re Sc
Sh Sc Re Re Re Re
Sc
= + − >
⎡ ⎤
⎛ ⎞
+
⎢ ⎥
⎜ ⎟
⎝ ⎠
⎢ ⎥
⎣ ⎦
(5.3.1.27)
Figure 5.3.1.2. Diffusion boundary layer thickness in a mixed laminar and
turbulent regions for free stream flow over a smooth isothermal flat plate
Finally, having average Sherwood number using equation (5.3.1.26) and equation
(5.3.1.27) would lead us to obtain average mass transfer coefficient according to
equation (5.3.1.23):
,
L
vapor
m vapor
Sh D
h
L
= (5.3.1.28)
This average mass transfer coefficient can be used directly by equation (5.3.1.6)
to obtain average rate of evaporation at low rates:
142
, , ,
( )
vapor vapor m vapor vapor s vapor e
m J h c c ′′ = = − ` (5.3.1.29)
5.3.2. Mass Transfer at High Rates
For assuming low rate mass transfer, it required that the velocity field be
unaffected by mass transfer and that the transferred components be dilute. When
those conditions are not met, the mass transfer coefficient will be different than
the value given for low rate mass transfer. The difference can be either an
increase or a decrease and can range from a few percent to an order of magnitude
or more, depending upon the concentrations of the diffusing components. In
addition to the diffusive transport represented by the mass transfer coefficient,
convective transport can contribute substantially to the total mass flux.
Mass transfer driving force at high rates
Figure 5.3.2.1 shows a boundary layer over a flat surface through which
there is a net mass transfer, m′′ ` [kg.m
2
.s
1
] of the various components in the
direction normal to the surface [ 181]. The mass flux of component i with
concentrations of c
i,e
[kg.m
3
] and c
i,s
[kg.m
3
] at the free stream and the surface is:
, , i i s i s s
m J v ρ ′′ = + ` (5.3.2.1)
where,
i
m′′ `
is the total mass flux of component i [kg.m
2
.s
1
], J
i,s
is the mass
diffusional flux at the interface [kg.m
2
.s
1
], ρ
i,s
is the density of the component i
at the interface [kg.m
3
], and v
s
is the mass average velocity of all components
[m.s
1
].
143
As defined earlier, the mass transfer coefficient for component i, h
m,i
[kg.m
2
.s
1
]
is:
, , , ,
( )
i s m i i s i e
J h c c = − (5.3.2.2)
Figure 5.3.2.1. The mass concentration boundary layer in the highrate mass
transfer problem
Total mass flux of component i in terms of the concentrations c
i,s
and c
i,e
are:
, , , ,
,
( )
i m i i s i e i s s
J
i s
m h c c v ρ ′′ = − + `
¸¸¸_¸¸¸
(5.3.2.3)
Since the density of the component i at the interface, ρ
i,s
, can be presented as ρ
i,s
= m
i,s
ρ
s
where m
i,s
is the mass fraction of component i at the surface, and ρ
s
is the
total density of all components at the interface [kg.m
3
], equation (5.3.2.3) can be
rewritten as
, , , ,
,
( )
i m i i s i e i s s s
J
i s
m h c c m v ρ ′′ = − + `
¸¸¸_¸¸¸
(5.3.2.4)
144
ρ
s
v
s
is the definition of total mass flux [kg.m
2
.s
1
], m′′ ` , which is inserted in
equation (5.3.2.4).
, , , , , ,
( )
i m i i s i e i s i s i s
m h c c m m J m m ′′ ′′ ′′ = − + = + ` ` ` (5.3.2.5)
Although the mass transfer coefficient is based on diffusive transfer from the
water surface, it may now differ considerably from the value for lowrate
transport. Equation (5.3.2.5) may be rearranged as:
, , , , , , , ,
( ) ( / ) ( )
i s i m i i e i s i s i m i i e i s
m m m h c c m m m m h c c ′′ ′′ ′′ ′′ ′′ − = − ⇒ − = − ⇒ ` ` ` ` `
, ,
,
,
/
i e i s
m i
i s i
c c
m h
m m m
⎛ ⎞ −
′′ =
⎜ ⎟
⎜ ⎟
′′ ′′ −
⎝ ⎠
`
` `
(5.3.2.6)
Equation (5.3.2.6) expresses the total mass flux of all components through the
interface, m′′ ` , as the product of the mass transfer coefficient and a ratio of
concentrations. This ratio is called the mass transfer driving force for component
i:
, ,
,
,
/
i e i s
m i
i s i
c c
B
m m m
⎛ ⎞ −
=
⎜ ⎟
⎜ ⎟
′′ ′′ −
⎝ ⎠
` `
(5.3.2.7)
The ratio of mass fluxes in the denominator is called the mass fraction in the
transferred state, denoted as m
i,t
[181]:
,
/
i t i
m m m ′′ ′′ ≡ ` ` (5.3.2.8)
The mass fraction in the transferred state is simply the fraction of the total mass
flux, m′′ ` , which is made up of component i. It is not really a mass fraction since it
can have any value from −∞ to +∞, depending on the relative magnitudes of
145
m′′ ` and
, i s
m′′ ` . Equation (5.3.2.6) and equation (5.3.2.7) provide a formulation of
mass transfer problems in terms of the mass transfer coefficient, g
m,i
, and the
driving force for mass transfer, B
m,i
:
, , m i m i
m h B ′′ = ` (5.3.2.9)
In many situations such as evaporation, only one component (i.e. water vapor) is
transferred through the interface, since the second component (i.e. air) is not
strongly absorbed into liquid (i.e. water) under normal conditions. If component i
is the only one passing through the surface, then
i
m m ′′ ′′ = ` ` , so that c
i,t
= 1. The
mass transfer driving force is simply
, ,
,
,
1
i e i s
m i
i s
c c
B
m
⎛ ⎞ −
=
⎜ ⎟
⎜ ⎟
−
⎝ ⎠
(5.3.2.10)
The effect of mass transfer rates on the mass transfer coefficient
Mass transfer coefficient, g
m,i
at high rates mass transfer problem is often
found through applying solution of the momentum and component equations or
through correlation of mass transfer data which are complicated approaches since
the flow field and mass transfer rate are coupled [181]. g
m,i
can be obtained from
the lowrate mass transfer coefficient by applying a correction for the effect of
finite mass transfer rates. Lienhard IV and Lienhard V [181] isolated effect of
m′′ ` on the mass transfer coefficient by approximating the boundary layer as a
stagnant film and allowing for the possibility of the counterdiffusion, so that both
diffusion and vertical convection may occur. They solved the partial differential
146
equation of component conservation in the molar form with appropriate molar
boundary condition assuming the mass transfer as a onedimensional steady mass
diffusion in stationary media without reactions. The solution of the rate of mass
transfer considering vertical convective transport in the massbased form is:
, ,
,
ln 1
i e i s
i
d i s i
c c
D
m
m m m δ
⎛ ⎞ −
′′ = +
⎜ ⎟
⎜ ⎟
′′ ′′ −
⎝ ⎠
`
` `
(5.3.2.11)
Recalling the definition of mass transfer driving force from equation (5.3.2.7),
equation (5.3.2.11) is presented as:
( )
,
,
,
ln 1
m i
i
m i
d m i
B
D
m B
B δ
⎡ ⎤
+
′′ = ⎢ ⎥
⎢ ⎥
⎣ ⎦
` (5.3.2.12)
Equation (5.3.2.9) and equation (5.3.2.12) lead to:
( )
,
,
,
ln 1
m i
i
m i
d m i
B
D
h
B δ
⎡ ⎤
+
= ⎢ ⎥
⎢ ⎥
⎣ ⎦
(5.3.2.13)
Now lowrate mass transfer coefficient of vapor can be obtained when
m′′ ` approaches zero.
0 0
,
, , ,
lim lim
m vapor
vapor
m vapor m vapor m vapor
m B
d
D
h h h
δ
→ →
∗
′′
≡ = =
`
(5.3.2.14)
Therefore, equation (5.3.2.13) can be rewritten as
( )
,
, ,
,
ln 1
m vapor
m vapor m vapor
m vapor
B
h h
B
∗
⎡ ⎤
+
= ⎢ ⎥
⎢ ⎥
⎣ ⎦
(5.3.2.15)
Since the group of
( )
, ,
ln 1
m vapor m vapor
B B
⎡ ⎤
+
⎣ ⎦
is called the blowing factor (BF), the
total mass flow rate of water vapor due to both diffusion and convective mass flux
147
is calculated according to lowrate mass transfer coefficient, blowing factor and
mass transfer driving force using
, ,
. .
vapor m vapor m vapor
m h BF B
∗
′′ = ` (5.3.2.16)
5.3.3 Algorithm for determination of evaporation rate
Considering the evaporation as a mass transfer problem, the following
algorithm for determination of evaporation rate is suggested:
1) Calculate temperature of the film (boundary layer) as
2
c a
film
T T
T
+
=
2) Determine the required air properties at the film temperature (μ, ρ, and ν)
[182]
3) Calculate binary diffusion coefficient for the water vapor in the air (D
vapor
) by
Chapman and Enskog formula [183]
4) Compute Schmidt number as:
vapor
Sc
D
ν
=
5) Compute Reynolds numbers as:
u L
Re
ρ
μ
∞
=
6) Calculate average Sherwood number L Sh from equation (5.3.1.26) and
equation (5.3.1.27)
7) Calculate average lowrate mass transfer coefficient
,
L
vapor
m vapor
Sh D
h
L
∗
=
148
8) Compute mass transfer driving force
, ,
,
,
1
vapor e vapor s
m vapor
vapor s
c c
B
m
⎛ ⎞ −
=
⎜ ⎟
⎜ ⎟
−
⎝ ⎠
from the
change of water vapor concentration near the surface and the mass fraction of
vapor at surface
9) Compute blowing factor as
( )
, ,
ln 1
m vapor m vapor
BF B B = +
10) Calculate total mass flux of water vapor due to evaporation according to
, ,
. .
vapor m vapor m vapor
m h BF B
∗
′′ = `
5.4. Parametric Studies
Effects of boundary layer temperature, relative humidity and wind velocity
on the rate of evaporation were studied and shown in Figure 5.4.1ac. As shown
in the figures, increasing ambient temperature leads to increase in the rate of
evaporation, which more significantly increases by increasing wind speed and
lowering ambient relative humidity. Increasing wind velocity, as shown in Figure
5.4.1a, results in decreasing boundary layer thickness which leads to an increase
in the evaporation rate, e.g., in an ambient temperature and relative humidity of
40
o
C and 50%, a decrease in the wind velocity from 10 to 0.5 [m.s
1
] translate into
a reduction of evaporation rate of a water surface with 0.3 m characteristic length
from 1.3 to 0.3 [kg.m
2
h
1
]. According to the proposed mass transfer model, in
both cases, the Schmidt number, Sc, is constant and equal to 0.72. However, the
Reynolds number, Re
L
, decreases from 174622 to 8731 leading to a decrease in
the average Sherwood number, L Sh , from 244 to 55 and therefore, the average
mass transfer coefficient,
m
h , falls from 0.0194 to 0.0043 [m.s
1
]. Having the
149
same values for ambient and interface moisture concentration in both cases as c
s
is
0.0514 and c
e
is 0.0313, the difference in the average mass transfer coefficient can
be held accountable for the reduction of evaporation rate by more than four times.
Increasing the ambient relative humidity, on the other hand, causes a
drastic reduction in the rate of moisture evaporation. As shown in Figure 5.4.1b,
assuming a wind velocity of 4 [m.s
1
], the evaporation rate at 40
o
C from a surface
with 0.3 m characteristic length decreases from 2.0 to 0.2 [kg.m
2
h
1
] when the
relative humidity increases from 10% to 90%. In these cases, the Schmidt (Sc),
the Reynolds (Re
L
) and the Sherwood ( L Sh ) numbers are constant and equal to
0.72, 70632 and 155, respectively, leading to a calculation of an average mass
transfer coefficient (
m
h ) of 0.0122 [m.s1]. While the mass concentration of water
vapor at the interface, c
s
is also constant and equal to saturated concentration of
0.0514 [kg.m
3
], the mass concentration of water vapor at the air stream, c
e
,
increases from 0.0051 to 0.0463 [kg.m
3
] due to increasing relative humidity
which causes a 10 time reduction in the value of the evaporation rate.
The size of sample in the direction of wind flow (characteristic length)
also plays a significant role in the change of evaporation rates. The reduction of
evaporation rates at the relative humidity of 50% and the wind velocity of 4 [m.s

1
] due to increasing characteristic length from 0.1 to 1 m was shown in Figure
5.4.1c. For example the results indicate that the evaporation rate at 40
o
C and
relative humidity of 50% decreases from 2.0 to 0.6 [kg.m
2
h
1
] when the
characteristic length increases from 0.1 to 1 m. In both cases, the Schmidt
150
number, Sc, and the Reynolds number, Re
L
, are constant and equal to 0.72 and
23544, while the average Sherwood number, L Sh , increases from 90 to 284
resulting in a decrease of the average mass transfer coefficient,
m
h , from 0.0211 to
0.0067 [m.s
1
]. When ambient and interface moisture concentrations are constant
in both cases as c
s
is 0.0514 and c
e
is 0.0257, the difference in the average mass
transfer coefficient leads into the reduction of evaporation rate by more than three
times.
5.5. Model Comparison with Experimental Results of Concrete Evaporation
Results of recent evaporation tests from several references [8, 12, 19, 149]
and one set of data from the authors tests were used to validate the model. The
characteristic length which is defined as the length of specimen in the direction of
the air flow was not given in most cases and therefore, an equivalent dimension
which is the diameter of the equivalent circle to that area was used in calculations.
Since the evaporating surface temperature was not reported, the rate of
evaporation was calculated, first assuming the evaporating surface has the same
temperature as the air and second, assuming surface cooling occurrence by the
order of 34
o
C. On the other hand, since the model cannot accept the air flow
velocity of 0, it was approximated with a very small wind velocity of 0.5 m/s in
those cases.
151
(Figure 5.4.1a) (Figure 5.4.1b)
(Figure 5.4.1c)
Figure 5.4.1. Effect of temperature, relative humidity, wind velocity and
characteristic length on the rate of evaporation from water surfaces.
The environmental parameters and predicted evaporation rates, as well as the
parameters used for calculation of evaporation rates based on mass transfer and
u
∞
0 10 20 30 40 50
Temperature of Boundary Layer (
o
C)
0
1
2
3
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
m
2
h
ϕ = 0%
10
20
40
50
60
70
80
30
90
= 4 m/s
L = 0.3 m u
∞
0 10 20 30 40 50
Temperature of Boundary Layer (
o
C)
0
0.4
0.8
1.2
1.6
2
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
(
m
2
h
)
0.5
6
(m/s) = 10
8
2
4
ϕ = 50%
L = 0.3 m
0 10 20 30 40 50
Temperature (
o
C)
0
1
2
3
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
m
2
h
L(m) = 0.10
0.25
0.50
0.75
1.00
= 4 m/s
ϕ = 50%
u
∞
152
boundary layer theory are shown in Table 5.5.1 and 5.5.2. As shown in the Table
5.5.1, calculated evaporation rates assuming identical temperature for the air and
evaporative surface overestimates the experimental results by 30300%. However,
evaporation rates estimated assuming surface cooling are very close to the
experimental reported data. Except in two cases of wind velocity of zero, the
other estimations are well within 030% of the experimental data. According to
Poole [184], the assumption that concrete temperature represents the temperature
at the evaporating surface is not strictly true because of the surface cooling
occurrence during evaporation. A dynamic equilibrium between the rate of heat
supply and the rate of liquid removal can explain the phenomena of evaporative
cooling. The surface temperature can be compared with the wetbulb temperature
defined as the equilibrium temperature which liquid water attains when heat is
transferred to it by contact with the air as an infinitely large quantity of gas.
Taking psychrometry into consideration, the mass transfer from the liquid by
evaporation is exactly balanced by the heat supply from the gas and can be
represented by a heat and mass transfer equilibrium [185].
( )
= ( ) / ( )
( )
s a c
vapor c a wb H s a
a wb H
h
m h T T K
T T K
ϕ ϕ
λ ϕ ϕ
λ
− −
′′ − = − ⇒ =
−
` (5.5.1)
where h
c
is the heat transfer coefficient by convection [W.m
2o
C
1
], T
a
is the
temperature of the air [
o
C], T
wb
is the wetbulb temperature of water [
o
C], λ is the
latent heat of evaporation at T
wb
[kJ.kg
1
], K
H
is the drying coefficient per unit
humidity difference [kg.s
1
.m
2
], φ
s
is the saturation humidity of water vapor, i.e.
1, and φ
a
is the humidity of the air.
153
To verify the surface cooling, the input data of the evaporation tests by
Azenha et al (2007) is compared with the data on usual psychrometric charts
[186] and as a conclusion a wetbulb temperature of 16
o
C is obtained. This result
is in agreement with the magnitude of assumed surface cooling of about 34
o
C.
Finally, the experimental data of Azenha et al. (2007) have been used to verify the
effect of wind velocity. As shown in Figure 5.5.1, the experimental data are well
predicted by the theory of boundary layer mass transfer. The comparison between
experimental data and the theoretical modeling is shown in Figure 5.5.2 for all 9
cases. Considering all results and analyses, it may be concluded that the suggested
model can well predict the initial rate of evaporation in the cementbased
materials in the range of 00.5 [kg.m
2
h
1
]. While the results of modeling for
higher evaporation rates were also verified using the test data reported by Lura
(2007), more experimental data are required for further verification of the model
in the evaporation rates higher than 0.5 [kg.m
2
h
1
].
154
Table 5.5.1.
Results of the evaporation rate in comparison with the experiments
Reference
Surface
Dimensio
ns,
(m)
Characteristic
Length, L
(m)
Air
Temp.,
T
a
(
o
C)
Relative
Humidity
, φ
()
Wind
Velocit
y, u
∞
(m/s)
Evaporation
Rate
assuming
T
c
=T
a
,
m′′ `
(kg/m
2
/h)
Evaporation
Rate
assuming
surface
cooling, m′′ `
(kg/m
2
/h)
Exp. Rate
of
Evaporatio
n,
m′′ `
(kg/m
2
/h)
Case1:
Azenha et
al (2007)
0.051x0.0
73
0.0688 20 0.6 0 0.21 0.01 0.07
Case 2:
Azenha et
al (2007)
0.051x0.0
74
0.0688 20 0.6 1.8 0.40 0.19 0.17
Case 3:
Azenha et
al (2007)
0.051x0.0
75
0.0688 20 0.6 3.3 0.54 0.25 0.26
Case 4:
Azenha et
al (2007)
0.051x0.0
76
0.0688 20 0.6 4.2 0.61 0.28 0.31
Case 5:
Lura et al
(2007)
d=0.1 0.1 30 0.50 6.8 1.65 1.17 1.30
Case 6:
Slowik et
al. (2008)
0.3x0.3 0.3385 20 0.45 4.8 0.40 0.32 0.24
Case 7:
Slowik et
al. (2008)
0.3x0.3 0.3385 22 0.3 4.8 0.58 0.44 0.43
Case 8:
Hammer
(2001)
0.1x0.1 0.1128 20 0.5 0 0.29 0.12 0.08
Case 9:
Authors
0.1x0.1 0.1128 24 0.34 0.5 0.34 0.23 0.20
155
Table 5.5.2.
Parameters used for calculation of evaporation rates
Parameters Case1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9
boundary layer
temperature,
T
film
[K]
291 291 291 291 303 291 294 291 295
dynamic
viscosity, μ
[kg.m
1
.s
1
]
1.81E
05
1.81E
05
1.81E
05
1.81E
05
1.87E
05
1.81E
05
1.82E
05
1.81E
05
1.83E
05
density of the
air, ρ [kg.m
3
]
1.212 1.212 1.212 1.212 1.166 1.212 1.202 1.212 1.196
kinematic
viscosity, ν
[m
2
.s
1
]
1.49E
05
1.49E
05
1.49E
05
1.49E
05
1.60E
05
1.49E
05
1.52E
05
1.49E
05
1.53E
05
diffusivity, D
[m
2
.s
1
]
2.04E
05
2.04E
05
2.04E
05
2.04E
05
2.20E
05
2.04E
05
2.08E
05
2.04E
05
2.10E
05
Schmidt
number, Sc
0.7315 0.7315 0.7315 0.7315 0.7262 0.7315 0.7303 0.7315 0.7296
Reynolds
number, Re
L
2304 8293 15204 19350 43762 108803 107166 3777 3686
avergae
Sherwood
number,
L Sh
28 54 73 82 123 194 192 36 36
average lowrate
mass transfer
coefficient,
m
h
[m.s
1
]
8.38E
3
1.59E
2
2.15E
2
2.43E
2
2.7E2
1.17E
2
1.18E
2
6.54E
3
6.63E
3
saturated water
vapor pressure,
p
sat
[Pa]
1825 1825 1825 1825 3796 1825 2206 1825 2347
partial pressure
of the vapor in
air, p
v
[Pa]
1408 1408 1408 1408 2256 1056 796 1174 1019
vapor
concentration at
the interface, c
s
[kg.m
3
]
0.0137 0.0137 0.0137 0.0137 0.0273 0.0137 0.0164 0.0137 0.0174
mass fraction of
the vapor at the
interface, m
s
0.0113 0.0113 0.0113 0.0113 0.0234 0.0113 0.0136 0.0113 0.0145
vapor
concentration in
the air stream, c
e
[kg.m
3
]
0.0104 0.0104 0.0104 0.0104 0.0161 0.0078 0.0058 0.0087 0.0074
mass transfer
driving force,
B
m
[kg.m
3
]
0.0033 0.0033 0.0033 0.0033 0.0115 0.0060 0.0107 0.0051 0.0101
blowing factor,
BF
0.998 0.998 0.998 0.998 0.994 0.997 0.995 0.997 0.995
156
0 1 2 3 4 5
Wind Velocity (m/s)
0
0.1
0.2
0.3
0.4
0.5
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
(
m
2
h
)
Boundary Layer Model
Azenha et al. 2007a,b
T
a
= 20
o
C T
c
= 16
o
C
ϕ = 60% L = 0.069 m
Figure 5.5.1. Prediction of experimental data of Azenha et al. (2007a,b) by a
method based on a boundary layer mass transfer concept
0 0.4 0.8 1.2 1.6
Theoretical Rate of Evaporation, kg/(m
2
h)
0
0.4
0.8
1.2
1.6
E
x
p
e
r
i
m
e
n
t
a
l
R
a
t
e
o
f
E
v
a
p
o
r
a
t
i
o
n
,
k
g
/
(
m
2
h
)
Azenha et al. 2007a,b
Slowik et al. 2008
Lura et al. 2007
Hammer 2001
Authors
Figure 5.5.2. The comparison between the experimental data and the theoretical
modeling of initial rate of evaporation based on a boundary layer mass transfer
theory
157
5.6. Conclusion
Plastic shrinkage crack, a threat to structural integrity of concrete
structures, are often observed in the first few hours after placing which is a time
period well within the initial stage of drying. During stage I drying, rate of
evaporation from concrete surfaces is roughly constant and similar to the external
potential evaporation rate from water surfaces which controls the drying process.
The theory of water evaporation based on the boundary layer theory, mass
transfer, diffusion, and convection is described and a driving force of
concentration gradient is introduced as the main cause of mass transfer. Mass
transfer at low and high rates is discussed separately along with other related
parameters such as mass transfer coefficient and a method for determination of
evaporation rate is suggested. Results of parametric studies show that increasing
ambient temperature leads to an increase in the rate of evaporation, which more
significantly increases by increasing wind speed and lowering ambient relative
humidity. Results of parametric studies also show that the size of sample in the
direction of wind flow (characteristic length) plays a significant role in the change
of evaporation rates. The proposed method is verified by the recent experimental
data from several researchers on cementbased materials. The results show that
given the environmental parameters as inputs, the model is capable of predicting
the evaporation rates with a good accuracy.
158
APPENDIX A
FRee Shrinkage Test Results
159
An experimental study was conducted to evaluate the effects of adding
low content of ARglass fibers (5, and 10 lb/yd
3
) on the free shrinkage of cement
pastes. The free shrinkage tests were performed according to ASTM standard C
15704. This test method covers the determination of the length changes that are
produced by causes other than externally applied forces and temperature changes
in hardened hydrauliccement mortar and concrete specimens made in the
laboratory and exposed to controlled conditions of temperature and moisture.
Measurement of length change permits assessment of the potential for volumetric
contraction (shrinkage) of concrete due to drying. This test method is particularly
useful for comparative evaluation of this potential in different hydrauliccement
mortar or concrete mixtures. In this test, specimens of 1” [25mm] square cross
section and approximately 111⁄4” [285 mm] in length are casted for each
mixture. The test specimens are cured in the molds covered with a plastic sheet
for 24 hours while water dripping from specimen is protected. Upon removal of
the specimens from the molds, the initial Length reading is taken using a digital
comparator dial. After the initial comparator reading, the specimens stored in the
drying room, so that the specimens have a clearance of at least 1 in. [25 mm] on
all sides. Comparator readings of each specimen are taken every one or two days
after initial reading. These readings were done in a room maintained at a relative
humidity of 30% while the specimens are at a temperature of 73°F [23°C]. The
comparator dial were read with both test specimen and the reference bar in the
comparator and the difference between the two readings and initial length of
specimens were used to calculate shrinkage strain.
The effect of ARglass fibers on free shrinkage was investigated by
ASTM standard C 15704. Three different cement paste mixtures were designed
including 2 mixtures with ARglass fibers and one control mixture with only plain
cement paste and no fiber. All mixtures were designed with the watercement
ratio of 0.45 and a high slump (6 in.) such as slurry. The cement content was 2080
lb/yd
3
(1450 Kg/m
3
) and only Portland cement, type I/II and water were used to
cast the paste. The only difference between these mixtures, Control, ARG5.0, and
ARG10 was due to adding 0, 5, and 10 lb/yd
3
fibers respectively. In all ARG
mixes, AntiCrak® HPglass fibers with 24 mm length were used. The mix
proportions of sample are resented in Table A.1.
Table A.1.
Mixture proportions of the ARGs and Control samples (lb/ft
3
)
Mix ID Control ARG5.0 ARG10
Portland Cement 90 90 90
Water 41 41 41
Glass Fiber (HPGlass) 0.0 0.19* 0.37*
w/c 0.45 0.45 0.45
* The fiber content of 0.19, and 0.37 lb/ft
3
is equivalent to 5, and 10 lb/yd
3
160
Tow prism specimens of 1” [25mm] square crosssection and
approximately 111⁄4” [285 mm] in length are cast for each mixture. For making
each sample, Portland cement were added to the mixer and blended for 60
seconds. Then, water poured gradually to the mixer and blended for 5 minutes
before adding the glass fibers only to the ARG mixtures and mixing for 5
additional minutes to thoroughly mix all the ingredients. After mixing concrete in
a suitable laboratory mixer, the concrete is placed in the mold in two
approximately equal layers and each layer compacted with the tamper. We
worked with the concrete until a homogeneous specimen is obtained. After the top
layer has been compacted, we strike off the mortar flush with the top of the mold,
and smooth the surface with a few strokes of a trowel. The test specimens are
cured in the molds covered with plastic sheet and then removed from the molds at
an age of 24 hours after mixing. After length measurement at the end of the curing
period using a digital comparator dial, the samples are stored in a drying room
until next reading. The lengths of samples are read in 1 or 2 days intervals up to
28 days and the difference between the reading and initial reading (at 24h) used to
calculate free shrinkage. The results of experimental tests are shown in Figures
A.1. The results show that adding low content ARglass fibers (5 to 10lb/yd
3
) to
the Portland cement paste reduces the free shrinkage by about 5% and 10%,
respectively.
Experimental data were analyzed using ACI 209R92 shrinkage model.
This model is an empirical model developed by Branson and Christiason [187],
with minor modifications introduced in ACI 209R82 [188]. ACI Committee 209
incorporated the developed model in ACI 209R92 [87]. The model for predicting
shrinkage strains as a function of time have a simple principle: a hyperbolic curve
that tends to an asymptotic value called the ultimate value. The shrinkage strain
ε
sh
(t,t
c
) at age of concrete t (d), measured from the start of drying at t
c
(d), is
calculated by equation (A.1).
( )
( , )
( )
c
sh c shu
c
t t
t t
f t t
α
α
ε ε
−
= ⋅
+ −
(A.1)
where, f (in days) and α are considered constants for a given member shape and
size that define the timeratio part, ε
shu
is the ultimate shrinkage strain, and
(t – t
c
) is the time from the end of the initial curing. The shape and size effect can
be totally considered on the timeratio by replacing α = 1.0, and f as given by
equation (A.2) in inlbs units.
[ ] 0.36 ( / )
26.0
V S
f e = (A.2)
where, V/S is the volumesurface ratio in inches.
161
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Sample 1
Sample 2
Control : V
f
= 0 lb/yd
3
w/c = 0.45
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Sample 1
Sample 2
ARG5.0 : V
f
= 5 lb/yd
3
w/c = 0.45
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Sample 1
Sample 2
ARG10 : V
f
= 10 lb/yd
3
w/c = 0.45
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Control (1)
Control (2)
ARG5.0 (1)
ARG5.0 (2)
ARG10 (1)
ARG10 (2)
w/c = 0.45
Figure A.1. Free shrinkage results of Control and ARG samples and their
comparison
Hence, equation (A.1) is transformed to:
1.5
1.5
( 1)
( , )
28 ( 1)
sh c shu
t
t t
t
ε ε
−
= ⋅
+ −
(A.3)
ACI 209R92 suggests multiplying ε
shu
by seven factors, depending on particular
conditions
6
shu sh
= 780 × 10 in./in. ε γ (A.4)
with
sh sh,tc sh,RH sh,vs sh,s sh, sh,c sh,
=
ψ α
γ γ γ γ γ γ γ γ (A.5)
where,
sh
γ represents the cumulative product of the applicable correction factors.
, sh tc
γ represents the initial moist curing coefficient for curing times different from
162
7 days for moistcured concrete.
, sh RH
γ is the ambient relative humidity coefficient
while Coefficient
, sh vs
γ allows adjustment for the size of the member in terms of
the volumesurface ratio.
, sh s
γ and
, sh ψ
γ are the slump the fine aggregate factors
respectively.
, sh c
γ represents effect of cement content and
, sh α
γ is air content
factor.
In order to use ACI 209R92 shrinkage model to analyze our free
shrinkage data, α is chosen as 1.5,
c
t equals to 1 day, and f is chosen as 28 days
since V/S ratio is equal to 0.25 in. The values of
, sh tc
γ ,
, sh RH
γ ,
, sh vs
γ ,
, sh s
γ ,
, sh c
γ ,
, sh α
γ are considered as 1.2, 1.094, 1.164, 1.136, 1.498 and 0.99 respectively since
c
t equals to 1 day, RH is 30%, V/S ratio is equal to 0.25 in, slump is 6 in. and air
content is about 5%.
sh,ψ
γ referring to fine aggregate content factor is not
applicable here since no aggregate was used in cement paste samples. With a
good adjustment, the value of 1.5 is replaced
, sh ψ
γ factor and as shown in Figure
A.2, the data show a good correlation with ACI model.
shu
ε for plain cement
paste or Control samples was calculated as 3012 x 10
6
in/in.
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Sample 1
Sample 2
ACI 209R92 Model
Control : V
f
= 0 lb/yd
3
w/c = 0.45
ε
shu
= 3012 x 10
6
in./in.
Figure A.2. Modeling free shrinkage data of Control sample using ACI 209R92
model
To model ARG samples, a factor as
f
γ is presented to include in cumulated
correction factors,
sh
γ in order to reduce ultimate shrinkage,
shu
ε and modify ACI
model for fiber effect. Then the shrinkage formula turned out as;
( )
6
shu sh,tc sh,RH sh,vs sh,s sh, sh,c sh, f
= 780 × 10 in./in.
ψ α
ε γ γ γ γ γ γ γ γ (A.6)
163
Using
f
γ as 0.96 and 0.9 for paste samples with 5 lb/yrd
3
and 10 lb/yrd
3
respectively, our modified ACI model fitted experimental data appropriately.
Using these fiber correction factors
shu
ε is calculated as 2890 x 10
6
in/in for
ARG5.0 and 2710 x 10
6
in/in for ARG10 samples. The results of fitting data with
modified ACI model are shown in Figure A.3.
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Sample 1
Sample 2
ACI 209R92 Model
V
f
= 5 lb/yd
3
w/c = 0.45
ε
shu
= 2890 x 10
6
in./in.
0 10 20 30
Time (Days)
0
500
1000
1500
2000
2500
3000
F
r
e
e
S
h
r
i
n
k
a
g
e
S
t
r
a
i
n
(
m
i
c
r
o
s
t
r
a
i
n
)
Sample 1
Sample 2
ACI 209R92 Model
ARG10 : V
f
= 10 lb/yd
3
w/c = 0.45
ε
shu
= 2710 x 10
6
in./in.
Figure A.3. Fitting experimental free shrinkage data of ARG samples with
modified ACI 209R92 model
Experimental and modeling free shrinkage data indicate that adding low
dosage of ARglass fibers to concrete samples reduced the free shrinkage strains
so slightly. The results show that ultimate free shrinkage of cement pastes can
reduce by almost 4% and less than 10% when adding 5 and 10 lb/yrd
3
ARglass
fibers to the batch. The final conclusion is that free shrinkage parameters do not
significantly differentiate the contribution of fiber.
164
APPENDIX B
Vacuum Evaporation Test Method and Data Transfer Description
165
The author developed a new test method which is capable of
characterizing evaporation parameters and simulating severe conditions to create
2D shrinkage cracks. This test method is based on the vacuum drying technique
and test results can be used to analyze evaporation and cracking. A schematic of
test setup is shown in Figure B.1.
Figure B.1. 2D shrinkage cracking test setup based on vacuum technology
Note that before running an evaporation test, the following items have to be
checked.
1. Oil level and quality (oil has to be changed after 45 complete tests)
2. Dry ice (1015 lbs of dry ice is required for a single 48 h test)
3. Mold (to make sure is ready and cleanedup)
4. Initial pressure inside the desiccator (adjust it at 20 in Hg before starting
the test)
5. Load cell (Offset the load cell before starting a test)
6. Output file header (to be filled up in details and completed in GCTS
software)
Testing equipments
Various components of the setup are described.
Equipped Vacuum Vessel
166
A Pyrex vacuum vessel covered with a transparent Acrylic lid was used to
maintain vacuum pressure. As shown in the Figure B.1, the lid has a valve to
induce vacuum pressure. There is another hole on the other side of lid used to pass
load cell wires from inside of the vessel to the outside. The vessel was equipped
with a scale system. In order to weigh samples continuously, an accurate load cell
with capacity of 5440 g (12 lb) and accuracy of 0.05g (0.0001 lb) was hired. Two
Plexiglas plates were attached to both end of the load cell to make weight
measurement as convenient as possible. The load cell was calibrated by a precise
digital scale (0.01g accuracy) using more than 30 different weights up to 1500 g.
The calibration formula was derived and the obtained voltage gain was used for
recording moisture loss.
Mold
As shown in Figure 3.2.1b, Plexiglas parts were used to make a
10×10×1.3 cm (4”×4”×0.5”) lowweight mold. This thin mold is proper to make
samples with high specific surface area exposed to the external evaporation
condition. In order to restrain the sample and cause earlyage 2D cracks, two pairs
of holes were drilled in each side of the mold and paper clips were passed through
them and bended at the outer side by the right angle to stay fixed during the test.
Vacuum Pump
A beltdrive twostage pump was used as the vacuum equipment operated
at its ultimate blankoff pressure, 1700 Pa (0.017 atm) for a long period of time.
The pump is designed to remove air from a vacuum system at atmospheric
pressure to a deep vacuum.
Condenser System
A condenser system including a vessel, alcohol, dry ice and flask was used
in this test method. Condenser vessel is a 15.2×30.5 cm (6”×12”) cylinder made
up of double layer glass materials with a cylindrical hole of 12.7×25.4 cm
(5”×10”) inside it. The dry ice was fallen first inside the trap up to 1/3 of the
height. Then alcohol was added to the trap to decrease the atmosphere
temperature to 70
o
F to condense the moisture. Double layer glass vessel used as
heat insulator to prevent fast evaporation of the dry ice. A 1000mL glass flask
was acquired as a trap to collect condensed moisture and prevent water vapor to
flow through the pump.
TBore Stopcock, Pressure Gage, Regulator and Hose
TBore stopcock was used to split and conduct vacuum flow of the vessel
between condenser and pressure gage. The pressure gage was required to control
the negative pressure inside the vessel and insure of ultimate pressure. All the
components of the vacuum drying system including vacuum pump, condenser
system, TBore stopcock, pressure gage and vacuum vessel are connected by high
pressure hoses.
167
Data Collection System
The load cell was connected to a computer using a strain gage transducer
amplifier and a simple computer interface unit. Computer has an acquisition card
and equipped with a testing software to read and collect test data. The software is
capable of getting calibration formula and read the data as weight in grams. The
procedure was set to collect the data every single minute and stop collecting the
data at the specified 24 or 48hrs.
Camera
In order to analyze several characteristics of earlyage cracks developed
by drying of the cement paste and clay samples, a digital camera (Pulnix
TM1325CL) was provided. The camera mounted 10 cm away from the top
surface of vessel lid facing the center of the sample inside the vessel. The camera
was set automatically to take pictures of the top surface of the sample every 15
min through end of the test.
Testing procedure
After a number of preliminary trial tests, the following procedure was
established to run evaporation test using vacuum technique. First, the mold was
prepared and weighed using an accurate digital scale (0.01g accuracy). The
cement paste was then made according to the mix design table and poured into the
mold. Total weight of the sample including weight of mold and cement paste was
measured using the digital scale. Note that the recorded weight was the same as or
within the accuracy range of the weighing system inside the vessel. The
condenser vessel was filled up to 1/3 of the height with alcohol. 1012 pounds of
dry ice was hooked up and crushed into small chunks before placing inside the
condenser vessel. After some hours, the initial added dry ice may evaporate being
exposed to the air in room temperature. In such a case, more dry ice is needed
during the test to maintain its equilibrium with alcohol. TBore stopcock was
opened to let the negative pressure flow from vacuum valve at top of the vessel
when the pump is on. The cement paste sample was then placed inside the
vacuum vessel attached to the load cell. The weight was measured by continuous
recording of the load cell and the camera was powered on to capture images at 15
minutes interval during the test.
At the beginning of drying tests, a vacuum reaching 34 kPa (10” Hg)
absolute was applied to each specimen for a short period of time. Then the
pressure decreased in the vessel to 1.7 kPa (0.5” Hg) absolute. Since the fresh
paste is so fluid at the beginning of the test, reaching ultimate blankoff pressure of
the vacuum pump leads to creation of many permanent voids on the surface of the
sample. Voids affect on crack characteristics of the sample as obstacles against
crack propagation. Also the load cell placed inside the vacuum vessel shows
erratic results at the very beginning of the test due to effect of vacuum pressure. A
short period of time (i.e. 30 min  1 h) is required to get stable data from load cell.
Hence, the first 30 minutes to one hour was assumed as an offset time supported
168
by the fact that only a mass loss of about 1g was recorded during this period due
to very low vacuum pressure.
Data Collection and Transfer Procedure
Step 1 Open the GCTS package and click on the ‘Tests’.
Step 2 Click on the ‘Universal’.
169
Step 3 Click on the ‘TestID’ when in the Transducer column ‘Scale’ is chosen.
Step 4 Type the Project info and Sample description inside ‘TestID’, and press
Enter.
170
Step 5 Click on the ‘Command’ and press on ‘F2’ to load a predefined command
procedure. You can load ‘MEHDI_1.CMD’ for test duration of 1440 min (24 h).
Step 6 In case a new procedure with longer duration or higher frequency is
required to be run, a command procedure has to be defined. Click on the ‘Servo
1’ and change the parameters to the desired values. Then click on ‘F3’ to save the
command.
Step 7 Click on the ‘Execute’ to start the data collection procedure.
171
Step 8 Click on ‘F5’ to start the data collection. A name has to be typed for the
name of the generated DAT file.
172
Step 9 After the test is complete, click on the ‘Report’ to convert the format of
the output file from DAT (only readable inside the GCTS package) to Text
(ASCII) format.
Step 10 The DAT file has to be loaded by clicking on or typing the DAT file
name and then pressing Enter.
173
Step 11 Choose ‘File’ and press Enter.
Step 12 Type in the file of the ASCII file and press ‘Enter’.
174
Step 13 The data transfer is complete and you can find the ASCII file on: C:\gcts
175
APPENDIX C
Finite Element Analysis of Evaporation Test Data
176
Evaporation Problem
A thin concrete slab with the lateral surfaces insulated from moisture
removal, is initially saturated with the moisture concentration of 520 kg/m
3
. The
bottom end of slab is impervious and insulated from moisture removal as well.
The top face is exposed to a lowpressure condition using a vacuum pump
simulating two different stages of drying. During stage I, the test simulates a
roughly constant evaporation rate of 0.42 kg/(m
2
h) from the top face. However,
during stage II drying which starts after 11.5 hours, evaporation can be observed
as a convectiontype process with zero ambient moisture concentration. The
problem is a 1D moisture transport from bottom end to the top face and
therefore, a uniform mesh of n=1,2,.. linear elements is used for this problem. The
forward difference method is applied to calculate several steps of the transient
solution to the moisture diffusion problem. L=0.0125 m, D(t) and h
2
(t) are
presented as a function of time according to optimization works on analytical
solutions as shown in Figure C.2.
A code is written to find the critical step for stable solutions. Moisture loss at each
time step is calculated as the difference between the integration of moisture
concentration inside the domain, and the initial moisture content. The results are
compared with the experimental data and the results of Abaqus analysis.
Figure C.1. Governing differential equations and boundary conditions of the
problem
0 gy c τ = + =
n
c τ =
0 gy c τ = + =
n n
hy c h y τ = + =
177
2x10
7
4x10
7
6x10
7
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 10 20 30 40
Drying Time, h
K
D'
0.001
0.0015
0.002
M
o
i
s
t
u
r
e
T
r
a
n
s
f
e
r
C
o
e
f
f
i
c
i
e
n
t
,
m
/
h
6x10
9
Figure C.2. Input diffusivity and moisture transfer coefficient as a function of
time
Solution
General 1D timedependent problem:
D.E.:
2
2
1
( , ) ( , ) ( , )
( ) ( ) ( ) ( ) ( , ) ( , )
time dependent terms
D BVP
u x t u x t u x t
x x x x u x t f x t
t t x x
ρ μ α β
−
∂ ∂ ∂ ∂ ⎡ ⎤
+ − + =
⎢ ⎥
∂ ∂ ∂ ∂
⎣ ⎦
¸¸¸¸¸¸_¸¸¸¸¸¸
¸¸¸¸¸¸¸¸¸_¸¸¸¸¸¸¸¸¸
B.C.s: mixed gy c τ ⇒ = +
I.C.s:
0 0
( , ) ( ) u x t u x =
Moisture transport problem:
D.E.:
( , ) ( , )
0
C x t C x t
D
t x x
∂ ∂ ∂ ⎡ ⎤
− =
⎢ ⎥
∂ ∂ ∂
⎣ ⎦
( ) 0, ( ) 1, ( , ) 0, ( ) 0, ( ) x x f x t x x D ρ μ β α ⇒ = = = = =
Galerkin’s method of minimizing residuals:
178
Trial solution D.E.:
1
( , ; ) ( ) ( )
n
j j
j
C x t a a t x φ
=
=
∑
( , ) ( , )
( ) 0
1, 2,...,
i
C x t C x t
D x dx
t x x
i n
φ
Ω
⎡∂ ∂ ∂ ⎤ ⎛ ⎞
− =
⎜ ⎟ ⎢ ⎥
∂ ∂ ∂
⎝ ⎠ ⎣ ⎦
=
∫
Integration by parts & substituting in residual equation
1
( )
( ) ( )
1, 2,...,
n
x
i
i i
x
d x C C C
x dx D dx D x
t dx t x
i n
φ
φ φ
Ω Ω
∂ ∂ ⎡ ∂ ⎤ ⎛ ⎞
+ = − −
⎜ ⎟ ⎢ ⎥
∂ ∂ ∂
⎝ ⎠ ⎣ ⎦
=
∫ ∫
1
( )
( )
n
j
j
C d j
a t
x x
φ
=
∂
=
∂ ∂
∑
1
( )
n
j
j
da
C
j
t dt
φ
=
∂
=
∂
∑
[ ]
1
1 1
( )
( ) ( ) ( , ; ) ( )
n
n n
x
j j
i
i j j i
x
j j
f
c k
da d
d x
x x dx D dx a x t a x
dt dx x
φ
φ
φ φ τ φ
= =
Ω Ω
⎛ ⎞ ⎛ ⎞
+ = −
⎜ ⎟ ⎜ ⎟
∂
⎝ ⎠ ⎝ ⎠
∑ ∑
∫ ∫
¸¸¸¸_¸¸¸¸
¸¸¸¸_¸¸¸¸ ¸¸¸¸¸_¸¸¸¸¸
Forward difference method
In forward difference method all quantities are evaluated at backward end of time,
t
n1
1 1 1
{ } { }
n n n
C a K a F
− − −
+ = `
1
1
{ } { }
{ }
n n
n
n
a a
a
t
−
−
−
=
Δ
`
Moisture transport in forward difference method
1 1
1 1
n n n
n n
C a F C K a
t t
− −
⎛ ⎞ ⎛ ⎞
= + −
⎜ ⎟ ⎜ ⎟
Δ Δ
⎝ ⎠ ⎝ ⎠
Capacity matrix
C can be diagonalized on the LHS.
1
, 1, 2,..., , 0
n
ii ij ij
j
CL = C i n C wheni j
=
= = ≠
∑
179
[ ]
11
22
1/ 0 0 0
0 1/ 0 0
0 0 0
0 0 0 1/
L
1
L
Lnn
C
C
CL
C
⎡ ⎤
⎢ ⎥
⎢ ⎥
=
⎢ ⎥
⎢ ⎥
⎣ ⎦
[ ] ( )
1
1
n n1 n n1 n
a =a + t CL F Ka
−
Δ −
Solution for 1 linear element problem
linear interpolation:
2
1
x x
L
φ
−
= ,
1
2
x x
L
φ
−
=
( )
j
i
ij
d D L D L
d x
k D dx K
D L D L dx x
φ
φ
Ω
− ⎡ ⎤
= ⇒ =
⎢ ⎥
− ∂
⎣ ⎦
∫
[ ]
1
2 1 1/ 3 0
6
( ) ( ) / 6
1 2 0 1/ 3
ij i j
c x x dx C L CL
L
φ φ
−
Ω
⎡ ⎤ ⎡ ⎤
= ⇒ = ⇒ =
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
∫
1
2
0
[ ]
n
x
i i x stageI
f F
c
τφ
⎡ ⎤
= − ⇒ =
⎢ ⎥
⎣ ⎦
2 2
0
stageII
F
h a
⎡ ⎤
⇒ =
⎢ ⎥
⎣ ⎦
but is modified by addition of h
2
term to [K]
1
φ
2
φ
180
Matrix equation at stage I for 1 linear element
2
2
0 0 4.12 5 4.12 5
6
0 1/ 3 4.12 5 4.12 5
0.0125, 0.42/ 3600 1.667 4
1 1 1
n
2 2 2
n n1 n1
a a a 1/3 E E
= + t
a a c a E E L
L c E
⎛ ⎞ − − − ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎛ ⎞
Δ −
⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎜ ⎟
− − −
⎝ ⎠
⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
⎝ ⎠
= =− = −
Matrix equation at stage II for one linear element
2 2
0 0 4.12 5 4.12 5
6
0 1/ 3 4.12 5 4.12 5
4.12 5 4.12 5 0
6
4.12 5 4.12 0 1/ 3
1 1 1
n
2 2 2
n n1 n1
1 1
n
2 2
n n1
a a a 1/3 E E
= + t
a a h a a E E L
a a E E 1/3
= + t
a a E E L
⎛ ⎞ − − − ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎛ ⎞
Δ −
⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎜ ⎟
− − −
⎝ ⎠
⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
⎝ ⎠
− − − ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎛ ⎞
Δ −
⎜ ⎟ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
− −
⎝ ⎠
⎣ ⎦ ⎣ ⎦ ⎣ ⎦ 2
2
5
0.0125,
1
2
n1
a
h a
L h is given as a function of time
⎛ ⎞
⎡ ⎤ ⎡ ⎤
⎜ ⎟
⎢ ⎥ ⎢ ⎥
⎜ ⎟
− −
⎣ ⎦ ⎣ ⎦
⎝ ⎠
=
Assembled matrices for n elements analysis
0
2 0
0 2 0
0 0
0 2
0
stageI
D L D L
D L D L D L
D L D L D L
K
D L D L
D L D L D L
D L D L
− ⎡ ⎤
⎢ ⎥
− −
⎢ ⎥
⎢ ⎥ − −
=
⎢ ⎥
− −
⎢ ⎥
⎢ ⎥
− −
⎢ ⎥
−
⎣ ⎦
2
0
2 0
0 2 0
0 0
0 2
0
stageII
D L D L
D L D L D L
D L D L D L
K
D L D L
D L D L D L
D L D L h
− ⎡ ⎤
⎢ ⎥
− −
⎢ ⎥
⎢ ⎥ − −
=
⎢ ⎥
− −
⎢ ⎥
⎢ ⎥
− −
⎢ ⎥
− −
⎢ ⎥
⎣ ⎦
181
[ ]
1
1/ 3 0
0 1/ 6 0
0 1/ 6 0
6
0 0
0 1/ 6 0
0 1/ 3
CL
L
−
⎡ ⎤
⎢ ⎥
⎢ ⎥
⎢ ⎥
=
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎣ ⎦
2
0
0
0
,
0
stageI stageII
F F
c
⎡ ⎤
⎡ ⎤
⎢ ⎥
⎢ ⎥
⎢ ⎥
= =
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎣ ⎦
⎣ ⎦
.
.
since convection terms have been already added to
the [K]
Critical Time Steps
To find out whether the solution technique is stable or not, the behavior of
the solution as time t → ∞is monitored. If the behavior is oscillatory divergence,
the solution technique is unstable. In this work, a forward difference method was
used. This method is conditionally stable and the critical time step need to be
found. The critical time step can be found through trial and error starting by large
numbers of time steps and then decreasing it to a number which gives the stable
results with no oscillation. The monitored results are the moisture concentration
on the bottom, a
1
and moisture concentration on top, a
n
which are selected as
representative values of moisture concentration through the space domain.
As shown in below Figure C.3, the critical time steps for stage I drying is
152<Δt
crit
<153. Therefore Δt = 150 s was chosen for the rest of analysis.
Similarly, the critical time step for number of element (n=2) was found to be
between 37 and 38 s and therefore, 35 s was chosen for the rest of analysis. As
shown in Figures C.3 to C.6, these critical time steps decrease by descretizing the
space domain to more but smaller elements. At stage I for n=3, 16<Δt
crit
<17, and
for n=4, 9<Δt
crit
<10. Similarly at stage II drying, for n=1, 151<Δt
crit
<152, for n=2,
40<Δt
crit
<41, for n=3, 17<Δt
crit
<18, and for n=4, 9<Δt
crit
<10. The results are in
accordance with the formula for critical time step in the forward difference
method which is:
( )
2
2
2
( / )
crit
t
d
μ α δ
π
Δ ≅ which can be reduced to
2
2
2
crit
L
t
D π
⎛ ⎞
Δ ≅
⎜ ⎟
⎝ ⎠
where, d is the spatial dimension of the problem, μ and α are from D.E. and δ is
the distance between two adjacent nodes. This formula estimates the critical time
step for n=1 as:
182
( )
2
2
2 0.0125
1 76.85s <152s
4.12 7
crit
stageI
n t
E π
⎛ ⎞
= ⇒ Δ ≅ =
⎜ ⎟
−
⎝ ⎠
and for number of elements n = 2, estimates the critical time step as:
( )
2
2
2 0.0125
2 4.12 7 19.21s < 37 s
2
crit
stageI
n t E
π
⎛ ⎞
= ⇒ Δ ≅ − =
⎜ ⎟
⎝ ⎠
100
200
300
400
500
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=1
Δt = 153 s
100
200
300
400
500
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=1
Δt = 152 s
Figure C.3. Finding the critical time step (152<Δt
crit
<153) by trial and error for
FE analysis with 1 element
100
200
300
400
500
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=2
Δt = 38 s
100
200
300
400
500
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=2
Δt = 37 s
Figure C.4. Finding the critical time step (37<Δt
crit
<38) by trial and error for FE
analysis with 2 elements
183
3E+015
2E+015
1E+015
0
1E+015
2E+015
3E+015
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=3
Δt = 17 s
100
200
300
400
500
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=3
Δt = 16 s
Figure C.5. Finding the critical time step (16<Δt
crit
<17) by trial and error for FE
analysis with 3 elements
2e+169
1e+169
0
1e+169
2e+169
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=4
Δt = 10 s
100
200
300
400
500
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
,
k
g
/
m
3
0 4 8 12
Time, h
a1
an
Stage I Drying
Number of elements: n=4
Δt = 9 s
Figure C.6. Finding the critical time step (9<Δt
crit
<10) by trial and error for FE
analysis with 4 elements
184
Sensitivity of the 1D Analysis to the Number of Elements
The results of this FE analysis are intended to compare with the
experimental data. However, the experimental data are measured as the
cumulative moisture loss and not the direct measurement of moisture
concentration. Therefore, the integration of moisture concentration over the space
domain at each time step is compared with the initial moisture content to find the
cumulative moisture loss. For the purpose of analyzing the sensitivity of the
results to the number of elements, the cumulative moisture loss at the end of the
test is used. The number of elements increases from n=1 to n=5 which translates
into increase in the number of nodes from 2 to 6. As shown in Figure C.7, the
final cumulative moisture loss are approaching to the asymptotic results from the
upper side, but the difference is so insignificant that it may attributed to the noise
of the system.
64.09
64.095
64.1
64.105
64.11
64.115
F
i
n
a
l
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
2 3 4 5 6
Number of Nodes
Linear 1D elements
Figure C.7. Tracking the values of cumulative moisture loss at end of the test by
increasing number of elements
Results of 1D FE analysis with the MATLAB code
As shown in Figure C.8, the moisture concentration through the thickness
of the sample can be predicted by this method. The FE analysis results show that
there is not any significant difference between results of modeling with 1 or 2
elements. Also analyzing with smaller time steps does not have any effect on the
predicted results. To verify the results with the experiments, the experimental
moisture loss data are compared with the analytical results calculated as the
difference between initial total moisture content and the integration of moisture
concentration in the domain. The results show that the prediction is within 2% of
the experimental results. Slight difference between the results of analysis and
experiments is observed at the beginning of the stage II drying which can be due
to the transition from stage I to stage II and the nature of proposed different
boundary conditions. The comparison between results of 1D FE analysis by the
MATLAB code and the analytical results are shown in Figure C.9.
185
FE analysis by ABAQUS
The finite element analysis is carried out using ABAQUS v.6.101 [70]. The
technique utilizes the analogy between the heat transfer and the moisture
diffusion. The governing nonlinear differential equation for the diffusional
moisture transport in the domain of the problem can be written as:
2 2
2 2
( )
C C C
D C
t x y
⎡ ⎤ ∂ ∂ ∂
= +
⎢ ⎥
∂ ∂ ∂
⎣ ⎦
0
200
400
600
M
o
i
s
t
u
r
e
C
o
n
c
e
n
t
r
a
t
i
o
n
o
n
T
o
p
F
a
c
e
,
k
g
/
m
3
0 10 20 30 40
Time, h
n = 1
Δt = 150 s
n = 1
Δt = 5 s
n = 2
Δt = 35 s
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
Experiment
FE Analysis (n=1)
Δ
t
= 150 s
FE Analysis (n=1)
Δ
t
= 5 s
FE Analysis (n=2)
Δ
t
= 35 s
w/c = 0.45
No curing
T = 24
o
C
(a) (b)
Figure C.8. (a) Results of FE analysis with different elements and time steps. (b)
Comparing the results with experimental data
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
Experiment
MATLAB 1D Analysis
Analytical
w/c = 0.45
No curing
T = 24
o
C
Figure C.9. Comparing results of 1D FE analysis by the MATLAB code with
the analytical results
186
where C(x,y,t) is the moisture concentration varying in the domain with time, and
D(C) is the isotropic moisture diffusivity coefficient which is a function of C
itself. Moisture diffusivity at the boundary/surface of the sample during stage I
drying is in the form of J=F
0
and during stage II drying is in the form of a
convective boundary condition and can be written as:
*
( ) ( )
s
C
J D C k C C
n
∂
= − = − −
∂
where
C
n
∂
∂
is the moisture gradient at the drying surface with a unit normal “n”,
k is the convective moisture transfer coefficient,
Cs is the moisture concentration at the surface and
C* is the ambient moisture concentration.
Figure C.10. Cement paste slab subjected to different boundary conditions during
stag I and II drying
In equivalent transient thermal problem in 2D space, the corresponding
equations are given by:
2 2
2 2
T k T T
t c x y ρ
⎡ ⎤ ⎛ ⎞ ∂ ∂ ∂
= +
⎜ ⎟ ⎢ ⎥
∂ ∂ ∂
⎝ ⎠ ⎣ ⎦
whrere, T(x,y,t) is the temperature varying in the domain with time, and k is the
isotropic thermal conductivity which may be a function of T, ρ is the density and
c is the specific heat. While corresponding boundary conditions for heat transfer
during stage I can be in the same form as diffusion problem (J=F
0
), the heat
convection at the boundary for the stage II drying can be written as:
187
*
( )
fT s
T
h T T
n
∂
= −
∂
where
T
n
∂
∂
is the thermal gradient with a unit normal “n”,
fT
h is the convective transfer coefficient for thermal problem,
T
s
is the temperature at the boundary,
T
*
is the ambient temperature.
Comparing governing and boundary equations of heat transfer and diffusion
problems lead us to the conclusion of onetoone analogy between these two
boundary value problems under following correspondences:
Table C.1.
Analogy and corresponding terms between two boundary value problems
Diffusion Problem Heat Transfer Problem
C(x,y,t) T(x,y,t)
D(C) k/ρc
k h
fT
C
*
T
*
C
s
T
s
Descriptions and details of the modeling
Due to 1D flux, a thin slices like the one shown in the Figure C.11 can be
a representative of all other slices perpendicular to top surface. In that slice, the
concentration and flux will only changes according to the distance from the
bottom/surface of sample.
Figure C.11. A representative thin slice cut perpendicular to top surface for
modeling 1D moisture transport
An implicit heat transfer finite element analysis is carried out. The finite element
domain is shown in Figure C.12.
188
Figure C.12. Finite element domain
Material Properties
The material values used in the analysis are as follows:
(a) The moisture diffusivity of the slab as a function of concentration. The input
values come from the analytical works shown in Figure C.2. However, in that
figure, the values of diffusivity are shown as a function of time. Considering an
average moisture concentration at each time step, the values of diffusivity can be
plotted versus moisture concentration. This relationship is shown in Figure C.13
which is wellfitted by the CEBFIP model [154]. These values can be introduced
to Abaqus as isotropic concentrationdependent data in a tabular format.
0 200 400 600 800
Moisture Concentration, kg/m
3
1.0E008
1.1E007
2.1E007
3.1E007
4.1E007
5.1E007
6.1E007
7.1E007
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
Calculated from Experiment
CEBFIP Model
D
1
= 5.15 x 10
7
m
2
/s
D
0
= 1.86 x 10
8
m
2
/s
α = D
0
/D
1
= 0.03, n = 18
C
0
= 630 kg/m
3
C
c
= 120 kg/m
3
Figure C.13. Input values of diffusivity as a function of moisture concentration
(b) The density and specific heat are 1. Since in the mentioned analogy, D(C)
corresponds to k/ρc and the values of moisture diffusivity are introduced for
conductivity, the values of density (ρ) and specific heat (c) should be taken as 1.
Steps
Different steps defined for this specific problem are as follows.
189
(a) Initial step: In the initial step, a predefined field is created to be used as initial
condition. The initial value of C
i
= 520 kg/m
3
is applied to the whole body as the
initial condition.
(b) Step1: This step is defined to simulate stage I drying. Since the top boundary
conditions are different in the stage I from the stage II, these two steps have to be
created separately.
(c) Step2: This step is defined to simulate stage II drying. The initial condition in
this step is selected as “computed” which refers to the results computed in the
previous step.
Loadings
As shown in Figure C.14, the boundary condition of stage I drying
(constant flux: J=F
0
) is defined as a load with the magnitude of 0.42 kg/(m
2
h).
This value comes from the experimental analysis. This load is assigned to step1
as the boundary condition for stage I. The other boundary conditions which are
impermeable surfaces on the sides and bottom of the sample are ignored, since
zero flux corresponds to zero load.
Figure C.14. The boundary condition of stage I drying defined as a load
Interaction
A single interaction is created for the purpose of imposing a convective
boundary condition for the step2. Convective transfer coefficient and ambient
temperature are introduced into ABAQUS through defining a “surface film
condition” type interaction with timevariable film coefficient presented in a
tabular format and ambient temperature as a “sink temperature”. In this problem,
the moisture transfer coefficient shown in Figure C.2 is inputted in the table for
convective transfer coefficient and the sink temperature is taken 0 corresponding
to ambient moisture concentration in the vacuum desiccator. The interaction then
assigned to step2 and the load boundary condition (constant flux for stage I) is
inactivated for the step2.
Convergence Studies on Different FEM Meshes
190
Figure C.15a. Mesh1: 2 layers
Figure C.15b. Mesh2: 4 layers
Figure C.15c. Mesh3: 8 layers
Figure C.15d. Mesh4: 16 layers
Figure C.15e. Mesh5: 32 layers
Figure C.15f. Mesh6: 64 layers
Figure C.15. Different mesh sizes used for convergence studies
191
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
Figure C.16. Convergence studies on the results of evaporation tests with Q4
elements
Table C.2.
Results of FE analysis with ABAQUS using Q4 elements comparing to the
experimental results
Model
Element
type
Number
of nodes
Number
of
elements
Initial flux
(kg/(m
2
.h))
Final
Cumulative
moisture
loss (g)
Q4Slab1 DC2D4 63 40 0.30401 49.2899
Q4Slab2 DC2D4 105 80 0.36166 57.4078
Q4Slab3 DC2D4 189 160 0.39079 61.4349
Q4Slab4 DC2D4 357 320 0.40539 63.4345
Q4Slab5 DC2D4 693 640 0.41269 64.4335
Q4Slab6 DC2D4 1365 1280 0.41635 64.9328
Experiment    0.42 65
Experiment
Mesh1
Mesh2
Mesh3
Mesh4
Mesh5
Mesh6
192
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
Figure C.17. Convergence studies on the results of evaporation tests with Q8
elements
(DC2D8: An 8node quadratic heat transfer quadrilateral)
Table C.3.
Results of FE analysis with ABAQUS using Q8 elements comparing to the
experimental results
Model
Element
type
Number
of nodes
Number
of
elements
Initial flux
(kg/(m
2
.h))
Final
Cumulative
moisture loss
(g)
Q8Slab1 DC2D8 165 40 0.39298 61.7196
Q8Slab2 DC2D8 289 80 0.40662 63.5864
Q8Slab3 DC2D8 537 160 0.41335 64.5142
Q8Slab4 DC2D8 1033 320 0.41669 64.9723
Q8Slab5 DC2D8 2025 640 0.41835 65.2012
Q8Slab6 DC2D8 4009 1280 0.41918 65.3162
Experiment    0.42 65
Experiment
Mesh1
Mesh2
Mesh3
Mesh4
Mesh5
Mesh6
193
48
52
56
60
64
68
F
i
n
a
l
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 500 1000 1500 2000 2500
Number of Nodes
Q4
Q8
0.28
0.32
0.36
0.4
0.44
I
n
i
t
i
a
l
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
Figure C.18. Final cumulative moisture loss and initial evaporation rates versus
number of nodes
Comparing FE Analysis and the Experimental Results
Mesh and Element types
For the purpose of comparison with the analytical analysis which was
done in 100 layers, the same number of layers is used for meshing along y
direction (along the direction of external flux). As shown in previous section, this
number of element ensures the convergence of the results as well. Since there is
no moisture transfer along the xdirection (perpendicular to the direction of
external flux), the domain is divided by 20 layers. The type of analysis is an
implicit heat transfer analysis. Since the shape of domain is very simple and basic,
a 4node linear heat transfer quadrilateral element (DC2D4) is used in this
analysis.
(a) The results of FE analysis regarding moisture concentration at nodes are
shown in Figure C.19 in different times ranging from 2 to 40 h.
194
(a) t = 2 h
(b) t = 10 h
(c) t = 14 h
(d) t = 20 h
(e) t = 40 h
Figure C.19. Distribution of moisture concentration (kg/m
3
) at: a) t=2 h, b) t=10
h, c) t=14 h, d) t=20 h, e) t=40 h
195
(b) The results of FE analysis regarding the flux at integration points are shown in
Figure C.20 at different times ranging from 2 to 40 h.
(a) t = 2 h
(b) t = 10 h
(c) t = 14 h
(d) t = 20 h
(e) t = 40 h
Figure C.20. Distribution of moisture flux (kg/(m
2
h)) at: a) t=2 h, b) t=10 h, c)
t=14 h, d) t=20 h, e) t=40 h
(c) The results of FE analysis regarding normalized moisture loss can be obtained
by the integration of flux at the surface over the time. The total moisture loss
196
however, can be calculated as the product of normalized moisture loss and the
surface area. This result along with the results of experimental analysis, analytical
works and 1D FE analysis (MATLAB code) is shown in Figure C.21 and C.22.
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
Experiment
Analytical
Abaqus Results
w/c = 0.45
No curing
T = 24
o
C
Figure C.21. Comparing results of FE analysis by ABAQUS with the analytical
results
As shown in the Figure C.22, the results of 1D FE analysis are very close
to the ABAQUS results. Both results along with analytical results are within 2%
of the experimental data (Figure C.9 and C.21). It seems that the proposed 1D FE
method and Abaqus analysis can be used to analyze the drying of cementbased
materials in conditions where no analytical solution is available. The examples
are:
• Analysis of drying of internally cured concrete materials by presoaked
lightweight aggregate. In this case the lightweight aggregate acts as an
internal volume moisture source.
• Analysis of the slow process of drying in cementbased materials
consideration the aging factor. In this case, some amounts of moisture is
consumed by the hydration process which can be modeled as an internal
volume moisture sink.
• Analysis of drying of concrete element with complex shape, such as a
concrete samples with oblique surfaces, etc...
• Analysis of 2D/3D moisture transport through concrete elements.
197
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
Experiment
MATLAB 1D Analysis
Abaqus Results
w/c = 0.45
No curing
T = 24
o
C
Figure C.22. Comparing results of analysis by ABAQUS with the experimental
results and results of 1D analysis by the MATLAB code
Matlab Code
close all;clc;clear all;
% inputs 

% file and version
fname = 'n1_delta_t150s.dat'; % result file
colWidth = 14; % format column width
digit = 6; % format digit used
progVer = 'Forward_8 04/10/11'; % program version
% element numbers and size
n =1;
L = 0.0125/n;
% test constants
Area = 0.01; % surface are (m2)
F0 = 0.42/3600; % evaporation potential at stage I (kg/m2/s)
D_I = 5.15e7; % diffusion coefficient (m2/s)
t_trn = 10.5; % transition time from stage I to II
t_final = 40; % time of end of the test
ini_content_I = 530; % in kg/m3
% input D_II and h2 as a function of time (from analytical works)
198
t_DII = linspace(0,((t_finalt_trn)*3600),60);
D_II_time =[5.14722E07,4.86389E07,4.59444E07,4.34167E
07,4.10278E07,...
3.87778E07,3.66389E07,3.46389E07,3.27222E07,3.09167E
07,...
2.92222E07,2.76167E07,0.000000261,2.46639E07,2.11417E
07,...
1.99778E07,1.74417E07,1.54083E07,1.37528E07,1.29972E
07,...
1.17917E07,1.07611E07,9.84722E08,9.30556E08,8.79444E
08,...
8.31111E08,7.85278E08,7.42222E08,7.01389E08,6.62778E
08,...
6.26389E08,5.77778E08,5.46111E08,5.15833E08,4.875E08,...
4.60833E08,4.35556E08,4.11389E08,3.88889E08,3.675E08,...
3.47222E08,3.28056E08,0.000000031,2.93056E08,2.76944E
08,...
2.61694E08,2.47306E08,2.33694E08,2.20861E08,2.08694E
08,...
1.97222E08,1.86389E08,1.76139E08,1.66444E08,1.57278E
08,...
1.48639E08,1.40472E08,1.32722E08,1.25444E08,1.18528E08];
h2_time=
[0.001727,0.001727,0.001727,0.001727,0.001727,0.001727,...
0.001727,0.001727,0.001727,0.001727,0.001727,0.001727,0.001727,..
.
0.001727,0.001648,0.001545,0.00146,0.001392,0.001339,0.001297,...
0.001264,0.001238,0.001215,0.001195,0.001178,0.001166,0.001162,..
.
0.001162,0.001162,0.001161,0.001152,0.001142,0.001137,0.001137,..
.
0.001137,0.001137,0.001137,0.001137,0.001137,0.001137,0.001137,..
.
0.001137,0.001137,0.001137,0.001137,0.001137,0.001137,0.001137,..
.
0.001137,0.001137,0.001137,0.001137,0.001137,0.001137,0.001137,..
.
0.001137,0.001137,0.001137,0.001137,0.001137]/3600;
% Analysis of stage I 

% initial condition for stageI
a_I = ones([(n+1) 1])*ini_content_I;
% time steps for stageI (s)
delta_t_I=150; %n=1:152 %n=2:37 %n=3:16 %n=4:9 %n=5:6 %n=6:4
%n=8:2
199
% inverse of lumping capacity matrix (CL1)
CLm1 = zeros([(n+1) (n+1)]);
for i = 1:n+1
CLm1(i,i) = 6/L*1/6;
end
CLm1(1,1)=6/L*1/3;
CLm1(n+1,n+1)=6/L*1/3;
% internal force is zero, however because of the mixed boundary
condition
% (i.e. tau1=g1y1+c1 & tau2=h2y2+c2), F_i refers to [c1;..;cn]
% at stage I [c1,...,cn1] are zeros but cn is equal to outside
flux F0
F_I =zeros ([(n+1) 1]);
F_I(n+1) = F0;
% K matrix includes the convection coefficient in addition to
alpha terms
% at stage I there is no convection and therefore it just has
alpha terms
K_I = zeros([(n+1) (n+1)]);
for i = 1:n+1
K_I(i,i) =2*D_I/L;
if i1>0
K_I(i,i1)=D_I/L;
end
if i+1<n+2
K_I(i,i+1)=D_I/L;
end
end
K_I(1,1)=D_I/L;
K_I(n+1,n+1)=D_I/L;
% number of rounds for time domain of stageI
rounds_I = floor(t_trn*3600/delta_t_I);
% calculate {a}n based on {a}n1 according to forward difference
a_I_store = [];
for m = 1:rounds_I
M(m)=m; % store round numbers in an array
a_I = a_I+delta_t_I*CLm1*(F_IK_I*a_I);
a_I_store = [a_I_store a_I];% extend the {a}n matrix
D_I_t(m)=D_I;
h2_I_t(m)=h2_time(1);
end
% plot the results of {a}n vs. time
T_M=M*delta_t_I/3600; % convert rounds to actual time
figure(1);plot(T_M,a_I_store(n+1,M),'r',T_M,a_I_store(1,M),'b')
legend('a_n','a_1');
title('StageI: Moisture content on top and bottom of sample');
xlabel('Time, h')
ylabel('moisture content, kg/m3')
200
% cumulative moisture loss
sum_step_I = sum(a_I_store);
for i = 1:rounds_I
Integ_a_I(i) = (sum_step_I(i)*2a_I_store(1,i)
a_I_store(end,i))*L/2*Area*1000;
M_loss_I (i) = ini_content_I*(n*L)*Area*1000  Integ_a_I(i);
end
% plot the CML vs. rounds
figure(2);plot(T_M,M_loss_I(M),'r')
title('stageI: Cumulative Moisture Loss');
legend('Stage I:CML');
xlabel('rounds')
ylabel('Cumulative Moisture Loss, g')
fprintf(1, 'moisture loss at the end of stage I (g): %5.4f\n',
M_loss_I(end));
%n=1:43.801, %n=2:44.030, %n=3:44.072 %n=4:44.090 %n=5:44.093
%n=6:44.0953
%%n=8:44.098
% Analysis of stage II 

% initial condition for stageI
a_II = a_I_store(:,end);
% time steps for stageII (s)
delta_t_II=150; %n=1:160 %n=2:50 %n=3:18 %n=4:9 %n=5:6 %n=6:4
%n=8:2
% internal force is zero, and [c1,...,cn] are also zeros
F_II =zeros ([(n+1) 1]);
% number of rounds for time domain of stageII
rounds_II = floor((t_finalt_trn)*3600/delta_t_II);
% calculate {a}n based on {a}n1 according to forward difference
a_II_store = [];
for q = 1:rounds_II
D_II = interp1(t_DII,D_II_time,(q*delta_t_II),'spline'); %
diffusivity m2/s
%D_II = 5.15e7*(0.03+(1.03)/(1+((1
mean(a_II)/ini_content_I)/(1120/ini_content_I))^18));
% diffusion m2/s
h2 = interp1(t_DII,h2_time,(q*delta_t_II),'spline'); %
convection coefficient m/s
% K matrix includes the convection coefficient in addition to
alpha terms
% at stage II there is a convection term as "h2" which is
added to the last
201
% term in the K matrix
K_II = zeros([(n+1) (n+1)]);
for j = 1:n+1
K_II(j,j) =2*D_II/L;
if j1>0
K_II(j,j1)=D_II/L;
end
if j+1<n+2
K_II(j,j+1)=D_II/L;
end
end
K_II(1,1)=D_II/L;
K_II(n+1,n+1)=D_II/Lh2;
Q(q)=q; % store round numbers in an array
a_II = a_II+delta_t_II*CLm1*(F_IIK_II*a_II);
a_II_store = [a_II_store a_II];
D_II_t(q)=D_II;
h2_II_t(q)=h2;
end
% plot the results of {a}n vs. time
T_Q=Q*delta_t_II/3600+T_M(end); % convert rounds to actual
time
figure(3);plot(T_Q,a_II_store(n+1,Q),'r',T_Q,a_II_store(1,Q),'
b')
title('StageII: Moisture content on top and bottom of sample');
legend('a_n','a_1');
xlabel('Time, h')
ylabel('moisture content, kg/m3')
% cumulative moisture loss
sum_step_II = sum(a_II_store);
for l = 1:rounds_II
Integ_a_II(l) = (sum_step_II(l)*2a_II_store(1,l)
a_II_store(end,l))*L/2*Area*1000;
M_loss_II (l) = ini_content_I*(n*L)*Area*1000 
Integ_a_II(l);
end
% plot the CML vs. rounds
figure(4);plot(T_Q,M_loss_II(Q),'r')
title('satge II: Cumulative Moisture Loss');
legend('Stage II: CML');
xlabel('Time, h')
ylabel('Cumulative Moisture Loss, g')
fprintf(1, 'moisture loss at the end of stage II (g): %5.4f\n',
M_loss_II(end))
%n=1:
% combine two stages 

202
a_store =[a_I_store a_II_store];
M_loss = [M_loss_I M_loss_II];
MQ = [M Q+M(end)];
T_MQ = [T_M T_Q];
D_t = [D_I_t D_II_t];
h2_total = [h2_I_t h2_II_t];
% plot the input D vs. time
figure(5); plot(T_MQ,D_t(MQ),'r')
title('Input moisture diffusivity as a function of time');
legend('D(t)');
xlabel('Time, h')
ylabel('moisture diffusivity, m2/s')
% plot the input h2 vs. time
figure(6); plot(T_Q,h2_II_t(Q),'b')
title('Input convective moisture transfer coefficient as a
function of time');
legend('h2(t)');
xlabel('Time, h')
ylabel('convective moisture transfer coefficinet, m/s')
% plot the results of {a}n vs. time
figure(7); plot(T_MQ,a_store(n+1,MQ),'r',T_MQ,a_store(1,MQ),'
b')
title('Moisture content on top and bottom of sample');
legend('a_n','a_1');
xlabel('Time, h')
ylabel('moisture content, kg/m3')
% plot the CML vs. rounds
figure(8);plot(T_MQ,M_loss(MQ),'b')
title('Cumulative Moisture Loss During the Test');
legend('Cumulative Moisture Loss');
xlabel('Time, h')
ylabel('Cumulative Moisture Loss, g')
%  plot output results

% (1.1) print input and output response
fid1 = fopen(fname,'w');
fprintf (fid1,' program version : %30s\n\n',progVer);
% (1.2) print heading of result file
textHeading = {'n';'L';'Area';'F0';'t_trn';'tfinal'};
valHeading = [n,L,Area,F0,t_trn,t_final];
PrintHeading(fid1,textHeading,valHeading,colWidth,digit);
fprintf
(fid1,'**********************************************************
********************************************\n\n');
% (1.3) print table of result file
203
textData =
{'Time';'D_t';'h2_t';'top_moisture';'bott_moisture';'moist_loss'}
;
data = {T_MQ,D_t,
h2_total,a_store(n+1,:),a_store(1,:),M_loss};
NJtableResult = StoreTable(data);
NJ = NJtableResult{1};
tableResult = NJtableResult{2};
PrintTable(fid1,textData,tableResult,NJ,colWidth,digit);
fclose all;
204
APPENDIX D
Drying Tests on Clays: An Application of the Developed Test Method
205
Results of LowPressure Drying Experiments on Clay
Wet clay has different microstructure than Portland cement paste
including different particle size and distribution, and has no hydration
characteristics to change the microstructure by the time. To evaluate the
efficiency and applicability of the test method in capturing transport properties
and evaporation characteristics of geological materials, two oversaturated clay
samples were prepared and subjected to the vacuum drying. According to
Attenberg test method, the clay used in this study has Liquid Limit (LL) of 38.16,
Plastic Limit (PL) of 21.43, and Plastic Index (PI) of 16.73. Casagrade's plasticity
chart distinguishes the mineral as Montmorillonite clay. The clay was passed
through sieve No. 30 before placing in an oven maintaining a uniform
temperature of 110°C for 24 h. The clay was removed after the material has been
dried to a constant mass according to ASTM D 221605. The clay was then mixed
with water with watertosolid ratios of 0.45 and 0.6 and labeled by C0.45 and
C0.60 samples compared to Portland cement paste P0.45. The mix design is
shown in Table D.1.
Table D.1.
Mix proportions of the test series (kg/m
3
)
Test series
Portland
cement
Water
watertosolid
ratio
P0.45 1450 650 0.45
C0.45 1450 650 0.45
C0.60 1450 650 0.60
As shown in Figure D.1, after 48h drying the cumulative moisture loss of
cement paste sample P0.45 was 66 g while the cumulative moisture loss of C0.45
and C0.60 samples were 74 and 77 g. This shows that 95% and 82% of the
moisture initially added to the clay samples was evaporated at the end of the tests
which confirms the capability of the test method in simulating the drying process.
206
0 10 20 30 40 50
Drying Time, h
0
10
20
30
40
50
60
70
80
M
o
i
s
t
u
r
e
L
o
s
s
,
g
P0.45
C0.60
C0.45
0.1 1 10 100
Drying Time, h
0
0.2
0.4
0.6
0.8
D
r
y
i
n
g
R
a
t
e
,
K
g
/
(
m
2
.
h
)
P0.45
C0.60
C0.45
(a) (b)
Figure D.1. Results of lowpressure evaporation tests on clay samples and their
comparison with plain cement paste sample: (a) cumulative moisture loss curves;
(b) drying rate curves
The calculated evaporation rates during drying test are shown in Figure
D.1b. The average calculated drying rate at initial stages of drying for clay
samples with w/s ratio of 0.45 and 0.6 were 0.25 kg/(m
2
h) and 0.28 kg/(m
2
h),
respectively. Comparing these rates with 0.42 kg/(m
2
h) for the plain cement paste
sample (P0.45) indicates that clay decreases the initial evaporation rates during
early stage of drying. It may be related to the mass transfer at the boundary of the
material which appears to be taken place in a thin film inside porous medium
under the vacuum condition. This transition layer is defined as the location where
the pressure drops from the saturated water vapor pressure to the pressure of the
vacuum chamber [189]. This pressure drop induces a gas flux between porous
medium and the vacuum chamber, and determines the mass transfer at the
boundary of the material. During the drying process, the thickness of this
transition layer which controls the intensity of the mass flux is regulated by the
physical properties and microstructure of the material [189]. Therefore, the
difference in the physical properties and microstructure of clay and Portland
cement could be the reason of different initial evaporation rates.
This observation indicates that similar to cement pastes the drying of clay
under constant lowpressure conditions occurs in two distinct stages of a constant
drying rate period (stage I), and a falling drying rate period (stage II). During
stage I drying, the main phase transitions occur at the boundary surface and vapor
phase diffusion into the air is determining rate of evaporation. Reaching a critical
value of moisture content at the surface, the material is unable to support the
necessary flux to the surface and rate of evaporation starts to fall. During stage II
drying, the removal of moisture is mainly controlled by diffusion which
dominates the capillary forces [106]. The transition time from stage I to stage II
207
drying was significantly altered from 9.7 h for cement paste to 26.8 h and 24.1 h
for clays with w/s ratios of 0.45 and 0.60 respectively.
Simplified analysis of drying experiments on clay
In order to simulate the results of drying tests using a dualstage drying
concept, two different boundary conditions at the top surface are required. As
shown in Figure D.1, first, a boundary condition of constant flux is applied
followed by the boundary condition of constant concentration, that is set to
ambient moisture concentration (i. e. zero in this case). The moisture transfer
inside the matrix is simulated by Fick’s second law of diffusion which governs a
majority of moisture transfer problems in porous media [58, 108].
Figure D.2. Boundary conditions and Fick’s 2nd law of diffusion for the
presented drying problem
The analytical solution for the initial and boundary conditions of stage I
drying, assuming a constant diffusivity is given by Eq. (1) [109].
208
( )
2 2 2 2
0
2 2 2 2 2
1
3 2 ( 1)
, exp( )cos
6
n
I I
i
I
n
F L D t x L D n t n x
C t x C
D L
L L n L
π π
π
∞
=
⎫
− − − ⎪ ⎧
= + + −
⎨ ⎬
⎩
⎪
⎭
∑
(D.1)
where C is the moisture concentration (kg/m
3
), D is the moisture (liquid and
vapor) diffusion coefficient (m
2
/s) and t is the time (s). Also, C
i
is the initial
moisture concentration (kg/m
3
), F0 is the constant flux in stage I drying
(kg/(m
2
s)), and L is the thickness of sample (m). Using this solution, moisture
concentration at the top surface is obtained providing constant F0 and D
I
values.
Since stage I drying ends when the surface reaches equilibrium moisture
concentration (i. e. zero), moisture diffusivity at stage I drying is calculated as
long as the time of transition from stage I to stage II has not lapsed.
For stage II drying, analytical solution can be integrated over the thickness
of sample. The total amount of diffusing moisture in stage II drying, M
t
(kg), can
be obtained from the corresponding quantity after infinite time,
M
∞
(kg).
Therefore, moisture loss during the advanced phase of drying is given by [109]:
2 2 2
2 2
0
8
1 exp{ (2 1) / 4 }
(2 1)
t
II
n
M
D n t L
M n
π
π
∞
=
∞
= − − +
+
∑
(D.2)
Considering only first term of the series, moisture diffusivity, D
II
(m
2
/s), can be
derived by
2
2 ln
4
II
L
D a
π
−
= ⋅ when
2 ln
8
ln 1 ln
t
M
a t
M π
∞
⎛ ⎞
− − =
⎜ ⎟
⎝ ⎠
(D.3)
As shown in Table D.23 and Figure D.3, the diffusivities of clay samples C0.45
and C0.60 at the stage I drying were 1.4x10
7
and 1.1x10
7
m
2
/s which are 4 times
smaller than the initial diffusivity of cement paste. However, the diffusivities at
stage II drying for C0.45 and C0.60 samples were higher by 20% and 140%
compared to cement paste. While the difference between the results of clays and
cement pastes may be related to their physical and microstructural differences, the
difference between the results of two clay samples may be discussed by different
amount of moisture concentrations. Unlike the stage I drying, diffusion
coefficients in Stage II drying depend on moisture concentration [110]; however,
in this analysis diffusivity in stage II is taken constant resulting in calculation of
an average diffusivity, which is averaged over space and time.
209
Table D.2.
Initial evaporation rates, transition time, evaporation rate at 24 h, cumulative
moisture loss, and diffusivities of tested samples
Test
series
Initial
evaporation
rate
(kg/(m
2
.h))
Transitio
n time of
drying
stages
(h)
Evaporation
rate at 24 h
(kg/(m
2
.h))
Cumulative
moisture
loss at 48 h
(g)
Moisture
diffusivity
at stage I
(m
2
/s)
Moisture
diffusivity
at stage II
(m
2
/s)
P0.45 0.42 9.7 0.024 66 5.1E07 3.3E09
C0.45 0.28 26.8 0.167 74 1.4E07 8.0E09
C0.60 0.25 24.1 0.176 77 1.1E07 4.0E09
P0.45
0
2E007
4E007
6E007
8E007
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
C0.60
0
2E009
4E009
6E009
8E009
1E008
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
C0.45
Stage I Stage II
Figure D.3. Results of diffusion analysis on wet clay and cement paste samples
Advanced Moisture Transport Analysis on the Drying Clay
The advanced modeling of drying of cementitious materials as porous
media is presented in the Appendix. It first provides an overview of existing
theories for the drying of porous media using a single nonlinear diffusion
equation, which is accounted for all processes namely, Darcian transport of liquid
and vapor water, and diffusion of vapor in dry air. The main mechanisms of
drying are identified accordingly in two separate stages of drying based on
210
connectivity of liquid and vapor phases inside the pores, observed qualitatively
through experiments. Then a quantitative model of drying is proposed, in which
each mechanism is accounted for by appropriate equations. It is shown that the
developed model can well predict the experimental data of the drying of cement
based materials, and moisture distribution as well as moisture diffusivity and its
variations with time. The same analytical method is applied to clay samples.
Results of analysis are shown in Figure D.4 which presents a fairly well
prediction of the drying data by the variation of diffusivity and slight variation of
moisture transfer coefficient. Results of analysis show that the diffusion
coefficients vary in the range of 2.2x10
7
to 1.34x10
8
m
2
/s. These results are in
good agreement with the results of Ketelaars [ 190] who reported diffusivity
values in the range of 1x10
7
to 1x10
9
m
2
/s for different clay samples.
Image Analysis on EarlyAge Cracked Samples
Timelapse photography shows that visible earlyage cracks were detected
in the clay samples sample as early as 6 h and it developed throughout the test.
The cracking area versus time is shown in Figure D.6 and Figure D.7 shows a
cracked clay sample after 48 h of vacuum drying.
0
20
40
60
80
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30 40
Drying Time, h
Experiment
Simulation
0
0.1
0.2
0.3
0.4
0.5
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, h
Drying of Clay
w/s = 0.45
(Figure D.4(a))
211
1x10
7
2x10
7
3x10
7
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 10 20 30 40
Drying Time, h
K
D'
0.001
0.0015
0.002
M
o
i
s
t
u
r
e
T
r
a
n
s
f
e
r
C
o
e
f
f
i
c
i
e
n
t
,
m
/
h
1x10
8
(Figure D.4(b))
Figure D.4. (a) Prediction of cumulative moisture loss and evaporation rate as a
function of time for the experimental data of an evaporation test, (b) Back
calculated D and k for the best fit of data
drying time: 6 h 7 h 8 h 9 h
drying time: 10 h 11 h 12 h 13 h
Figure D.5. 2D Crack development during drying of a clay sample
212
0 4 8 12 16
Drying Time, h
0
1
2
3
C
r
a
c
k
A
r
e
a
,
i
n
2
Clay (w/s=0.45)
Cement Paste (w/c=0.45)
Figure D.6. Comparing cracking area of a clay sample vs. time with a Portland
cement paste
Figure D.7. Image of a 2D crack pattern in a clay sample after vacuum drying for
48 h
213
APPENDIX E
Image Analysis on Earlyage Cracks: Quantification and Orientation
214
Quantification of Crack Parameters
Several parameters for the quantification of crack patterns are presented.
The parameters are used to compare shrinkage cracks in different samples. First,
these parameters are introduced and then their values for different samples using
an automatic procedure developed by a computer code are compared.
Interesting parameters for quantification of cracks in concrete samples
1. Total area of cracks, A (mm
2
): A is the total area of cracks on the image.
2. Areal fraction, A
A
(mm
2
/mm
2
): A
A
is the ratio between the areas of the
counted cracks to the entire area of the image.
3. Total length of cracks, L (mm): L is the total dendritic length of all the
skeletonized cracks in the image.
4. Average crack length, L
mean
(mm): L
mean
is the mean value of the dendritic
length of all cracks in the image.
5. The crack density, L
A
(mm/mm
2
): L
A
is the total dendritic length of cracks
per image area.
6. Average width of cracks, W (mm): W is the total area of cracks (A)
divided by total dendritic length (L).
7. Total number of cracks, N
total
(): N
total
is the total number of objects in the
binary image of cracks.
8. Number of cracks per unit area, N
A
(1/mm
2
): N
A
is the total number of
features (cracks) divided by the area of image.
9. Total number of crack intersections, N
Int
(): N
Int
is the count of
intersection points in the image of crack.
10. Average nearest neighbor distance in plane, Δ
2
(1/mm): Δ
2
presents a
measure for the average centretocentre distance of randomly dispersed
nearest neighbors in a plane which is closely related to the configuration
and can be demonstrated as
1/ 2
2
0.5
A
N Δ = .
Beside above parameters, length and width of all detected cracks in the image are
calculated. The graphs of normal probability density and normal cumulative
density present more information of the distribution of crack length and crack
width in the image.
215
Automatic procedures for the quantification of crack parameters
Step 1: make a binary image of crack pattern Step2: Detect cracks intersection
points from skeletonized image
of crack
Step 3: Dilate the intersection points Step 4: Subtract dilated
intersection points from the
initial binary image
Figure E.1. The process of image analysis on a cracked paste sample
216
0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 1.2
0
0.5
1
1.5
2
2.5
3
normal probability density  crack width (mm)
0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 1.2
0
0.2
0.4
0.6
0.8
1
normal cumulative density  crack width (mm)
10 5 0 5 10 15
0
0.05
0.1
0.15
0.2
0.25
normal probability density  crack length (mm)
10 5 0 5 10 15
0
0.2
0.4
0.6
0.8
1
normal cumulative density  crack length (mm)
Figure E.2. Normal probability and cumulative density curves of crack width and
crack length
217
Table E.1.
Results of crack analysis on a plain and FRC samples
Parameters obtained from crack analysis Plain Sample
Glass Fiber
Sample
Total length of cracks (mm) 621.0 582.3
Total cracking area of image (mm
2
) 200.4 81.7
Density of cracks in image (1/mm) 0.062 0.058
Mean crack length (mm) 1.13 0.60
Total areal fraction of cracks () 0.020 0.008
Average width of cracks (mm) 0.32 0.14
Number of cracks per unit area (mm
2
) 0.05 0.10
Average nearest neighbor distance (mm) 2.13 1.61
Total number of crack intersections () 77 55
Total number of cracks in image () 551 969
Degree of orientation of cracks (01) 0.099 0.084
Average specific number of intersections
(1/mm)
0.066 0.053
218
Determination of Crack Orientations in EarlyAge Cracked Cement Pastes
Concept
The degree of crack orientation is determined on the basis of the classical
stereological technique of oriented secants on a plane [191, 192, 193, 194]. In this
technique, parallel equidistant straight lines cover the view to be analyzed. A
count of the intersections of these lines and the cracks network is performed. The
number of intersections is divided by the total length developed by the straight
lines in order to get the specific number of intersections, which is characteristic of
the specific surface of cracking.
N
L
= specific number of intersections
S
v
= specific surface of cracking = 2N
L
Figure E.3. Analysis by counting number of intersections of cracks network with
the oriented secants (parallel equidistant lines)
In case of an anisotropic cracking, these characteristics are calculated by
averaging the different values obtained by rotating the family of straight lines.
N
L
(Θi) = specific number of intersections in the orientation Θi
N
L
=
n
L
1
1/n N ( i) Θ
∑
; where n = number of orientations
S
v
= 2N
L
For the purpose of emphasizing the anisotropy, the different values of N
L
(Θi) are
reported in a polar referenced plane, in a figure which is named "rose of the
number of intersections". The anisotropy is characterized by a scalar, called
degree of orientation, ω, given as
max min
max min
( )
( 1)
2
L L
L L
N N
N N
ω
π
−
=
+ − ⋅
219
where; N
Lmax
and N
Lmin
are the maximum and minimum values taken by N
L
respectively. Degree of orientation, ω, takes the value 0 for a perfectly isotropic
crack pattern and the value 1 when all the cracks have the same orientation
(extreme anisotropy).
Automatic Process of Orientation Analysis
The automatic procedure elaborated for the estimation of crack pattern
orientation includes;
• Converting grayscale (8bit unit) images of cracked sample into binary
images.
• Reducing the cracks to single pixel lines by binary thinning, known as
skeletonization.
• Providing families of equidistant parallel lines in different angles (from 0
o
to 180
o
, here steps are 15
o
) in the format of binary images.
• Superimposition of the image of crack network and image of parallel
equidistant lines with a given direction using the logical binary operation
“AND”.
• Counting the number of intersections of the cracks with the set of lines.
• Measuring total length of parallel lines covering the image.
• Calculating specific number of intersections defined as the number of
intersections divided by the total length of parallel lines.
• Doing all above steps for the next incremental angles.
• Calculating average specific number of intersections and degree of
orientation for the crack network as given in the previous section.
Testing and Calibration of Crack Orientation Analysis Process
In order to insure the reliability of this image analysis process, a series of
trial shapes and figures are analyzed. These trials include; a circle, a series of
parallel lines, and four different rectangles. Since the circle is perfectly isotropic
and parallel lines are completely oriented to one direction and have extreme
anisotropy, the expected results for the orientation degrees are 0 and 1
respectively. The rectangles tested here have different values of length to width
ratio as; 1 (square), 2, 3 and 6. The analysis should be capable of capturing the
effect of increase in the values of orientation degree by the elongation of
rectangle. Here are the results;
220
1) For a circle:
Degree of
orientation: 0.0519
Rose of the
number of
intersections:
2) For parallel lines:
Degree of
orientation: 1.0000
Rose of the
number of
intersections
3) For a rectangle
with length to width
ration of 1 (square):
Degree of
orientation: 0.3328
Rose of the
number of
intersections:
221
4) For a rectangle
with length to width
ration of 2:
Degree of
orientation: 0.5078
Rose of the
number of
intersections:
5) For a rectangle
with length to width
ration of 3:
Degree of
orientation: 0.6136
Rose of the
number of
intersections:
6) For a rectangle
with length to width
ration of 6:
Degree of
orientation: 0.7466
Rose of the
number of
intersections:
222
Crack Orientation Analysis on Cement Paste Samples
The process of image analysis on an earlyage cracked cement paste
sample (plain sample 1) after 24 hours of vacuum drying is shown as below;
original grayscale image (8bit unit) Æ binary image (black and white) Æ opposite of
binary image
skeletonization of crack network Æ superimposing secants with crack image Æ points of
intersection
0
45
90
135
180
225
270
315
0 0.02 0.04 0.06 0.08
Specific Number of Intersections,1/mm
Rose of the number of intersections:
ω: 0.0998
N
L
: 0.0666 mm
1
S
v
: 0.1333 mm
1
223
The results of analysis for the other samples are as follow;
0
45
90
135
180
225
270
315
0 0.02 0.04 0.06 0.08
Specific Number of Intersections,1/mm
Plain cement paste (sample 2) Rose of the number of intersections
0
45
90
135
180
225
270
315
0 0.02 0.04 0.06 0.08
Specific Number of Intersections,1/mm
ARglass fiber cement paste (sample 1) Rose of the number of intersections
ARglass content: 3
kg/m
3
ω: 0.0922
N
L
: 0.0727 mm
1
S
v
: 0.1455 mm
1
ω: 0.1376
N
L
: 0.0677 mm
1
S
v
: 0.1354 mm
1
224
ω: 0.0844
N
L
: 0.0528 mm
1
S
v
: 0.1056 mm
1
0
45
90
135
180
225
270
315
0 0.04 0.08 0.12
Specific Number of Intersections,1/mm
ARglass fiber cement paste (sample 2) Rose of the number of intersections
ARglass content: 3
kg/m
3
0
45
90
135
180
225
270
315
0 0.02 0.04 0.06 0.08
Specific Number of Intersections,1/mm
ARglass fiber cement paste (sample 1) Rose of the number of intersections
ARglass content: 6
kg/m
3
ω: 0.1272
N
L
: 0.1070 mm
1
S
v
: 0.2140 mm
1
225
0
45
90
135
180
225
270
315
0 0.02 0.04 0.06 0.08
Specific Number of Intersections,1/mm
ARglass fiber cement paste (sample 2) Rose of the number of intersections
ARglass content: 6
kg/m
3
Table E.2.
Results of crack analysis and quantifications
Sample
Degree of
Orientation
, ω
Specific
Number of
Intersections,
N
L
(mm
1
)
Specific
Surface of
Cracking, S
v
(mm
1
)
Plain cement paste
0.096
0.0697
0.1393
ARglass cement paste (3 kg/m
3
) 0.1324
0.0874
0.1747
ARglass cement paste (6 kg/m
3
) 0.1097
0.0538
0.1076
ω: 0.1350
N
L
: 0.0584 mm
1
S
v
: 0.1168 mm
1
226
APPENDIX F
Modeling Earlyage Shrinkage of Concrete Due to High Rate of Evaporation
227
Drying of CementBased Materials
Governing Differential Equations
Drying of cementbased materials can be considered as a moisture flow
through porous media. Darcy's law can be used to obtain the flux of liquid water
in a saturated porous medium. [195]:
l
l
l
k
q p
μ
= − ∇ (F.1)
where q
l
is the flux of liquid water, k
l
is the liquid permeability, μ
l
is the dynamic
viscosity of liquid water and p is the macroscopic pressure. However, except in
very early stage of drying, the concrete is not fully saturated. In nonsaturated
state, the liquid water is transported because of gradient of the capillary pressure
[160]. The liquid water flux for transport through a nonsaturated medium can be
obtained by the extended Darcy’ law as presented in equation (F.2) In this
equation ( )
l
k θ , the effective permeability of liquid water, is a function of
moisture content.
( )
( )
l
l c
l
k
q p K
θ
θ
μ
= − ∇ = − ∇Ψ (F.2)
K is the hydraulic conductivity defined as ( ) /
l l
K g k ρ θ μ = and Ψ is the
hydraulic potential presented by
c l
p ( ) / g Ψ θ ρ = . Defining the liquid diffusivity
as
,
( / )
l l
D K
θ
θ = ∂Ψ ∂ , liquid water transport is presented as
, l l l
q D
θ
θ = − ∇ (F.3)
Conservation of mass for liquid water expressed in volumetric quantities can be
written as:
( )
,
/ . .
l l l v l l l v
t q E D E
θ
θ θ
− −
∂ ∂ = −∇ − = ∇ ∇ − (F.4)
On the other hand, the vapor transport under isothermal conditions at the
microscopic level is given by Fick's law:
1
v v v
l
M
q D p
RT ρ
= − ∇ (F.5)
where, q
v
is the flux of water vapor, ρ
l
is the density of liquid water,
v
D is the
diffusion coefficient of water vapor in air, M is the molecular mass of water, R is
the gas constant, T is the absolute temperature, and p
v
is the vapor pressure. The
term M/RT results from the ideal gas law, whereas ρ
l
results from the
transformation of the mass flux for vapor into an equivalent volumetric liquid
water flux [141]. Pel (1995) showed that at the macroscopic level, macroscopic
volumetric flux through a porous medium with only air and vapor is obtained
from equation (F.6).
, v v l
q D
θ
θ = − ∇ (F.6)
where
*
,
( )
( )
l
v l v vs
l l
n M h
D T D P
RT
θ
θ
θ
ρ θ
⎛ ⎞ − ∂
=
⎜ ⎟
∂
⎝ ⎠
(F.7)
228
where n is the porosity, (nθ
l
) is the voidspace available for vapor diffusion, T*
is the tortuosity accounting for the extra path length resulting from the tortuous
pores, D
v
is the vapor diffusivity, p
vs
is the saturation pressure of water vapor, and
h is the relative humidity. The macroscopic conservation of mass for water vapor
expressed in volumetric quantities can be written as:
.
v
v l v
q E
t
θ
−
∂
= −∇ +
∂
(F.8)
where
l v
E
−
is the rate of evaporation.
To find a single nonlinear diffusion equation for water in both phases,
equation (F.4) and equation (F.8) are combined.
( )
, ,
( )
.
v l
l v l
D D
t
θ θ
θ θ
θ
∂ +
= ∇ + ∇
∂
(F.9)
Defining moisture diffusivity as
, , l v
D D D
θ θ θ
= + , equation (F.9) is represented as:
( )
( )
.
v l
l
D
t
θ
θ θ
θ
∂ +
= ∇ ∇
∂
(F.10)
Assuming equilibrium between water and vapor, the mass density of the vapor in
the pores and total moisture content are given as:
1
l v v l
v l l
l v
n
n
θ θ ρ ρ
θ θ θ θ
ρ ρ
⎛ ⎞
= ⇒ = + = +
⎜ ⎟
⎝ ⎠
(F.11)
At the room temperature, the density of water vapor is 0.74 kg/m
3
.
4
7.4 10
1
l
n
θ θ
−
⎛ ⎞ ×
= +
⎜ ⎟
⎝ ⎠
(F.12)
As one can see from equation (F.12), for liquid water content to vapor content
ratio greater than about 10
3
, the total moisture content can be appropriately
approximated by the liquid water content. Therefore, equation (F.10) can be
represented as a nonlinear diffusion equation for total moisture as:
( ) . D
t
θ
θ
θ
∂
= ∇ ∇
∂
(F.13)
Dualstage Drying Methodology
Studies of drying behavior of clay brick ceramics by Hall et al. [99],
Platten [146], Cooling [100], and different wet capillaryporous materials by
Kowalski [101] indicate that evaporative flux at the boundary is roughly constant
for a period of time at the beginning of the drying referred to as stage I and then
falls significantly by the time in the second period or stage II. The same drying
behavior was observed for the Portland cement pastes in recent experimental
works by authors [148], as well as in mortar [149] and lightweight concrete [66].
In most of earlyage drying tests, the moisture transport through the sample is
assumed to be onedirectional (1D) due to high surface areatodepth ratio, and
exposing only top surface to the drying condition. For modeling the internal
229
moisture transport in stage I and stage II drying, 1D form of equation (F.13) is
used with the replacement of dimensionless moisture concentration variable, θ (),
with the dimensional variable C (kg/m
3
).
2
2
2
( )
C C dD C C
D C D
t x x dC x x
∂ ∂ ∂ ∂ ∂ ⎡ ⎤ ⎛ ⎞
= = +
⎜ ⎟
⎢ ⎥
∂ ∂ ∂ ∂ ∂
⎣ ⎦ ⎝ ⎠
(F.14)
The only difference in applying this formula in stage I and II, is the constant
diffusivity at stage I, while the diffusivity is variable at stage II drying. The
assumption is based on experimental observations that concrete is capillary
saturated during stage I [105] but unsaturated at later stages of drying and its
value will be a strong function of moisture concentration. Therefore, the time of
drying is divided into several time steps at stage II and constant diffusivity is
applied at each time step with respect to the time and the location. Using this
method, the timevarying diffusivity values at stage II are only averaged over the
space. This can be acceptable considering high evaporative surfacetodepth ratio
of the samples. Consequently, equation (F.14) can be simplified to equation
(F.15) for stage I as well as each time step for stage II.
2
2
C C
D
t x
∂ ∂
=
∂ ∂
(F.15)
Hydration of cement and its implications on transport properties need to
be taken into account during earlyage drying of cementitious materials. With
time, liquid water will be chemically bounded into hydration products as well as
physically bound water to increasing internal solid surfaces. Both chemically and
physically bound water may be assumed to be in local equilibrium with free water
due to rapid rate of hydration at the early age. To simplify the concept, the
concentration of bound water, S, is directly taken proportional to the
concentration of the free water, C, the substance free to diffuse.
S = RC (F.16)
During earlyage drying when the moisture diffusion is accompanied by
immobilization of bound water, equation (F.15) is modified as:
2
2
C C S
D
t x t
∂ ∂ ∂
= −
∂ ∂ ∂
(F.17)
Equation (F.16) can be presented in a new form by substituting for S from
equation (F.16).
2 2
2 2
1
C D C C
D
t R x x
∂ ∂ ∂
′ = =
∂ + ∂ ∂
(F.18)
Replacing the term D/(1+R) with D’ as the effective moisture diffusion
coefficient, the equation is seen in the usual form of diffusion equations.
Therefore, D’ presents the potential for conductivity of free water through the
fresh cementitious materials.
The cement paste is assumed to be homogeneous at initial conditions with an
impermeable bottom surface. Therefore, the initial and boundary conditions are
0, 0
i
t C C x L = = ≤ ≤ (F.19)
230
at 0, 0
C
x J D
x
∂
′ = = − =
∂
(F.20)
where C
i
is the initial moisture concentration (kg/m
3
), and J is the moisture flux
(kg/m
2
).
Moisture transport during stage I drying is considered as an external mass transfer
control process with a constant drying rate. Thus, the boundary condition at the
top surface is
0
at , = = x L J F (F.21)
where, F
0
is the constant flux in stage I drying (kg/m
2
s). Solution to moisture
concentration values as a function of time and location during stage I drying
based on equations (F.1821) is [109]
( )
2 2 2 2
0
2 2 2 2 2
1
3 2 ( 1)
, exp( ) cos
6
n
i
n
F L D t x L D n t n x
C t x C
D L
L L n L
π π
π
∞
=
⎫
′ ′ − − − ⎪ ⎧
= + + −
⎨ ⎬
′
⎩
⎪
⎭
∑
(F.22)
In this model transition time from stage I to stage II, referring to the loss of
continuity in the liquid phase, is defined by the user which may be obtained from
experimental data. Prediction of the transition time based on physical grounds
requires further studies. For modeling stage II drying based on the concept of
convectivediffusive transfer of vapor at the surface, the exchanged flux of
moisture, J, can be characterized by the convection equation .
*
( )
s
C
J D k C C
x
∂
′ = − = −
∂
(F.23)
where k (m/h) is convective moisture transfer coefficient or surface factor, C
s
is
the surface moisture concentration and C
*
is the ambient moisture concentration
(kg/m
3
). k parameter takes into account the wind velocity, the surface
temperature, and the relative humidity for drying to the atmosphere or in the case
of lowpressure drying encompasses vacuum condition and temperature. The
analytical solution for equations (F.1821) during each time step at stage II drying
with the convective boundary condition as of equation (F.23) is given by [153]
( ) ( )
2
2
1
2sin( )
, cos( )
sin( )cos( )
D t
n
n L
i n
n n n
n
x
C t x C C C e
L
λ
λ
λ
λ λ λ
⎛ ⎞ ′
−
∞ ⎜ ⎟
⎝ ⎠
∗ ∗
=
= + − +
+
∑
(F.24)
Where,
tan( )
n n
kL
D
λ λ =
′
(F.25)
Geometry, governing equations and boundary conditions during stages I and II
drying are presented in Figure (F.1).
231
Figure F.1a
Figure F.1b
Figure. F.1. Geometry, governing equations and boundary conditions during: (a)
stage I drying, (b) stage II drying assuming convective moisture flow on the top
surface
232
BackCalculation Procedures
A back calculation model is used to predict diffusivity and moisture
transfer coefficient at any time step based on given cumulative moisture loss and
evaporation rate curves. Optimization is implemented to find two best fit
parameters, D’(t) and k(t) to predict the experimental results. The objective
function for minimization is the weighted sum of two errors of cumulative
moisture loss and flux. In optimization algorithm, normalized design variables (D’
and k) are used to avoid numerical problems caused by mixing big and small
numbers in numerical operations. Thus, diffusion coefficient at the current time
step is normalized with its value at previous time step, D
norm
=D’
i
/D’
i1
, so does
the mass transfer coefficient k
norm
= k
i
/k
i1
. Once the optimal solution D
norm
and
k
norm
are found, they are reverted back to D’ and k. Inequality constraints are
imposed to the design variables D’ and k in terms of lower and upper bounds.
Initial values for D’ are obtained from analysis of stage I drying, while initial
values for k are estimated from equation (F.26) which is derived by rearranging
equation (F.23).
(exp)
( )
*
( )
est
trans
crit
J t t
k
C C
=
=
−
(F.26)
where, k
(est)
is the estimated surface factor (m/h) and
(exp)
( ) =
trans
J t t is the
experimental rate of evaporation at the transition time which is roughly equal to
initial evaporation rate. C
crit
is the surface moisture concentration at end of stage I
or critical moisture concentration (kg/m
3
) and C
*
is the ambient moisture
concentration. For other time steps, initial D’ values for optimization process,
( ) est
D are estimated by the assumption of proportional decrease of diffusivity
with the flux, while initial k values, k
(est)
are estimated by a convection transport
formula.
(exp) ( )
( )
1
(exp) ( 1)
( )
( )
−
−
=
i
est
i
i
J t
D D
J t
(F.27)
(exp) ( )
( )
( 1)
*
( )
( , )
i
est
i
J t
k
C t x L C
−
=
= −
(F.28)
In equation (F.27) and (F.28),
( ) i
t is the current time step, while
( 1) − i
t is the
previous time step. At each time step, these two initially estimated parameters,
D
(est)
and k
(est)
are passed to the optimizer and the optimizer calls the function
evaluation to calculate the error and update these two parameters several times
until optimal solution for D’ and k are found. Results of back calculations for the
reported data by Selih (1996) are shown in Figure (F.2) and (F.3).
233
0
20
40
60
C
u
m
u
l
a
t
i
v
e
M
o
i
s
t
u
r
e
L
o
s
s
,
g
0 10 20 30
Drying Time, d
Selih (1996)
Simulation
0
0.05
0.1
0.15
E
v
a
p
o
r
a
t
i
o
n
R
a
t
e
,
k
g
/
(
m
2
.
h
)
0.1 1 10 100
Drying Time, d
w/c = 0.40
T = 22
o
C
RH = 50%
No Initial Curing
Figure F.2a
6x10
9
1x10
7
2x10
7
3x10
7
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
0 10 20 30
Drying Time, d
K
D'
0
0.0005
0.001
0.0015
M
o
i
s
t
u
r
e
T
r
a
n
s
f
e
r
C
o
e
f
f
i
c
i
e
n
t
,
m
/
h
Figure F.2b
Figure F.2. (a) Prediction of cumulative moisture loss and evaporation rate as a
function of time for the experiment performed by Šelih (1996), (b) Back
calculated D’ and k for the best fit of data
234
0.4 0.6 0.8 1
Normalized Moisture Concentration
7.00E009
1.07E007
2.07E007
3.07E007
4.07E007
M
o
i
s
t
u
r
e
D
i
f
f
u
s
i
v
i
t
y
,
m
2
/
s
BackCalculated Diffusivities
CEBFIP: n = 10
CEBFIP: n = 15
CEBFIP: n = 20
D
1
= 2.78 x 10
7
m
2
/s
D
0
= 6.94 x 10
9
m
2
/s
α = D
0
/D
1
= 0.025
C
c
= 0.7
Analysis on
Experiments by
Selih (1996)
Figure F.3a
0 40 80 120 160 200
Moisture Concentration, kg/m
3
0
0.2
0.4
0.6
0.8
1
t
/
h
Exp
Sim
Exp
Sim
Exp
Sim
Exp
Sim
1 d
3 d
7 d
28 d
Time
h = 0.1 m
t
h
Figure F.3b
Figure F.3. Results of analysis on drying data by Šelih (1996): (a) Moisture
diffusivity as a function of normalized moisture concentration expressed by CEB
FIP models, (b) moisture concentration distributions through the thickness of the
sample in comparison with experimental results
235
Shrinkage Strains and Drying
Simplified Models
Some researchers have focused on developing linear relationship between
shrinkage strains and the relative humidity. Wittmann and Roelfstra [ 196 ]
described the unrestrained hygral length change,
sh
ε , as a function of relative
humidity, h, by a linear relationship such as:
.
h
sh sh
h ε α = Δ (F.29)
where
h
sh
α is an empirical shrinkage coefficient depending on concrete
composition and maturity. Alvaredo and Wittmann [197] have used this linear
relationship with the
h
sh
α ranging from 1.52.5x10
3
. Bazant and Xi [198] and
Bazant et al. [199] also introduced the same concept into their models. According
to van Zijl [ 200], the experimental measurement indicate that the empirical
shrinkage coefficient,
h
sh
α is constant in the range 0.4 < h < 0.99.
Similar efforts have been made on linearly relating moisture content to the
shrinkage strain assuming a quasilinear relationship between the moisture
content and the relative humidity in the range of 50100% in desorption isotherm
curves [201, 202]. Thelandersson et al. [203] modeled the developed strain in the
drying concrete according to equation (F.29).
.
C
sh sh
C ε α = Δ (F.30)
where C is the moisture content (), and
C
sh
α is the coefficient of hygral
contraction which can be obtained from the linear part of unrestrained shrinkage
versus weight loss curve [204].
The major reason for dealing with water content rather than relative humidity is
that water content can be more easily measured by weight loss monitoring [55].
Therefore, there is no requirement for desorption isotherms of the concrete to
relate humidity, h, to the water content, C. Granger et al. [205] and Verbeck [206]
showed that particularly when h remains above 50 percent, this relationship is
confirmed fairly well by experiments. This model has been used in hydro
mechanical models accounting for the full coupling of drying, shrinkage, creep
and cracking [207]. Benboudjema et al. [208, 209, 210] used the coefficient of
hygral contraction as
5
=1.22×10
C
sh
α .
The advantage of these simplified models (equations F.2930) is an easy
implementation in drying models to obtain developed strains and related stresses.
Having moisture distributions at different time in drying of a concrete sample, and
using a Finite Element Method (FEM), the moisture concentration can be easily
translated into an internal stress distribution. This stress distribution can be finally
used to calculate the total deformation of a drying specimen [196]. However, the
drawback is the lack of a physical interpretation between nature of the porosity in
the material and the developed strains.
236
Capillary PressureBased Models
Experimental results [205, 211] and models based on the mechanics of porous
media [212, 213, 214] reveal the nonlinearity between the shrinkage strain and
relative humidity or water content. This nonlinearity which is more pronounced at
the beginning and at the end of the drying process [215] led to suggestions of
bilinear and nonlinear relationships [ 216, 200]. Capillary tension, disjoining
pressure, and solid surface tension (GibbsBangham) are among several
mechanisms recognized for shrinkage of cementbased materials [217]. However,
much of the shrinkage that is observed in earlyage concrete can be attributed
stresses associated with capillary tension [218]. Kelvin and Laplace equations are
used to present the magnitude of capillary pressure developed in the concrete
pores due to drying phenomena [219]. According to Baron [219] these equations
can be derived from the equilibrium of chemical potentials between liquid and
vapor phases of water, under certain assumptions. These assumptions include
capillary water forms a single continuous phase, water vapor is an ideal gas,
capillary water is homogenous and incompressible, velocity is zero at any point,
capillary water is not under any external loads such as gravity, the gas pressure is
equal to the atmospheric pressure and the water is in the chemical equilibrium
with ambient atmosphere. Kelvin equation presents the capillary pressure as:
ln( )
l
l g c
v
RT
p p p h
M
ρ
− = = (F.31)
where
l
p is the pressure of liquid water,
g
p is the pressure of the gas phase (dry
air and water vapor),
c
p is the capillary pressure,
l
ρ is the water density (1000
kg/m
3
), R is the universal gas constant, T is temperature, M
v
is the molar mass of
water (0.01801 kg/mol), and h is the fractional relative humidity. This relationship
allows the use of internal relative humidity (h) as a parameter for modeling
developed stress and strain in concrete due to drying [220]. According to this
formulation, the capillary pressure at 300 K (26.8
o
C) and relative humidity of 0.5
(RH = 50%) is 96 MPa, which shows the high magnitude of capillary pressure in
the pores. The capillary pressure between humid air (dry air and water vapor) and
liquid water should be balanced by the surface tension
lg
σ at the interface of
liquid/gas phase. Assuming a circular meniscus with the radius r
m
, the first law of
Laplace can be derived as:
2
lg
l g c
m
p p p
r
σ
− = = (F.32)
where σ
lg
is surface tension, and r
m
is the radius of assumed circular meniscus.
237
(a) (b)
Figure F. 4. a) Equilibrium between liquid water and humid air at the interface of
liquid/gas phase inside a pore in concrete, b) different pressures applied on solid
skeleton inside capillary pores [221]
As a reaction to this negative pressure within the pore fluid, compression
develops in the solid microstructure [222]. An average pressure on the solid
skeleton due to the capillary pressure, p
sol_c
is calculated as
_
(1 ) sin
s c l l l g lg lg atm
p p p p α α α γ σ = + − + − (F.33)
where, α
l
is the volumetric fraction of liquid water in the capillary pores, α
lg
is a
coefficient incorporating effect of surface tension (σ
lg
). Inserting equation (F.32)
into equation (F.33) and rearranging gives
_
sin
( )
2
lg m
s c l l g g c atm
r
p p p p p p
α γ
α = − + + − (F.34)
According to Baron’s assumption
atm g
p p = and therefore
_ _
sin
2
lg m
s c l c sol c c
r
p p p
α γ
α α
⎛ ⎞
= + =
⎜ ⎟
⎝ ⎠
(F.35)
Gray and Schrefler [223] replaced the coefficient α
solc
by the saturation degree of
liquid phase, S
l
, and therefore, the pressure on solid skeleton can be presented as
_ s c l c
p S p = (F.36)
While equation (F.36) shows only the microscopic pressure on the solid skeleton
due to the capillary pressure, Bazant [ 224] presented variation of disjoining
pressure in a similar fashion to capillary pressure.
ln( )
a
d
a
RT
p h
M
ρ
Δ = (F.37)
where, p
d
is the disjoining pressure, and ρ
a
and M
a
are the density and molar mass
of adsorbed water, respectively. Similar to the capillary pressure, the variation of
238
the applied pressure on the solid skeleton due to the disjoining pressure can be
presented as:
_ _ s d sol d c
p p α Δ = Δ (F.38)
Considering equation (F.36) and (F.38), the total applied pressure on the solid
skeleton, p
sol
due to the capillary and the disjoining pressures can be obtained
using the saturation degree of liquid phase, S
l
, and a homogenization factor, α
s
[215] as:
s d l c
p S p α = (F.39)
The homogenization factor incorporates effects of both capillary and disjoining
pressures as well as the effect of scale change from microscopic to macroscopic
levels, and can be obtained from shrinkage tests.
Equation (F.39) can be used to determine the strains induced by the pore pressures
due to the drying process. A unit cell approach can be taken to study this effect on
porous media such as concrete materials. As shown in Figure (F.5), concrete can
be approximated by a square in which the pores are placed in the center and the
rest of area is the solid phase. Effective stress σ¯ is defined as the stress acting on
the solid skeleton which can be obtained from the equilibrium.
(1 )
s
p σ φ σ φ = − − ¯ (F.40)
where σ is the external stress (restraining effects) and φ is the porosity. In
unrestrained shrinkage tests, the external stress is zero ( 0 σ = ). On the other
hand, σ¯ can be related to the strains due to elastic behavior of solid skeleton.
s
E σ ε = ¯ (F.41)
where, E is the young’s modulus, and
s
ε is the shrinkage stain.
The shrinkage strain can therefore be obtained by combining equation (F.40) and
(F.41).
(1 )
s s
p
E
φ
ε
φ
=
−
(F.42)
To calculate shrinkage strains based on equation (F.42), capillary porosity, φ and
modulus of elasticity, E has to be modeled, initially. The capillary porosity, φ ,
has been predicted with different models [ 225, 226, 227, 228]. The model
proposed by Bentz and Garboczi [226] is used in this study with the following
equation:
/ 0.41
/ 0.313
w c
w c
α
φ
−
=
+
(F.43)
where w/c is the watertocement ratio, and α is the degree of hydration.
239
Figure F.5. Effective stress on solid skeleton based on unit cell modeling [221]
The degree of hydration is modeled as a function of time using a power function.
a
f
f
t
t
α α
⎛ ⎞
=
⎜ ⎟
⎜ ⎟
⎝ ⎠
(F.44)
where, t
f
is the final drying time, and
f
α is the degree of hydration at the final
drying time. Results of simulation of degree of hydration are in good agreement
with the experimental data reported by Lamond and Peilert [229] showing that the
degree of hydration of ordinary Portland concrete with w/c of about 0.45 at 24 h is
about 0.2 (see Figure F.6). This model is used for studying earlyage drying
shrinkage tests.
a
f
f
t
t
α α
⎛ ⎞
=
⎜ ⎟
⎜ ⎟
⎝ ⎠
Figure F.6. Simulation of degree of hydration in comparison with experimental
data
240
Elastic modulus development in the early age concrete can be modeled using a
twostage methodology.
;
( ) ;
a
i i
i
b
i
i f i i
f i
t
E t t
t
E
t t
E E E t t
t t
⎧
⎛ ⎞
⎪ ≤
⎜ ⎟
⎪
⎝ ⎠
⎪
=
⎨
⎛ ⎞
⎪ −
+ − >
⎜ ⎟
⎪
⎜ ⎟
−
⎪ ⎝ ⎠ ⎩
(F.45)
where E,
i
E , and
f
E are the elastic modulus at different times. t,
i
t and
f
t are the
time, initial time of setting, and the final time of drying, respectively. a and b are
constants. However, a should be more than 1 and b less than 1 to conform to the
experimental data [Dao et al. 2009] and [Van Heyningen and Boon 1973]. As
Shown in Figure (F.7), such a model is capable of simulating the development of
modulus of elasticity at very early age (<6 h) as well as from very early hours to
few days (0150 h).
a
i
i
t
E E
t
⎛ ⎞
=
⎜ ⎟
⎝ ⎠
:
a
i i
i
t
t t E E
t
⎛ ⎞
≤ =
⎜ ⎟
⎝ ⎠
: ( )
b
i
i i f i
f i
t t
t t E E E E
t t
⎛ ⎞
−
> = + − ⎜ ⎟
⎜ ⎟
−
⎝ ⎠
Figure F.7. Modeling development of modulus of elasticity in early ages (0150
h)
Modeling EarlyAge Shrinkage from Drying Data
In order to confirm the soundness of this method in calculating earlyage
shrinkage from the drying data, the method can be applied to different sets of data
with available experimental moisture loss curves and shrinkage versus time
curves. The experimental data of Wongtanakitcharoen and Naaman [20] were
chosen for this purpose which involve onedimensional drying of linear prismatic
specimens 1000 mm in length and 100×60 mm in cross sectional area, under
temperature of 3540
o
C provided by a radiator heater, and relative humidity of
241
22.5%. As shown in Figure (F.8) rapid rate of evaporation was insured using a
high velocity fan provided constant high flow of air over the specimen’s surface.
Figure F.8. Unrestrained earlyage shrinkage test setup (Wongtanakitcharoen and
Naaman 2007)
Among different earlyage shrinkage tests performed by Wongtanakitcharoen and
Naaman, test results on plain concrete and PVA fiber reinforced concrete with
0.2% volumetric fraction were chosen for analysis. The water to cement ratio by
weight was 0.53. ASTM C150 Type III Portland cement was used in the concrete
mix and the coarse aggregate was crushed limestone of 9.525mm maximum size,
and the fine aggregate was NS2 sand. The weight ratio of Portland cement to
water, sand and coarse aggregate in this mixture were 10.5311 and the
properties of PVA fiber is shown in Table F.1.
Table F.1.
Properties of PVA fiber used in this study
The ambient moisture concentration, C
*
, can be obtained from the desorption
isotherm curves which relate internal relative humidity to the water content. Using
the BSB model [110], C
*
was obtained 56 kg/m
3
. As shown in Figure (F.9a), the
evaporation rates were computed from cumulative moisture loss data using a
numerical differentiation procedure. The drying results were used to back
calculate diffusion and convective moisture transfer coefficients as shown in
Figure (F.9b). Similar to author’s drying results on plain and glass fiber reinforced
concrete [148], the moisture loss curves for these two samples (plain and 0.2%
242
PVA) are very similar and therefore a single cumulative moisture loss curves is
used for the rest of analysis.
A desorption isotherm model, the BSB model [230], is used to relate internal
relative humidity to the water content.
*
(1 )[1 ( 1) ]
m
AkV H
C
kH A kH
=
− + −
(F.46)
where, C* is moisture content (kg of water / kg of solid), H is relative humidity
(fraction) and A, k, V
m
are three empirical parameters defined as:
0
0
exp( ), 855
A
A A
T
= =
0.024(0.85 0.45 / ) ; for t< 5 d and 0.3 < w/c < 0.7
m ct
V w c V = +
V
ct
= 0.9, 1.0, 0.85 and 0.60 for cement type 1 to 4, respectively.
1
(1 ) 1
1
A
n
k
A
− −
=
−
5.5(0.33 2.2 / ) ; for t > 5 days and 0.3 < w/c < 0.7
ct
n w c N = +
N
ct
= 1.1, 1.0, 1.15 and 1.5 for cement type 1 to 4, respectively.
Results of analysis including simulation of profiles moisture
concentration, relative humidity, capillary pores pressure, macroscopic pressure
on solid skeleton are shown in Figures (F.10) and (F.11).
(a) (b)
Figure F.9. (a) Prediction of cumulative moisture loss and evaporation rate as a
function of time for the reported experimental data, (b) Backcalculated D’ and k
for the best fit of data
243
(a)
(b)
Figure F.10. (a) Simulation of moisture concentration profiles during drying of
plain and PVA0.2% concrete, (b) simulation of relative humidity profiles from
desorption isotherms (BSB model)
244
(a)
40 20 0
Macroscopic Pressure on Solid Skeleton, MPa
0
0.2
0.4
0.6
0.8
1
t=1.5 3
t = 24 h
d
α = 1.6
(b)
Figure F.11. (a) Simulation of capillary pores pressure profiles, (b) simulation of
macroscopic pressure on solid skeleton during earlyage drying of plain and
PVA0.2% concrete
245
Simulation results for the development of capillary porosity are shown in
Figure (F.12). Also elastic modulus development curves introduced the model is
shown in Figure (F.13). Based on these data, the profile of shrinkage strains
during the drying process is simulated as shown in Figure (F.14).
0 5 10 15 20 25
Time, h
0.5
0.54
0.58
0.62
0.66
0.7
Simulation
Figure F.12. Change in Capillary porosity according to porosity model by Bentz
and Garboczi [226]
:
a
i i
i
t
t t E E
t
⎛ ⎞
≤ =
⎜ ⎟
⎝ ⎠
: ( )
b
i
i i f i
f i
t t
t t E E E E
t t
⎛ ⎞
−
> = + −
⎜ ⎟
⎜ ⎟
−
⎝ ⎠
Figure F.13. Introduced elastic modulus for plain and PVA 0.2% concrete
246
(a)
(b)
Figure F.14. Simulated earlyage shrinkage strain profiles during drying of: a)
plain concrete, b) PVA0.2% concrete
247
Since the reported data by Wongtanakitcharoen and Naaman (2007) are
the average shrinkage strains, the simulated shrinkage strains in Figure (F.14) are
averaged over the thickness of samples and shown in Figure (F.15). As shown in
the figure, simulation results indicate that the proposed method is capable of
predicting experimental shrinkage strains during earlyage drying process of
cementbased materials and presenting the variation of moisture concentration,
internal relative humidity, capillary pores pressure and applied macroscopic
pressure on the solid skeleton within the time and space domain of the test.
0 5 10 15 20 25
Drying Time, h
0
1000
2000
3000
4000
5000
Wongtanakitcharoen and Naaman (2007)
Figure F.15. Simulation of earlyage shrinkage strain averaged over the thickness
of the sample in comparison with the experimental data reported by
Wongtanakitcharoen and Naaman (2007)
248
References
[1] Horvath A. Construction materials and the environment. Annu Rev
Environ Resour 2004; 29: 181204.
[2] US Central Intelligent Agency. The world factbook. 2003. Washington,
DC: CIA.
[3] CEMBUREAU (the European Cement Association), Rue d’Arlon 55, BE
1040 Brussels, Belgium.
[4] Kwon SJ, Na UJ, Park SS, Jung SH. Service life prediction of concrete
wharves with earlyaged crack: Probabilistic approach for chloride
diffusion. Struct Saf 2009;31(1):7583.
[5] Huang XM, Yang CY. Earlyage concrete cover crack and its effects on
concrete cover. Key Eng Mater 2006;302303:6306.
[6] Tietz, T. The cement industry and climate change challenges. California
Nevada Cement Association. 2008. A presentation at American Concrete
Institute’s Spring Convention, March 30April 3, Los Angeles, Calif.
[7] Yoon IS, Schlangen E, de Rooij MR, van Breugel K. The effect of cracks
on chloride penetration into concrete. Key Eng Mater 2007;348349:769
72.
[8] Lura P, Pease B, Mazzotta GB, Rajabipour F, Weiss J. Influence of
shrinkagereducing admixtures on development of plastic shrinkage
cracks. ACI Mater J 2007;104(2):18794.
[9] Banthia N, Gupta R. Plastic shrinkage cracking in cementitious repairs and
overlays. Mater Struct 2009;42(5):56779.
[10] MoraRuacho J, Gettu R, Aguado A. Influence of shrinkagereducing
admixtures on the reduction of plastic shrinkage cracking in concrete.
Cem Concr Res 2009;39(3):1416.
[11] Cohen MD, Olek J, Dolch WL. Mechanism of plastic cracking in Portland
cement and portland cementsilica fume paste and mortar. Cem Concr Res
1990;20(1):10319.
[12] Slowik V, Schmidt M, Fritzsch R. Capillary pressure in fresh cement
based materials and identification of the air entry value. Cem Concr
Compos 2008;30:557–565.
249
[13] Radocea A. A study on the mechanism of plastic shrinkage of cement
based materials. PhD thesis, Göteborg, Chalmers University of
Technology, Göteborg;1992.
[14] Wittmann FH. On the action of capillary pressure in fresh concrete, Cem
Concr Res 19766;6(1):49–56.
[15] Esping O. Effect of limestone filler BET(H2O)area on the fresh and
hardened properties of selfcompacting concrete, Cem Concr Res
2008;38:938–944.
[16] Morris P, Dux P. Cracking of plastic concrete, Aust J Civ Eng
2003;1(1):1721.
[17] Aly T, Sanjayan JG. Mechanism of early age shrinkage of concretes,
Mater Struct 2009;42:461–468.
[18] Slowik V, Schlattner E, Klink T. Experimental Investigation into early age
shrinkage of cement paste by using fiber Bragg gratings. Cem Concr
Compos 2004;26(5):473–9.
[19] Hammer TA. Effect of silica fume on the plastic shrinkage and pore water
pressure of highstrength concretes, Mater Struct 2001;34:273278.
[20] Wongtanakitcharoen T, Naaman AE. Unrestrained early age shrinkage of
concrete with polypropylene, PVA, and carbon fibers, Mater Struct
2007;40:289300.
[21] Dao VTN, Dux PF, Morris PH. Tensile properties of earlyage concrete.
ACI Mater J 2009;106(6): 483492.
[22] Benboudjema F. Modélisation des déformations différées du béton sous
sollicitations biaxiales. Application aux enceintes de confinement de
bâtiments réacteurs des centrales nucléaires, PhD dissertation, Université
de MarnelaVallée, 2002.
[23] ASTM C15708. Standard test method for length change of hardened
hydrauliccement mortar and concrete, ASTM Book of Standards
2008;04.02.
[24] Grzybowski M, Shah SP. Shrinkage Cracking of Fiber Reinforced
Concrete, ACI Mater J 1990; 87(2):138148.
250
[25] Paillere AM, Buil M, Serrano JJ. Effect of fiber addition on the
autogenous shrinkage of silica fume concrete, ACI Mater J
1989;86(2):139144.
[26] Kovler, K. Testing system for determining the mechanical behavior of
early age concrete under restrained and free uniaxial shrinkage, Mater
Struct 1994;27(6):324330.
[27] Toma G, Pigeon M, Marchand J, Bissonnette B, Bercelo L. Earlyage
autogenous restrained shrinkage: stress build up and relaxation,
Proceedings of the Fourth International Research Seminar on ‘Self
Desiccation and its Importance in Concrete Technology’, Ed. B. Persson
and G. Fagerlund, Lund University. 1999:61–72.
[28] Altoubat AS, Lange DA. Gripspecimen interaction in uniaxial restrained
test, ACI SP206 on Concrete: Material Science to Application, A Tribute
to Surendra P. Shah. 2002:189204.
[29] Aly T, Sanjayan, JG, Collins F. Effect of polypropylene fibers on
shrinkage and cracking of concretes, Mater Struct 2008;41(10):17411753.
[30] Weiss WJ, Shah SP. Recent trends to reduce shrinkage cracking in
concrete pavements, Aircraft/Pavement Technology: In the Midst of
Change, Seattle, WA, 1997:217228.
[31] Kraai PP. A proposed test to determine the cracking potential due to
drying shrinkage of concrete. Concr Constr 1985;30(9):775778.
[32] Shaeles CA, Hover KC. Influence of mix proportions and construction
operations on plastic shrinkage cracking in thin slabs. ACI Mater J
1988;85(6):495504.
[33] Opsahl OA, Kvam SE. Betong med EEstalfiber (Concrete with FEsteel
fibers), Report No. STF 65 A82036, SINTEF dep, Cement and Concrete
Research Institute, Technical University of Norway, Trondheim, June
1982.
[34] Padron I, Zollo RF. Effect of synthetic fibers on volume stability and
cracking of Portland cement concrete and mortar. ACI Mater J
1990;87(4):327–332.
[35] Carlson RW, Reading TJ. Model study of shrinkage cracking in concrete
building walls. ACI Struct J 1988;85(4):395404.
251
[36] Shah SP, Karaguler ME, Sarigaphuti M. Effects of shrinkage reducing
admixture on restrained shrinkage cracking of concrete. ACI Mater J
1992;89(3):289295.
[37] Shah SP, Marikunte S, Yang W, Aldea C. Control of cracking with
shrinkagereducing admixtures. Transp Res Rec 1997;1574:2536.
[38] Bentur A. Early age cracking in cementitious systems, Chapter 6.5 –
RILEM State of the Art Report TC 181EAS, ed. A. Bentur, RILEM
Publications, Bagneux, France. 2003.
[39] Hossain AB, Weiss J. Assessing residual stress development and stress
relaxation in restrained concrete ring specimens. Cem Concr Compos
2004;26:531540.
[40] Turcry P, Loukili A, Haidar K, PijaudierCabot G, Belarbi A. Cracking
tendency of selfcompacting concrete subjected to restrained shrinkage:
experimental study and modeling, J Mater Civ Eng 2006;18(1):46–54.
[41] AASHTO PP 3499. Standard practice for estimating the crack tendency
of concrete. 2004.
[42] ASTM C 158104. Standard test method for determining age at cracking
and induced tensile stress characteristics of mortar and concrete under
restrained shrinkage. 2004.
[43] Turcry P, Loukili A. Evaluation of plastic shrinkage cracking of self
consolidating concrete. ACI Mater J 2006;103(4):2729.
[44] Sanjuan MA, Moragues A. A testing method for measuring plastic
shrinkage in polypropylene fibre reinforced mortars. Mater Lett
1994;21:23946.
[45] Nanni A, Ludwig DA, Mcgillis MT. Plastic shrinkage cracking of
restrained fiberreinforced concrete. Transp Res Rec 1993;1382:6972.
[46] Banthia N, Gupta R. Test method for evaluation of plastic shrinkage
cracking in fiberreinforced cementitious materials. Exp Tech
2007;31(6):448.
[47] Naaman AE, Wongtanakitcharoen T, Hauser G. Influence of different
fibers on plastic shrinkage cracking of concrete. ACI Mater J
2005;102(1):4958.
252
[48] Ma Y, Zhu B, Tan M, Wu K. Effect of Ytype polypropylene fiber on
plastic shrinkage cracking of cement mortar. Mater Struct 2004;37:925.
[49] Jacobsen S, Aarseth LI. Effect of wind on drying from wet porous
building materials surfacesA simple model in steady state. Mater Struct
1999;32:3844.
[50] Shimomura T, Maekawa K. Analysis of the drying shrinkage behaviour of
concrete using a micromechanical model based on micropore structure of
concrete. Mag Concr Res 1997;49(181):303322.
[51] Wang K, Shah SP, Phuaksuk P. Plastic shrinkage cracking in concrete
materialsInfluence of fly ash and fibers”, ACI Mater J 2001;98(6):458
64.
[52] Samman TA, Mirza WH, Wafa FF. Plastic shrinkage cracking of normal
and highstrength concrete: A comparative study”, ACI Mater J
1996;93(1):3640.
[53] Berhane Z. Evaporation of water from fresh mortar and concrete at
different environmental conditions. ACI J Proc 1984;81(6):5605.
[54] Hall C, Hoff WD. Water transport in brick, stone, and concrete. New
York: Taylor & Francis, 2002.
[55] Torrenti JM, Granger L, Diruy M, Genin P. Modeling concrete shrinkage
under variable ambient conditions, ACI Mat. J. 1997;96 (1):3539.
[56] West RP, Holmes N. Predicting moisture movement during the drying of
concrete floors using finite elements, Constr Build Mater 2005;19:674
681.
[57] Garbalinska H. Application of √ttype, logarithmic and halftime methods
in desorptive measurements of diffusivity in narrow humidity ranges. Cem
Concr Res 2006;36 :1294303.
[58] Huldén M, Hansen CM. Water permeation in coatings. Prog. Org.
Coatings. 1985;13(3):171194.
[59] Kim, JK, Lee CS. Prediction of differential drying shrinkage in concrete.
Cem Concr Res 1998;28 (7):985994.
[60] Bazant ZP, Najjar LJ. Drying of concrete as a nonlinear diffusion problem.
Cem Concr Res 1971;1(5):46173.
253
[61] Bazant ZP, Najjar LJ. Nonlinear water diffusion in nonsaturated concrete.
Mater Struct 1972;5(1):320.
[62] Wilson GW, Fredlund DG, Barbour SL. Coupled soilatmosphere
modelling for soil evaporation. Can Geotech J 1994;31(2):151161.
[63] Kodikara JK, Chakrabarti S. Modelling of moisture loss in cementitiously
stabilised pavement materials. ASCE J Geomech 2005;5 (4):295303.
[64] Chen D, Mahadevan S. Cracking analysis of plain concrete under coupled
heat transfer and moisture transport processes. J Struct Eng
2007;133(3):400410.
[65] Walton JC, Plansky LE, Smith RW. Models for estimation of service life
of concrete barriers in lowlevel radioactive waste disposal, Rep. Prepared
for U.S. Nuclear Regulatory Commission, NUREG/CR5542 EGG
2597.1990.
[66] Šelih J, Bremner TW. Drying of saturated lightweight concrete: An
experimental investigation. Mater Struct 1996;29(7):401405.
[67] Young JF. Physical mechanisms and their mathematical descriptions.
Mathematical modelling of creep and shrinkage of concrete, Ed. ZP
Bazant, Wiley, Chichester.1988:6398.
[68] ACI 305, Hot weather concretingACI 305R99, Farmington Hills,
Michigan: American Concrete Institute. 1999.
[69] Uno PJ. Plastic shrinkage cracking and evaporation formulas, ACI Mater J
1998;95(4):365375.
[70] Dassault Systemes, SIMULIA, Rhode Island, USA, 2010.
[71] Bissonnette B, Therrien Y, Pleau R. Steel fiber reinforced concrete and
multiple cracking, Can J Civil Eng 2000;27 (4):774784.
[72] Swamy RN, Stavarides H. Influence of fiber reinforcement on restrained
shrinkage cracking, ACI J 1979;76 (3):443460.
[73] Hossain AB, Pease B, Weiss WJ. Quantifying restrained shrinkage
cracking potential in concrete using the restrained ring test with acoustic
emission, Transp Res Rec 2003; Concr Mater Const 1834:2433.
254
[74] Mane SA, Desai TK, Kingsbury D, Mobasher B. Modeling of restrained
shrinkage cracking in concrete materials, ACI SP20614 2002:219242.
[75] Shah HR, Weiss WJ. Quantifying shrinkage cracking in fiber reinforced
concrete using the ring test, Mater Struct 2006;39 (9):887899.
[76] Kim B, Weiss WJ. Using acoustic emission to quantify damage in fiber
reinforced cement mortars, Cem Concr Res 2003;33 (2):207214.
[77] Soranakom C, Bakhshi M, Mobasher B. Role of alkali resistant glass
fibers in suppression of restrained shrinkage cracking of concrete
materials, Proc 15th International Glass Fibre Reinforced Concrete
Association Congress (GRC 2008), Prague, Prague Czech Republic, April
2023, 2008.
[78] ACI Committee 544.5R10. Report on physical properties and durability
of fiberreinforced concrete, American Concrete Institute, Farmington
Hills, MI, 2010.
[79] Majumdar AJ, Laws V. Glass fibre reinforced cement, Oxford BSP
Professional Books, London, UK,1991.
[80] Litherland KL, Oakley DR, Proctor BA. The use of accelerated aging
procedures to predict the long term strength of GRC composites, Cem
Concr Res 1981;11:455466.
[81] Bentur A, Diamond S. Fracture of glass fiber reinforced cement, Cem
Concr Res 1984;14:3134.
[82] Bentur A. Mechanisms of potential embrittlement and strength loss of
glass fiber reinforced cement composites, Proc Durability of glass fiber
reinforced concrete symposium organized by PCI, November 1215, 1985,
Chicago , ed. S. Diamond, 1986, pp. 109123.
[83] Grzybowski M, Shah SP. A model to predict cracking in fiber reinforced
concrete due to restrained shrinkage, Mag Concr Research 1989), Vol. 41,
No. 148, pp. 125135.
[84] Shah SP, Ouyang C, Marikunte S, Yang W, BecqGiraudon E. A method
to predict shrinkage cracking of concrete, ACI Mater J 1998;95 (4): 339
346.3
[85] Tuma JJ. Engineering mathematics handbook, McGrawHill, New York,
1970.
255
[86] Bazant ZP, Panula L. Practical prediction of timedependent deformation
of concrete, part 1: shrinkage, part2: creep, Mater Struct 1978;11 (5): 307
328.
[87] ACI Committee 209. Guide for modeling and calculating shrinkage and
creep in hardened concrete (ACI 209.2R08), American Concrete Institute,
Michigan, USA. 2008.
[88] Sultan E, Boudaoud AI, Amar MB. Evaporation of a thin film: diffusion
of the vapour and Marangoni instabilities. J Fluid Mech 2005;543:183
202.
[89] Arya SP. Introduction to micrometeorology. San Diego: Academic Press,
2001.
[90] Huang CH. Pasquill’s influence: On the evaporation from various liquids
into the atmosphere, J Appl Meteorol 1997;36 (8):10216.
[91] Brighton PWM. Evaporation from a plane liquid surface into a turbulent
boundary layer. J Fluid Mech 1985;159:32345.
[92] Prata AT, Sparrow EM. Diffusiondriven nonisothermal evaporation. J
Heat Transfer 1985;107(1):23942.
[93] Brutsaert W. A model for evaporation as a molecular diffusion process
into a turbulent atmosphere”, J Geophys Res 1965;70:501724.
[94] Carmeliet J, Hens H, Roels S, Adan O, Brocken H, Cerny R, Pavlik Z,
Hall C, Kumaran K, Pel L. Determination of the Liquid Water Diffusivity
from Transient Moisture Transfer Experiments. J Therm Envelope Build
Sci. 2004;27:277305.
[95] Janz M. Methods of measuring the moisture diffusivity at high moisture
levels. Report TVBM3076, University of Lund, Lund Institute of
Technology, Division of Building Materials, 1997.
[96] Mehta PK, Monteiro PJM. Concrete: Microstructure, Properties and
Materials. Boston: McGrawHill, 2006.
[97] Soroushian P, Ravanbakhsh S. Control of plastic shrinkage cracking with
specialty cellulose fibers, ACI Mater J 1998;95(4):42935.
256
[98] Copeland LE, Hayes JC. Determination of nonevaporable water in
hardened portlandcement paste. ASTM Bull 1953;194:704.
[99] Hall C, Hoff WD, Nixon MR. Water movement in porous building
materialsVI. Evaporation and drying in brick and block materials. Build
Environ 1984;19(1):1320.
[100] Cooling LF. Contribution to the study of florescence. II. The evaporation
of water from bricks. Trans Br Ceram Soc 1930;29:3954.
[101] Kowalski SJ. Thermomechanics of drying processes. New York: Springer,
2003.
[102] Scherer GW. Theory of drying. J Am Ceram Soc 1990;73(1):314.
[103] Plumb OA, Gu L, Webb SW. Drying of porous materials at low moisture
content. Drying Technol 1999;17(10):19992011.
[104] Kowalski SJ. Thermomechanical approach to shrinkage and cracking
phenomena in drying. Drying Technol 2001;19(5):73165.
[105] Garrabrants AC, Kosson DS. Modeling moisture transport from a portland
cementbased material during storage in reactive and inert atmospheres.
Drying Technol 2003;21(5):775805.
[106] Vu TH. Influence of pore size distribution on drying behaviour of porous
media by a continuous model. MS Thesis, OttovonGuerickeUniversität
Magdeburg, Germany, 2006.
[107] Simunek J, Van Genuchten M, and Sejna M. The Hydrus1d Software
Package for Simulating the OneDimensional Movement of Water, Heat,
and Multiple Solutes in VariablySaturated Media. Version 3.0,
Agriculture Research Service, US Department of Agrriculture, Riverside,
CA, 2005.
[108] Blandin HP, David JC, Vergnaud JM, Illien JP, Malizewicz M. Modelling
of drying of coatings: effect of the thickness, temperature and
concentration of solvent. Prog Org Coat 1987;15:16372.
[109] Crank J. The mathematics of diffusion. New York: Oxford Science
Publications, 1989.
257
[110] Xi Y, Bažant Z, Molina L, and Jennings HM. Moisture diffusion in
cementitious materials Moisture capacity and diffusivity. Adv Cem Bas
Mater 1994;1:258266.
[111] Cussler EL. Diffusion: mass transfer in fluid systems. 3rd edition, New
York: Cambridge University Press, 2009.
[112] American Society for Testing Materials. Standard Specification for Fiber
Reinforced Concrete, ASTM C1116/C1116M10. Annual Book of ASTM
Standards 2010. Section 4. Constrcution. 04.02.
[113] Tsotsas E, Mujumdar AS. Modern drying technology: Computational tools
at different scales. Darmstadt, Germany: WileyVCH, 2007.
[114] Sebastian P, Turner IW. An investigation of the boundary conditions for a
vacuum drying problemIntroducing the transition layer concept. Drying
Technol 1994;12(4):71760.
[115] Ramon G, Agnon Y, Dosoretz G. Heat transfer in vacuum membrane
distillation: Effect of velocity slip. J Membr Sci 2009;331:11725.
[116] Bažant ZP, Raftshol WJ. Effect of cracking in drying and shrinkage
specimens. Cem Concr Res 1982;12:209226.
[117] Aldea CM, Shah SP, Karr A. Effect of cracking on water and chloride
permeability of concrete. J. Mat. in Civ. Engrg 1999; 11(3):181187.
[118] Bažant ZP, Sener S, Kim JK. Effect of cracking on drying permeability
and diffusivity of concrete, ACI Mater J 1987;84(5):351–358.
[119] Vejmelková E, Padevet P, Cerny. Effect of cracks on hygric and thermal
characteristics of concrete. Bauphysik 2008;30(6):438–44.
[120] Torrijos MC, Giaccio G, Zerbino, R. Internal cracking and transport
properties in damaged concretes. Mater Struct 2010;43:109–21.
[121] Brutsaert W. Evaporation into the atmosphere: theory, history, and
applications, D. Reidel Publishing Co., Boston, 1982.
[122] Hisatake K, Tanaka S, Aizawaa Y. Evaporation rate of water in a vessel, J
Appl Phys 1993;73 (11):7395401.
[123] Kimpton DD, Wall FT. Determination of diffusion coefficients from rates
of evaporation, J Phys Chem 1952;56 (6):715717.
258
[124] Lugg GA. Diffusion coefficients of some organic and other vapors in air,
Anal Chem 1968;40 (7):10721077.
[125] Kwon KC, Ibrahim TH, Park Y, Simmons CM, Pseudobinary molecular
diffusion of vapors into air, Adv Environ Res 2004; 8:667678.
[126] Kwon KC, Park Y, Simmons CM, Tibere GL, Ibrahim TH, Molecular
diffusion of volatileliquid vapors into air, Chem Eng Commun
2003;190:144967.
[127] Prandtl L. On fluid motions with very small friction (in German), Third
International Mathematical Congress, Heidelberg, Germany, (1904) 484
491.
[128] Hisatake K, Fukuda M, Kimura J, Maeda M, Fukuda Y. Experimental and
theoretical study of evaporation of water in a vessel, J Appl Phys 1995;77
(12): 666474.
[129] Davies JT, Rideal EK. Interfacial phenomena, Academic Press, New
York, 1961.
[130] Wiederhold PR. Water vapor measurement: Methods and instrumentation,
Marcel Dekker Inc, New York, 1997.
[131] Tuve GL, Bolz RE. Handbook of tables for applied engineering science,
2nd edition, CRC Press, Cleveland, 1976.
[132] Nellis G, Klein S, Heat Transfer, Cambridge University Press, New York,
2008.
[133] Buckingham E. Studies on the movement of soil moisture, Bulletin No.
38, US Department of Agriculture, Washington, 1907.
[134] Philip JR. Flow in porous media, Annu Rev Fluid Mech 1970; 2:177204.
[135] Sposito G. The ‘Physics’ of Soil Water Physics, Water Resour Res 1986;
22 (9S):83S88S.
[136] Bear J, Bachmat Y. Introduction to modeling of transport phenomena in
porous media. Kluwer, Dordrecht, 1991.
[137] Lewis RW, Schrefler BA. The finite element method in the static and
dynamic deformation and consolidation of porous media. Wiley, New
York, 1998.
259
[138] Coussy O. Poromechanics. Wiley, New York, 2004.
[139] Mainguy M, Coussy O, BaroghelBouny V. Role of air pressure in drying
of weakly permeable materials. ASCE J Eng Mech 2001;127:582–592.
[140] Idiart AE, Lopez CM, Carol I. Modeling of drying shrinkage of concrete
specimens at the mesolevel, Mater. Struct. 2011; 44:415–435.
[141] Pel L. Moisture transport in porous building materials, PhD thesis,
Technische Universiteit Eindhoven, Eindhoven, Netherlands, 1995.
[142] Samson E, Marchand J, Snyder K, Beaudoin J. Modeling ion and fluid
transport in unsaturated cement systems in isothermal conditions, Cem
Concr Res 2005;35:141–153.
[143] Richards LA. Capillary conduction of liquids through porous mediums,
Physics 1931;1:318333.
[144] Sposito G. Lie group invariance of the Richard equation, in J.H. Cushman
(ed.), Dynamics of Fluids in Hierarchical Porous Media, Academic Press,
London, 1990: 9398.
[145] Coussot P. Scaling approach of the convective drying of a porous medium.
Eur. Phys. J. B 2000;15:557566.
[146] Platten AK. A study of evaporation and drying in porous building
materials, PhD thesis, University of Manchester Institute of Science and
Technology (UMIST), Manchester, UK, 1985.
[147] Kowalski SJ. Thermomechanics of drying processes, Springer, New York,
2003.
[148] Bakhshi M, Mobasher B. Experimental observations of earlyage drying
of Portland cement paste under lowpressure conditions, Cem Concr
Compos 2011;33 (4): 474484.
[149] Azenha M, Maekawa K, Ishida T, Faria R. Drying induced moisture losses
from mortar to the environment Part I experimental research, Mat.
Struct. 2007;40 (8) :801811.
[150] Rogers JA, Kaviany M. Funicular and evaporativefront regimes in
convective drying of granular beds, Int. J. Heat Mass Transfer 1992;35
(2): 469480.
260
[151] Rahman MK, Baluch MH, AlGadhib AH, Modeling of shrinkage and
creep stress in concrete repair, ACI Mater J 1999;96 (5):542550.
[152] Akita H, Fujiwara T, Ozaka Y, A practical procedure for the analysis of
moisture transfer within concrete due to drying, Mag Concr Res 1997;49
(179): 129137.
[153] Demirkol E, Erdoğdu F, Palazoğlu TK, Analysis of mass transfer
parameters (changes in mass flux, diffusion coefficient and mass transfer
coefficient) during baking of cookies, J Food Eng 2006;72:364371.
[154] Comite EuroInternational du Beton, CEBFIP model code 1990, EPF
Lausanne, Switzerland, 1998.
[155] Kim JK, Lee CS. Moisture diffusion of concrete considering self
desiccation at early ages, Cem Concr Res 1999;29 (12):1921–1927.
[156] Baluch MH, Rahman MK, AlGadhib AH, Raza A, Zafar S. Crack
minimization model for Hot Weather Concreting, Arabian J Sci Eng
2006;31 (1C):7791.
[157] Sakata K. A study on moisture diffusion in drying and drying shrinkage of
concrete, Cem Concr Res 1983;13:216224.
[158] Šelih J, Sousa ACM, Bremner TW. Moisture transport in initially fully
saturated concrete during drying, Transp Porous Media 1996;24 (1):81
106.
[159] Šelih J. Movement of water during drying of fully saturated concrete, PhD
Dissertation, The University of New Brunswick, Canada, 1995.
[160] Bakhshi M, Mobasher B, Soranakom C. Earlyage drying of cementitious
materials: theory and modeling, Construct Build Mater, under Review,
2011.
[161] Azenha M, Maekawa K, Ishida T, Faria R. Drying induced moisture losses
from mortar to the environment. Part II: numerical implementation, Mater
Struct 2007;40:813–825.
[162] TM 58227, Standard Practice for Concrete Pavements, U.S. Army Corps
of Engineers, USACE, 1987, http://www.usace.army.mil/inet/usacedocs/.
261
[163] CSA Standard A23.1. Design and control of mixes, 5th Canadian Metric
ed., Canadian Portland Cement Association, 1991.
[164] Australian Technical Bulletin, Cracks in concrete due to plastic shrinkage
and plastic settlement, Technical Bulletin 95/1, Australian PreMixed
Concrete Association, Oct. 1995.
[165] Hover KC. Evaporation of water from concrete surfaces, ACI Mater J
2006;103 (5):384389.
[166] Menzel CA. Causes and prevention of crack development in plastic
concrete, Proc Portland Cement Association Annual Meeting, 1954:130
136.
[167] Dalton J. Experimental essays on evaporation, Proc Manchester Literary
and Philosophical Society, 1802;5:536602.
[168] ACI Committee 308, Guide to curing concrete, Committee Report ACI
308R01, ACI Manual of Concrete Practice, American Concrete Institute,
Farmington Hills, MI, 2001.
[169] AlFadhala M, Hover KC. Rapid evaporation from freshly cast concrete
and the Gulf environment, Const Build Mate 2001;15:17.
[170] Kohler, MA. Lake and pan evaporation, Water Loss Investigations: Lake
Hefner Studies, Technical Report, Geological Survey Professional Paper
269, U.S. Government Printing Office, Washington, D.C., 1954:127148.
[171] Kohler, MA, Nordenson, TJ, Fox WE. Evaporation from pans and lakes,
Research Paper No. 38, U.S. Department of Commerce, Washington,
D.C., May 1955.
[172] Shuttleworth WJ. Evaporation, Handbook of Hydrology, DR Maidment,
ed, McGrawHill, 1993:4.14.53.
[173] Veihmeyer FJ. Evapotranspiration, Handbook of Appl Hydrol, VT Chow,
ed, McGraw Hill, 1964:111 to 1138.
[174] Churchill SW, Ozoe H, A correlation for laminar free convection from a
vertical plate, J Heat Transfer 1973;95C:540–541.
[175] Incropera FP, DeWitt DP. Introduction to heat transfer, 5th Ed, Wiley,
John Wiley & Sons, New York, USA, 2007.
262
[176] Thibodeaux LJ. Chemodynamics: environmental movement of chemical in
air, water, and soil; John Wiley and Sons: New York, 1979:160163.
[177] Crawford M. Air pollution control theory; McGraw Hill: New York, 1976.
[178] Friedlander SK. Smoke, dust, and haze: Fundamentals of aerosol behavior;
John Wiley and Sons: New York, 1977.
[179] Reist PC. Aerosol science and technology, 2nd ed; John Wiley and Sons,
Inc: New York, 1993.
[180] Yi SM, Holsen T, Noll K. Comparison of dry deposition predicted from
models and measured with a water surface sampler. Environ Sci Technol
1997; 31:272278.
[181] Lienhard IV JH, Lienhard V JH. A heat transfer textbook, 3rd Edition,
Phlogiston Press, Cambridge, MA, 2008.
[182] Touloukian YS, Saxena SC, Hestermans P. Thermophysical properties of
matter;11: Viscosity, IFI/Plenun, NY, 1970.
[183] Poling BE, Prausnitz JM, O'Connell J.P. The properties of gases and
liquids, 5th ed, McGrawHill, New York, 2000.
[184] Poole TS. Guide for curing Portland cement concrete pavements;II:
FHWAHRT05038. Federal Highway Administration, McLean, VA,
2006.
[185] Nonhebel G, Moss AAH. Drying of solids in the chemical industry,
Butterworth & Co Ltd, London, England, 1971.
[186] Green D, Perry R. Psychrometry, evaporative cooling, and solids drying,
Perry's Chemical Engineers' Handbook, 8th Ed, McGrawHill, New York,
US, 2007.
[187] Branson DE, Christiason ML. Time dependent concrete properties related
to designstrength and elastic properties, creep and shrinkage, SP
27: Creep, Shrinkage and Temperature Effects, American Concrete
Institute, Farmington Hills, MI, 1971: 257277.
[188] ACI Committee 209. Prediction of creep, shrinkage and temperature
effects in concrete structures, ACI SP76: Designing for Creep and
Shrinkage in Concrete Structures  A Tribute to Adrian Pauw, American
Concrete Institute, Farmington Hills, MI, 1982:193300.
263
[189] Perre P. Drying with internal vaporisation: Introducing the concept of
identity drying card (IDC), Drying Technol 1995;13 (57):10771097.
[190] Ketelaars AAJ. Drying deformable media: kinetics, shrinkage and stresses.
PhD dissertation, Eindhoven University of Technology, Netherland, 1992.
[191] Salytkov SA. Stereometrische metallographie, VEB Deutscher Verlag für
Grundstoffindustrie, Leipzig, 1974.
[192] Stroeven P. Some aspects of micromechanics of concrete, PhD
Dissertation, Stevin Laboratory, Technological University of Delft,
Netherland, 1973.
[193] Stroeven P. Geometric probability approach to the examination of
microcracking in plain concrete, J Mater Sci 1979;14 (5):11411151.
[194] Litorowicz A. Identification and quantification of cracks in concrete by
optical fluorescent microscopy, Cem Concr Res 2006;36:15081515.
[195] Darcy HPG. Les fontaines publiques de la ville de Dijon. Exposition et
application á suivre et des formules á employer dans les questions de
duistribution d’eau. Victor Dalamont, Paris, France, 1856.
[196] Wittmann FH, Roelfstra PE. Total deformation of loaded drying concrete,
Cem Concr Res 1980;10: 601610.
[197] Alvaredo AM, Wittmann FH. Shrinkage as influenced by strain softening
and crack formation, RILEM Proc 5
th
Int Symp Creep and Shrinkage of
ConcreteMathematical Modeling, 1993:103113, Barcelona, Spain.
[198] Bazant ZP, Xi Y. Drying creep of concrete: Constitutive model and new
experiments separating its mechanisms, Mater Struct 1997;27: 3–14.
[199] Bazant ZP, Hauggaard AB, Baweja S, Ulm FJ. Microprestress
solidification theory for concrete creep. I: aging and drying effects, J
Engrg Mech 1997;123(11):11881194 .
[200] van Zijl GPAG. Computational modelling of masonry creep and
shrinkage, PhD dissertation, Delft University, 1999.
[201] Pihlajavaara SE. A review of some of the main results of a research on the
aging phenomena of concrete: Effect of moisture conditions on strength,
shrinkage and creep of mature concrete, Cem Concr Res 1974;4:761771.
264
[202] BaroghelBouny V, Mainguya M, Lassabatereb T, Coussy O.
Characterization and identification of equilibrium and transfer moisture
properties for ordinary and highperformance cementitious materials, Cem
Concr Res 1999;29(8):12251238.
[203] Thelandersson S, Matensson A, Dahlblom O. Tension softening and
cracking in drying concrete, Mater Struct 1988;21:41624.
[204] Granger L. Comportement différé du béton dans les enceintes de centrales
nucléaires: analyse et modélisation. Ph.D. thesis, 1996. Ecole Nationale
des Ponts et Chaussées, MarneLaVallée, France [in French].
[205] Granger L, Torrenti JM, Acker P. Thoughts about drying shrinkage:
Experimental results and quantification of structural drying creep, Mater
Struct 1997;30:588598.
[206] Verbeck GJ, Hehnuth RH. Structures and physical properties of cement
paste, Proc. 5th Int. Symp. on Chemistry of Cement, Tokyo, Japan, 1968,
132.
[207] Benboudjema F, Meftah F, Torrenti JM. Interaction between drying,
shrinkage, creep and cracking phenomena in concrete, Eng Struct
2005;27: 239–250.
[208] Benboudjema F, Heinfling G, Meftah F, Sellier A, Torrenti JM. On the
prediction of delayed strains for concrete subjected to drying and loading
simultaneously, Proc. of CONCREEP 6: Creep, shrinkage and durability
mechanics of concrete and other quasibrittle materials 2001:245–250.
[209] Benboudjema, F, Meftah F, Torrenti, JM. Drying creep: An elastoplastic
damage approach of the structural effect, Fracture Mechanics of Concrete
Structures, R. de Borst, J. Mazars, G. PijaudierCabot, and J. G. M. van
Mier, eds. Lisse, The Netherlands, 2001;1:169–176.
[210] Benboudjema F, Meftah F, Torrenti JM. Structural effects of drying
shrinkage, J Eng Mech 2005;131 (11):11951199.
[211] Day RL, Cuffaro P, Illston JM. The effect of drying on the drying creep of
hardened cement paste. Cem Concr Res 1984;14(3):329–338.
265
[212] Bourgeois F, Burlion N, Shao JF. Modelling of elastoplastic damage in
concrete due to desiccation shrinkage. Int J Numer Anal Methods
Geomech 2002; 26:759–774.
[213] Gawin D, Pesavento F, Schrefler BA. Modelling of deformations of high
strength concrete at elevated temperature. Mater Struct 2004;37:218–236.
[214] Grasberger S, Meschke G. Thermohygromechanical degradation of
concrete: From coupled 3D material modeling to durability oriented
multifield structural analyses. Mater Struct 2004;37:244–256.
[215] Benboudjema F, Meftah F, Torrenti JM. A viscoelastic approach for the
assessment of the drying shrinkage behaviour of cementitious materials,
Mater Struct 2007;40:163–174.
[216] Obeid W, Mounajed G, Alliche A. Experimental identification of Biot’s
hydromechanical coupling coefficient for cement mortar. Mater Struct
2002;35:229–236.
[217] Grasley ZC, Lange DA. Thermal dilation and internal relative humidity of
hardened cement paste, Mater Struct 2007;40:311–317.
[218] Mindess S, Young JF. Concrete, PrenticeHall, New Jersey, 1981.
[219] Baron J. le béton hydraulique, ENPC Press, Paris, France, 1982.
[220] Grasley ZC, Lange DA. Modeling drying Shrinkage stress gradients in
concrete, ASTM J Test Eval 2003;26(2): 115122.
[221] Benboudjema F. Modélisation des déformations différées du béton sous
sollicitations biaxiales. Application aux enceintes de confinement de
bâtiments réacteurs des centrales nucléaires, PhD dissertation, Université
de MarnelaVallée, 2002 [in French].
[222] Bissonnette B, Marchand J, Charron JP, Delagrave A, Barcelo L. Early age
behavior of cementbased materials, Mater Sci Concr VI, J. Skalny, S.
Mindess, Ed, American Ceramic Society, 2001:243326.
[223] Gray WG, Schrefler B.A. Thermodynamic approach to effective stress in
partially saturated porous media, Eur J Mech A/Solids 2001; 20:521–538.
[224] Bažant ZP. Thermodynamics of hindered adsorption and its implications
for hardened cement paste and concrete, Cement and Concrete Research,
2(1), 1972:116.
266
[225] Gallé C. Effect of drying on cementbased materials pore structure as
identified by mercury intrusion porosimetry: A comparative study between
oven, vacuum, and freezedrying, Cem Concr Res 2001;31 (10):1467
1477.
[226] Bentz DP, Garboczi EJ. Percolation of phases in a threedimensional
cement paste microstructural model, Cem Concr Res 1991;21(23):325
344.
[227] Powers TC, Brownyard TL. Studies of the physical properties of hardened
Portland cement paste, J Am Concr Inst 1947; 22(9):971992.
[228] Hansen TC. Physical structure of hardened cement paste. A classical
approach, Mater Struct 1986;19 (6):423436.
[229] Lamond J, Peilert J. Significance of tests and properties of concrete and
concretemaking materials STP169D, ASTM International, Philadelphia,
PA, 2006.
[230] Xi Y, Bažant ZP, Molina L, Jennings HM. Moisture diffusion in
cementitous materialsAdsorption isotherms, Adv Cem Based Mater
1994;1:248257.
ABSTRACT Earlyage cracks in fresh concrete occur mainly due to high rate of surface evaporation and restraint offered by the contracting solid phase. Available test methods that simulate severe drying conditions, however, were not originally designed to focus on evaporation and transport characteristics of the liquidgas phases in a hydrating cementitious microstructure. Therefore, these tests lack accurate measurement of the drying rate and data interpretation based on the principles of transport properties is limited. A vacuumbased test method capable of simulating earlyage cracks in 2D cement paste is developed which continuously monitors the weight loss and changes to the surface characteristics. 2D crack evolution is documented using timelapse photography. Effects of sample size, w/c ratio, initial curing and fiber content are studied. In the subsequent analysis, the cement paste phase is considered as a porous medium and moisture transport is described based on surface mass transfer and internal moisture transport characteristics. Results indicate that drying occurs in two stages: constant drying rate period (stage I), followed by a falling drying rate period (stage II). Vapor diffusion in stage I and unsaturated flow within porous medium in stage II determine the overall rate of evaporation. The mass loss results are analyzed using diffusionbased models. Results show that moisture diffusivity in stage I is higher than its value in stage II by more than one order of magnitude. The drying model is used in conjunction with a shrinkage model to predict the development of capillary pressures. Similar approach is implemented in i
drying restrained ring specimens to predict 1D crack width development. An analytical approach relates diffusion, shrinkage, creep, tensile and fracture properties to interpret the experimental data. Evaporation potential is introduced based on the boundary layer concept, mass transfer, and a driving force consisting of the concentration gradient. Effect of wind velocity is reflected on Reynolds number which affects the boundary layer on sample surface. This parameter along with Schmidt and Sherwood numbers are used for prediction of mass transfer coefficient. Concentration gradient is shown to be a strong function of temperature and relative humidity and used to predict the evaporation potential. Results of modeling efforts are compared with a variety of test results reported in the literature. Diffusivity data and results of 1D and 2D image analyses indicate significant effects of fibers on controlling earlyage cracks. Presented models are capable of predicting evaporation rates and moisture flow through hydrating cementbased materials during earlyage drying and shrinkage conditions.
ii
To my family for their support, patience and understanding during the past four years
iii
ACKNOWLEDGMENTS I would like to start by expressing my gratitude to my advisor Professor Barzin Mobasher for giving me the opportunity to work with him and study at Arizona State University. His guidance and support are the reasons I was able to accomplish this work. Thank you to Professor Subramaniam D. Rajan and Dr. Claudia E. Zapata for serving on my defense committee. Thank you to Dr. M. Shekarchi and late Professor Ghalibafian for starting me on this trek of concrete at University of Tehran. I would like to gratefully acknowledge financial support for this project received from the Arizona Department of Transportation under project number SR633. This project was conducted in the Structural Mechanics and Materials Testing Laboratories at Arizona State University and as such I would like to acknowledge the support that has made these experiments possible. I would also like to acknowledge the assistance of Mr. Peter Goguen, the laboratories manager, Mr. Danny Clevenger and Mr. Jeff Long, for assistance in performance and preparation of experimental testing. In addition, I want to acknowledge the assistance of my colleagues at Arizona State University including, Dr. Amir Bonakdar, Dr. Deju Zhu, Dr. Chote Soranakom, Dr. Flavio Silva, Geoffrey Minor, Christopher Barsby, Sean Krauss, and Nathan Rodriguez. Thank you to all my friends who helped keep me sane during the process including Jonathan Fein, Erblina Vokshi, Vikram Dey, Aditya Vaidya, Kapil Krishnan, Pedram Shafieian, Shahrzad Badvipour and Kaveh Behbahani.
iv
And finally I would like to thank my brothers for being a constant reminder to worry about the important things in life.I would finally like to express my deepest appreciation and gratitude to my family. I would like to thank my parents for their love and support through my path of selfdiscovery. v .
..............10 1..........5.................3..3............22 2..................................25 vi ..................................4...4.......................... Restrained Shrinkage Cracking.........1 1...... Restrained Shrinkage Testing of Cementbased Materials ...................... Earlyage Shrinkage Cracking Mechanisms .1 1.........4.2........14 2................................1.. Mixture Properties.... Modeling the Drying of Cementbased Materials.TABLE OF CONTENTS Page LIST OF TABLES…………………………………………………………..... Introduction .....20 2......11 1.17 2... Objectives of the Dissertation ........................17 2........................ Overview of Thesis ..........……............. RESTRAINED SHRINKAGE CRACKING........................... Motivation ......xi NOMNECLATURE ………………………………………………......................2.3.5 1........3.......................................................... Results of Strain Gauges Attached to Steel Ring....................... INTRODUCTION ..4..........xx CHAPTER 1...............................……............…...8 1........................13 1.. Earlyage (Plastic) Shrinkage Testing of Cementbased Materials ..........4.............3.................2..................3.....3............17 2..................21 2. Free Shrinkage Testing of Cementbased Materials ....................................x LIST OF FIGURES…………………………………………………..........2.6 1..... Restrained Drying Shrinkage Test Methodology ............................................................. Review of Drying Shrinkage Testing Methods .................. Background in Shrinkage Testing and Modeling ....................................6 1.................1..Casting and Specimen Preparation ..............1........................1...................................................................
................................................................44 2............ Materials..................................68 3....4...............36 2......68 3.............................56 3...............1............................... Conclusion .......70 3..................................................... Analysis of typical evaporation data under low pressure test condition ......2.................43 2....................5..........7.......50 3..............................1...............................6........................ Capturing Shrinkage Cracks on the Ring Samples ...5.........................4............4..3...........2.....51 3..................................... Age Dependent Concrete Strength........................5........................ Moisture Diffusion and Free Shrinkage ... Comparing Experimental Data and Simulations............. Cumulative moisture loss and evaporation rate versus time ...........................CHAPTER Page 2..................5...........29 2...........1.............3............51 3..............38 2.71 vii .....5..48 2......... Calculation of Moisture Diffusivity ............ Restraining Effect .............................................54 3................... Testing methodology .............3........ Algorithm for Strain History in Steel Ring and Crack Width at Concrete Surface .........47 2...... Experimental program ........ Image Analysis of Shrinkage Cracks ............... Introduction .... Stress Strain Development .5................ Placing and Curing procedures .................4........................... Mixing......... Parameter Estimation ..........5.. Theoretical Modeling of Restrained Shrinkage .. Scope of Test Program .6......3........................................1................................................. EXPERIMENTAL OBSERVATIONS OF EARLYAGE DRYING .....56 3...............32 2............5..4.. Creep of Concrete ....35 2.............4......39 2...............................62 3...........................................5...................2.......................................................3.....2......4.....................................
.................................3........................77 3................1..................................98 4....... Evaporation as a Mass Transfer Problem ...............126 5.... Theory of evaporation from cementitious materials .............. Comparison with Experimental Data and Finite Element (FE) Analysis ....... Effect of sample size .. Effect of w/c ratio ................89 4............ Introduction ............................5...............133 viii ......... Mass Transfer at Low Rates ....2....................................129 5.................2...... THEORY AND MODELING OF EARLYAGE DRYING ........6......122 5.93 4..87 4...126 5...101 4.....100 4..5........1.................6...............3......... Available Method to Estimate Initial Rate of Evaporation .............. Stage I Drying ..3..........4................................. Simulation and Parametric Studies ......5....................................................................................................87 4...................... MODELING RATE OF EVAPORATION POTENTIAL ....75 3.........131 5...1........................................... Effect of fiber content and cracking.........115 4..... Conclusion ........................................1..........2................................... Stage II Drying .............101 4..... Effect of duration of initial curing ...............................................1...............2................................................................... Introduction ...................CHAPTER Page 3..106 4............................................... TwoStage Modeling of the Drying of Cementitious Materials .....3...........5.......................................................................3............4.5........4......... Physics of water evaporation ......79 3................84 4..4...4......................... Governing Equations and Geometry of the Problem .............................................3......................1.................. BackCalculation Procedures for Modeling of Stage II Drying ...113 4..............................................3...72 3............ Conclusion ......2.........
CHAPTER
Page
5.3.2. Mass Transfer at High Rates ..............................................................142 5.3.3 Algorithm for determination of evaporation rate ................................147 5.4. Parametric Studies ....................................................................................148 5.5. Model Comparison with Experimental Results of Concrete Evaporation 150 5.6. Conclusion ................................................................................................157 APPENDIX A B C D E F Fee Shrinkage Test Results ..........................................................................158 Vacuum Evaporation Test Method and Data Transfer Description.............164 Finite Element Analysis of Evaporation Test Data......................................175 Drying Tests on Clays: An Application of the Developed Test Method .....204 Image Analysis on Earlyage Cracks: Quantification and Orientation ........213 Modeling Earlyage Shrinkage of Concrete Due to Evaporation………….227
REFERENCES ....................................................................................................248
ix
LIST OF TABLES Table Page
2.4.1.1. Mixture proportions of the Control and GRC samples (kg/m3) ................23 2.4.4.1. Mean Crack width and standard deviation of samples (mm) ...................34 2.5.3.1. Backcalculated parameters of Control and GRC3 samples .....................42 3.4.1.1. Scope of the test program .........................................................................70 3.4.2.1. Mix proportions of the test series..............................................................71 3.5.1. Initial evaporation rates, transition time, evaporation rate at 24 h, cumulative moisture loss, and diffusivities of tested samples .......................72 5.5.1. Results of the evaporation rate in comparison with the experiments ........154 5.5.2. Parameters used for calculation of evaporation rates ................................155 A.1. Mixture proportions of the ARGs and Control samples (lb/ft3)...................159 C.1. Analogy and corresponding terms between two BVPs ................................187 C.2. Results of FE analysis with ABAQUS using Q4 elements comparing to the experimental results .....................................................................................191 C.3. Results of FE analysis with ABAQUS using Q8 elements comparing to the experimental results .....................................................................................192 D.1. Mix proportions of the test series (kg/m3) ...................................................205 D.2. Initial evaporation rates, transition time, evaporation rate at 24 h, cumulative moisture loss, and diffusivities of tested samples ........................................209 E.1. Results of crack analysis on a plain and FRC samples ................................217 E.2. Results of crack analysis and quantifications ...............................................225 F.1. Properties of PVA fiber used in this study………………………………...241
x
LIST OF FIGURES Figure Page
1.1.1. Greenhouse gases trap some of the heat causing global warming ................4 1.1.2. Sustainable development principle ...............................................................4 1.2.1. Equilibrium between liquid water and humid air at the interface of liquid/gas phase inside a pore and different pressures applied on solid skeleton ............................................................................................................6 1.3.1.1. Molds, samples and digital comparator dial used in free shrinkage tests ...8 1.3.1.2. The configuration and geometry of restrained shrinkage ring specimen ..10 2.4.1.1. Experimental setup of shrinkage test and recording strain gauge results .24 2.4.2.1. Typical result of an attached strain gauge for a plain concrete sample ....26 2.4.2.2. (a) Results of two strain gauges mounted on a plain concrete sample, (b) Location of the crack and strain gauges .........................................................27 2.4.2.3. Smoothing typical results of a strain gauge attached to steel ring ............27 2.4.2.4. Effect of glass fibers on results of strain gauges attached to the rings .....28 2.4.3.1. Imaging with the digital camera for crack investigations .........................30 2.4.3.2. Transverse cracks due to restrained drying shrinkage; a) control sample, b) GRC3 sample after 14 days of drying in the shrinkage chamber ..............30 2.4.3.3. Reconstructed shrinkage crack images of control and GRC3 samples.....31 2.4.4.1. Image analysis of a single image taken from a Control sample ...............33 2.4.4.2. Mean and SD of shrinkage crack widths at different times of drying ......35 2.5.1.1. Fick's law of diffusion is used to simulate humidity profile through the thickness of the concrete section using an error function ..............................37
xi
Figure
Page
2.5.1.2. Fitted free shrinkage results by the modified ACI 209R92 model ..........38 2.5.2.1. Creep model for Control and GRC mixes.................................................40 2.5.3.1. Flexural response of concrete samples under threepoint bending tests ..40 2.5.3.2. Tensile stressstrain and crack width model .............................................41 2.5.3.3. Backcalculated tensile stress strain model at different ages ....................43 2.5.5.1. Schematic drawing for the proposed drying shrinkage model; (a) ring specimen; (b) strain components ...................................................................45 2.6.1. Comparison of experimental results and model simulation; (a) history of strain in steel ring; and (b) crack width dimension history at the outer concrete surface .............................................................................................49 3.2.1b. Plan view of the mold (numbers in mm) ...................................................56 3.3.1.1. Typical cumulative moisture loss and evaporation rate of a cement paste sample versus time, (a) in linear scale, (b) in log scale .................................58 3.3.1.2. Schematics of moisture transport during drying process of porous media representing solid (S), liquid (L) and vapor phases (V): (a) initial condition , (b) capillary saturation, (c) hygroscopic state (continuous vapor phase) ......59 3.3.1.3. Evaporation test results from water surface comparing to cement paste ..60 3.3.1.4. 2D Crack development during drying of a fresh plain cement paste ......61 3.3.2.1. (a) Deriving aln by fitting a straight line to the curve of
ln (1 − ΔM t / ΔM max ) vs. time in stage II, (b) Simulation of cumulative
moisture loss vs. experimental data for a cement paste sample .....................69
xii
Figure
Page
3.5.1.1. Effects of surface area and thickness on results of evaporation tests on plain cement paste ..........................................................................................74 3.5.2.1. Effects of w/c ratio on evaporation results of plain cement pastes ...........76 3.5.3.1. Effect of curing duration on results of evaporation tests ..........................78 3.5.4.1. Results of drying tests on Portland cement paste with different content of ARglass fibers...............................................................................................81 3.5.4.2. Crack pattern of cement paste specimens with and without ARglass fibers after 24 hours of drying under lowpressure test condition .................82 4.2.1. Schematics of the interfacial boundary layer during evaporation from free water surface ..................................................................................................91 4.2.2. Variation of the diffusion coefficient and saturated vapor pressure versus temperature ....................................................................................................92 4.2.3. Parametric study on the effects of temperature, boundary layer thickness and relative humidity on evaporation rate from water surface ......................94 4.3.1. Two stages of drying during evaporation of capillary porous materials .....97 4.3.1.1. (a) Fully saturation state of porous material, (b) capillary saturation of porous material, (c) variation of hydraulic potential and vapor pressure from the interior of the material through ambient in stage I drying. ......................99 4.3.2.1. (a) Low saturation state of porous material, (b) variation of hydraulic potential and vapor pressure using simplified model (model A), (c) the variations using convection model (model B) .............................................101
xiii
..................................3...4....112 4..4....................... governing equations and boundary conditions during: (a) stage I drying.2..........2.....106 4.....110 4.............. (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the experimental data of an evaporation test.......... Parametric study of moisture diffusivity at stage II drying and convective moisture transfer coefficient on cumulative moisture loss and evaporation rates of cement paste samples ......................... .............. on cumulative moisture loss and evaporation rates during drying of cement paste samples .....1......5.....................1......3.... on cumulative moisture loss and evaporation rates during drying of cement pastes ............... and transition time.... Parametric study on the effect of moisture diffusivity at stage I on profiles of moisture concentration drawn at 1 h intervals during drying ........2........... (b) stage II drying assuming convective moisture flow ........... ttrans..........1...... (b) Backcalculated D and k for the best fit of data ...... Geometry.......................3.......116 4......4........ Parametric study on the effect of initial evaporation rate............ k.......4....113 4............................2................5.. Distribution of moisture flux (kg/(m2h)) at different drying times...........................118 4......Figure Page 4.2.. Parametric study on the effect of moisture transfer coefficient.... Backcalculated moisture diffusivity of an earlyage noncured cement paste as a function of normalized concentration expressed by CEBFIP models ..2............121 xiv ...109 4................. F0..4...........2.....2......4...
Effect of temperature.....1...2.......................143 5....4... (2007a.....3...156 5........ The comparison between the experimental data and the modeling values of evaporation potential based on a boundary layer mass transfer theory .106) ....141 5................ Diffusion boundary layer thickness in a mixed laminar and turbulent regions for free stream flow over a smooth isothermal flat plate .......................3..2..134 5...124 5................................. (b) Backcalculated D’ and k for the best fit of data .. relative humidity... Results of analysis on drying data by Šelih (1996)..Two stages of drying during evaporation of capillary porous materials ...... Schematic representation of the concentration distribution near an interface...3..3...............................1...2......b) by a method based on a boundary layer mass transfer concept ...............156 xv ............. ACI nomograph for estimating rate of evaporation of surface moisture from concrete (ACI 305..........1.....130 5.....................................5...........1... (a) Prediction of cumulative moisture loss and evaporation rate as a Page function of time for the experiment performed by Šelih (1996).............5......1........2...................... The mass concentration boundary layer in the highrate mass transfer problem ..................... wind velocity and characteristic length on the rate of evaporation from water surfaces..........5......123 4...................1.......2............................Figure 4..128 5...1.....151 5...........4..............1....................127 5......1......5..................1.......... ...... Prediction of experimental data of Azenha et al................ Development of 2D plastic shrinkage cracks during drying of a fresh plain cement paste ..................................................................
..............182 C.....8... Input diffusivity and moisture transfer coefficient as a function of time .................182 C............... Finding the critical time step (37<Δtcrit<38) by trial and error for FE analysis with 2 elements ........4....................... Finding the critical time step (152<Δtcrit<153) by trial and error for FE analysis with 1 element .....1.........2........Figure Page A................................ (a) Results of FE analysis with different elements and time steps...... Free shrinkage results of Control and ARG samples and their comparison 161 A......185 C...................177 C....5................................................... Tracking the values of cumulative moisture loss at end of the test by increasing number of elements ......1. Governing differential equations and boundary conditions of the problem....................................................3. Fitting experimental free shrinkage data of ARG samples with modified ACI 209R92 model.......176 C..... (b) Comparing the results with experimental data......163 B... Finding the critical time step (16<Δtcrit<17) by trial and error for FE analysis with 3 elements ......................................183 C.............................................183 C..............................................185 xvi ...165 C.........................184 C... Finding the critical time step (9<Δtcrit<10) by trial and error for FE analysis with 4 elements ........................... Modeling free shrinkage data of Control sample using ACI 209R92 model162 A....................................1.............3.................................................6.......2.................................. 2D shrinkage cracking test setup based on vacuum technology .........7..................9........................... Comparing results of 1D FE analysis by the MATLAB code with the analytical results...................
....189 C............. c) t=14 h..................14........206 xvii ................186 C......... d) t=20 h......................................10... Input values of diffusivity as a function of moisture concentration . Different mesh sizes used for convergence studies ..17...187 C..................13...... Comparing results of analysis by ABAQUS with the experimental results and results of 1D analysis by the MATLAB code ............. A representative thin slice cut perpendicular to top surface for modeling 1D moisture transport ..............Figure Page C..............19....1..............................190 C................................194 C............................ b) t=10 h......193 C.....188 C.................195 C.188 C....... Final cumulative moisture loss and initial evaporation rates versus number of nodes ...11............................................. c) t=14 h.... The boundary condition of stage I drying defined as a load ......................... e) t=40 h .................. (b) drying rate curves.............................21. Comparing results of FE analysis by ABAQUS with the analytical results196 C....197 D.22.......................20......... Distribution of moisture concentration (kg/m3) at: a) t=2 h...................... Results of evaporation tests on clay and their comparison with cement paste sample: (a) cumulative moisture loss curves... Convergence studies on the results of evaporation tests with Q4 elements191 C........... d) t=20 h.... e) t=40 h ........... Finite element domain ..................................15.......... Convergence studies on the results of evaporation tests with Q8 elements192 C.........................................12......... b) t=10 h...............16..18............................. Distribution of moisture flux (kg/(m2h)) at: a) t=2 h................... Cement paste slab subjected to different boundary conditions during stag I and II drying .....
.................................... (b) Backcalculated D and k for the best fit of data .. Normal probability and cumulative density curves of crack width and crack length..5....207 D.......3.................................... Geometry................. Equilibrium between liquid water and humid air at the interface of liquid/gas phase ……………………………………………………………..............1............211 D. Effective stress on solid skeleton based on unit cell modeling……………239 xviii ......................218 F........... Analysis by counting number of intersections of cracks network with the oriented secants (parallel equidistant lines) .2. governing equations and boundary conditions during………...212 D...........1.............5......………237 F..................231 F...........................2..........................6...................... time with a Portland cement paste ............... Image of a 2D crack pattern in a clay sample after vacuum drying for 48 h212 E...................... 2D Crack development during drying of a clay sample .215 E......216 E.....211 D....7..... Results of diffusion analysis on wet clay and cement paste samples .... Results of analysis on drying data by Šelih (1996)……………………….............................. 4.2........3...................................Figure Page D. Comparing cracking area of a clay sample vs............................209 D............. Prediction of cumulative moisture loss and evaporation rate as a function of time (Selih 1996) ………………………………………………………........ (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the experimental data of an evaporation test.....................3.........234 F.....233 F... The process of image analysis on a cracked paste sample .............. Boundary conditions and Fick’s 2nd law of diffusion for the presented drying problem ....4..............
9. Introduced elastic modulus for plain and PVA 0.…241 F.10. Simulation of earlyage shrinkage strain averaged over the thickness of the sample in comparison with the experimental data reported by Wongtanakitcharoen and Naaman (2007) ………………………………247 xix .244 F.245 F.. Simulated earlyage shrinkage strain profiles during drying ………….2% concrete …………245 F.246 F.7.239 F.240 F..13.……..8.11.242 F. Simulation of moisture concentration relative humidity profiles ………. Unrestrained earlyage shrinkage test setup (Wongtanakitcharoen and Naaman 2007)…………………………………………………………. Simulation of capillary pores pressure and macroscopic pressure on solid skeleton profiles…………………………………………………………. Prediction of cumulative moisture loss and evaporation rate (Wongtanakitcharoen and Naaman 2007)………………………….. Modeling development of modulus of elasticity in early ages (0150 h)…. Simulation of degree of hydration in comparison with experimental data.Figure Page F..14. Change in Capillary porosity during aerlyage………………………….6.12..243 F.15.
m2] la.s1] ′′ ma : total mass flux of all components in the humid air [kg.m3] D: diffusivity of the water vapor in the air [m2.m3] ce: mass concentration of water vapor at the air stream [kg.s1] patm: atmospheric pressure [kPa] p: water vapor pressure [kPa] xx .s1] KH: drying coefficient per unit humidity difference [kg.mol1] ms: mass fraction of water vapor at the surface m′′ : total mass flux of water vapor [kg.m2.s1] E: Rate of evaporation based on Menzel’s equation [kg.m3] ccrit: critical water vapor concentration for transition from stage I to II [kg. 3.176x1010 [m] L: total characteristic length in the direction of air flow [m] M: molar mass of the water [kg.m3] cs: mass concentration of water vapor at the interface [kg.m2.h1] hc: heat transfer coefficient by convection [W.m3] BF: blowing factor c: mass concentration of water vapor [kg.e.m2.m2.v: characteristic length of the mixture of air and water vapor.s1] Js: rate of diffusion of water vapor per unit area of the surface [kg.NOMENCLATURE Bm: mass transfer driving force for water vapor [kg.mol1] Ma: molar mass of the air [kg.oC1] hm: mass transfer coefficient for water vapor [m.s1] hm : average lowrate mass transfer coefficient for water vapor [m. i.m2.s1.
psat: saturated water vapor pressure [kPa] pv: partial pressure of the water vapor in the air far from the surface [kPa] Recrit: critical Reynolds number ReL: Reynolds number for the characteristic length in the direction of flow Rex: local Reynolds number Sc: Schmidt number Shx: local Sherwood number Sh L : average Sherwood number T: absolute temperature [K] Ta: temperature of the air far from the water surface [oC] Tc: temperature of the concrete surface [oC] Tfilm: temperature of the boundary layer (film) Twb: wetbulb temperature of water [oC] t: transport time for the mass diffusion [s] ttrn: time of transition from stage I to stage II drying [h] u∞ : freestream velocity [m.s1] vs: velocity of water vapor at the interface [m.h1] x: the distance from the leading edge [m] xcrit: distance from leading edge at which critical Reynolds number is reached [m] y: distance in the direction of diffusion [m] δ d : diffusion boundary layer thickness [m] xxi .s1] va: mass average velocity of all components in the humid air at interface [m.s1] V: wind velocity [km.
s1] ρ: density of the air [kg.m3] ρa: total density of all components of the humid air at the interface [kg.s1] ν: kinematic viscosity of the air [m2.m3] ρs: density of the water vapor at the interface [kg.m3] φ: relative humidity of the air far from the surface Ω D : dimensionless collision integral for diffusion xxii .λ: latent heat of evaporation at Twb [kJ.kg1] μ : dynamic viscosity of the air [kg.m1.
57 million paid employees in 2000. Although residential construction is still growing in the industrialized countries primarily 1 .. gravel.1.S. Among these sectors construction materials is very important as for almost a century.4% of the U. Construction involves numerous manufacturing sectors such as construction materials. and service sectors. and fishing.7% to the GDP. gross domestic product (GDP) that year. It is one of the largest industries in any country of the world with a vast number of participants. i. over three times more than agriculture. Introduction 1. These numbers show that construction industry’s contribution to the GDP in 2000 was larger than the GDP of 212 countries of the world [2]. Heavy civil construction is increasingly faced with considerable public opposition due to construction activities’ environmental impacts.5 billion in 2001. the construction industry uses the most materials by weight comparing to other industries [1]. crushed rock. and cement are the ones which are used the most by volume. building systems. The value of new construction put in place was $842. Among all construction materials. and almost four times more than motor vehicles and equipment manufacturing [1]. These data indicate that the materials that make up the bulk of Portland cement concrete and asphalt concrete are used in the largest volume and a special attention has to be given to concrete and cement industry. there were 709. Only in the United States. Motivation The construction industry has to support continuing population growth and social and economic development.e. Construction services contributed 4. or 8. sand. forestry.1.590 establishments with 6.
Altogether. the combustion of fossil fuels to heat up the clinker is another way of CO2 production. Fossil fuels are hydrocarbons and burning process is the oxidation of carbon that leads to CO2 emission (C + O2 CO2+ thermal energy) by as much as 165 kg CO2 per 1 ton of clinker [3]. finding alternative fuels and energy.owing to renovations and retrofits. CO2 production due to calcination of limestone is one way of CO2 production (CaCO3 CaO+ CO2) which leads to CO2 emission of 525 kg per 1 ton of clinker [3].69 ton CO2. extending service life of cementbased materials and enhancing performance of cement systems also result in reduction of CO2 emissions as well as reducing materials demand and 2 . It is now wellknown that CO2 as one of the main greenhouse gases traps some of the sun’s heat energy causing a significant climate challenge called as ‘global warming’. the growth cannot be maintained without tackling global economical and environmental issues namely the rising cost and demand of natural resources and energy and increasing environmental concerns due to industrial waste products. reformulation. the cement production contributes approximately 5% of the earth’s CO2 emissions which is a very significant number. capturing CO2. suburban sprawl. and automobility [1]. production of 1 ton clinker emits 0. Therefore. concrete and cement industry is considered as a large polluter of environment due to production of CO2. Different methods to reduce CO2 emissions include but not limited to: improving energy efficiency. While these solutions are directly related to the production of cement. and establishing tough regulations against CO2 emissions. Focusing on building material sector. On the other hand.
The main topic is the drying which is a major cause of earlyage shrinkage. These cracks are the main routes through which 3 . and accelerate deterioration.economical benefits. and therefore they crack when restrained against shrinkage strains. The main thrust of this effort is to use novel characterization and simulation techniques in the development of a range of materials for applications in various infrastructure systems. Using sustainable construction materials will eventually lead to structures which are more durable and economical. especially during early hours after casting. 5]. This task can be achieved through a fundamental scientific understanding of degradation mechanisms in cement systems. The focus is on increasing durability of cementbased materials by controlling earlyage shrinkage cracks. resulting in increased maintenance costs and reduced service life [4. while meeting the ever increasing demand of the public. This research program will focus on addressing sustainability metrics through prolonging the service life of products currently used. Cementitious materials are weak against tensile stresses. Earlyage shrinkage cracks in concrete reduce load carrying capacity. This is a sustainable solution since it leads to environmental and social responsibility as well as economical performance and therefore meets the needs of the present without compromising the ability of future generations to meet their own needs.
Greenhouse gases trap some of the heat causing global warming [6] Figure 1.1.1. Although attributed to several driving forces such as differential settlement. and autogenous 4 .Figure 1. Sustainable development principle aggressive agents such as chloride ions penetrate into the concrete mass and threaten the longterm durability of structures [7].1.2. thermal dilation. Plastic shrinkage in presence of restraints leads to tensile stresses which may easily exceed the low tensile strength of fresh concrete and result in cracking.
This pressure acts on the solid particles resulting in the contraction of the still plastic material [18] ranging from strain levels of 14x103 [12.deformation [8]. 14] and continues to rise as the evaporation proceeds. the solid particles at the surface of fresh concrete are no longer covered by the water and water menisci are formed in the interparticle spaces because of adhesive forces and surface tension [ 12 ]. 17]. 10. Earlyage Shrinkage Cracking Mechanisms Earlyage cracking is a threat to structural integrity of concrete structures and if not inhibited would lead to service life reduction and unsightliness. wind speed and relative humidity increase the risk of early age cracking [11]. The developed capillary pressure in drying fresh concrete can reach 50 kPa in less than few hours [10. 5 . A negative pressure in the capillary water is built up due to the curvature of the water surface [13. Due to evaporation. 14. 20 ]. 11].2. 15] and exceed 1MPa [16. High rates of evaporation which depends on air and concrete temperature. the developed shrinkage strain may easily exceed low tensile strength of the fresh concrete [21] and cracking occurs. 12. The drying as the main cause of shrinkage is studied thoroughly. plastic shrinkage cracking in concrete occurs principally due to a high rate of water evaporation from the concrete surface [9. 19 . 1. If concrete is restrained.
1. Equilibrium between liquid water and humid air at the interface of liquid/gas phase inside a pore and different pressures applied on solid skeleton [22] 1.1. This test method is particularly useful for comparative evaluation of shrinkage potential in different hydrauliccement mortar or concrete mixtures. Assuming the length of the specimen is much larger than the cross 6 . This test method covers the determination of the length changes that are produced by causes other than externally applied forces and temperature changes in hardened hydrauliccement mortar and concrete specimens made in the laboratory and exposed to controlled conditions of temperature and moisture. specimens of 1” [25mm] square crosssection and approximately 111⁄4” [285 mm] in length are casted for each mixture. In this test. Measurement of length change permits assessment of the potential for volumetric contraction (shrinkage) of concrete due to drying.3. Background in Shrinkage Testing and Modeling 1. Free Shrinkage Testing of Cementbased Materials The free shrinkage test is performed according to ASTM standard C 157 [23].Figure 1.3.2.
7 .sectional dimensions. this test cannot be an indicator of cracking performance of cement systems against shrinkage and does not differentiate the contribution of different materials such as fibers in controlling shrinkage cracks. After the initial comparator reading. In this method. the test specimens are cured in the molds covered with a plastic sheet for 24 hours while water dripping from specimen is protected. Upon removal of the specimens from the molds. These readings are done in a room maintained at a relative humidity of 30% while the specimens are at a temperature of 73°F [23°C]. so that the specimens have a clearance of at least 1” [25 mm] on all sides. the initial length reading is taken using a digital comparator dial. Since the specimens are not restrained in this test method. Comparator readings of each specimen are taken every one or two days after initial reading. The measurement of change in length with time can then provide a measure of onedimensional shrinkage of the material. the specimens are stored in the drying room. The difference between the average of readings on at least two specimens and the initial length of specimens is used to calculate shrinkage strain. then shrinkage takes place only in the length direction.
however. especially when cross sectional dimensions are large [24]. samples and digital comparator dial used in free shrinkage tests 1. Paillere. uniaxial tensile stresses are produced which is similar to a uniaxial tensile test. Buil.(a) (c) (b) Figure 1.5 m) with flared ends that were restrained.3. Restrained Shrinkage Testing of Cementbased Materials If the prismatic specimen is restrained on the length direction. Molds. 27. Other researchers have used similar methods and linear specimens to assess the shrinkage cracking potential of concrete mixtures [26. Unfortunately 8 .2. and used small crosssectional dimensions (70 x 100 mm) to produce shrinkage cracking. The linear specimens have the advantage of the relatively straightforward data interpretation. and Serrano [25] used long specimens (1.3.1.1. 28. it is difficult to provide sufficient restraint to produce cracking with linear specimens. just as it is difficult to conduct a uniaxial tensile test for concrete [24]. It is also difficult to restrain linear specimens. 29].
When restraint to shrinkage is provided in two directions. Opsahl and Kvam [33]. 40]. instrumented rings have been used by researchers to measure the magnitude of tensile stresses that develop inside the material [24. Although this test method can be used to study cracking tendency of different materials. to better quantify earlyage cracking tendency of cementitious material. More recently. Shaeles and Hover [32]. 36. these test methods are generally not used for quality control procedures [28. 9 . the ring test has been developed into both AASHTO [41] and ASTM [42] standards. 38. Platetype specimens have been used by Kraai [31]. 37. but the steel ring would prevent this and cracking occurs. The main difference between these standards is the relative ratio of the concrete to steel ring thickness which influences the degree of restraint provided to the concrete. They used a 25 mm thick and 38 mm wide concrete ring cast around a 25 mm thick steel ring with an external diameter of 175 mm. a biaxial state of stress is produced. the test is not applicable to fresh concrete when plastic cracking is the main concern. A restrained shrinkage test using a steel ring was done as early as 1939 to 1942 by Carlson and Reading [35]. and Padron and Zollo [34]. 30]. As a result of drying.due to difficulties associated with providing sufficient end restraint. 39. Other investigators have used different types of specimens to simulate cracking due to restrained shrinkage. the results obtained from platetype specimens may depend on specimen geometry in addition to the material properties [24]. Consequently. Due to its simplicity and economy. a concrete ring would tend to shrink.
or heat lamps [48] to simulate severe evaporation conditions causing plastic shrinkage cracks in fresh concrete. parameters addressing rates of evaporation were not utilized in their work.3. and silica fume on the normalized cumulative moisture loss and maximum rate of 10 . Samman et al. Earlyage (Plastic) Shrinkage Testing of Cementbased Materials Several test methods have been implemented using fans [43. fans and heaters [46. While Wongtanakitcharoen and Naaman [20] and Wang et al.3. [51] studied the effect of fiber content. 44. results of test methods in which the drying of fresh concrete have been studied lack accurate measurement of the evaporation rate and data interpretation based on transport characteristics.Groove cut in plywood Threaded rod and a bolt Steel tube Cardboard form 267 mm 289 mm 423 mm Figure 1. The configuration and geometry of restrained shrinkage ring specimen 1.1. Wind tunnel [49] and vacuum drying [50] have also been used to expedite the drying process. heater. However.3.2. w/c ratio. and methanol treatment respectively. and fly ash on normalized cumulative moisture loss using fan. admixtures. these tests were not designed to focus on evaporation characteristics of the drying material as the principal driving force of plastic shrinkage. [52] used an electric fan to investigate the effect of w/c ratio. Also. 45]. 47].
Huldén and Hansen [58]. Using an approach based on soil science.4. Modeling the Drying of Cementbased Materials In existing literature. Chen and Mahadevan [64] referred to Walton et al [ 65 ] in using diffusion as the dominant moisture transport 11 . however. Kim and Lee [59] modeled the moisture movement by the Fick’s second law of diffusion without explanation of the physics of the problem. However. However. Analysis of the evaporation rates of cementbased materials with direct attention to stages of drying process has not been thoroughly investigated. 1.3. Kodikara and Chakrabarti [62] expressed the moisture movement during drying as a two separate stage mechanism [63]. Berhane [53] evaluated the effect of w/c ratio and ambient relative humidity using evaporation rates obtained from moisture loss curves. the moisture movement during drying of concrete is often described as a diffusion process where the diffusivity depends highly on the moisture content. They then extended the results from porous media to stones and concrete.evaporation. Hall and Hoff [54] analyzed the drying of clay brick ceramics using two distinct stages: a constant drying rate period and a falling drying rate period. 61] who considered the drying of concrete as a nonlinear diffusion problem. the scale of data requires the evaporation rates to be interpreted based on mass transport parameters. West and Holmes [56]. maximum rate of evaporation is not sufficient to characterize different materials characteristics. they modeled the phenomena as a onestage diffusion process. Garbalińska [57]. Most of these works refer to Bazant and Najjar [60. Torrenti et al [55].
and Garbalinska [57] applied a constant moisture concentration equal to ambient moisture concentration as the boundary condition in their models. In another approach. Torrenti et al [55] and Shimomura and Maekawa [50] also used evaporative boundary conditions similar to convection. 12 . The evaporation through the surface as a boundary condition is applied differently in available models. Šelih and Bremner [66] expressed the drying process as two different types of mechanisms of movement which can be described by a Darcytype equation near saturation. Although Shimomura and Maekawa [50] discussed the movement of liquid water as governing mechanism regarding moisture movement near the saturated state. Šelih and Bremner [66] concluded from the experimental results that the diffusiondriven moisture transport approach is appropriate only for the late stage of drying when the material is predominantly unsaturated. Bazant and Najjar [60]. Shimomura and Maekawa [50] used mass conservation of the vapor and liquid water to derive a diffusiontype equation. the assumed boundary conditions are not supported by the physics and driving forces of surface evaporation process. succeeding by a diffusiontype equation. Nonetheless. While ACI 305R99 [ 68 ] and Uno [ 69 ] recommended a constant flux equal to evaporation rate of water surface for earlyage hot weather concreting. 59]. their work was only limited to experimental observations. However. referring to Young [67]. 56. 58. a convective boundary condition proportional to the difference of internal and ambient moisture concentration was most often employed [50. the concept was not implemented in their model. and the moisture movement in the form of vapor flux is dominant.mechanism in concrete.
Objectives of the Dissertation The objectives of this research are as follows: (1) Build and calibrate an evaporation tests system to investigate evaporation characteristics of the drying cementbased material as the principal driving force of plastic shrinkage. (3) Investigate two distinct stages of drying which are generally proposed for drying of various capillary porous media. (6) Propose a simplified nonlinear diffusion equation. (5) Develop an image analysis technique to perform crack surface morphology and distinguish various surface crack patterns.1. 13 . (7) Propose appropriate boundary conditions for different stages of drying based on the physics of the problem which is useful when modeling the drying process.4. which is accounted for all drying processes namely. Darcian transport of liquid water and diffusional transport of vapor water. (4) Equip the tests system with restraining methods and imaging device to impose and track 2D shrinkage cracks from the early initiation stage to crack propagation and stabilization stages using timelapse photography. (2) Accurate measurement of the evaporation rates and data interpretation based on mass transport parameters.
Chapter 3 presents a test method for characterizing evaporation parameters and simulating the sequential formation of shrinkage cracks in twodimensional cement paste samples under lowpressure condition. Results of the tests including strain history in restraining steel captured by strain gauges attached to the midheight of the steel ring is monitored as concrete is exposed to an arid environment. 1. Overview of Thesis Chapter 2 describes testing methodology of a ringtype restrained drying shrinkage test. and restraining effect is presented. Along with the experimental results.(8) Model earlyage shrinkage from the drying model using the intermediate tools namely. Effects of sample size. aging material properties. an analytical approach that relates key influential parameters of modeling including diffusion. creep. desorption isotherms and capillary pressurebased models. initial curing and fiber content on weight loss and evaporation rates are explained. Two major stages of drying including a constant drying rate period 14 . w/c ratio. Also a systematic image analysis approach is followed to measure crack width growth in concrete ring specimen. shrinkage. (9) Predict cracking tendency of cementbased materials from material models by comparing predicted developed shrinkage with strain capacity of the material. Materials and mixture properties of different concrete samples including plain and fiber reinforced concrete samples are investigated experimentally.5.
In this chapter theory of water evaporation is described based on the boundary layer theory. A dualstage methodology for modeling the drying is introduced based on surface moisture transfer and internal moisture transport characteristics. The Schmidt number.(stage I) and a falling drying rate period (stage II) is investigated through experimental results. and fluid mechanics. the methodology is applied to the results of drying experiments under low and normal pressure conditions. Chapter 4 summarizes physics of water evaporation followed by a description of theory of evaporation from cementitious matrix that is represented as a capillary porous media. Sc. which is a dimensionless number defined as the ratio of momentum diffusivity (ν) and mass diffusivity (D). 15 . On the other hand. mass transfer. The results are interpreted based on mass transfer parameters using continuity of liquid and vapor phases. The FE method is further extended to 2D analysis using ABAQUS v. After studying the effect of different parameters on the model. Mass transfer at low and high rates is discussed separately.6. A driving force of concentration gradient is introduced as the main cause of mass transfer during the drying process. diffusion. Chapter 5 suggests a new approach in the absence of other theoretical methods for calculation of initial rate of evaporation from cementbased materials.101 [ 70 ] utilizing the analogy between the heat transfer and the moisture diffusion. physically relates the relative thickness of the hydrodynamic (velocity) and masstransfer (concentration) boundary layers. A 1D Finite Element (FE) analysis code is also presented in order to verify the results of proposed analytical method. the local Sherwood number.
Shx is a dimensionless number representing the effectiveness of mass convection
at the surface. Using these parameters and Reynolds number, Re, which gives a measure of the ratio of inertial forces to viscous forces, mass transfer coefficient can be derived. An algorithm for determination of evaporation rate is suggested. Considering similarity between rate of evaporation from water surface and concrete surface during initial stage of drying, predicted evaporation rates are verified by the recent experiments. The results show that given the environmental parameters as inputs, the model is capable of predicting the evaporation rates with a good accuracy assuming surface cooling occurrence.
16
2. Restrained Shrinkage Cracking
2.1. Introduction
This chapter deals with the role of fibers in extending the cracking resistance of concrete subjected to drying shrinkage. In hot and low humidity environments, concrete shrinks due to loss of moisture from capillary and gel pore microstructure. When concrete is restrained from free shrinkage, tensile stresses develop and may result in cracking if the stress exceeds materials’ low tensile strength. This is more dominant when at an early age the tensile strength is quite low and rate of moisture evaporation is high. The main objective of this chapter is to study effect of fibers on responses of a wellknown restrained shrinkage test. The methodologies of restrained drying shrinkage tests are described. Materials and mixture properties of different concrete samples including plain and fiber reinforced concrete samples are investigated experimentally. The experiments are performed according to the ringtype restrained shrinkage test method and strain history in the steel ring is monitored. A systematic methodology based on image analysis approach is used to measure crack width growth in concrete ring specimen. An analytical approach that relates key influential parameters of modeling including diffusion, shrinkage, creep, aging material properties, and restraining effect is presented.
2.2. Review of Drying Shrinkage Testing Methods
Understanding the mechanism of shrinkage cracking is essential to design of durable structures. Plastic shrinkage occurs during the early age period when
17
the strength of the paste is quite low and drying volume changes occur due to depletion of water due to evaporation, chemical reactions, in addition to chemical shrinkage. While the fibers may not affect the evaporation rate, their addition increases the strength and strain capacity sufficiently during the early ages so that the potential for tensile cracking is minimized. Fiber addition to concrete
therefore reduces cracking potential due to restrained shrinkage. As the concrete hardens, high stiffness fibers such as steel and macro synthetic fibers serve to increase the strength, crack growth resistance, and strain carrying capacity, providing a mechanism for additional restraint to distribute volumetric shrinkage and maintain a low crack width. To measure free drying shrinkage, ASTM C 157 [23] recommends a prismatic specimen of 25, 75 or 100 mm square cross section and 285 mm in length. If it is assumed that the length of the specimen is much larger than the cross sectional dimensions, then shrinkage takes place only in the length direction. The measurement of change in length with time can then provide a measure of onedimensional shrinkage of the material. In this test method no tensile stress and consequently no crack develops in the specimen. Therefore, this test method does not differentiate the contribution of fibers which reveals in the resistance of concrete to cracking. If the prismatic specimen is restrained on the length direction, uniaxial tensile stresses are produced which is similar to a uniaxial tensile test. The linear specimens have the advantage of the relatively straightforward data 18
interpretation, however, it is difficult to provide sufficient restraint to produce cracking with linear specimens, especially when cross sectional dimensions are large [24]. It is also difficult to restrain linear specimens, just as it is difficult to conduct a uniaxial tensile test for concrete [24]. Paillere et al. [25] used long specimens (1.5 m) with flared ends that were restrained, and used small crosssectional dimensions (70 x 100 mm) to produce shrinkage cracking. Other researchers have used similar methods and linear specimens to assess the shrinkage cracking potential of concrete mixtures [26, 27, 28, 29]. Unfortunately due to difficulties associated with providing sufficient end restraint, these test methods are generally not used for quality control procedures [28, 30]. Other investigators have used different types of specimens to simulate cracking due to restrained shrinkage. Platetype specimens have been used by Kraai [31], Shaeles and Hover [32], Opsahl and Kvam [33], and Padron and Zollo [34]. When restraint to shrinkage is provided in two directions, a biaxial state of stress is produced. Consequently, the results obtained from platetype specimens may depend on specimen geometry in addition to the material properties [24]. A restrained shrinkage test using a steel ring was done as early as 1939 to 1942 by Carlson and Reading [35]. They used a 25 mm thick and 38 mm wide concrete ring cast around a 25 mm thick steel ring with an external diameter of 175 mm. As a result of drying, a concrete ring would tend to shrink, but the steel ring would prevent this and cracking occurs. More recently, to better quantify earlyage cracking tendency of cementitious material, instrumented rings have 19
been used by researchers to measure the magnitude of tensile stresses that develop inside the material [24, 36, 37, 38, 39, 40]. Due to its simplicity and economy, the ring test has been developed into both AASHTO [41] and ASTM [42] standards. The main difference between these standards is the relative ratio of the concrete to steel ring thickness which influences the degree of restraint provided to the concrete. A ringtype restrained shrinkage testing method similar to AASHTO PP 3499 is used in this chapter to study effects of fibers on shrinkage cracking.
2.3. Restrained Shrinkage Cracking
Uniaxial restrained shrinkage tests and tensile tests conducted on largescale steel FRC specimens with fiber contents ranging from 0100 kg/m3 indicate that multiple cracking affects the overall response of the steel FRC in the hardened state [71]. The Ring test however is unable to capture the effect of multiple cracking simply because as a single crack forms the boundary conditions on the wall change. Several approaches to evaluate the influence of ring
geometry and drying direction on the behavior of the restrain shrinkage test using the steel ring specimen have been proposed [ 72 , 73 , 39]. These studies
demonstrate use of the steel ring in measuring the residual stress development as well as the stress relaxation once cracking occurs. Mane el al. [74] developed an experimental and analytical simulation algorithm to study the restrained shrinkage cracking in plain and FRC. A constant humidity chamber holding the restrained shrinkage specimens was utilized with a fan providing constant flow of air around the specimens. The strain in the restraining steel and the crack width in the
20
Shah and Weiss [75] demonstrated that prior to cracking. the stress transfer across the crack. and creep characteristics.5 and 6 kg/m3) on shrinkage cracking [77]. 3. The results are correlated with the specimen geometry. aging. 4. One possible method to reduce the adverse effect of cracking is the addition of randomly distributed short fibers to the concrete mixture. and concrete stiffness.4. 2. An instrumented ring specimen similar to AASHTO PP3499 [41] has been used to quantify the restrained shrinkage and tensile creep behavior of 21 . They also developed an analytical procedure for stress development in the steel ring. Acoustic emission measurements indicated similar energy release in plain and fiber reinforced specimens. and crack size [75].concrete samples were monitored continuously. Restrained Drying Shrinkage Test Methodology Restrained shrinkage due to drying environment may cause cracking in concrete and cementitious materials. ductility. it is possible to calibrate and interpret the experimental test results. humidity and temperature conditions. an analytical approach is needed to incorporate influential parameters of shrinkage.5. Shrinkage cracking can be a critical problem in concrete construction. fibers appear to delay the development of a localized. and microcracking in the stress analysis of a restrained concrete section. creep. shrinkage. In order to eliminate the influence of test conditions. stiffness of the steel ring. however. Using the theoretical models. An experimental study is conducted to evaluate effects of adding low contents of ARglass fibers (1. visible crack [76]. the stresses that develop in a plain and a FRC are very similar.
GRC1. Mixture Properties. 0.55 and slump of 65.3. GRC3 and GRC4.5.concrete. Gobain Vetrotex America Inc. Casting and Specimen Preparation To understand the drying shrinkage and cracking properties of glassfiber reinforced concrete (GRC). A schematic configuration and geometry of shrinkage ring specimen is shown in Figure 1. it is assumed that uniform shrinkage takes place along the height of specimen. as multifiber strand of 100 round filaments bonded together.90 mm.5 are designed with 1. In order to cast this sample. respectively.06%. Since the height of specimen is two times to its thickness.5. elastic modulus and specific gravity of fibers are 24 mm. The cement content was 680 Kg/m3 and only fine aggregates were used to cast concrete. The specimen consists of a 66. four concrete mix designs were developed with watercement ratio of 0.2 mm in thickness with outer diameter of 289.5 kg/m3 of glass fibers (24 mm long) respectively. These dosages are equivalent to 0.1. Four mixtures of Control. with filament diameter of 14 micron.3 mm. The length.11%.4.6 mm thick annulus of concrete cast around a rigid steel ring 11.2. 0.1. a cardboard form is used to hold the concrete from the outer sides. 72 22 .17% and 0% volumetric fractions. 2. 3 and 4. The width of the shrinkage cracks are measured at different ages using a highresolution digital camera and implementing a systematic procedure of image analysis. Two strain gauges are mounted on interior surface of steel ring at the mid height level and 90 degrees apart to measure developed strains in steel due to shrinkage of concrete. The AlkaliResistant (AR) Glass fibers were provided by St.4 mm and a height of 133.
glass fibers are added gradually to the GRC mixer and blended for 3 minutes. 80.1. First half of dry materials including cement and fine aggregates are introduced into the mixer and blended for 2 minutes.0 GRC1. respectively. Mixture proportions of the Control and GRC samples (kg/m3) Mix ID Portland cement Fine aggregates Water AR glass fibers w/c Sand/cement Control 680 1360 374 0 0. The AR glass fibers contain zirconium dioxide (ZrO2) which makes the glass composition resistant against alkaline environment. 82]. The mixture proportions of all mixes are shown in Table 2.0 GRC4.55 2.5 0. water is added to the mixer thoroughly and blended for 5 minutes before adding the rest of dry materials and mixing for an extra 5 minutes.5 680 1360 374 4.0 The following mixing procedure is adopted in this study. All molds are filled in three layers with proper compaction by a thin rod in 23 . and thus stable in concrete [78.4.55 2.5 0.68.1.1 and include Type I/II Portland cement as the primary binding agent.4.0 GRC3 680 1360 374 3 0.5 680 1360 374 1. 79. 81.GPa and 2.55 2. Finally.55 2. Then.17%. The dosage of AR glass fibers in GRC mixes are limited to the weight fractions of 4. Table 2.1.5 kg/m3 which is equivalent to very low volume fraction of 0.
the specimens are demolded and top surface of ring samples are sealed by a silicon rubber adhesive sealant to allow drying to occur only from the outer circumferential surface of the concrete specimen.1. The strain gauges attached to the inner surface of the steel ring are connected to the interface module and a dummy temperature compensating gauge in the form a halfbridge. no special technique such as vibration is used to help with the compaction of the fresh mixture in molds.1. Samples are placed in the shrinkage chamber shown in Figure 2.1. covered with sealed plastic sheets for 24 hours. Experimental setup of shrinkage test and recording strain gauge results The rings are placed in a constant humidity chamber at the temperature of 40º C.between the layers.1. The response from the strain gauges is collected using 24 . Figure 2. Due to high fluidity of all mixes. The samples are kept in a moist curing room. Four replicate specimens are made from each batch.4.4. After 24 hours.
4. 6 days after drying in the chamber.4. in the Figure 2. Two strain gauges mounted on each sample give different responses especially after cracking.g. concrete cracks and its effect on steel ring in plain concrete is a drop in the strain values. SCXI1000 carrier and NI data acquisition device and transferred into a LABVIEW programmed computer. Different responses before cracking can be related to different distances of strain gauges relative to heater location or source of 25 . As shown in the figure. When shrinkage strain in concrete exceeds ultimate tensile strength. This drop in strain gauge results shows time of cracking due to very low residual postcrack tensile strength in plain concrete. e. three stages can be seen in the response of strain gauge vs.4. In this stage (stage 3).2. the strain gauge response reflects effect of shrinkage in concrete at stage 2. Results of Strain Gauges Attached to Steel Ring The results of a typical strain gauge attached to steel ring for a plain concrete sample is shown in Figure 2. The program records the strain gauges readings at specified intervals.2.SCXI1321(16channel.1. The first stage which lasts 3 hours to less than 24 hours shows the effect of rising temperature and expansion of steel ring due to placement in chamber.1. time. Half Bridge) module. time response can be used to detect time of cracking in the specimen. 2. After steel ring reaches the same constant temperature as the chamber. Strain gauge vs. In this stage the response of concrete cast around the steel ring is linear elastic.2. strain in steel drop approximately to zero in plain concrete samples and visible crack width increases by the time.
4. In the postcrack response.2. The raw and smoothened results of a strain gauge attached to steel ring are shown in Figure 2.temperature.3. days Figure 2. microstrains Stage 1 Stage 3 Stage 2 25 0 25 50 75 100 PostPeak Cracking Prepeak region (elastic loading) Expansion equilibrium Corresponding to ultimate tensile strength 0 2 4 6 8 10 Drying Time.4.2. Typical result of an attached strain gauge for a plain concrete sample 26 .1. Results of two strain gauges mounted on a steel ring along with location of crack and strain gauges are shown in Figure 2.2 representing the strain responses of a plain concrete sample. The smoothened results of strain gauges were used to compare results of different mixtures.2. A simple code is used to reduce and smooth the data.4. the difference can be related to different distances of strain gauges from crack location in the sample. 50 Strain in Steel Ring.
micro strains 25 0 25 50 75 100 w/c=0.4. Smoothing typical results of a strain gauge attached to steel ring 27 . micro strains 25 0 25 w/c=0.2.55 curing time = 1 day Raw Strain Gage Data Smoothed Strain Gage Data 0 2 4 6 8 10 12 14 Drying Time. days (a) (b) Figure 2.50 Strain in Steel Ring.2. (b) Location of the crack and strain gauges 50 Strain in Steel Ring. curing time = 1 day strain gauge 1 strain gauge 2 50 75 100 Strain Gage 1 Strain Gage 2 0 2 4 6 8 10 12 14 Drying Time.2.3.55. days Figure 2.4. (a) Results of two strain gauges mounted on a plain concrete sample.
micro strains 0 20 40 60 80 100 0 2 Control GRC1.5. and the postcrack response of GRC samples with higher fiber dosages are significantly more than Control sample.2. According to the figure.5 GRC3 GRC4. The figure clearly indicates that using ARglass fibers delayed cracking in concrete samples for nearly 1 or 2 days.4. Strain in Steel Ring.5 4 6 8 10 12 14 Drying Time.4.2. which are the samples with no or very low amount of fibers show abrupt drop in the strain gauge data due to cracking. Cracking happens between 68 days for all different samples. The results of strain gauges of different mixtures are shown in Figure 2. Effect of glass fibers on results of strain gauges attached to the rings 28 . the results of the one showing more stability and not so close to the crack location was used to be compared with other samples. only Control sample and somehow GRC1.4.From two strain gauges attached on each sample’s steel ring.4. days Figure 2.
The samples were mounted on a traveling round plate and a 10x digital camera is used to take pictures.3. As one can see. 29 . overlapping portion of images is cut off and images are attached along the crack length.3. In Figure 2. Capturing is performed for all specimens after 14. 21 and 28 days of drying in the chamber. Figures 2.3.2. To do reconstruction of whole image. but in order to quantitatively measure them. Images taken by this camera are grayscale images with RGB color format.2.3. 21 and 28 days are chosen as convenient ages that provide sufficient time for shrinkage cracks to extend along the height of samples.3 show that crack width in GRC3 sample is almost 3 times smaller than in control. Using this setup. A view of the digital camera when taking picture of cracks of a concrete sample is shown in Figure 2. capturing digital images followed by image analysis approach is adopted in this study. images are acquired along the crack length using a series of 812 images.4.4.1.4. the difference between crack widths of different samples is obvious. wide images of transversal shrinkage cracks in control and GRC3 samples are shown. 14. Since the average cracking time is 7 days.4. Capturing Shrinkage Cracks on the Ring Samples In order to measure crack width. the entire surface of the specimen can be scanned and examined for new cracks and the measurements of the widths of already existing cracks.
a) control sample.2. b) GRC3 sample after 14 days of drying in the shrinkage chamber 30 .3.4.3.1.4. Imaging with the digital camera for crack investigations (a) (b) Figure 2. Transverse cracks due to restrained drying shrinkage.Figure 2.
Reconstructed shrinkage crack images of control and GRC3 samples 31 .3.4.3.Control (Sample 1) Control (Sample 2) GRC3 (Sample 1) GRC3 (Sample 2) Figure 2.
crack width is automatically measured by pixel counting at 64 different lines perpendicular to crack length and implementing the scale factor.1. To report crack width of a sample.4.4.5 mixes at 14. Then a code is developed to read the 8bit images. ask the user of scale factor and crack boundaries and then makes a binary image of crack boundaries.4. respectively. These values for four replicates of Control. standard deviation. GRC1. and demonstrates crack width histogram and normal probability density. Finally code calculates mean crack width. Effect of fiber addition is so significant on reducing crack width dimension. Crack width for each mixture is reported as an average crack width of four replicate samples. Figure 2. In the next step. average crack width and Standard Deviation (SD) obtained from 812 discrete images along the crack length are calculated. First all images are converted from RGB color to 8bit format.5. These results are shown in Table 2.4. crack width dimension increases by increasing drying time.5. 21 and 28 days are shown in Table 2. 3 and 4.4.4. 32 .5 kg/m3 ARglass fiber to the plain concrete (Control mix) crack width dimension at 14 days is reduced by 51%. GRC3 and GRC4. By adding 1. Image Analysis of Shrinkage Cracks Image analysis technique is used to measure crack width of concrete samples precisely. Results indicate that in all shrinkage samples.4.1. 72% and 82%.4.1 shows image analysis process and output statistical results of a single crack image.2. Approximately the same trend is observed after 21 and 28 days of drying in shrinkage chamber.
1. Image analysis of a single image taken from a Control sample 33 .(a) Input grey scale image (b) Binary image of crack region (c) Crack width measurement along the crack length (d) Crack width histogram probability (e) Crack width normal Figure 2.4.4.
065 0.404 0.4.244 0.766 0.405 0.048 0.303 0.340 0.391 1.043 0. Mean Crack width and standard deviation of samples (mm) Crack Width at 14 Days Average SD 1.227 0.038 0.066 0.053 0.427 0.173 0.504 0.168 0.5 1 GRC4.250 0.229 0.333 0.348 0.048 0.5 2 GRC4.307 0.010 0.823 0.157 0.216 0.046 0.386 0.363 0.153 0.046 0.5 3 GRC4.074 0.146 1.364 1.026 Sample Code Control 1* Control 2* Control 3 Control 4 Average GRC1.079 0.069 0.470 0.062 0.553 0.041 0.147 0.038 0.Table 2.020 0.061 0.300 0.020 0.246 0.104 0.025 0.742 1.5 4 Average * Crack width dimension of these samples were not measured at 28ays.196 1.528 0.351 0.064 0.061 0.094 0.104 0.406 0.045 0.993 0.084 0. 34 .316 0.046 0.4.5 1 GRC1.069 0.020 0.528 0.043 0.5 2 GRC1.114 0.461 1.041 0.051 0.179 0.688 0.028 0.041 0.043 0.5 4 Average GRC3 1* GRC3 2* GRC3 3 GRC3 4 Average GRC4.168 1.073 0.445 0.323 0.034 Crack Width at 21 Days Average SD 1.025 0.051 0.488 1.046 0.480 0.557 0.534 1.551 0.020 0.046 0.015 0.046 0.1.280 1.5 3 GRC1.033 0.229 0.189 0.033 0.046 0.017 1.218 0.023 0.048 0.048 0.704 0.343 0.427 1.089 0.028 Crack Width at 28 Days Average SD 1.
4.5 samples is higher than that of Control and other GRC mixtures.4.2 Control GRC1.2.2 and Table 2. 84]. standard deviation of the crack widths for GRC1.5 1.5 kg/m3 is not enough to ensure of crack controlling.4.4. The resistance of GRC1.5 samples against crack extension can be similar to both plain concrete (plain sample) and other GRC samples with higher fiber dosage.5 Crack Width. 2.4.2. These models try to 35 . days Figure 2. Mean and SD of shrinkage crack widths at different times of drying According to Figure 2. mm 1 0.4.5. 83. Theoretical Modeling of Restrained Shrinkage Various theoretical approaches for modeling the restrained shrinkage cracking of concrete have been developed [40.5 0 0 5 10 15 20 25 30 Time.5 GRC3 GRC4. addition of 3 kg/m3 glass fibers or more to the plain mixture results in much lower deviation in the results and reduction of crack width dimension by 70%. However. It is concluded that glass fiber dosage of 1.
which provides better understanding on how drying environment generates tensile stress and cracks in a restrained concrete specimen.1. The theoretical model is used to calibrate and interpret the experimental test results.1) 36 . Meanwhile. Finally a summary of results concluding significant effect of ARglass fibers on minimizing crack width dimension supported by both experimental and analytical data are presented. A research attempt to develop an analytical approach that relates key influential parameters of modeling including diffusion. The simulation of crack width opening which is of great importance from the aspects of durability and serviceability is also lacking. free shrinkage is taken place and the humidity profile h(z) through the thickness of the concrete section is simplified to follow the Fick's law of diffusion: h( z ) = hs − (hs − hi )erf ( z ) (2.1.5. aging material properties and restraining effect is presented in this section [74]. shrinkage. This model is capable of showing effects of adding fiber on the restrained shrinkage and cracking of concrete. there is no model to properly predict postcrack behavior of plain concrete and GRC in the restrained shrinkage test based on the experimentally obtained tensile stressstrain response of the material.address the interaction between materials properties and the shrinkage characteristics of concrete.5. Moisture Diffusion and Free Shrinkage As concrete loses its moisture to the environment. creep. 2.
5. ε sh ( z ) = ε sh (t ) (1 − h( z )3 ) (2.1. while an average value of 1.3) where f and ξ are constants and t is time in days.5. Free drying shrinkage can be obtained by experiments or from the empirical relationships proposed by BazantPanula [86] for drying shrinkage at infinite time (εsh∞) and timedependent shrinkage formulation suggested by ACI 209 [87] as. hs and hi represent the humidity (fraction) at the outside surface and interior section and erf(z) represents the error function [85].where z is the distance measured from the outside surface inward the specimen.1.1. steel depth (z) humidity concrete hi h( z ) evaporation hs Figure 2. tξ f + tξ ε sh(t ) = ε sh∞ (2. ACI 209R92 recommends an average value for f of 35 for 7 days of moist curing. Fick's law of diffusion is used to simulate humidity profile through the thickness of the concrete section using an error function A cubic function is used to relate the free shrinkage strain as a function of the humidity profile throughout the thickness.1.0 37 .2) where εsh(t) is the free drying shrinkage strain at any time.5.
5 0 0 7 14 21 28 Time .5.5. As shown in the figure.2. Fitted free shrinkage results by the modified ACI 209R92 model 2. respectively.5. the best fitted values for drying shrinkage at infinite time (εsh∞) for Control. Due to lacking of early age tensile creep model. 2000 Free Shrinkage Strain .2. respectively.days Figure 2. it produces creep strain with time and offset the strain due to free shrinkage. it is 38 .micro strain 1500 1000 Control GRC1.5 500 GRC3 GRC4.5. and the proper values of f and ξ are selected as 28 days and 1. Creep of Concrete As concrete is subjected to tensile stress.5.1. GRC1. Meanwhile.is suggested for ξ. the experimental data of free shrinkage is fitted by the ACI equation as shown in Figure 2. resulting in relaxation of the elastic stresses in the material. 1944 and 1836 micro strains.5 samples are 2160.1. GRC3 and GRC 4. free drying shrinkage is obtained by experiments performed according to ASTM standard C 15704. 2052.2. In this study.
1. Note that the presented creep model needs further studies to be verified by experimental data. monotonic threepoint bending tests are performed on the Control and GRC3 samples at 1. 3 and 4.1) where νu is the ultimate creep coefficient. For the case of GRC1.6 for d and ψ. 3. 7 and 28 days.5.5. tψ d + tψ v(t ) = vu (2.5 and values of ψ are chosen as 1.3.5. ACI209R92 recommends an average value of 10 and 0. 3 and 1.5. 2. respectively. In this study νu is assumed as 12 for all samples. However. since the recommendation is for the plain concrete under compression tests.5.3 and 0. the values of d and ψ are modified to 6 and 1 respectively for the tensile creep behavior of Control mixture.5. Figure 2. respectively.3.2. Age Dependent Concrete Strength To obtain tensile strength properties of the mixtures along the time. 39 .1. d and ψ are constants and t is time in days.assumed that the creep coefficient of concrete in compression ν(t) used in ACI 209 report [87] is applicable to the present tensile mode of loading.1 shows the flexural responses of these two mixtures at 1 and 7 days. values of d are chosen as 4.
75 1 1.75 1 1.5 GRC3 GRC4. N Flexural Load.5 4 0 0 7 14 21 28 Time .2.1. mm Figure 2. Flexural response of concrete samples under threepoint bending tests 40 .1.5.12 Free Shrinkage Strain .25 0. Creep model for Control and GRC mixes 4000 Age: 1 day 5000 Flexural Load. mm CMOD (Crack Opening).25 CMOD (Crack Opening).5 0.25 0.5 0.3. N 3000 Control GRC3 4000 Age: 7 days Control GRC3 3000 2000 2000 1000 1000 0 0 0.micro strain 8 Control GRC1.5.25 0 0 0.days Figure 2.
a + bt t G f (t ) = G f 28 ( ). which is properly applicable to both plain and GRC samples is hired.3. Similar to ACI timedependent compressive strength development function. Figure 2.1) 41 . 7 and 28 days are summarized in the following Table 2. back calculated tensile strength. a + bt ft (t ) = ft 28 ( st 3 (t ) = st 3− 28 ( t ) a + bt t Ec (t ) = Ec 28 ( ) a + bt (2.1. residual strength. 3.2. a model shown in below figure.5. st3. fracture energy. Ec are assumed to follow these functions: t ).3.5. ft.3.To calculate tensile stressstrain parameters of the samples from threepoint bending tests. Gf and young’s modulus.5. Tensile stressstrain and crack width model The parameters of stressstrain backcalculations model of Control and GRC3 samples at 1.
02286 0.048 0.02286 0. ε t2 (t) = a ⋅ ebt + c wt2 (t) = a ⋅ ebt + c (2. 42 .0254 wt2 (mm) 0.1.019 15.3.00011 0. The best values for the coefficients a.00011 0.00012 0.00017 where.00011 0.019 15. Gf28 and Ec28 are tensile strength.00012 0.758 2.043 0. a and b are material parameters.289 Age Mix ID (days) 1 Control 3 7 28 1 GRC3 3 7 28 εt1 εt2 wt1 (mm) 0.315 0.3.770 8.3.264 13.889 0.770 0.1) and equation (2. wt1(t) = a ⋅ ebt + c . a.529 0.310 0. are the transition and ultimate crack widths.3. residual strength.141 15.00020 0.269 0. and c after fitting equation (2.0254 0.00012 0.1143 0.00017 0.5. b and c are constant material parameters.00011 0.Table 2. b.0668 0.3.00019 0. ft28.5.5.141 15.00019 0.03048 0.1 are being used in this modeling approach.246 0.538 0.034 2.5.046 0.04064 0. are assumed to follow these functions: ε t1(t) = a ⋅ ebt + c .00017 0.275 0.0254 0.103 1. respectively.00017 0. and wt1 and wt2.2) In the above mentioned functions.10922 1.5.1524 0.475 0.034 2. fracture energy and Young’s modulus of concrete at 28 days.1524 0. st328.255 0. t is the time in days.103 St3 (MPa) 0. εt1 and εt .041 0.00026 0.270 0.8128 0.2) with the values of Table 2.448 13. are the tensile strains at limit of 2 proportionality and peak stress respectively.068 2.547 Ec (GPa) 11.02794 0.762 ft (MPa) 1.068 2.00017 0. mm) 0. t is time in days. Backcalculated parameters of Control and GRC3 samples at different ages Gf (MPa.758 2.
5. Symbol Δ is used for the incremental change of quantities between previous time step tj1 and the current time step tj and index i is used for the layer number at particular location zi. MPa 1 0 2 1 0 7. 28 GRC3 . 28 da ys Control 0 400 800 1200 Tensile Strain.3 : ge 14.1 s ay Tensile Stress. ΔFc + ΔFst = 0 (2.1 .3.5.1) 43 . 7 A .3 2 d 3.4. e: Ag . Restraining Effect In this model. Backcalculated tensile stress strain model at different ages 2.3. The restraining effect provided by steel ring that prohibits concrete to shrink freely can be determined by equilibrium of force between the tension force in concrete ΔFc (positive) and compression force in steel (negative) ΔFst.5. microstrain Figure 2. Other related terms such as forces and stresses follow the same sign as strains.4. tensile strain is positive while compressive and shrinkage strains are negative. 14.
Es is the young modulus of steel. Ac and As are area of concrete and steel.4.2) * where.1) can be written as * Δε el Ec (t j −1 ) Ac + Δε st Es As = 0 (2. which has elastic compressive strain Δεst.1(b).5. Ec (t j −1 ) is the secant modulus of the concrete at previous time step averaged from all discretized concrete layers. Stress Strain Development Figure 2. An arch segment of the ring as shown in a dash box can be approximated to 1D problem depicted in Figure 2.5.4.5.3) 2. only the elastic tensile strain component of the concrete Δεel produces stress and it is balanced with the compressive stress in steel.5. The figure represents a schematic drawing of strain components developed in the concrete ring specimen. The free shrinkage strain Δεsh (negative) will be restrained by steel ring creating compressive strain in the steel Δεst (negative) and the steel compressive force must be balanced with the tension force in concrete. Rearrange the terms in equation (2.5. respectively. the incremental compressive steel strain Δεst is obtained: Δε st = − * Ec (t j −1 ) Ac Es As Δε el (2.In the equilibrium.5.2). which immediately generates elastic strain in concrete 44 . Using modulus of materials. This is only an approximation since the modulus at the current time step is not yet known.5. the thickness is assumed to be flat.4.4.5.5.5.1(a) shows a concrete ring specimen subjected to moisture loss until it cracks. the equilibrium of force in (2.
In addition. For time interval Δt. the previous stress history before time tj1 also ( prev contributes to the creep strain increment between time tj1 and tj. Schematic drawing for the proposed drying shrinkage model. (a) ring specimen.5. which can be calculated by using Bolzmann’s superposition principle: ( prev Δε cp σ ) ⎧ j −1 1 σ elk − σ elk −1 ⎫ ⎪ ⎪ v(t j − tk −1 ) ⎬ − ε cp (t j −1 ) = ⎨∑ Ec 28 ⎪k =2 2 ⎪ ⎩ ⎭ (2. Δε cp σ ) . the concrete tensile stress will generate creep strain Δεcp (positive). (b) strain components 45 .4) Figure 2.5.4.Δεel (positive).1.5.
4.5).4.5.6) results in: −Δε sh = * Ec (t j −1 ) Ac + Es As Es As Δε el + Δσ el v(dt ) ( prev + Δε cp σ ) Ec 28 (2.8) into equation (2.5.7) To obtain stress and strain distribution.5.4.5.4. the equilibrium of incremental strains developed between time step tj1 and tj can be written as: ( prev −Δε sh = Δε el + Δε cp + Δε cp σ ) − Δε st (2. t j −1 ) * = Δε el ( zi ) Ec ( zi .4.5.6) Use the relationship between tensile stress and creep coefficient Δεcp = ΔσelC(dt)/Ec28 to determine creep strain. t j −1 )]Ec ( zi .5.4.4.5) Substitute the restraining steel strain Δεst from equation (2.8) Substitute incremental stress Δσel(zi) defined in equation (2.4.3) into equation (2. yields −Δε sh = * Ec (t j −1 ) Ac Es As ( prev Δε el + Δε el + Δε cp + Δε cp σ ) (2.8) and rearrange the terms. t j −1 ) (2. the thickness of concrete is discretized into Nc sub layers while steel has only one layer.5. and substitute in equation (2. The incremental elastic stress in layer i located at zi can be expressed as: * Δσ el ( zi ) = [ε el ( zi .5. the incremental concrete elastic strain at each sub layer can be expressed as: 46 .4.Therefore.5. t j ) − ε el ( zi .
t j ) (2.1) 2) Calculate incremental elastic tensile strain at each layer due to free shrinkage. t j )] from Figure 2. t j −1 ) (2.9) It should be noted that the concrete is considered as a whole thickness with an * average secant modulus Ec (t j −1 ) in accounting for the restraining effect from steel while concrete is treated individually with its secant modulus at each layer * Ec ( zi . t j −1 ) + Δε el ( zi ) (2. 2.Δε el ( zi ) = ( prev −Δε sh ( zi ) − Δε cp σ ) ( zi ) Q+ * Ec ( zi .5. t j −1 ) . and its increment by: Δε sh ( zi ) = ε sh ( zi .3) * Ec ( zi .1. 3) Update total elastic strain.5.9). t j −1 ) for the effect of free shrinkage and creep. restraining effect and creep from equation (2. stress and secant modulus using concrete model described in Figure 2. ε t ( zi .3.6.1.2 (2. Q= * Ec (t j −1 ) Ac + Es As Ec 28 v(dt ) Es As (2.3). t j ) ε t ( zi . Algorithm for Strain History in Steel Ring and Crack Width at Concrete Surface 1) Calculate free shrinkage strain distribution at each layer i from equation (2. t j ) = function [ε t ( zi .2.2) σ t ( zi .3. t j ) = ε t ( zi . t j ) − ε sh ( zi .6.4.5.5.6.4) 47 .5.2) and equation (2.5. t j ) = σ t ( zi .6.5.5.5.4.5.5.6.
tj) exceeds the strain at peak stress εt2(tj).4) If the updated strain εt(zi. 5) Sum concrete force at each concrete layer Aci to obtain the total tensile force. t j ) = ⎡ε t ( zi .6.5. t j ) − ε t _ unload ( zi . 2.tj) is the unloading strain in the prepeak stress strain curve corresponding to the same stress level as the strain in post peak response εt(zi.5. The results show a good correlation between the experimental data including strain in steel ring and crack width opening histories and simulation data. While the best correlation of the simulation with experimental data is for Control sample.6.tj) and D(zi) is the diameter of the concrete ring at location zi.1.5) where εt_unload(zi. Fc (t j ) = ∑ Aciσ t ( zi .6) 6) Calculate nominal stress and strain in steel. Comparing Experimental Data and Simulations Results of simulation by this model and comparison with the experimental data are shown in Figure 2.7) 7) A complete strain history at steel ring and crack width at concrete surface can be obtained by repeating steps 16 until the time tj reaches the specified age.6.6. defined by: σ ns (t j ) = Fs (t j ) As . t j ) and Fs (t j ) = − Fc (t j ) i =1 Nc (2. the crack width at each layer of concrete is calculated by: wt ( zi . which the magnitude equals to compressive the force in steel. ε ns (t j ) = σ ns (t j ) Es (2.5. t j ) ⎤ π D( zi ) ⎣ ⎦ (2.6. the model overestimates the crack widths 48 .
6. mm 1 GRC4. days Figure 2.of GRC samples after 14 days.5 4 6 8 10 12 14 Drying Time. Strain in Steel Ring. Comparison of experimental results and model simulation.5 0 0 5 10 15 20 25 30 Time. the model is capable of capturing significant effects of adding glassfibers to control shrinkage cracking in concrete. days 2 Control 1. microstrains 0 20 40 60 Control 80 100 0 2 ARG2. and (b) crack width dimension history at the outer concrete surface 49 . (a) history of strain in steel ring.5 GRC1.5 ARG5.5 GRC3 Crack Width. However.0 ARG7.1.5 0.
This model is capable of capturing significant effects of fiber addition in order to control cracking of shrinkage. This effect in addition to delaying crack occurrence by 1 or 2 days represent the low dosage glassfiber concrete as an appropriate construction material to be used in dry and hot regions with the risk of high drying shrinkage. 50 .7. Collectively speaking.2. The analytical model for predicting steel strain history and concrete crack width dimension for restrained shrinkage test shows a relatively good correlation with the experimental data. Conclusion Reducing crack width dimension by 3 times by using low content of ARglassfibers is the most significant effect of adding fibers to the mixture in controlling shrinkage cracking. ringtype restrained shrinkage test method is a good method to capture the behavior of concrete materials against restrained drying shrinkage.
47]. and accelerate deterioration. Although attributed to several driving forces such as differential settlement. 45]. results of test methods in which the drying of fresh concrete have been studied lack accurate measurement of the evaporation rate and data interpretation based on transport characteristics. Plastic shrinkage in presence of restraints leads to tensile stresses which may easily exceed the low tensile strength of fresh concrete and result in cracking.1. and methanol treatment respectively. Introduction Plastic shrinkage cracks in concrete reduce load carrying capacity. or heat lamps [48] to simulate severe evaporation conditions causing plastic shrinkage cracks in fresh concrete. 5]. Also. these tests were not designed to focus on evaporation characteristics of the drying material as the principal driving force of plastic shrinkage. thermal dilation. 44. Wind tunnel [49] and vacuum drying [50] have also been used to expedite the drying process. However.3. and autogenous deformation [8]. fans and heaters [46. and fly ash on normalized cumulative moisture loss using fan. parameters 51 . [51] studied the effect of fiber content. Experimental Observations of EarlyAge Drying 3. These cracks are the main routes through which aggressive agents such as chloride ions penetrate into the concrete mass and threaten the longterm durability of structures [7]. Several test methods have been implemented using fans [43. w/c ratio. 10. resulting in increased maintenance costs and reduced service life [4. plastic shrinkage cracking in concrete occurs principally due to a high rate of water evaporation from the concrete surface [9. 11]. While Wongtanakitcharoen and Naaman [20] and Wang et al. heater.
95]. Hall and Hoff [54] analyzed the drying of clay brick ceramics using two distinct stages: a constant drying rate period and a falling drying rate period. 90 .addressing rates of evaporation were not utilized in their work. 52 . 92 . Garbalińska used desorptive measurement techniques to determine moisture diffusivity [57]. the scale of data requires the evaporation rates to be interpreted based on mass transport parameters. 89 . moisture flows through the porous material toward the evaporative surface to supply the necessary surface flux. and silica fume on the normalized cumulative moisture loss and maximum rate of evaporation. Analysis of the evaporation rates of cementbased materials with direct attention to stages of drying process has not been thoroughly investigated. However. They then extended the results from porous media to stones and concrete. By analyzing moisture loss data during the drying of well hydrated cement paste samples. admixtures. maximum rate of evaporation is not sufficient to characterize different materials characteristics. A Similar analytical method is proposed to measure moisture diffusivity in drying fresh cement pastes. During the drying phase. [52] used an electric fan to investigate the effect of w/c ratio. However. Berhane [53] evaluated the effect of w/c ratio and ambient relative humidity using evaporation rates obtained from moisture loss curves. The relationship between unsaturated flow within concrete pores and evaporation rate on the surface can be used to measure moisture diffusivity [94. 91 . 93 ]. Samman et al. Evaporation is viewed as the diffusion of water molecules through the boundary layer at the surface to maintain the equilibrium of liquidgas phase transition [ 88 .
53 . The test method was applied to a variety of samples in order to evaluate the impact of sample thickness. This test method allows measuring material properties that characterize mass transfer during the early stages of fresh paste drying. w/c ratio. 47. Characterization of images taken from a drying surface helps in understanding crack initiation. duration of initial curing and fiber content on the evaporation characteristics being tested. and stabilization during the drying process. a majority of recent publications have focused on onedimensional cracking of the concrete by utilizing special restraints [8. 97 ].Plastic shrinkage cracks due to a high rate of evaporation are twodimensional [96]. surface area. The development of crack patterns during drying is documented using timelapse photography. 9. A drying technique based on a lowpressure condition was employed to cause a rapid rate of evaporation while weight loss was measured. This drying method imposes a onedimensional moisture flow through the thickness and facilitates parameter estimation from the mass transfer data. 10. An analytical method to determine moisture diffusivity using experimental data is presented and the effects of different testing parameters on diffusion coefficients are investigated. However. The images have been used to determine the ability of the fibers in controlling shrinkage cracking in drying cement pastes. propagation. A test method to monitor the drying of fresh cement paste and the formation of plastic twodimensional cracks is presented in this work.
Similar preliminary results from comparison of the weight of the condensed water in the flask with the total moisture loss at the end of the test show the efficiency of the condensing system. To avoid excessive moisture removal at the beginning of the 54 . Testing methodology Apparatus of the developed lowpressure drying test method is shown in Figure 3. and the entire assembly is placed inside a glass desiccator. providing shrinkage restraint in two directions. Using a vacuum pump and a pressure regulator the air pressure inside the desiccator is lowered to absolute 1700 Pa (0. A prismatic sample is filled with fresh paste such that its face is exposed and all other sides are sealed.2. The weight loss measured was recorded using a computer interface unit. This level of air pressure was selected as a system level parameter after extensive pressure calibration procedures.2. The mold consists of interlocking pieces made of polycarbonate as shown in Figure 3.1a. A condensing system including a DDrying apparatus is used similar to Copeland and Hayes [98] to remove the water vapor from the desiccator. and if it is too high. If the pressure is too low. A controlled test condition therefore requires evaporation rates which simulate the range of severe atmospheric conditions.2.1b and uses anchor hooks to connect the fresh paste with the mold.3. microstructural damage due to cavitation occurs. The sample is placed on a load cell which serves as a digital scale. The specimen surface was photographed at 15 min intervals using a digital camera mounted 10 cm above the sample. evaporation occurs at normal rates. The weight of the sample is continuously monitored throughout the drying cycle.5 inHg) and maintained at this pressure throughout the test.
This indicated the free water evaporation rate of the test setup.1a.25 cm Diameter) sample dry ice alcohol load cell vacuum pump condenser computer unit interface strain gage transducer amplifier camera PC data acquisition system Figure 3.test. Based on this procedure. The test procedure was subjected to additional calibration using liquid water as the evaporating material. followed by a decrease in pressure to 1700 Pa (0. a test procedure consisting of 30 minutes at 34 kPa (10 inHg).2. pressure gage pressure regulator vacuum Tbore stopcock vacuum vessel (25 cm height .5 inHg) for the remainder of the test was selected. Schematic of vacuum drying test setup 55 . a calibration procedure was developed using a cement paste exposed to several vacuum pressures such that the rate of evaporation was maintained at a constant rate.
Figure 3.1.3.1.3. and adjusted by means of a baseline calibration curve.1) 56 . Results were expressed with respect to the exposed surface area in accordance to equation (3.1).2. J = ΔM 1 dM = A Δt A dt (3. Cumulative moisture loss and evaporation rate versus time The cumulative moisture loss curve was obtained for periods of up to 24 hours or more. Cumulative moisture loss data were subjected to a numerical differentiation procedure in order to measure the evaporation rate.3. Analysis of typical evaporation data under low pressure test condition 3.3.1b.1. Plan view of the mold (numbers in mm) 3.
3.05 kg/(m2hr) afterwards. and Kowalski [101] for wet capillaryporous materials such as paper. ceramics and clay. Similar to their observations.42 kg/(m2hr) during first 10 hours. leather.1.3.where J is the evaporation rate (kg/m2s). Cumulative moisture loss and evaporation rate curves for a typical cement paste sample are shown in Figure 3. and a falling drying rate period (stage II). 57 . Figure 3. the drying rate of cement paste at the beginning is roughly constant with an evaporation rate of 0. ΔM is the mass change at specified time steps (kg).1a.1. [99]. and gradually reduces to 0. Under constant drying condition. Cooling [100] for clay brick ceramics. drying of cement pastes under constant external conditions occurs in two distinct stages of a constant drying rate period (stage I). Δt is the time step (s) and A is the surface area of the original sample (m2). This change in the evaporation rate is also presented in log format.1b shows the evaporation rate versus log of drying time which is in accordance with results obtained by Hall et al.
1.6 Evaporation Rate.h) .2 0. (a) in linear scale.1.h) 0.3 0.3. g 80 60 40 20 0 0 10 Raw Data Smoothened Data 20 30 Drying Time.3.1a 0.3.0.5 0.4 0.4 0.2 0 Cumulative Moisture Loss. kg/(m2.1b Figure 3. Typical cumulative moisture loss and evaporation rate of a cement paste sample versus time. (b) in log scale 58 Evaporation Rate. h 40 50 Figure 3.1. kg/(m2.1 0 0.6 (A) (B) (A) Stage I Drying (constant drying rate period) (B) Stage II Drying (falling drying rate period) 0. h 100 Figure 3.1.1 transition time from stage I to stage II: 10 h Stage I Drying Stage II Drying 1 10 Log of Drying Time.
Different stages of drying can be described based on the degree of continuity between liquid and vapor phases. 99]. This was verified by the evaporation tests on water samples. Results of water surface evaporation compared with a plain cement paste sample are shown in Figure 3. (c) hygroscopic state (continuous vapor phase) 59 . Hall and Hoff [54] referred to several experimental data showing that stage I behavior is independent of capillary processes inside the material.1.2.3. Schematics of state phases and moisture transport during drying process of porous media according to Scherer [102] and Plumb [103] are presented in Figures 3. (a) (b) (c) Figure 3.3. (b) capillary saturation (continuous liquid phase). liquid (L) and vapor phases (V): (a) initial condition (full saturation). the main phase transitions occur at the boundary surface and vapor phase diffusion into the air is determining rate of evaporation. Drying rate at early time is constant and about the same as rate of evaporation from water surface exposed to the same conditions [8.3.2ac.3.1.1. During stage I drying. Schematics of moisture transport during drying process of porous media representing solid (S).
3.3a 0.4 0.3b Figure 3. h 40 50 Figure 3. g 60 40 water cement paste (w/c=0.2 0 0. h 100 Figure 3. kg/(m2.h) 0. Evaporation test results from water surface comparing to cement paste 60 .1.1.3.3.1.1 1 10 Drying Time.6 0.8 water cement paste (w/c=0.80 Cumulative Moisture Loss.45) Drying Rate.45) 20 0 0 10 20 30 Drying Time.3.
It is therefore postulated that within stage I drying.45 kg/(m2hr). and fully developed up to 4 h. 24h Figure 3.Both samples show evaporation rates in the range of 0. 4h. drying takes place under external mass transfer control. 3h. potential cracking could have no significant effect in increasing evaporation rate which was the same before and after cracking. a continuous liquid phase is present in the pore structure and external evaporative flux away from the surface causes a gradient in capillary pressure in the pores which is the main mechanism of liquid migration from inside toward the boundary [104]. Drying time: 15 min.1.3.4. a time period well within the stage I drying. Since in this stage. Note that visible plastic cracks were seen as early as 3 h as shown in Figure 3. 3h 15min Drying time: 3h 30min. 2D Crack development during drying of a fresh plain cement paste 61 .3.1.3.
2.1. this transition in drying stage is gradual.1. The moisture content at the surface reaches a critical value at and below which the material is unable to support the necessary flux to the surface to satisfy the potential evaporation [105]. Thus. An empirical definition of transition time is defined by the intersection point of two asymptotic linear approximations to the evaporation rate data at stage I and stage II.3. the removal of moisture is mainly controlled by diffusion which dominates the capillary forces [106]. Referring to the results presented in Figure 3.3. Calculation of Moisture Diffusivity Evaporation from water surface can be treated as a boundary layer problem.2c shows that in stage II drying. the moisture content decreases slowly until it reaches the equilibrium value below which the material cannot be dried. In stage II drying. phase transitions take place within capillary pores and mass transfer begins to occur through the pore vapor phase.3. At this time. the liquid phase becomes discontinuous while vapor phase becomes continuous within the matrix. the moisture content drops continuously. stage II drying begins and rate of evaporation starts to fall.1b. Figure 3.1b. During this period. unsaturated capillary flow determines the rate of evaporation and drying happens under internal mass transfer control [105].7 h. this transition time for a plain cement paste sample is determined as 9. 3.1.3.As the liquid phase water moves to the surface and evaporates with a constant rate during stage I drying. Therefore. in which the water vapor diffuses from a saturated state on the surface to 62 . As shown in Figure 3.
2. It consists of a boundary layer identical to water evaporation boundary. ψ is the hydraulic or capillary potential (m). and an unsaturated flow for transfer of liquid water from within the material to supply the flux.the ambient moisture concentration [8893]. ∂C ∂ 2C =D 2 ∂t ∂x (3. The focus of this work is on calculation of diffusivity of water within the cement paste rather than airwater vapor diffusivity at the boundary layer. Rate of evaporation is controlled by two parameters of binary airwater vapor diffusivity and the concentration gradient. Definition of capillary diffusivity (m2/s) as D = Κ (∂ψ ∂θ ) leads us to Fick’s second law of diffusion which governs a majority of moisture transfer problems in porous media [50.3. ceramics. 104] which in the case of a onedimensional expression is written as: ∂θ ∂ ⎛ ∂ψ ⎞ = ⎜Κ ⎟ ∂t ∂x ⎝ ∂x ⎠ (3. Flow within the porous medium can be described by the extended Darcy equation for unsaturated flow [99. 58. Evaporation from a fresh cement paste at stage I is quite similar to the case of clay.2) 63 .2. 108]. and other capillary porous media [107]. and t is the time (s). The calculated diffusivities at stage I and II show the potential for moisture conductivity through the fresh cement paste.3. Κ is moisture conductivity (m/s). x is the pass of moisture transfer (m).1) where θ is the volume fraction moisture content ().
Boundary condition for the second stage which is a constant concentration at the surface is set to ambient moisture concentration. moisture concentration can be assumed to be zero as a reasonable approximation. In order to simulate the drying test results using a dualstage drying concept. water vapor pressure can be estimated as 2% of barometric pressure which is 34 Pa. Using the experimental evaporation rate during stage I drying.where C is the moisture concentration (kg/m3). F0 is the constant flux in stage I drying (kg/(m2s)). At ambient pressure of 1 atm. the water vapor pressure is the product of relative humidity and the saturated water vapor pressure. and zero moisture concentration at stage II. measured from the bottom face (m). boundary condition of constant flux at the surface can be applied which is obtained directly from experimental data. Therefore. and C* is the ambient moisture concentration (kg/m3). Assuming the percentage of water vapor remains the same after running the vacuum pump. the two required boundary conditions are: Stage I drying: Stage II drying: at x = L.2.0127 kg of Water/ Kg of Air. The corresponding moisture content defined as ω = Mwater /MDry Air = 0.3. For the first stage. However. at x = L. low pressure (3.622 x Pwater /(1700Pwater) is equal to 0. in this low pressure test condition. 64 . L is the thickness of sample (m). x is the pass length of diffusion. two different boundary conditions at the top surface are required. J = F0 Cs = C∗ = Cw. the sum of the partial pressure of all the species in the air including dry air and water vapor is 1700 Pa. D is the moisture (liquid and vapor) diffusion coefficient (m2/s) and t is the time (s).3) where.
C = Ci 0≤ x≤ L (3.3.2).2.4) Since the bottom surface is impermeable.2. DI is the diffusivity in stage I drying.2.6) where.6).6) [109].3. C (t . x ) = Ci + F0 L ⎧ D I t 3 x 2 − L2 2 − 2 ⎨ 2 + DI ⎩ L 6 L2 π n =1 ∑ ∞ ( − 1) n n2 exp( − D I n 2π 2 t L2 ) cos nπ x ⎫ ⎪ ⎬ L ⎪ ⎭ (3. at the surface is [109]: 65 . assuming a constant diffusivity is given by equation (3.3. which is averaged over space and time. diffusion coefficients in Stage II drying depend on moisture concentration [ 110 ]. Since stage I drying ends when the surface reaches equilibrium moisture concentration.2. the boundary condition at the bottom is J = −D ∂C =0 ∂x at x = 0 (3. for the initial and boundary conditions of stage I drying. Analytical solution for the boundary conditions of stage II drying subjected to constant concentration. Using equation (3. For simplicity. C*. moisture diffusivity at stage I drying is calculated as long as the time of transition from stage I to stage II has not lapsed.2. Applying this method on a typical result of a plain cement paste sample results in calculating a diffusivity value of 5.3.3. the initial boundary condition is t = 0.5) The analytical solution of equation (3. moisture concentration at the top surface is obtained providing constant F0 and DI values.3. diffusivity in stage II is taken constant resulting in calculation of an average diffusivity.2.Considering Ci as the initial moisture concentration (kg/m3).15x107 m2/s at stage I drying. Unlike the stage I drying.
M ∞ (kg).2. moisture diffusivity. Therefore. (3. Mt (kg).3. can be used to determine moisture diffusivity during stage II drying.8) as ⎛ M ⎞ 8 ln ⎜1 − t ⎟ − ln 2 M∞ ⎠ π π = ⎝ DII ⋅ 2 −4 L t 2 (3.8) Results of analyses performed by Garbalińska [57] on cement based materials show that in the expansion of equation (3. DII is the diffusivity in stage II drying. By considering only first term of the series. DII (m2/s).3. aln .2. can be derived by rearranging equation (3.2.7) By integrating equation (3.2.3.2. 66 . which has left the sample at time t. stage II drying begins when more than 60% of the moisture is lost.C − Ci 4 ∞ (−1) n (2n + 1)π x exp{ − DII (2n + 1) 2 π 2t / 4l 2 }cos = 1− ∑ π n = 0 2n + 1 2L C∗ − Ci where.3.2.8).3. slope of the curve. can be obtained and related to its corresponding quantity after infinite time.7) over the thickness of sample.9) Assuming the logarithmic terms on the right hand side of the equation are following a linear relationship with the time. moisture loss during the advanced phase of drying is given by ∞ Mt 8 exp{ − DII (2n + 1)2 π 2t / 4 L2 } = 1− ∑ (2n + 1)2 π 2 M∞ n =0 (3. for almost all samples tested in this study. As discussed in the following sections. the higher terms with n>0 can be neglected after releasing 40% of the total mass change.3. total amount of diffusing moisture in stage II drying.
⎛ M ⎞ 8 ln ⎜1 − t ⎟ − ln 2 = aln t π ⎝ M∞ ⎠ DII = (3. The reason is that diffusion rates of water vapor in low pressure are much higher than that under normal room pressure.10) −4 L2 π2 ⋅ aln (3. aln for a plain cement paste sample is shown in Figure 3.1b. total amount of moisture loss at any time can be predicted. This figure shows that dualstage drying model and corresponding moisture diffusivity values can be used to predict drying characteristics of cementbased materials in this low pressure condition. the continuity of the slope is not completely achieved as shown by the slight bump in the simulation data. Comparison of the simulation and experimental data for a drying cement paste sample is shown in Figure 3.3.2. where rate of water vapor diffusion in the desiccator controls the rate of vaporization and drying. because the diffusion coefficient of gasses is roughly inversely proportional to the gas pressure at constant temperature [ 111 ].33x109 m2/s at stage II drying.3. Since two different boundary conditions are applied in stage I and stage II.3.3. Further studies on variable diffusion coefficients and applying different boundary conditions such a convectivediffusive boundary condition at stage II are required in order to better fit the results.2.1a.11) Procedures for obtaining the slope. Using calculated values of diffusivity from stage I and stage II drying. Note that the evaporation rate in the normal room pressure is different than that in the low pressure desiccator.2. 67 .2. This process results in calculation of diffusivity as 3.
0. surface area. w/c ratio.5 and 6 kg/m3 Alkali Resistant (AR) glass fibers (St.1 and includes a test matrix consisting of twelve sets of evaporation tests conducted on different portland cement pastes.45. w/c ratios of 0. 4. Gobain Vetrotext) to the plain cement paste. and two surface dimensions of 127 mm x 127 mm and 100 mm x 100 mm were used. 3. 68 .Therefore.5.4. duration of initial curing and fiber content. Scope of Test Program Effects of different variables including sample size.6 were chosen as material variables. Two sample thicknesses of 11 and 21 mm. the analysis is applicable to drying at normal room pressure providing evaporation rates and ambient moisture concentration of that test condition. equivalent to 0. 0. The scope of the test program is shown in Table 3.1.06%. and fiber content were studied. However.55 and 0.11%.5. 0. Effect of fibers was investigated by adding 1. duration of initial curing. 3.4.4. the calculated diffusion coefficients for stage I and II drying are only representatives for this low pressure drying condition. w/c ratio. Experimental program 3.1. Variables of the study included sample thickness.
1b Figure 3.15 × 107 m2/s Dstage II = 3. (a) Deriving aln by fitting a straight line to the curve of ln (1 − ΔM t / ΔM max ) vs.3.3. time in stage II.1a 80 Moisture Loss.0 Stage II Drying 2 ln (1Δmt/Δmmax) 1 aln 4 Evaporation Data of a Cement Paste Sample Experimental Data Straight Line Fitting 6 8 0 10 20 30 Drying Time.33 × 109 m2/s 20 0 0 10 20 30 40 Drying Time.2. g 60 Transition Point 40 Experimental Data Simulation Dstage I = 5.2.3. h 40 Figure 3.2. hr Figure 3.1. (b) Simulation of cumulative moisture loss vs. experimental data for a cement paste sample 69 .
01 11 0. The mixture proportions for cement pastes are provided in Table 3.45 0.2.5 P0.016 P0. depth.01 0.45 0. Placing and Curing procedures The test program consisted of variations in the physical parameters of testing such as sample area.4. specific gravity of 2.01 11 0.45 Initial curing (h) 0 0 0 0 0 0 3 24 0 0 0 0 Fiber content (kg/m3) 0 0 0 0 0 0 0 0 1.45 0. Mixing.45IC0F3 P0.700 MPa.01 11 0.45IC0F0SA0.01 11 0.016 11 0.45 0. Materials. and materials mixture variations. 3 and 24 hours were also tested to evaluate effect of curing duration.01 11 0.01 11 0.55 0. Scope of the test program Test series P0.01 21 0.23% volumetric fractions.50IC0F0 P0.45IC0F1.45IC24F0 P0. 3.1.45IC3F0 P0. Cement paste samples with initial curing of 0. The fibers are multifiber strand of 100 round filaments bonded together.68 and tensile strength of 1.Table 3.45 0.1. AlkaliResistant (AR) glass fibers of 24 mm in length and aspect ratio of 150 were used. The filament diameter is 14 microns and fiber has elastic modulus of 72 GPa.55IC0F0 P0.01 11 0.01 11 0.45 0.45 0.45 0.5 6 Surface Thickness area (mm) (m2) 11 0.4.45IC0F0Th21 P0.2.1 and include Type I/II portland cement as the primary binding agent.4. In fiber reinforced cement pastes.5 3 4.5 P0.45IC0F4.17% and 0.45IC0F0 (control) P0.01 11 0.45IC0F6 w/c 0.50 0.60 0. After mixing in 70 .60IC0F0 P0.
45IC24F0.50 0.45IC0F0. Parameter Estimation Parameters obtained from evaporation tests includes aspects of initial rate of evaporation.45 0.accordance with ASTM C 1116 [ 112 ]. Table 3.45 0.45IC3F0.5 3 4.5. cumulative moisture loss at 24 h and moisture diffusivities. P0.1 and are discussed in the following sections.5.016 samples 3.45IC0F4.45 0. Results are compiled in Table 3. evaporation rate at 24 h.55 0. P0.4.45IC0F6 Portland cement 1450 1400 1355 1310 1450 1450 1450 1450 Water 650 700 745 790 650 650 650 650 ARglass w/c fiber 0 0 0 0 1.1. No curing was applied except for samples of two test series which were covered with a plastic sheet for 3 and 24 h at the room temperature before running tests.5 P0. time of transition from stage I to stage II drying.5 6 0.60 0. samples were cast in the molds immediately and subjected to test within 15 minutes.50IC0F0 P0.45 0. 71 .45IC0F3 P0.5 P0. P0.45IC0F0Th21 and P0.2.60IC0F0 P0.45IC0F0SA0.45IC0F1.55IC0F0 P0.45ICxxF0* P0. Mix proportions of the test series Test series P0.45 * This mix proportion was used for P0.
0 11. and diffusivities of tested samples Test series P0.08E09 3.033 0.61E09 3.Table 3.34 0.031 0.64E09 2.45IC3F0 P0.27E07 1.043 0.8 11.27E07 5.46 0.3 7.1.h)) (h) 0.5.45IC0F0 P0.45IC24F0 P0. Effect of sample size Evaporation characteristics of samples with the two different thicknesses and surface areas were studied.25E09 2.079 0.36E10 3.39 0.2 11.017 0.0 10.8 9.44 0. Initial evaporation rates.33E07 5.15E07 5.94E07 2.45 0.41 0.1.55IC0F0 P0.44 0.45IC0P0.0 12.5 11.039 0.91E07 5.45IC0F0P0.15E07 5.50E09 4.38 9.45IC0F0P0.33E09 3.h)) (g) (kg/(m2.45IC0F3 P0.27E07 4.019 0.7 18.5.27E09 2.05E07 5.45IC0P0.024 0.300 0.9 12.03E09 4. transition time.42 0.45IC0F6 Transition Cumulative Initial time of Evaporation moisture evaporation drying rate at 24 h loss at 24 h rate stages (kg/(m2. cumulative moisture loss.11E07 Moisture diffusivity at stage II (m2/s) 3.60IC0F0 P0.25E09 3.19E07 5.40 0.78E08 5.087 65 127 108 68 74 78 49 26 64 65 65 66 Moisture diffusivity at stage I (m2/s) 5.3 3.16E09 1.22E09 2.50IC0F0 P0.42 0.028 0. Two sample thicknesses of 11 mm and 21 mm in 72 . evaporation rate at 24 h.3 0.37 0.036 0.
Therefore the transition time changes from 9.44 and 0.addition to two exposed surface dimensions of 100 mm x 100 mm and 127 mm x 127 mm were used.1.1.hr) to 0. Larger sample however has a lower evaporation rate by 12% during stage I. The difference may be attributed to the side and edge effects. A 66% increase in the total moisture loss was observed as the surface area was increased by 62%. Since control and thicker samples have approximately the same initial evaporation rates as shown in Table 3. then the initial evaporation is primarily a surface phenomenon. as in the vacuum evaporation the mass flux is only a function of total pressure gradient and the water vapor diffusion coefficient is not related to the size of evaporative surface [113. As expected. Thicker sample has a slightly higher initial evaporation rate (~4%) and if this is attributed to within sample variation. This indicates that the critical moisture concentration was reached after 73 . respectively.5. 114.hr) for thicker sample and sample with larger surface area. critical moisture concentration reached later for thicker sample due to higher initial moisture content.5.1. increasing sample thickness and surface area.42 kg/(m2. Cumulative moisture loss and evaporation rates of these test series are shown in Figure 3.7 h for control to 18. Initial evaporation rates changed slightly from 0.37 kg/(m2. Results indicate a 95% increase in the total moisture loss due to doubling of the thickness. increased the cumulative moisture loss at the end of the test.9 h for the thicker sample. 115].
7 mm 25.1.1 1 10 100 0.1.6 0. h 40 50 Figure 3.5.h) .6 0. h 40 50 Figure 3.1 1 10 100 0.4 Sample Thickness 12. kg/(m2.5.1.45IC0F0 Thickness: 10 mm 0.1b Figure 3.7 mm x 12. g 150 120 90 60 30 0 0 10 P0.2 0 Cumulative Moisture Loss.45IC0F0 Surface Dimension: 100 mm x 100 mm 20 30 Drying Time.5.1a Drying Time.4 mm 0.4 P0.Drying Time. h 0. h 0.h) Evaporation Rate.2 0 Cumulative Moisture Loss. kg/(m2.7 mm 0 10 20 30 Drying Time. Effects of surface area and thickness on results of evaporation tests on plain cement paste 74 Evaporation Rate. g 120 80 40 0 Surface Dimension 10 mm x 10 mm 12.1.
however higher w/c ratio resulted in a higher cumulative moisture loss. Similarly.44. initial drying rate as shown in Figure 3.5. hence increasing w/c ratio does 75 . 74 and 78 g respectively. the initial rate of evaporation is rather the same for all samples. 0.hr) for samples with w/c ratio of 0. The values of diffusivity at stage I and stage II drying determined for these three samples are shown in Table 3. 0. 0. the transition time for larger sample was obtained as 12.5.2. 68.losing 63%65% of the initial moisture content.6x109 and 3.6 led to 20% increase in total moisture loss.55 and 0.5.60.45 and 0. An increase in w/c from 0.45. Such similar values for moisture diffusivity of Portland cement paste mixture validate the range of accuracy of the analytical method.4.5.hr)) as the upper limit.25x109 m2/s. On the other hand.016) is determined as 3. while for thicker (P045IC0F0Th21) and larger (P045IC0F0SA0. 0. respectively.2.1. Values of moisture diffusivities at stage I drying is within range of 4.45 to 0. which range from 0 to 65. 3. and 0.45 kg/(m2.45. 0. These magnitudes are sufficiently close to the rate of evaporation from water surface (i.27x107 m2/s. 0. 0.8 h which refers to the time when 68% of initial moisture is lost.33x109 m2/s for the control sample (P045IC0F0). Moisture diffusivity at stage II is calculated as 3.e.42. As expected.46 kg/(m2.5. respectively. Average rates of evaporation at first drying stage were 0. Effect of w/c ratio Figure 3.2.6 after a 24 h drying period.91x107 to 5.1a shows the cumulative moisture losstime curves for water cement ratios of 0.1a was slightly affected by increasing w/c ratio.50.
55 0.2E007 4. h 20 25 Figure 3.1a 4.45 0.h) Cumulative Moisture Loss. kg/(m2.6E009 3.2.6 0. h 0.45 0.2E009 Stage I Drying 5.4E007 0.60 80 60 40 20 0 0 5 PxxxIC0F0 10 15 Drying Time.2.1.1 1 10 100 0.6 0.5.4E009 Stage II Drying 4E009 Diffusivity.5 w/c 0.2 0 Evaporation Rate.2.4 0.65 Figure 3.5.50 0.55 0.4 0. m2/s 3. g w/c 0.1b Figure 3.Drying Time. Effects of w/c ratio on evaporation results of plain cement pastes 76 .5.
45 to 0.45IC24F0.2. This parameter fell from 65 g for the noncured cement paste to 49 g and 26 g for cement pastes with initial curing of 3 and 24 h.55 and 0. respectively translating into 24% and 60% reduction in cumulative moisture loss of cement paste. After 24 h of drying.not significantly increase the initial evaporation rate. 3. The cumulative moisture loss after 24 h of drying was substantially reduced by increasing the duration of initial curing. respectively. to 3. 11 h.5.5.3.7 h for the series with w/c ratio of 0.1b and indicate that moisture diffusivities of different samples at stage I drying are very similar.45IC0F0.55 and 0.8 h.3. Effect of duration of initial curing Effect of duration of curing prior to initiation of drying was studied by test series P0.5.hr) for all samples. This transition can only be justified by the change of microstructure. Figure 3. evaporation rates dropped to 0. But the transition time changes from 9.6.5.5. 0.03 kg/(m2.45IC3F0 and P0. Diffusivity values determined for different w/c ratios are shown in Figure 3. These samples had the same mix proportions and differed only in duration of initial curing which ranged from 0.1 shows the results of the cumulative moisture loss and evaporation rates versus time.3 h for series with w/c ratios of 0.45 to 9. 21% and 25% when w/c ratio varied from 0. Rates of evaporation at the second drying stage indicate that samples with higher w/c ratios have consistently higher evaporation rates. However.6. P0. and to 24 h. and 12. pore size distributions and physical properties of the 77 . 0. at stage II drying moisture diffusivities increased by 5%.
1a 4E009 3E009 2E009 Diffusivity.1 0 Cumulative Moisture Loss. h 0.5.h) .5.4 Duration of Curing 0h 3h 24 h 0.5 0.45ICxxF0 25 Figure 3.2 0.1b Figure 3.3. g 80 60 40 20 0 0 5 10 15 Drying Time. h 25 Figure 3. Effect of curing duration on results of evaporation tests on plain cement pastes 78 Evaporation Rate.5.3 0. h 20 P0.1.1 1 10 100 0.3. kg/(m2. m2/s 1E009 0 Stage I Drying Stage II Drying 4E007 2E007 0 0 5 10 15 20 Duration of Initial Curing.3.Drying Time.
However. This is in agreement with the result of Naaman et al.5. Results of stage I drying indicate considerable reduction in initial evaporation rates from a control of 0.4.transition layer through which evaporation occurs [114].hr) for 3 hcured and 24 hcured samples.3 kg/(m2hr) for control sample to 0.7 and 18.34 kg/(m2. Fiber addition to the cement paste did not result in any significant reduction in cumulative moisture loss at the end of the test. evaporation rates dropped from 0. respectively. i.42 to 0. The reduction is clearly attributed to the reduction of moisture diffusivities at stage I by increasing curing duration. The effect of fiber reinforcement is compared with the plain cement paste in Figure 3.5. respectively. The transition time from stage I to stage II drying decreased from 9. after 12 h drying when all three test series were on stage II drying.39 and 0.5 times than diffusion coefficients of 3h. 3.2 kg/(m2.and 24hcured samples.5. As shown in Table 3.e.hr) and 0. respectively. Increasing curing duration led to decreasing stage II diffusion coefficient of 3 hcured and 24 hcured samples comparing to the control sample by as much as 62% and 93%.1. Effect of fiber content and cracking Effect of fiber addition and its volume fraction was studied by utilizing ARglass fiber in the mixture. the reduction of drying rates in second stage of drying is more significant.1.4.hr) for 3 and 24 h cured samples.3 h for the 3 hcured specimens and down to 3 h for the 24 hcured sample. 79 . moisture diffusivity of control sample is 2.7 h to 7.07 kg/(m2. [47] which shows fibers result in the reduction of the cumulative moisture loss by only 5% or less.
20%.4. This 80 . results support the minimal effect of cracks on drying rate in stage I.However. 0. Since earlyage cracks cause the evaporative surface to be more exposed to the low pressure condition in the desiccator. results show a decreasing trend for the diffusivities at stage II by increasing fiber dosage as shown in Figure 3. Since fiber controls plastic shrinkage cracking during early hours (see Figure 3.2.41.5 and 6 kg/m3 are 2%. 3. fiber addition reduced the maximum crack width and cracking area of the cement paste.1b. 3.5. 4.4.4.42.5.hr) associated with samples with 1.5. A more significant effect of fiber addition was observed in the manner of transition from the first stage of drying to the second.40 and 0.2). 0. 33% and 38% less than corresponding value for plain control sample.4. The transition time from stage I to stage II drying increased by the addition of fiber for all fiber reinforced paste samples to ranges of 10 to 11. 4.5. According to Figure 3.5 h compared to 9. The initial evaporation rate dropped from 0.5 and 6 kg/m3 ARglass fiber additions. diffusivity results at stage I drying do not vary significantly by the change of fiber dosage. Addition of fibers resulted in a gradual transition of the modes of drying which could be attributed to the effect of fibers in controlling earlyage cracks. As shown in the micrographs of Figure 3. ARglass fibers caused slight reduction of evaporation rates at the first drying stage. respectively.1b.5.7 h for the control sample. The diffusivities of fiber reinforced cement paste samples with fiber content of 1. respectively. However.42 kg/(m2hr) for plain sample to 0.38 kg/(m2. they may contribute to higher evaporation rates than uncracked surfaces.5.
4.1. m2/s 2.45IC0Fxx 10 15 Drying Time. g 80 60 40 20 0 0 5 P0.4 Fiber Content (kg/m3) 0 1.2 0. kg/(m2. h 20 25 Figure 3.5 0.1a 3. kg/m3 6 Figure 3. Results of drying tests on cement paste with different content of ARglass fibers 81 Evaporation Rate.3 0.5 6 0.1 1 10 100 0.Drying Time.5.4.8E009 Stage I Drying 5.2E007 4.5.1b Figure 3.5.h) . h 0.1 0 Cumulative Moisture Loss.6E009 Stage II Drying 3E009 Diffusivity.4.4E007 0 2 4 Fiber Content.4E009 1.5 3 4.
Crack pattern of cement paste specimens with and without ARglass fibers after 24 hours of drying under lowpressure test condition effect can be explained by the higher degree of surface cracking in the presence of fibers which reduces the mean path length of moisture transfer to reach the surface.45IC0F0 P0.4.P0.45IC0F3 P0. The result can thus be interpreted as the effect of cracking on the drying 82 .45IC0F6 Figure 3.5.2.
In addition.45 cm/cm2. 22% and 61% reduction in areal fraction of cracks were observed by adding 3 and 6kg/m3 fibers to cement paste. [ 119 ] show that moisture diffusivity of high performance concrete and cement pastes increased by one order of magnitude due to the cracking. the restraint of cracks by fibers.rate in stage II when unlike to stage I. the increased number of microcracks. drying happens under internal mass transfer control and the microstructure plays a significant role on the drying process.1 mm and a spacing of 70 mm. maximum crack widths of samples associated with 3 and 6kg/m3 fibers were 47% and 71% less than corresponding value for the control sample. Also. Also. [ 120 ] reported the increase of water permeability by one order of magnitude due to increase of crack density from 0. [117]. The length of microcracks however was increased by addition of fibers. the recent experimental data reported by Vejmelková et al. it can be concluded that fibers reduce moisture diffusivities of cracked samples at stage II by reducing crack 83 . respectively.2. Higher moisture diffusivities at stage II in plain pastes with wider cracks compared to fiber reinforced pastes with narrower but more plentiful cracks are in accordance with the studies performed by Bažant and Raftshol [116] and Aldea et al. This explanation can be supported by the fact that in stage II.25 to 0. and reduced crack widths are effective on lowering moisture diffusivity.5. While they found the moisture diffusivity and the water permeability are proportional to the crack width cubed. As shown in Figure 3. respectively. Torrijos et al.25 times by shrinkage cracks with a width of 0. results of another experimental study [118] indicated that diffusivity of concrete increased 2. Considering all these results.4.
During stage I drying.6. During stage II drying. following conclusions may be drawn: (1) Drying of cement pastes can be segmented into two distinct stages: constant drying rate period (stage I) and falling drying rate period (stage II). Conclusion A test method capable of characterizing evaporation parameters and simulating sequential formation of shrinkage cracks in twodimensional samples under lowpressure condition was developed. Parameters including thickness and surface area of samples. main phase transitions take place at the surface and vapor phase diffusion process is ratedetermining.width and crack density which results in the durability improvement of cementbased materials. 3. moisture content decreases slowly until reaching the equilibrium value below which the material cannot be dried. moisture content at the surface reaches a critical value. At the transition time. phase transitions take place within capillary pores and unsaturated capillary flow is ratedetermining. 84 . w/c ratio. The cumulative moisture loss and evaporation rates were calculated. evaporation rate is constant. (2) The cumulative moisture loss and evaporation rate curves can be used to differentiate evaporation characteristics of different cement pastes. Based on the results of experiments and analyses in this study. During this stage.
Also. effect of fibers in controlling the plastic crack width results in significant durability enhancement. This could be due to the restraint of cracks by fibers which is effective in stage II when internal mass transfer controls the drying. the presence of fibers reduces the diffusivity values. However. presence of fibers and cracking control do not affect the drying rate in stage I drying. Since there is a cubic relationship between permeability rate and the crack width. Results show that 22% and 61% reduction in areal fraction of cracks were observed by adding 3 and 6kg/m3 fibers to cement paste. Moisture diffusivity is higher in stage I by more than one order of magnitude than its values in stage II. In stage II drying.and duration of initial curing have significant effects on evaporation results. (3) A dualstage drying model based on two different boundary conditions at the top surface and a diffusional moisture transfer within the pores can be used to simulate experimental results. 85 . respectively. maximum crack widths of samples associated with 3 and 6kg/m3 fibers were 47% and 71% less than corresponding value for the control sample. The model can be used to determine moisture diffusivity of the samples during drying process. (4) Fibers contribute to the control of plastic shrinkage cracks. respectively.
(5) Among several different parameters investigated in this study. 86 . duration of curing has the most significant effect on the reduction of drying rate and moisture diffusivities.
With regard to longterm durability.1. West and Holmes [56]. Chen and Mahadevan [64] referred to Walton et al [65] in using diffusion as the dominant moisture transport mechanism in concrete.4. and the moisture movement in the form of 87 . however. Theory and Modeling of EarlyAge Drying 4. Šelih and Bremner [66] concluded from the experimental results that the diffusiondriven moisture transport approach is appropriate only for the late stage of drying when the material is predominantly unsaturated. Most of these works refer to Bazant and Najjar [60. the moisture movement during drying of concrete is often described as a diffusion process where the diffusivity depends highly on the moisture content. Shimomura and Maekawa [50] used mass conservation of the vapor and liquid water to derive a diffusiontype equation. Introduction Drying of cementitious materials is a preventable. In existing literature. Using an approach based on soil science. Garbalińska [57]. Therefore. Torrenti et al [55]. they modeled the phenomena as a onestage diffusion process. shrinkage cracking is the most anticipated consequence of drying due to low earlyage strength of cement based materials. understanding the physics of drying and modeling the process is of great importance. but often inevitable phenomenon during early age. Kodikara and Chakrabarti [62] expressed the moisture movement during drying as a two separate stage mechanism [63]. Huldén and Hansen [58]. 61] who considered the drying of concrete as a nonlinear diffusion problem. Kim and Lee [59] modeled the moisture movement by the Fick’s second law of diffusion without explanation of the physics of the problem.
vapor flux is dominant. After studying the effect of different parameters on the model. 56. succeeding by a diffusiontype equation. the concept was not implemented in their model. their work was only limited to experimental observations. Šelih and Bremner [66] expressed the drying process as two different types of mechanisms of movement which can be described by a Darcytype equation near saturation. and Garbalinska [57] applied a constant moisture concentration equal to ambient moisture concentration as the boundary condition in their models. In this chapter. referring to Young [67]. A dualstage methodology for modeling the drying is introduced based on surface moisture transfer and internal moisture transport characteristics. However. a convective boundary condition proportional to the difference of internal and ambient moisture concentration was most often employed [50. In another approach. Torrenti et al [55] and Shimomura and Maekawa [50] also used evaporative boundary conditions similar to convection. 59]. Bazant and Najjar [60]. Nonetheless. While ACI 305R99 [68] and Uno [69] recommended a constant flux equal to evaporation rate of water surface for earlyage hot weather concreting. the 88 . 58. The evaporation through the surface as a boundary condition is applied differently in available models. Although Shimomura and Maekawa [50] discussed the movement of liquid water as governing mechanism regarding moisture movement near the saturated state. physics of water evaporation is presented followed by a description of theory of evaporation from cementitious matrix that is represented as a capillary porous media. the assumed boundary conditions are not supported by the physics and driving forces of surface evaporation process.
Dw is the molecular diffusivity of water vapor in air (m2/s).g. Vapor is transferred through molecular exchange in the same manner as heat and momentum are transferred. the net flux of a material (e. 93]. According to Fick’s first law.2.methodology is applied to the results of drying experiments under low and normal pressure conditions and is further verified by a Finite Element (FE) analysis.1) where J is the mass flux (kg/(m2s)) or drying rate in evaporation problems. Cw is the water vapor concentration (kg/m3) and x is diffusion distance (m). since water vapor concentration at the water surface is higher than its concentration away from the surface [88. 90. vapor) in any direction is proportional to its concentration gradient in that direction and therefore the rate of evaporation at a horizontal surface is given by [122] J = − Dw dCw dx (4. It usually occurs at the water surface exposed to the atmosphere as the vapor is carried away by the air flow [121. This concentration gradient drives water vapor from higher concentration to lower. evaporation rate can be described by: 89 . Assuming water vapor as an ideal gas ( pw = Cw RT / M ). Physics of water evaporation Water evaporation is a phase transition process by which molecules are converted from the liquid state into a vapor state. as most of transfer takes place within a few molecular free path lengths of the surface [89]. This process in the atmosphere arises from the action of molecular diffusivity. 91]. 4. 92.2. 89.
local relative humidity at water surface is 1 (100%).2.2.J = − Dw M dpw RT dx (4.K) and T is the absolute temperature (oK). The thickness of the boundary layer is of the order of 1mm [128] and the movement through this boundary layer is presumed to be governed by the molecular diffusivity [129]. 124. 90 . as shown in Figure 4.1. Bringing the concept to the evaporation problem. If relative humidity of the air well away from the surface is below 100%. 125. water vapor moves upwards from the surface and evaporation rate can be obtained as J = − pw0 Dw M dH ⋅ RT dx (4.2) where pw is the water vapor pressure (Pa). since the air in direct contact with the water is saturated [54].2. the partial pressure of water vapor. At the water surface of an open dish. a thin interfacial layer covering the water surface is considered as the boundary layer through which the vapor molecules transfer to the surrounding air. Based on definition of fractional relative humidity as H = pw / pw0. a concept set forth by Prandtl [127] for the momentum transport in the neighborhood of a solid wall. 126]. M is the molar mass of water (Kg/mol). In mass transfer problems. the air near the surface may be regarded as a boundary layer.2.3) The approach based on equation (4. is fixed by the saturated vapor pressure of water. pw. pw0.3) has been suggested in studying the evaporation processes and determining diffusion coefficients of volatile liquids diffused into air [123. R is the universal gas constant (m3Pa/mol.
this equation reflects the effects of relative humidity. moisture flux can be expressed as J = Dw M pw0 (1 − H ) RT δ (4.Figure 4.2.5) This equation shows that rate of evaporation is a function of relative humidity.2. so that from equation (4. Pw0. and boundary layer thickness.2. H. Following equation proposed by Wiederhold [130] relate saturation water vapor pressure to the temperature.1.4) where pw0 is the saturation vapor pressure and pw* is the ambient vapor pressure and J is the mass flux or rate of evaporation (kg/m2s). Since H= pw/pw0. δ.2. Since saturated water vapor pressure is a function of temperature itself [ 130 ] and boundary layer thickness is affected by the air flow. 91 . temperature and wind velocity on the evaporation rate. Schematics of the interfacial boundary layer during evaporation from free water surface For simplicity we assume linear variation of vapor pressure across the boundary layer which is of thickness δ.2) we have J = Dw M pw0 − pw∗ δ RT (4. saturated water vapor pressure.
2. temperature 92 Saturated Vapor Pressure. This relationship is shown in Figure 4.2 as well.2.4 1. The relationship between saturated water vapor pressure and temperature is shown in Figure 4. kPa Temperature (oC) .18 (4. 132].2.502(T − 273.15) T − 32.7) where Dw is the diffusion coefficient of water vapor in the air (m2/s) and T is the absolute temperature (oK).2. x105 m2/s Water 50 25 4 AirWater Vapor 3. p is the ambient pressure (Pa).2.775 ×10−6 + (4.6) where pw0 is saturated vapor pressure of water (Pa). 1976) 280 300 320 340 360 380 Temperature (K) Figure 4.479 ×10−8 ) T + (1.656 ×10−10 ) T 2 (4. and T is the absolute temperature (K).6 0 (By Bolz and Tuve. diffusion coefficient of water vapor in air is a function of temperature which can be obtained by a regression curve fit to data from Bolz and Tuve [131. Variation of diffusion coefficient and saturated vapor pressure vs.2. Dw = − 2.2. 0 20 40 60 80 100 100 75 Binary Diffusion Coefficient. Also.pw0 = 611.0007 + 3.21(1.46 × 10−8 p) e 17.2 2.
4.3. relative humidity and boundary layer thickness on rate of evaporation is studied and shown in Figures 4.016 x 103 kg/mol and the universal gas constant as R= 8. effects of temperature.314472 m3Pa/mol. the evaporation rate at 40oC decreases from 6.4) to (4. As shown in the figures. e.2 mm to 3 mm translate into a reduction of evaporation rate from 13.Having molar mass of water vapor as M=18.7 kg/(m2h) when the relative humidity increases from 10% to 90%.7 mm. in an ambient temperature and relative humidity of 40oC and 50%. applicable to 93 .2.7). Theory of evaporation from cementitious materials Moisture evaporation of Portland cement paste and concrete can be explained by the concept of moisture transfer through porous media using an isothermal drying model. Increasing boundary layer thickness however results in decreasing the evaporation rate..2.g. For instance assuming a boundary layer thickness of 0.8 to 0.9 kg/(m2h). which more significantly increases by lowering ambient relative humidity and boundary layer thickness.K and using equations (4. Since drying of porous media includes complex moisture transfer in both liquid and vapor states.2. the driving force of vapor pressure gradient which is the case in water vapor evaporation cannot be used directly here. increasing the ambient relative humidity cause a drastic reduction in the rate of moisture evaporation. a general concept of hydraulic potential ψ. an increase in the boundary layer thickness from 0.3a and 4.4 to 0. Similar to the effect of the boundary layer thickness. Instead. increasing ambient temperature leads to increase in the rate of evaporation.3b.2.
kg/(m2. kg/(m2.7mm 1mm 3mm 0 0 10 20 30 o 40 50 Temperature ( C) (a) 10 RH = 10% Rate of Evaporation.h) 16 RH=50% δ = 0.20 Rate of Evaporation.3.7mm 20% 30% 40% 50% 6 4 60% 70% 2 80% 90% 0 0 10 20 30 o 40 50 Temperature ( C) (b) Figure 4.h) 8 δ = 0. boundary layer thickness and relative humidity on evaporation rate from water surface 94 . Parametric study on the effects of temperature.4mm 8 4 0.2mm 12 0.2.
Equation (4.3. 141 . ψ depends on only one state variable. u = − Κ ∇ψ (4. where unsaturated flow rate.1) together with continuity equation leads to fundamental Richards equation as [135. Note that in this approach of drying. u (m/s) is related to the hydraulic potential ψ (m) by a liquid conductivity coefficient K (m/s) [133.2) 95 . and t is the time (s). as well as K which also depends explicitly on θ .3. 139 ]. The validity of this hypothesis was discussed by Mainguy et al.2) where θ is the volume fraction moisture content (). and hence it can be called capillary potential as well. water vapor and dry air [136. 143. 135]. 144] ∂θ = ∇ ⋅ ( Κ ∇ψ ) ∂t (4. 134. 138 . 137 .3. However. it has been shown that by introducing certain assumptions above formulation in terms of one single driving force can be retrieved [ 140 .1) Hydraulic potential is indeed the energy required to transfer unit weight of liquid from the porous materials to a reservoir of the same liquid at the same temperature and elevation. namely θ. Equation (4. The concept comes from unsaturated flow which is governed by extended Darcy law. The drying of porous media has been recently analyzed with multiphase approaches considering the moisture transfer through porous space partially saturated by liquid water.both vapor phase transfer and water liquid transfer through unsaturated porous medium is used. [139].3. 142 ]. In this approach u is the massaveraged velocity which stands for the water in both liquid and vapor states.
Coussot [145] discussed the cause of moisture motion through the porous network by the difference in capillary pressure between the top and the bottom of the sample. Studies of drying behavior of clay brick ceramics by Hall et al. In this study. 141]. Nonetheless.shows that gradient of hydraulic or capillary potential is the main cause of moisture migration through capillary media [104].3. This equation turns out to be a nonlinear diffusion equation.3. and different wet capillaryporous materials by 96 .4) can also be derived from the mass conservation and transport laws for vapor and liquid water in concrete [50. Platten [146]. Equation (4.4) where D is the moisture (liquid and vapor) diffusion coefficient (m2/s).3) This relationship shows that unsaturated flow in porous media can be expressed either by hydraulic potential gradient or by vapor pressure gradient.3. with D depending on θ. the focus of the work is on moisture loss analysis and therefore to calculate moisture content. Kelvin equation relates capillary potential to capillary pressure or vapor pressure by: ψ= RT Pws ln M Pw0 (4. ψ is replaced by θ in equation (4. variation of both parameters as the driving force from the interior of material through ambient with no discontinuity is used to explain the drying process. Cooling [100]. Accordingly. [99].3.2) using the definition of capillary diffusivity as D = Κ (dψ dθ ) . and we have ∂θ = ∇ ⋅ ( D ∇θ ) ∂t (4.
Figure 4.Kowalski [147] as shown in Figure 4.1 indicate that evaporative flux at the boundary is roughly constant for a period of time at the beginning of the drying referred to as stage I and then falls significantly by the time in the second period or stage II. The same drying behavior was observed for the Portland cement pastes in recent experimental works by authors [148].3.3. These two stages of drying based on variation of driving forces and degree of continuity between liquid and vapor phases are discussed individually in the following sections. as well as in mortar [149] and lightweight concrete [66].1. Two stages of drying during evaporation of capillary porous materials 97 .
Results show that the external evaporation potential is independent of internal capillary microstructure. ψ s = ψ i .1. but the gradient of vapor pressure at the boundary layer. but the moisture concentration and hydraulic diffusivity are sufficiently high that an 98 . There is no vapor pressure or hydraulic potential gradient to drive the moisture inside the porous body.3. drives the moisture to the ambient. referring to equation (4.4.3. In analogy to vapor pressure.1. the porous media (i. cementitious materials) are fully saturated and therefore. Stage I Drying Mass loss experiments of cement paste subjected to low pressure has been recently conducted [148]. the hydraulic potential inside the porous body is also uniform and surface hydraulic potential is equal to initial value. Variation of hydraulic potential and vapor pressure from the interior of material through ambient is depicted in Figure 4.1a at the beginning of drying. the vapor pressure inside the open cell microstructure is equal to the saturated vapor pressure. The phenomenon can be described by the variation of hydraulic potential or vapor pressure and the degree of continuity between liquid and vapor phases.1c.3. As shown in Figure 4.3. A drying begins.1. which for a saturated material is ψ = 0 .3) and varies to ψ * as the ambient hydraulic potential. and this aspects controls the moisture transport within the material in stage I drying. Although the vapor pressure is still approximately equal to the saturated vapor pressure. which varies from saturated vapor pressure just on top of the material to the ambient vapor pressure. the moisture evaporates at the surface and the surface vapor pressure falls slightly from the initial value.e. ps ≈ pi .
(a) (b) (c) Figure 4.1. (a) Fully saturation state of porous material as the initial condition.1b but the liquid phase is still continuous. and therefore. The material is not fully saturated as shown in Figure 4.1. the porous material remains capillary saturated. Continuation of drying leads to continuous decrease of moisture concentration deep inside the material.3. This drying scenario is referred to as the funicular drying regime [150].adequate capillary flow to the surface can be generated by small value of vapor pressure gradient or water content gradient [54]. 105]. vapor pressure is approximately equal to saturated vapor pressure.1. The drying is in stage I while the material is capillary saturated. and the profile of internal water concentration remains flat [54. 99 . (c) variation of hydraulic potential and vapor pressure from the interior of the material through ambient in stage I drying. (b) capillary saturation of porous material as a later stage of drying.3.
water vapor pressure can fall almost to the ambient vapor pressure instantaneously to reach hygral equilibrium with the environment. Stage II Drying As drying continues. As shown in Figure 4. at this stage. the diffusion of water vapor through continuous vapor space becomes the dominant mechanism for moisture transport [105]. the vapor pressure falls from saturated vapor pressure to a lower value. On the other hand.1b. During this period. The former condition in terms of moisture concentration can be expressed as equal moisture concentration between surface and ambient levels [148]. At this stage.3. As shown in Figure 4. the evaporation continues until vapor pressure reaches ambient vapor pressure or hydraulic potential balance created which can be referred to as hygroscopic equilibrium. two different theories can be suggested for the variation of hydraulic potential or vapor pressure at the surface.1c and designated by stage II (B) modeling.3.2. based on first theory designated by stage II(A).3. the saturation level is lower than capillary saturation and therefore. From the driving force aspect. at this stage of drying.2. The latter condition is 100 . the second theory suggests a gradual decrease for the water vapor pressure from the saturated pressure to the ambient vapor pressure as shown in Figure 4. the evaporation rate falls and drying happens under internal mass transfer control [106].2. and not liquid water movement. when the liquid phase becomes discontinuous [105].3.1a. the moisture concentration in liquid phase decreases and reaches the insular saturation which is associated with the transition time from stage I to stage II. Since moisture transfer is limited only to the diffusion of water vapor generated by local evaporation of bound liquid water.4.2.
4.2. and exposing only top surface to the drying condition.4. The moisture transport through the sample ( 0 ≤ x ≤ L ) is assumed to be onedirectional (1D) because of high surface areatodepth ratio.4.1. (b) variation of hydraulic potential and vapor pressure from the interior of the material through ambient in stage II using simplified model (model A). (c) variation of hydraulic potential and vapor pressure from the interior of the material through ambient in stage II using convection model (model B) 101 . Governing Equations and Geometry of the Problem A model is developed to address the moisture transport properties of cement pastes during drying tests under low pressure condition [148].3.1. TwoStage Modeling of the Drying of Cementitious Materials 4.tantamount to imposing a convectivediffusive transfer of vapor at the surface for modeling stage II drying. (a) Low saturation state of porous material as the advanced stage of drying. (a) (b) (c) Figure 4.
By assuming an isothermal condition, effects of change in temperature during drying process is negligible. For modeling the internal moisture transport in both stage I and stage II drying, 1D form of equation (4.3.4) is used with the replacement of dimensionless moisture concentration variable, θ (), with the dimensional variable C (kg/m3) [55, 61, 63, 110, 151].
∂C ∂ ⎡ ∂C ⎤ dD ⎛ ∂C ⎞ ∂ 2C = D(C ) = ⎜ ⎟ +D 2 ∂t ∂x ⎢ ∂x ⎥ dC ⎝ ∂x ⎠ ∂x ⎣ ⎦
2
(4.4.1.1)
The only difference in applying this formula in stage I and II, is the assumption of constant diffusivity at stage I, while diffusivity can be variable at stage II drying. The assumption is based on Garrabrants and Kosson’s experimental observations [105] which support relatively constant moisture diffusivity in the early stage of drying, followed by variable moisture diffusivity at later stages of drying. While the first term in the right hand side of equation (4.4.1.1) is zero in the stage I drying, it is nonzero at stage II [141], although it looks very small considering higher order terms. For simplification, dD / dC can be assumed to be zero during stage II as well. However, since ∂D dD ∂C = ∂t dC ∂t the assumption would lead to (4.4.1.2)
∂D = 0 , which contradicts the experimental ∂t
observations. To take care of this problem, the time of drying was divided into several time steps at stage II and constant diffusivity was applied at each time step with respect to the time and the location. Therefore, the timevarying diffusivity values at stage II are only averaged over the space. This can be acceptable
102
considering high evaporative surfacetodepth ratio of the samples. Consequently, equation (4.4.1.1) can be simplified to equation (4.4.1.3) for stage I as well as each time step for stage II. ∂C ∂ 2C =D 2 ∂t ∂x (4.4.1.3)
Hydration of cement and its implications on transport properties need to be taken into account during earlyage drying of cementitious materials. With time, liquid water will be chemically bounded into hydration products as well as physically bound water to increasing internal solid surfaces. Both chemically and physically bound water may be assumed to be in local equilibrium with free water due to rapid rate of hydration at the early age. To simplify the concept, the concentration of bound water, S, is directly taken proportional to the concentration of the free water, C, the substance free to diffuse. S = RC (4.4.1.4)
During earlyage drying when the moisture diffusion is accompanied by immobilization of bound water, equation (4.4.1.3) is modified as: ∂C ∂ 2C ∂S =D 2 − ∂t ∂x ∂t (4.4.1.5)
Equation (4.4.1.5) can be presented in a new form by substituting for S from equation (4.4.1.4). D ∂ 2C ∂C ∂ 2C = = D′ 2 ∂t 1 + R ∂x 2 ∂x (4.4.1.6)
Replacing the term D/(1+R) with D’ as the effective moisture diffusion coefficient, the equation is seen in the usual form of diffusion equations. 103
Therefore, D’ presents the potential for conductivity of free water through the fresh cementitious materials. The cement paste is assumed to be homogeneous at initial conditions with an impermeable bottom surface. Therefore, the initial and boundary conditions are t = 0, C = Ci ∂C =0 ∂x 0≤ x≤ L at x = 0 (4.4.1.7) (4.4.1.8)
J = − D′
where Ci is the initial moisture concentration (kg/m3), and J is the moisture flux (kg/m2). Moisture transport during stage I drying is considered as an external mass transfer control process with a constant drying rate. Thus, the boundary condition at the top surface is J = F0 at x = L (4.4.1.9)
where, F0 is the constant flux in stage I drying (kg/m2s). Solution to moisture concentration values as a function of time and location during stage I drying based on equation (4.4.1.64.4.1.9) is [109]
F0 L ⎧ D ′t 3 x 2 − L2 2 − 2 C (t , x ) = Ci + ⎨ 2 + 2 D′ ⎩ L 6L π
n =1
∑
∞
( − 1) n
n2
exp(
− D ′n 2π 2 t
L2
) cos
nπ x ⎫ ⎪ ⎬ L ⎪ ⎭
(4.4.1.10) In this model transition time from stage I to stage II, referring to the loss of continuity in the liquid phase, is defined by the user which may be obtained from experimental data. Prediction of the transition time based on physical grounds requires further studies. For modeling stage II drying based on the 104
concept of convectivediffusive transfer of vapor at the surface, the exchanged flux of moisture, J, can be characterized by the convection equation [109, 152]. J = − D′ ∂C = k (Cs − C* ) ∂x (4.4.1.11)
where k (m/h) is convective moisture transfer coefficient or surface factor, Cs is the surface moisture concentration and C* is the ambient moisture concentration (kg/m3). k parameter takes into account the wind velocity, the surface temperature, and the relative humidity for drying to the atmosphere or in the case of lowpressure drying encompasses vacuum condition and temperature. The analytical solution for equations (4.4.1.64.4.1.9) during each time step at stage II drying with the convective boundary condition as of equation (4.4.1.11) is given by [153]
C ( t , x ) = C∗ + ( Ci − C∗ ) + ∑
2sin(λn ) x cos(λn ) L n =1 λn + sin(λn )cos(λn )
∞
⎛ 2 D′t ⎞ ⎜ − λn 2 ⎟ L ⎠ e⎝
(4.4.1.12) Where,
λn tan(λn ) =
kL D′
(4.4.1.13)
Geometry, governing equations and boundary conditions during stages I and II drying are presented in Figure (4.3.2.2).
105
Figure 4.3.2.2a
Figure 4.3.2.2b
Figure 4.3.2.2. Geometry, governing equations and boundary conditions during: (a) stage I drying, (b) stage II drying assuming convective moisture flow on the top surface
4.4.2. Simulation and Parametric Studies Modeling of the drying process is presented based on dualstage moisture movement through the porous media and imposing a constant flux and convective boundary conditions at stage I and II, respectively. Model simulates the evaporation process using equation (4.4.1.10) as the solution for stage I drying, and equation (4.4.1.12) as the solution for each time step at stage II drying. In the code, an extra function is developed to find sufficient number of solutions of λn in equation (4.4.1.13) and return them to the main code. The code calculates moisture concentration profile over the volume. The difference between integration of moisture concentration and initial total moisture yields the
106
and surface factor.4.45 is considered as the drying material. except for study of the depth in which the thickness is variable. including effective diffusivity. on moisture concentration profiles are studied at 1 h intervals with three orders of magnitude in diffusion coefficients of 5x109.4. F0. Parametric studies are performed to verify capability of the model in capturing effects of major moisture transport parameters. 107 . k. A drying surface area of 100 mm x 100 mm and thickness of 10 mm are used as the size of specimen.4. but it controls moisture concentration profiles. 5x108 and 5x107 m2/s as shown in Figure 4. Effect of diffusivity at stage I.2. and sample thickness. diffusivity does not control cumulative moisture loss and evaporation rates due to the imposed boundary condition at this stage.4.2. Throughout the parametric study. assuming an initial evaporation rate of 0. Parametric studies also include effects of initial external rate of evaporation. The code calculates flux at the surface according to equation (4. but funicular flow. transition time. a variable that is measured during a drying test. a cement paste with w/c ratio of 0.4 kg/(m2. As shown in Figure 4.2. D’I.1. Although in gravimetrical methods.11). As expected.1b. moisture concentrations reach negative values or large moisture concentration gradients show up imposing 0.1a and 4. L. the natural internal moisture flow in stage I. D’. It can be concluded that assumed diffusivities are too low to balance the flux on the top surface. on the cumulative moisture loss and evaporation rates.h) and a transition time of 10 h. at stage I drying.1.4 kg/(m2. ttrans. moisture concentration profiles are not obtained.h) as the initial evaporation rate.cumulative moisture loss.
4. is simulated after 40 h of drying.001 m/h. D’II.2. an increase in cumulative moisture loss from 55 g to 64 g.5 kg/(m2.2a effect of diffusivity at stage II drying. As shown in Figure 4.h) and ttrans as 5 h. assuming D’I as 5x107 m2/s. In Figure 4. k as 0. equivalent to an increase by 13% in total moisture loss.0005 m/h to 0. The results of simulation are in agreement with the concept of diffusivity which is moisture transport conductivity through capillary medium under driving force of a concentration gradient.2b.4. an increase in moisture diffusivity translates into higher evaporation rate and higher cumulative moisture loss at stage II drying.2a.leads to values of initial diffusivity of the order of 5x107 m2/s which satisfy the flux with small values of internal moisture concentration gradients.4. F0 as 0. F0 as 0. As shown in the figure.002 m/h assuming D’I as 5x107 m2/s.h) and ttrans as 5 h.2. A significant increase in evaporation rates at early hours of stage II drying is also simulated by increasing this parameter.4 kg/(m2. on cumulative moisture loss and evaporation rates is studied with five different diffusivity values increasing from 5x1010 m2/s to 5x108 m2/s. k.2.5x108 m2/s. is studied with increasing this parameter from 0. A much smoother fall in evaporation rates are observed at the beginning of stage II drying by assuming 108 . D’II as 1. and also significant increase in evaporation rates after 10 h of drying as shown in Figure 4. These results reflect increases in the cumulative moisture loss from 44 g to 63 g translating into a 28% increase in total moisture loss after 40 h of drying. Effect of convective moisture transfer coefficient.
2 t h 0 1200 800 400 0 400 Moisture Concentration.2.1b 0.2 t h t/h 0 0 200 400 600 Moisture Concentration.4.4 h = 10 mm 0.2.1 t = 10 h 2 t=1 1 t = 10 h 2 t=1 0.1.8 0. kg/m3 800 0 0 200 400 600 Moisture Concentration.4 DI = 5x108 m2/s h = 10 mm 0.2.4 h = 10 mm 0.2 t h 0.6 t/h t/h DI = 5x109 m2/s 0.8 2 t=1 0.2. kg/m3 800 Figure 4. kg/m3 800 Figure 4.1a 1 t = 10 h Figure 4.6 0. Parametric study on the effect of moisture diffusivity at stage I on profiles of moisture concentration drawn at 1 h intervals during drying of cement paste samples 109 .8 0.6 DI = 5x107 m2/s 0.4.1c Figure 4.4.4.
4 0. g 80 60 40 20 0 0 10 DI = 5x107 m2/s k = 0.002 0.5x108 5.h) Evaporation Rate.0x108 1.4 kg/(m2.2. g 80 ttrans = 5 h 60 40 20 0 0 10 DI = 5.0005 0.0x107 m2/s DII = 1.1 0 Cumulative Moisture Loss. h 40 Figure 4.5x108 m2/s F0 = 0.001 0.h) ttrans = 5 h 20 30 Drying Time.0x109 1.2.2 0.0x1010 0.h) 20 30 Drying Time.2a 0. Parametric study of moisture diffusivity at stage II drying and convective moisture transfer coefficient on cumulative moisture loss and evaporation rates of cement paste samples 110 Evaporation Rate.4. kg/(m2.0.4 0.5 DII (m2/s) 5.2 0 Cumulative Moisture Loss.0015 0.6 k (m/h) 0.4.001 m/h F0 = 0. h 40 Figure 4.2b Figure 4.5x109 5.4. kg/(m2.5 kg/(m2.3 0.h) .2.2.
2. Effect of transition time is studied by five different values from 6 to 14 h. which is due to lower surface moisture concentration.3a and Figure 4.0005 m/h results in a marked drop in evaporation rates at the beginning of stage II. cumulative moisture loss after 40 h drying increases from 32 to 92 g and also evaporation rates at the beginning of stage II drying increase by increasing sample thickness. 1.3b show parametric studies on initial external evaporation rate. k. lower evaporation rates along with a large fall from initial values are simulated by increasing F0.4. 0. As expected. but has minimal effects on cumulative moisture loss after 24 h.4. In both simulations.2.5x108 m2/s.4.001 m/h.2. make drying stage transition smoother. ttrans. 111 . F0 and ttrans are given as 5x107 m2/s.5 kg/(m2.5 kg/(m2. which is expected due to lower surface moisture concentrations.001 m/h.h) and 5 h. D’II.2. 0. D’I. respectively.3 to 0. assuming D’I. The specimen thickness is the last parameter studied by increasing from 5 to 25 mm in five intervals.4. Figure 4.5x108 m2/s. 1.higher values of k such as 0. and k are given as 5x107 m2/s. at stage II drying.002 m/h. and time of transition from stage I to stage II. and 0. respectively.4. while low k values such as 0. respectively. However.3b. increasing F0 from 0. As shown in Figure 4.h) reflects in higher cumulative moisture loss and evaporation rates at early hours of drying. F0. higher values result in longer period of stage I and more significant fall in evaporation rate between stage I and II. D’II. As shown in Figure 4.
2 0 Cumulative Moisture Loss. g 75 F0 = 0. h 40 Figure 4. g 80 ttrans = 10 h 60 40 20 0 0 10 DI = 5x107 m2/s DII = 1x108 m2/s k = 0.h) .h)) 0.4 0.h) 50 25 0 DI = 5.001 m/h 20 30 Drying Time.35 0.0x107 m2/s DII = 1. on cumulative moisture loss and evaporation rates during drying of cement paste samples 112 Evaporation Rate.6 F0 (kg/(m2. h 40 Figure 4. and transition time.0. kg/(m2.h) Evaporation Rate. h 14 12 10 8 6 0.001 m/h 0 10 20 30 Drying Time. ttrans.2 0.40 0.4 0.3 0.5 kg/(m2.3b Figure 4.3. kg/(m2.2.3a 0.4.50 0.5 ttrans .2.5x108 m2/s k = 0. F0.4.45 0.2.1 0 Cumulative Moisture Loss. Parametric study on the effect of initial evaporation rate.4.30 0.
0 0.5 0.001 m/h 40 Figure 4.3. on cumulative moisture loss and evaporation rates during drying of cement pastes 4. h DII = 1.4. D’(t) and k(t) to predict the experimental results. BackCalculation Procedures for Modeling of Stage II Drying A back calculation model is used to predict diffusivity and moisture transfer coefficient at any time step based on given cumulative moisture loss and evaporation rate curves.h) . The objective function for minimization is the weighted sum of two errors of cumulative moisture loss and flux. In optimization algorithm. Once the optimal solution Dnorm and 113 Evaporation Rate.2.5 kg/(m2. Parametric study on the effect of moisture transfer coefficient.0x107 m2/s L (cm) 2.h) ttrans = 5 h DI = 5.1 F0 = 0. k. Dnorm=D’i /D’i1.5 1. normalized design variables (D’ and k) are used to avoid numerical problems caused by mixing big and small numbers in numerical operations. Thus. g 100 75 50 25 0 0 10 20 30 Drying Time. so does the mass transfer coefficient knorm= ki /ki1. diffusion coefficient at the current time step is normalized with its value at previous time step. kg/(m2. Optimization is implemented to find two best fit parameters.5 0.4.75 0.25 0 Cumulative Moisture Loss.4.0 1.5 2.5x108 m2/s k = 0.
t is the current time step.4.3.11).1) and equation (4.1) where.1) which is derived by rearranging equation (4. while initial values for k are estimated from equation (4. Initial values for D’ are obtained from analysis of stage I drying. Ccrit is the surface moisture concentration at end of stage I or critical moisture concentration (kg/m3) and C* is the ambient moisture concentration. while t (i −1) is the previous time step.4.3). x = L) − C* (i ) (4.3) k ( est ) = J (exp) (t (i ) ) C (t (i −1) .3. these two initially estimated parameters. D(est) and k (est) are passed to the optimizer and the optimizer calls the function evaluation to calculate the error and update these two parameters several 114 .4.4.3) In equation (4. k (est ) = J (exp) (t = ttrans ) Ccrit − C* (4.4.3. (est) are estimated by a convection transport D (est ) = Di −1 J (exp) (t (i ) ) J (exp) (t (i −1) ) (4. For other time steps. Inequality constraints are imposed to the design variables D’ and k in terms of lower and upper bounds. k formula. they are reverted back to D’ and k. while initial k values.4.3. initial D’ values for optimization process. k (est) is the estimated surface factor (m/h) and J (exp) (t = ttrans ) is the experimental rate of evaporation at the transition time which is roughly equal to initial evaporation rate.3.knorm are found.1. D (est ) are estimated by the assumption of proportional decrease of diffusivity with the flux.4. At each time step.3.
3. Experimental data including cumulative moisture loss and evaporation rates of a drying Portland cement paste with w/c ratio of 0.3. diffusivity fell almost at the beginning of stage II drying.5. Results of prediction of experimental data and the best solutions for D’ and k parameters are shown in Figure 4.1b.3.4.0017 m/h to 0.4. Results are discussed in the following section and shown in Figures 4. This is in agreement with the literature where diffusivity after being relatively constant during early stages of drying. 115 . Comparison with Experimental Data and Finite Element (FE) Analysis An experimental procedure for drying under lowpressure conditions was developed by authors that is capable of characterizing evaporation parameters using measurement of mass change during drying process [148].times until optimal solution for D’ and k are found. The model fits cumulative moisture loss and evaporation rates accurately throughout the drying period.15x107 m2/s to 1.18x108 m2/s.4. while diffusion and moisture transfer coefficients are constant through the initial stage of drying and variable during stage II.4. As shown in Figure 4. falls by decreasing moisture content below midrange saturation [105].4.1a and 4.1b.3. 4.3.45 are used in the backcalculation procedures. The reduction of moisture diffusivity at stage II drying is much more significant when this parameter reduced from initial value of 5.0011 m/h. but the moisture transfer coefficient remains constant even in the first 7 h of stage II before it gradually reduced from 0.1a and 4.1b.
4. h 40 Figure 4.1a 0.3.001 4x107 D' 2x107 6x109 0 10 20 30 Drying Time.3 0. (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the experimental data of an evaporation test.4.0015 Moisture Diffusivity.3. h 0.1 Cumulative Moisture Loss.4. (b) Backcalculated D and k for the best fit of data 116 .5 0. m2/s 6x10 7 0. kg/(m2. h 40 Figure 4.4 Experiment Analytical data FE Analysis 0.1 1 10 100 0. g 80 60 40 20 0 0 10 w/c = 0. m/h 0.3.002 K Evaporation Rate.1.2 0.h) Moisture Transfer Coefficient.Drying Time.45 No curing T = 24oC 0 20 30 Drying Time.1b Figure 4.
5D1 characterizing the location of the drop in the curve D(Cnorm).86x108 to 5.1) where. ⎛ ⎜ ⎜ 1−α D(Cnorm ) = D1 ⎜ α + ⎛ 1 − Cnorm ⎜ 1+ ⎜ ⎜ ⎜ 1 − Cc ⎜ ⎝ ⎝ ⎞ ⎟ ⎟ n ⎟ ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ ⎠ (4.4x108 m2/s. In this equation. the results are well within the range of the data recently reported by Vejmelková et al. This relationship is presented in Figure 4. hence it was related to the average moisture concentration to obtain an intrinsic characterization. 155 ]. the moisture diffusivity is related to the relative humidity [56.5. According to Bazant and Najjar [60]. However. While in CEBFIP model. time is not the natural parameter for characterization of moisture diffusivity.78x108 to 2.1 which can be fitted by a model similar to CEBFIP [154]. Baluch et al [156] and Kim and Lee [155].15x107 m2/s. Cc is the normalized moisture concentration at D(Cnorm) = 0. [119] which range from 1. Comparing the values with the results of this analysis which ranges from 1. calculated moisture diffusivity for wellhydrated and cured concrete samples is within range of 3x1010 to 5.From the physical point of view. D1 represents D(Cnorm) when the samples is initially fully saturated and normalized moisture concentration is 1.53x108 m2/s 117 . in this work the moisture diffusivity is modeled versus average normalized moisture concentration. D0 is the minimum D(Cnorm) and α is the parameter that represents the ratio D0/D1. it may be concluded that very earlyage moisture diffusivity of a poorlycured cement paste is more than tenfold of the moisture diffusivity of a cured concrete.5.
6 0.2 0.for noncracked and 5. the convective moisture transfer coefficient. are within the range of the data reported in the literature [156.86 x 108 m2/s α = D0/D1 = 0.101 [70] to verify the results of proposed analytical method including the flux at the 118 . On the other hand. the calculated values of parameter k.0E008 D1 = 5. m2/s 5.8 Normalized Moisture Concentration 1 Figure 4. A Finite Element (FE) analysis was carried out using ABAQUS v. Backcalculated moisture diffusivity of an earlyage noncured cement paste as a function of normalized moisture concentration expressed by CEBFIP models As shown in Figure 4.1E007 6.4 0.6.003 m/h. among different values for variable n.1E007 1.1E007 1.5. 7. 157] varying from 0.5.15 x 107 m2/s D0 = 1.025 to 0.1E007 Moisture Diffusivity.1E007 3.036 which is in accordance with the results of Bazant and Najjar [60].0003 to 0. who reported values in the range of 0. was obtained as 0. n = 15 gives the best fit to the calculated diffusivity results.10 for different types of concrete.2 BackCalculated Diffusivities CEBFIP: n = 10 CEBFIP: n = 15 CEBFIP: n = 20 CEBFIP: n = 25 0 0. the moisture diffusivity ratio.03 Cc = 0. On the other hand.46x107 to 2. α.1.1.1E007 4.1E007 2.00x107 m2/s for cracked cement paste samples.
boundary conditions for stage I can be introduced as a constant heat flux. The other boundary conditions which are impermeable surfaces on the sides and bottom of the sample are ignored.top surface and the integration of the surface flux or cumulative moisture loss. Ts is the temperature at the boundary. Comparing equations (4. since no load corresponds to zero flux or impermeability. k/ρc and hfT in the heat transfer problem.3) with equations (4.3) ∂n whrere.5. the corresponding equations are ∂T ⎛ k ⎞ ⎡ ∂ 2T ∂ 2T ⎤ =⎜ + ⎟⎢ ⎥ ∂t ⎝ ρ c ⎠ ⎣ ∂x 2 ∂y 2 ⎦ (4.t) is the temperature varying in the domain with the time.1.y.1. the boundary condition of stage I drying is defined as a load at surface with the magnitude of initial evaporation rate.3) is for the heat convection at the boundary. k is the isotropic thermal conductivity which may be a function of T. T(x. In this analysis. Also an 119 . D. The technique utilizes the analogy between the heat transfer and the moisture diffusion.5. respectively.11) leads to the conclusion of onetoone analogy between these two boundary value problems where C.5. and T* is the ambient temperature. ∂T is the thermal gradient with a unit ∂n normal “n”. In equivalent transient thermal problem in 2D space.5. h fT is the convective transfer coefficient for thermal problem. While equation (4. and k in the diffusion problem are corresponding to T.4. analogous to moisture convection during stage II at the boundary.2) ∂T = h fT (T* − Ts ) (4.4.6) and (4. and ρ and c are the density and the specific heat.2 and 4.5.
[66. and since there was no interest on the edge effects.2. 158.4.3. Among different drying tests 120 . The comparison between results of the surface flux by FE analysis and the analytical and experimental results are shown in Figure 4.1a. Since there is no heat transfer corresponding to moisture transfer along the xdirection which is perpendicular to the direction of external flux. In order to further confirm the soundness of this analytical method. temperature of 22oC and relative humidity of 50%.3. the FE results show a good agreement with analytical and experimental data.1a. As shown in Figure 4. Cumulative moisture loss can also be obtained as the product of the integration of the flux obtained from FE analysis and the surface area. the domain is divided by only 20 layers. 159] were chosen for this purpose which involve onedimensional drying of concrete cylinders under normal room pressure.4. The type of analysis was an implicit heat transfer analysis. a uniform mesh with 4node linear heat transfer quadrilateral elements (DC2D4) was used. The experimental data of Šelih et al.5.interaction was created for the purpose of imposing a convective boundary condition during stage II drying. For the sake of comparison with the results of analytical method. Results of FE analysis for the flux at different drying times are shown in Figure 4. the method can be applied to different sets of drying data with available experimental moisture profiles. the same number of layers (100 layers) is used along ydirection which is along the direction of external flux.
Distribution of moisture flux (kg/(m2h)) at: a) t = 2 h. d) t = 20 h. b) t = 10 h. e) t = 40 h.2.5. c) t = 14 h.(a) t = 2 h (b) t = 10 h (c) t = 14 h (d) t = 20 h (e) t = 40 h Figure 4. 121 .
indicating a good agreement between the simulation and the experimental results.9 kg/m3. Using the BSB model [110].3 and 804. Conclusion Results of modeling moisture evaporation from the water surface and cementbased materials based on the physics of drying. C* was obtained 49 kg/m3.5. The drying results were used to backcalculate diffusion and convective moisture transfer coefficients as shown in Figure 4. The distribution of moisture concentration is compared with the experimental data as shown in Figure 4..5. along with saturated vapor pressure.5. 912. The ambient moisture concentration.8. coarse and fine aggregate in this mixture were 489.3b. can be obtained from the desorption isotherm curves which relate relative humidity to the water content. water.4b.5. respectively. 122 .6. Evaporation from free water surface is modeled implementing Prandtl’s interfacial boundary layer concept through which the water vapor molecules diffuse to the surrounding air due to gradient of moisture concentration.40 was chosen [159] as a representative data set for drying of earlyage cementitious materials. the evaporation rates were computed from cumulative moisture loss data using a numerical differentiation procedure. The model shows direct relationship of water evaporation rate to relative humidity.4a. As shown in Figure 4. 4.performed by Šelih et al. and expressed as a function of average moisture concentration in accordance to the CEBFIP model as shown in Figure 4.3a. results of drying tests on noncured normal aggregate concrete with w/c ratio of 0. 195. The amounts of Portland cement.5. C* .
g 60 40 20 0 Selih (1996) Simulation 0 10 20 Drying Time.5.3b Figure 4. m2/s 3x107 2x10 7 Evaporation Rate.1 0.0005 Moisture Diffusivity. (b) Backcalculated D’ and k for the best fit of data 123 . d 30 Figure 4.Drying Time.40 T = 22oC RH = 50% No Initial Curing 100 0.15 0.3a 0. d 30 Figure 4. (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the experiment performed by Šelih (1996).5.3.1 1 10 w/c = 0.0015 0.001 0.h) Moisture Transfer Coefficient. d 0.5. kg/(m2.05 0 Cumulative Moisture Loss. m/h K 0 D' 1x107 6x109 0 10 20 Drying Time.
4a 1 0.8 t h 0.4 0.6 h = 0. Results of analysis on drying data by Šelih (1996): (a) Moisture diffusivity as a function of normalized moisture concentration expressed by CEBFIP models.4 1d 3d 0. (b) moisture concentration distributions through the thickness of the sample in comparison with experimental results 124 .4.07E007 Analysis on Experiments by Selih (1996) D1 = 2.07E007 2.78 x 107 m2/s D0 = 6.6 0.00E009 0.4.94 x 109 m2/s α = D0/D1 = 0.5.5.1 m Time t/h 0.4b Figure 4.07E007 7. kg/m3 200 Figure 4.5.8 Normalized Moisture Concentration 1 Figure 4.7 1.2 7d 28 d Exp Sim Exp Sim Exp Sim Exp Sim 0 0 40 80 120 160 Moisture Concentration.07E007 BackCalculated Diffusivities CEBFIP: n = 10 CEBFIP: n = 15 CEBFIP: n = 20 Moisture Diffusivity.025 Cc = 0. m2/s 3.
The initial constant drying rate period (stage I) is modeled as a funicular liquid water transport stage and falling drying rate period (stage II) as a vapor diffusion transport stage. The agreement between the results of the proposed analytical method and an FE analysis and also experimental profiles of moisture distribution verifies the soundness of this analytical method. Water evaporation is used as a boundary condition for the initial stages of drying. Results of the analysis show that proposed analytical method is capable of predicting experimental data during drying process of cementbased materials and presenting the variation of introduced internal and surface moisture transfer parameters by the time.and boundary layer thickness. Results of applying the methodology to the drying experiments of cementbased materials under low and normal room pressure shows that calculated moisture diffusivities reduce drastically from stage I to stage II. which are functions of temperature and wind velocity themselves. respectively. rather than vapor diffusion at stage II. 125 . it can be concluded that the moisture diffusivity indicates liquid water conductivity at stage I. Physics of drying allows applying a constant flux boundary condition and a convective boundary condition at stage I and II. The moisture movement inside is modeled based on moisture transfer theories of capillary porous media. While Fick’s second law of diffusion is used for modeling moisture movement in both stages.
a time period well within the initial stage of drying [148]. Visible plastic cracks are observed few hours after placing.5. As shown in Figure 5.1. 14.2. the developed shrinkage strain may easily exceed low tensile strength of the fresh concrete [21] and cracking occurs (see Figure 5. The differences in crack tendency are most likely a consequence of the initial rate of evaporation [15]. results of drying experiments on fresh cementbased materials reveals that the initial evaporation rate remains roughly constant during a period. Modeling Rate of Evaporation Potential during Early Stage of Drying 5. 12. 66. wind speed and relative humidity increase the risk of early age cracking [11].1. 148. 14] and continues to rise as the evaporation proceeds. The initial rate of evaporation is very significant from the capillary pressure point of view as comparably small changes in the evaporation rate noticeably influence the capillary pressure versus time curve [12]. water menisci are formed in the interparticle spaces [12] and a negative pressure in the capillary water is built up [13. 15] and exceeding 1MPa in less than 8 h [16.1. 19]. If concrete is restrained. 126 .1). Introduction Earlyage cracking is a threat to structural integrity of concrete structures and if not inhibited would lead to service life reduction and unsightliness. The developed capillary pressure in drying fresh concrete reaching 50 kPa in few hours [10. 17] acts on the solid particles and results in the contraction of the still plastic material [18] ranging from strain levels of 1 to 3x103 [12. 149]. Due to evaporation. before beginning of a succeeding falling rate period [19. High rates of evaporation which depends on air and concrete temperature.
1.Drying time: 15 min. Development of 2D plastic shrinkage cracks during drying of a fresh plain cement paste It was found that the initial rate of evaporation from concrete surface is the same as rate of evaporation from water surface exposed to the same condition [8. the rate of water evaporation is required for modeling the moisture transport at this initial stage [148. 160] which can be found by either experiments [161] or using a mass transfer approach. 24h Figure 5. 10. Therefore. 127 . 4h.1. 3h. 148]. 3h 15min Drying time: 3h 30min. Moisture transport during initial stage of drying when concrete is still plastic is controlled by the external evaporation potential.
Considering similarity between rate of evaporation from water surface and concrete surface during initial stage of drying. A driving force of concentration gradient is introduced as the main cause of mass transfer during the drying process. An algorithm for determination of evaporation rate is suggested.Two stages of drying during evaporation of capillary porous materials In the absence of a theoretical method to estimate initial rate of evaporation from cementbased materials. In this paper theory of water evaporation based on the boundary layer theory. and fluid mechanics is described. 128 .2. predicted evaporation rates are verified by the recent experimental data from concrete surfaces. Mass transfer at low and high rates is discussed separately along with other related parameters such as mass transfer coefficient. a nomograph based on an empirical equation is used [68] which was calibrated by the data from Lake Hefner.Figure 5.1. diffusion. mass transfer.
1) where E is the rate of evaporation [kg. 129 (5. psat is the saturated water vapor pressure [kPa] at the surface temperature of concrete Tc .75.h1].06V) (5.253+0.1 [163] and Australian Technical Bulletin 95/1 [164] set requirements on evaporation rates below 1.1.2.1 [68].313(psat – pv) (0.75 and 0.5 kg/(m2h). respectively to prevent plastic shrinkage cracking. 0. rate of evaporation can be related to plastic cracking only when there is a model to predict capillary pressure and pressure on solid skeleton due to moisture loss and the earlyage tensile properties of the material is available.2. Canadian CSA Standard A23.h1]. the evaporation rate in the atmospheric pressure is commonly estimated using a nomograph in ACI 305 [68] given inputs for temperature. Available Method to Estimate Initial Rate of Evaporation From the mechanical point of view. available standards and specifications such as ACI 305. relative humidity.61exp(17. E = 0. 0.m2. As shown in Figure 5.2. In the absence of such a model. Uno [69] used the relationship between temperature and saturation vapor pressure to replace saturation vapor pressure in Menzel’s equation by the temperature. pv is the partial pressure of the water vapor in the air [kPa] and V is the wind velocity [km.2) .2. According to Hover [165] the underlying relationship for Menzel’s equation [166] which is the source of ACI nomograph is 200 years old Dalton’s law [167].3 Tc /(237. USACE TM 5822 [162].5. and wind velocity.3+ Tc)) where Tc is the temperature of the concrete surface [oC]. psat = 0.
2).Figure 5. Uno [69] developed a single equation which captures effect of temperature.1.5 .2. E = 5 [Tc + 18] ( 2.5 ) (V+4 ) × 10 6 (5.106) Using equation (5.2.1) and (5.2.[Ta + 18] 2. ACI nomograph for estimating rate of evaporation of surface moisture from concrete (ACI 305.3) 130 .ϕ .2. relative humidity and wind velocity.
AZ.h1]. 171 .e.m2. 6’ in diameter and 2’ deep [170].h1]. Menzel’s equation completely ignores this effect and therefore the predicted evaporation rate can be used only when evaporation surface is about the same size as of BPI (Bureau of Plant Industry) evaporation pan. Fort Collins. and Lake Kickapoo. 131 . cloudiness. However. 5. TX.where E is the rate of evaporation [kg. sunlight intensity and water quality [165]. On the other hand. have different evaporation equations [ 170 .. Ta is the air temperature [oC]. Evaporation as a Mass Transfer Problem Mass transfer is mass in transit as the result of a species concentration difference in a mixture. i. the rate of evaporation which is the diffusion of water vapor through the boundary layer covering the water surface is directly related to the characteristic length of the surface [122.3. Tc is the (surface) temperature of concrete [oC]. Evaporation is not a mass transport by bulk fluid motion. A Similar conclusion was reached by AlFadhala and Hover [169] who found that the nomograph is accurate for evaporation rates up to 0. 128].5 kg/(m2h).. 172 . φ is the relative humidity and V is the wind velocity [km. This inaccuracy can be due to the source of evaporation data used for deriving Menzel’s equation which is Lake Hefner in Oklahoma in 1950 to 1951 [165]. ACI 308R01 [168] reported that the accuracy of the prediction is within ±25% for actual evaporation rates up to 1 kg/(m2h) and consistently overestimates at higher evaporation rates by 50% or more. 173 ] which may be due to wind speed gradient. Different lakes such as Lake Meade. CO.
s vs = J i .3. and vi.3) 132 . s (5.1) and equation (5.3. s + ρi . s vi .s1].s is the velocity component i at the interface [m.s1]. s (vi .m2.3. s (vi . water vapor) is: mi′′ = ρi .2) where vs is the mass average velocity of all components [m. Further from the surface. ρi.m2. s vs diffusion convection (5. s = ρi . s − vs ) (5. humid air that occurs as result of a concentration gradient.3. s [kg.but rather a transport of one component. s − vs ) + ρi . nitrogen and other inert gases and the water vapor which is saturated very close to the water surface. the humid air has lower water vapor concentration and the nonuniformity of water vapor concentration near the surface causes a mass transfer due to the effect of diffusion. i. i.2) leads to a general equation for the total mass flux as: mi′′ = ρi . J i . water vapor within a mixture of components.s1] is the total mass flux. Combining equation (5. Mass flux due to diffusion.e.s is the density of the component i at the interface [kg.m3].e.3.e.1) where mi′′ [kg. The total rate of mass transfer of component i (i.s1] is related to the variation in the distance between molecules of the component i due to change in concentration of the component with time and can be expressed as: J i . Humid air is a mixture of dry air containing a mixture of oxygen.
the mass transfer due to the evaporation can be studied separately in regard to the flow rate.3.3. when the mass transfer rates are sufficiently low. Mass Transfer at Low Rates All mass transfer processes including evaporation generate flow at the interface.3).s1 1 ]. Ji. the mass flux is: 133 .1.1.3.1) According to Fick’s law of diffusion. Nonetheless. Ji is the rate of diffusion of component i per unit area of the surface [kg. when concentration of the diffusing component is low and diffusional mass flux is low enough that it does not affect the imposed velocity field. Considering these two terms. s (vi .m2. the velocities caused by mass transfer are negligible. Thus.3. diffusional flux can be represented by: J i = − Di ∂ci ∂y (5. Di is the diffusivity [m2.2) where.1.s]. s − vs ) = J i . s (5. Therefore in the low rate mass transfer condition. and y is the distance in the direction of diffusion [m].The first term on the righthand side of equation (5. the total mass flux is principally carried only by diffusion as given in the following equation: mi′′ = ρi . which is called the convective mass flux. ci is the concentration of component i [kg.s is the mass diffusional flux at the interface and the second term represents the transfer of component i by the accompanying flow due to diffusional flux at the interface. a stationary medium will remain at rest and a flowing fluid will have the same velocity field as if there were no mass transfer.m3]. More generally. 5.
s = − Di ∂ci ∂y (5.1. which is also shown in Figure 5.mi′′ = J i . the diffusion boundary layer thickness [m].1 for the case of evaporation due to laminar stream flow over a flat plate. it becomes more gradual in the region slightly above the interface and the concentration slowly approaches that in bulk fluid.3.1 shows a schematic representation of the concentration distribution near an interface for the case of evaporation from a flat plate.3.1.1.3. The gradient of concentration near the interface is directly related to δ d .4) 134 .1.3. Figure 5.e − ci . Schematic representation of the concentration distribution near an interface Although the variation in the concentration near the interface is very sharp. This gradient can be written as: ∂ci ∂y ≈ y =0 (ci .1.3.1.3) y =0 Figure 5. s ) δd (5.
we need to solve an “external forced convection mass transfer” problem.1.1. hm. can be expressed as the product of a mass transfer coefficient and the concentration difference between the surface and the fluid bulk (air flow).e ) ≈ Di (ci .4).e ) δd (5.where. respectively.1. This equation is not exact because the gradient is not constant throughout the boundary layer.3) and equation (5. ci.i ≈ δd Di (5. However.s are the concentration of component i at the air stream and interface [kg.i [m.i (ci . the mass transfer coefficient for component i.e in the region between the surface of the plate and the edge of the diffusion boundary layer.e ) δd (5. the rate of mass transfer can be taken to be proportional to the concentration difference between the interface and the bulk fluid.3.5) In mass transfer problems.6) Rearranging this equation leads to: hm. s − ci . using equation (5.3.1. Hence. J i ≈ Di (ci . Ji.3. Conceptually. the diffusional mass flux from a surface.s1] can be defined as: mi′′ = J i = hm . it does correctly reflect the fact that the concentration of component i (i.1.3. the free stream behaves much like a semiinfinite body that experiences a stepchange in its surface vapor concentration at the instant that 135 .s to ci. s − ci .7) To find the mass transfer coefficient for water vapor in the case of evaporation. water vapor) will change from ci.e and ci. Therefore.e.s.m3].3. s − ci .
1. as shown in Figure 5. u∞ [m. according to: t≈ x u∞ (5.1.10) To use dimensionless quantities to reduce number of independent variables. The disturbance associated with the change in the surface vapor concentration diffuses as a mass wave into the free stream.8) gives: Dvapor x u∞ δd ≈ 2 (5.3. and t is the time relative to the disturbance at the surface [s]. equation (5. the fluid motion transports the wave downstream from the leading edge. and the freestream velocity. The motion of mass diffusion wave can be characterized by a mass diffusion boundary layer thickness given approximately by δ d ≈ 2 Dvapor t (5.3. for the external flow problem.10) can be rearranged as: 136 .fluid encounters the leading edge of the plate. Therefore.3.1.9) Combining equation (5. The transport time for the mass diffusion is approximately related to the distance from the leading edge. This diffusion process takes time and.1. at x = x2.s1].3.1.1.s1].1.9) and equation (5. the mass wave has propagated further into the free stream than it had at x = x1.3.3.1.8) Where Dvapor is the diffusivity of the water vapor [m2.3. x [m].
1. Substituting the definition of kinematic viscosity into equation (5. Therefore.14) Inserting equation (5.1. the Reynolds number for the flow over a flat plate is defined according to: Rex = u∞ x ρ (5.3. and ρ is the density of water vapor [kg.m3].11) where ν is the kinematic viscosity of the water vapor [m2.1. Sc.3.13) μ Also the Schmidt number.12) leads to: 137 .3.δd x ≈ D D 2 Dvapor x 2 ν = 2 vapor = 2 vapor = x u∞ u∞ x u∞ x ν u∞ x ν ν Dvapor (5. x [m]. which provides a measure of the relative ability of the fluid to transport momentum and mass of component.1. the characteristic length is the distance from the leading edge.1.11) leads to: δd x ≈ 2 u∞ x ρ ν Dvapor (5.s1].3.3.3.14) into equation (5. Since Reynolds number is defined generally as Re = uchar Lchar ρ .m1.12) μ where μ is the dynamic viscosity [kg.13) and equation (5.1.1.s1].1. is defined as: Sc = ν Dvapor (5.3. for the flow over μ a flat plate.3. and the characteristic velocity is the freestream velocity.s1]. u∞ [m.
18) provides an approximate equation for the local Sherwood number associated with the laminar flow over a flat plate as: 138 .1.1. an approximate equation is provided for the local mass transfer coefficient associated with the laminar flow over a flat plate as: hm .δd x ≈ 2 Rex Sc (5.3. The Sherwood number is defined in general according to: Sh = hm . Therefore.1.vapor x Dvapor (5.1.3.3.3. the Sherwood number for flow over a flat plate is defined according to: Shx = hm .3.1.3.1. into equation (5.vapor ≈ Dvapor = Dvapor 2x δd Rex Sc (5.15).17) where Lchar is the characteristic dimension of the problem [m]. x [m].3.15) Combining equation (5.3. For stream flow over a flat plate.7) and equation (5.16) which is derived from conceptual model of the mass diffusion boundary layer thickness. Sh.1.16) The mass transfer is typically made dimensionless using the Sherwood number.i Lchar Dvapor (5.1. the characteristic dimension is the distance from the leading edge.18) Substituting equation (5.
3. the exact analytical solution of local Sherwood number for laminar and turbulent flows over a smooth isothermal flat plate which are correlated based on experimental data can be found elsewhere [132.1.Shx = hm.vapor dx L∫ 0 L (5.3387 Rex1 2 Sc1 3 ⎡ ⎛ 0. 5x105 was used for critical Reynolds number.3.22) and average Sherwood number is defined as: 139 . The reported data [176. In this study. 179.21) where Recrit is the critical Reynolds number for flow passing a flat plate of water. 177. The critical Reynolds number for the transition from laminar to turbulent flow varies with experimental conditions.3.21) are local.0468 ⎞ 2 3 ⎤ ⎢1 + ⎜ ⎟ ⎥ ⎢ ⎝ Sc ⎠ ⎥ ⎣ ⎦ 14 for Rex < Recrit (5. For a stream flow over a flat plate. the average mass transfer coefficient is given by: hm . The values of Sherwood numbers in equation (5. Shx = 0.1.vapor = 1 hm .20) Shx = 0. 175].3.20) and equation (5.1.vapor x Dvapor ≈ 0. while an average Sherwood number associated with average mass transfer coefficient is more useful to specify the boundary conditions.3.1.0296 Rex 4 5 Sc1 3 for Rex > Recrit (5. 174.3.5 Rex Sc (5. 180] are in the range of 104106.1. 178.1.19) Although above equation is only an approximate solution for a laminar flow.
1.3.1.25) in two 140 .25) can be integrated using any correlation or set of correlations for the local Sherwood number [132].3.1.13) and its derivative in equation (5.1.vapor L Dvapor (5.3.3.1.3.24) from x to Rex by substituting equation (5.1.1.3.18) leads to: Sh L = ∫ hm.3.3. it is convenient to change the coordinates of integration in equation (5.1.1.24) Since the Sherwood number correlations are expressed in terms of the Reynolds number.3. In the laminar region.20) is substituted into equation (5.1.3.1.Sh L = hm .23) Inserting equation (5.vapor dx 0 L Dvapor = ∫ 0 L Shx Dvapor x Dvapor dx =∫ 0 L Shx dx x (5.3.1.3.3.22) into equation (5.23) and using equation (5.2 is obtained by integrating equation (5.1.3.26) The average Sherwood number for the combined laminar and turbulent regions as shown in Figure 5.25) Equation (5.1.6774 ReL1 2 Sc1 3 ⎡ ⎛ 0. equation (5.3.1.25) in order to obtain: Sh L = 0.24): Sh L = ReL ∫ 0 Shx dRex Rex (5.0468 ⎞ 2 3 ⎤ ⎢1 + ⎜ ⎟ ⎥ ⎢ ⎝ Sc ⎠ ⎥ ⎣ ⎦ 14 for ReL < Recrit (5.
vapor = (5.3. equation (5.1.037 Sc1 3 ( ReL 0.8 ) for ReL > Recrit (5.1.3.1.3. having average Sherwood number using equation (5.21) is used in the turbulent region.27) would lead us to obtain average mass transfer coefficient according to equation (5.3.20) is used in the laminar region.3.8 − Recrit 0.3. Diffusion boundary layer thickness in a mixed laminar and turbulent regions for free stream flow over a smooth isothermal flat plate Finally.3. and equation (5.0468 ⎞ ⎤ ⎢1 + ⎜ ⎟ ⎥ ⎢ ⎝ Sc ⎠ ⎥ ⎣ ⎦ 23 14 Sh L = + 0.parts. from Rex < Recrit . for Recrit < Rex < ReL to obtain: 0.28) This average mass transfer coefficient can be used directly by equation (5.1.1.3.3.6) to obtain average rate of evaporation at low rates: 141 .1.2.23): Sh L Dvapor L hm .27) Figure 5.6774 Recrit1 2 Sc1 3 ⎡ ⎛ 0.1.26) and equation (5.1.1.
In addition to the diffusive transport represented by the mass transfer coefficient.s [kg. The difference can be either an increase or a decrease and can range from a few percent to an order of magnitude or more.m2.s1] of the various components in the direction normal to the surface [ 181 ].vapor (cvapor .2. s − cvapor .s is the density of the component i at the interface [kg. m′′ [kg.3. it required that the velocity field be unaffected by mass transfer and that the transferred components be dilute.e [kg.′′ mvapor = J vapor = hm . The mass flux of component i with concentrations of ci.m3] and ci. 142 . and vs is the mass average velocity of all components [m. depending upon the concentrations of the diffusing components.1) where.3.s1].2.29) 5.s is the mass diffusional flux at the interface [kg.3. When those conditions are not met.s1].e ) (5. s vs (5.1 shows a boundary layer over a flat surface through which there is a net mass transfer. Mass Transfer at High Rates For assuming low rate mass transfer.m2.m2. ρi. s + ρi .3. Ji.m3] at the free stream and the surface is: mi′′ = J i .1.m3]. convective transport can contribute substantially to the total mass flux. Mass transfer driving force at high rates Figure 5.s1]. the mass transfer coefficient will be different than the value given for low rate mass transfer. mi′′ is the total mass flux of component i [kg.2.
s (5. The mass concentration boundary layer in the highrate mass transfer problem Total mass flux of component i in terms of the concentrations ci.2.2.s (5.s1] is: J i . s ρ s vs Ji .s and ci.i [kg.s ρs where mi. s − ci .e are: mi′′ = hm . equation (5. and ρs is the total density of all components at the interface [kg.e ) + mi . the mass transfer coefficient for component i.2) Figure 5. s vs J i .e ) (5.3.s is the mass fraction of component i at the surface.i (ci .3.1.4) 143 . s − ci .3) can be rewritten as mi′′ = hm .3.2. s − ci .2.s = mi.i (ci .As defined earlier.m2. hm.3) Since the density of the component i at the interface.s . ρi.2.3.m3]. can be presented as ρi. s = hm .3.e ) + ρi .i (ci .
s ⎞ m′′ = hm .2. it may now differ considerably from the value for lowrate transport.s ⎠ (5. s − mi′′/ m′′ ⎠ (5.e − ci . s ) ⇒ ⎛ ci .i (ci .3. m′′ . s m′′ (5.3.e − ci .3.2. as the product of the mass transfer coefficient and a ratio of concentrations.2. denoted as mi.4).3. which is inserted in equation (5.ρs vs is the definition of total mass flux [kg.i (ci .3.2.6) Equation (5. s ⎞ Bm .i ⎜ ⎟ ⎜ ⎟ ⎝ mi . Equation (5.i (ci . which is made up of component i. This ratio is called the mass transfer driving force for component i: ⎛ ci .5) may be rearranged as: mi . s m′′ = J i . s − mi′′/ m′′) = hm. m′′ .t [181]: mi . s m′′ − mi′′= hm. It is not really a mass fraction since it can have any value from −∞ to +∞ .8) The mass fraction in the transferred state is simply the fraction of the total mass flux. s − ci .e ) + mi .2.2.s1].5) Although the mass transfer coefficient is based on diffusive transfer from the water surface. depending on the relative magnitudes of 144 .3. mi′′ = hm .e − ci . s + mi .7) The ratio of mass fluxes in the denominator is called the mass fraction in the transferred state. m′′ . s ) ⇒ m′′ (mi .e − ci .t ≡ mi′′/ m′′ (5.3.i = ⎜ ⎜ m − m′′/ m′′ ⎟ ⎟ i ⎝ i.6) expresses the total mass flux of all components through the interface.2.m2.
s ⎜ m −1 ⎝ i. since the second component (i.3. Equation (5.3.i = ⎜ i .m′′ and mi′′s . then m′′ = mi′′ .3. water vapor) is transferred through the interface. gm. gm. water) under normal conditions.t = 1. If component i is the only one passing through the surface.e.3.i.i Bm.10) The effect of mass transfer rates on the mass transfer coefficient Mass transfer coefficient. only one component (i.2. The mass transfer driving force is simply ⎛c −c Bm .e. so that ci. mass transfer problems in terms of the mass transfer coefficient. Bm.e.2.e i . air) is not strongly absorbed into liquid (i.9) In many situations such as evaporation.i (5. They solved the partial differential 145 . and the driving force for mass transfer.7) provide a formulation of .i at high rates mass transfer problem is often found through applying solution of the momentum and component equations or through correlation of mass transfer data which are complicated approaches since the flow field and mass transfer rate are coupled [181].6) and equation (5. Lienhard IV and Lienhard V [181] isolated effect of m′′ on the mass transfer coefficient by approximating the boundary layer as a stagnant film and allowing for the possibility of the counterdiffusion.2.s ⎞ ⎟ ⎟ ⎠ (5. gm.2. so that both diffusion and vertical convection may occur.i: m′′ = hm.i can be obtained from the lowrate mass transfer coefficient by applying a correction for the effect of finite mass transfer rates.
vapor = Dvapor δd (5.3.3.2.vapor ⎤ is called the blowing factor (BF).7).i = Di ⎡ ln (1 + Bm .12) Equation (5. h∗m. s − mi′′ m′′ Di ⎞ ⎟ ⎟ ⎠ (5.i δd ⎢ ⎥ ⎣ ⎦ (5. the ⎣ ⎦ total mass flow rate of water vapor due to both diffusion and convective mass flux 146 .equation of component conservation in the molar form with appropriate molar boundary condition assuming the mass transfer as a onedimensional steady mass diffusion in stationary media without reactions.2.vapor ) ⎤ ⎥ hm .12) lead to: hm .vapor ⎢ ⎥ ⎣ ⎦ (5.3.vapor ≡ lim hm.3.13) can be rewritten as ⎡ ln (1 + Bm .i δd ⎢ ⎥ ⎣ ⎦ (5.e − ci . equation (5.2.11) is presented as: m′′ = Di ⎡ ln (1 + Bm .vapor = m′′→ 0 Bm .3.3.3.2.2.13) Now lowrate mass transfer coefficient of vapor can be obtained when m′′ approaches zero. s ln ⎜1 + ⎜ δd ⎝ mi .vapor → 0 lim hm. equation (5.vapor ⎢ Bm.15) Since the group of ⎡ln (1 + Bm.3.2.14) Therefore.11) Recalling the definition of mass transfer driving force from equation (5.3.i Bm .3.2.9) and equation (5.vapor ) Bm. The solution of the rate of mass transfer considering vertical convective transport in the massbased form is: m′′ = ⎛ ci .i ) ⎤ ⎢ ⎥ Bm.vapor = h∗m .2.2.i ) ⎤ ⎢ ⎥ Bm .2.
26) and equation (5.is calculated according to lowrate mass transfer coefficient.3.1. ρ.3. BF .1.3.3.vapor (5.3 Algorithm for determination of evaporation rate Considering the evaporation as a mass transfer problem.2. the following algorithm for determination of evaporation rate is suggested: 1) Calculate temperature of the film (boundary layer) as T film = Tc + Ta 2 2) Determine the required air properties at the film temperature (μ.vapor = Sh L Dvapor L 147 . Bm .27) 7) Calculate average lowrate mass transfer coefficient h∗m .16) 5. blowing factor and mass transfer driving force using ′′ mvapor = h∗m.vapor . and ν) [182] 3) Calculate binary diffusion coefficient for the water vapor in the air (Dvapor) by Chapman and Enskog formula [183] 4) Compute Schmidt number as: Sc = ν Dvapor 5) Compute Reynolds numbers as: Re = u∞ L ρ μ 6) Calculate average Sherwood number Sh L from equation (5.
m2h1]. As shown in the figures.4. in an ambient temperature and relative humidity of 40oC and 50%.3 [kg.vapor = ⎜ vapor .3 to 0.vapor .e vapor .72. Sh L .s1]. Increasing wind velocity.5 [m.vapor 5. Bm .s1] translate into a reduction of evaporation rate of a water surface with 0. increasing ambient temperature leads to increase in the rate of evaporation. ReL. the Reynolds number. falls from 0.0043 [m. which more significantly increases by increasing wind speed and lowering ambient relative humidity. from 244 to 55 and therefore.g. results in decreasing boundary layer thickness which leads to an increase in the evaporation rate.4. a decrease in the wind velocity from 10 to 0. e. as shown in Figure 5. decreases from 174622 to 8731 leading to a decrease in the average Sherwood number. the Schmidt number.vapor 10) Calculate total mass flux of water vapor due to evaporation according to ′′ mvapor = h∗m .. hm . relative humidity and wind velocity on the rate of evaporation were studied and shown in Figure 5. is constant and equal to 0.4.1ac. s ⎜ mvapor . Parametric Studies Effects of boundary layer temperature. in both cases.vapor ) Bm .⎛c −c 8) Compute mass transfer driving force Bm . Sc.0194 to 0. According to the proposed mass transfer model.3 m characteristic length from 1. However. BF . s − 1 ⎝ ⎞ ⎟ from the ⎟ ⎠ change of water vapor concentration near the surface and the mass fraction of vapor at surface 9) Compute blowing factor as BF = ln (1 + Bm .1a. the average mass transfer coefficient. Having the 148 .
respectively.0122 [m. For example the results indicate that the evaporation rate at 40oC and relative humidity of 50% decreases from 2.2 [kg. the mass concentration of water vapor at the air stream.s1 ] due to increasing characteristic length from 0.0313.0051 to 0. on the other hand.4. the Reynolds (ReL) and the Sherwood ( Sh L ) numbers are constant and equal to 0. assuming a wind velocity of 4 [m.1 to 1 m was shown in Figure 5.6 [kg. increases from 0.0514 and ce is 0. The size of sample in the direction of wind flow (characteristic length) also plays a significant role in the change of evaporation rates. the difference in the average mass transfer coefficient can be held accountable for the reduction of evaporation rate by more than four times.0 to 0. the evaporation rate at 40oC from a surface with 0. As shown in Figure 5.0514 [kg. the Schmidt (Sc).1 to 1 m. In these cases. Increasing the ambient relative humidity.4. the Schmidt 149 . In both cases.m2h1] when the relative humidity increases from 10% to 90%.s1].m3] due to increasing relative humidity which causes a 10 time reduction in the value of the evaporation rate.0 to 0.72. The reduction of evaporation rates at the relative humidity of 50% and the wind velocity of 4 [m. causes a drastic reduction in the rate of moisture evaporation.m3]. leading to a calculation of an average mass transfer coefficient ( hm ) of 0.m2h1] when the characteristic length increases from 0.1b.same values for ambient and interface moisture concentration in both cases as cs is 0.1c. ce.3 m characteristic length decreases from 2.s1]. cs is also constant and equal to saturated concentration of 0.0463 [kg. While the mass concentration of water vapor at the interface. 70632 and 155.
149] and one set of data from the authors tests were used to validate the model.0514 and ce is 0. and the Reynolds number. from 0. Model Comparison with Experimental Results of Concrete Evaporation Results of recent evaporation tests from several references [8. 150 .72 and 23544. 5. it was approximated with a very small wind velocity of 0. first assuming the evaporating surface has the same temperature as the air and second. hm . Sc.number. the difference in the average mass transfer coefficient leads into the reduction of evaporation rate by more than three times. ReL. 12. The characteristic length which is defined as the length of specimen in the direction of the air flow was not given in most cases and therefore. while the average Sherwood number. On the other hand.5. 19.5 m/s in those cases. Sh L . the rate of evaporation was calculated.0067 [m.s1].0211 to 0.0257. Since the evaporating surface temperature was not reported. since the model cannot accept the air flow velocity of 0. assuming surface cooling occurrence by the order of 34oC. are constant and equal to 0. When ambient and interface moisture concentrations are constant in both cases as cs is 0. an equivalent dimension which is the diameter of the equivalent circle to that area was used in calculations. increases from 90 to 284 resulting in a decrease of the average mass transfer coefficient.
3 m u∞ (m/s) = 10 8 6 4 2 = 4 m/s L = 0.50 1 0.4.4.1c) Figure 5. relative humidity.4.5 0 0 0 50 Temperature of Boundary Layer (oC) 10 20 30 40 0 10 20 30 40 o 50 Temperature of Boundary Layer ( C) (Figure 5.8 2 1 60 70 80 90 0.4.1. wind velocity and characteristic length on the rate of evaporation from water surfaces. kg/(m2h) ϕ = 0% Rate of Evaporation. kg/m2h 2 u∞ = 4 m/s ϕ = 50% 0.1b) L(m) = 0.2 3 Rate of Evaporation.00 0 0 10 20 30 o 40 50 Temperature ( C) (Figure 5. The environmental parameters and predicted evaporation rates.6 ϕ = 50% 1.3 m u∞ 10 20 30 40 50 0.25 0. kg/m2h 1.2 L = 0. Effect of temperature. as well as the parameters used for calculation of evaporation rates based on mass transfer and 151 .4 0.75 1.10 Rate of Evaporation.1a) 3 (Figure 5.
the mass transfer from the liquid by evaporation is exactly balanced by the heat supply from the gas and can be represented by a heat and mass transfer equilibrium [185]. i.1) where hc is the heat transfer coefficient by convection [W. Ta is the temperature of the air [oC]. According to Poole [184]. ′′ mvapor = hc (Ta − Twb ) / λ = K H (ϕ s − ϕ a ) ⇒ (ϕ s − ϕa ) − hc = (Ta − Twb ) K H λ (5. the assumption that concrete temperature represents the temperature at the evaporating surface is not strictly true because of the surface cooling occurrence during evaporation. 1.5. the other estimations are well within 030% of the experimental data. λ is the latent heat of evaporation at Twb [kJ. φs is the saturation humidity of water vapor. Twb is the wetbulb temperature of water [oC]. A dynamic equilibrium between the rate of heat supply and the rate of liquid removal can explain the phenomena of evaporative cooling. The surface temperature can be compared with the wetbulb temperature defined as the equilibrium temperature which liquid water attains when heat is transferred to it by contact with the air as an infinitely large quantity of gas.1 and 5.5.e.m2oC1].2. Except in two cases of wind velocity of zero.boundary layer theory are shown in Table 5. However.kg1].1. Taking psychrometry into consideration. KH is the drying coefficient per unit humidity difference [kg.s1. As shown in the Table 5. and φa is the humidity of the air.5.m2].5. evaporation rates estimated assuming surface cooling are very close to the experimental reported data. calculated evaporation rates assuming identical temperature for the air and evaporative surface overestimates the experimental results by 30300%. 152 .
5 [kg.m2h1]. The comparison between experimental data and the theoretical modeling is shown in Figure 5. 153 .5. Considering all results and analyses.5 [kg. the experimental data of Azenha et al. Finally. the experimental data are well predicted by the theory of boundary layer mass transfer.1.To verify the surface cooling.5. it may be concluded that the suggested model can well predict the initial rate of evaporation in the cementbased materials in the range of 00.2 for all 9 cases. This result is in agreement with the magnitude of assumed surface cooling of about 34oC. While the results of modeling for higher evaporation rates were also verified using the test data reported by Lura (2007). As shown in Figure 5.m2h1]. the input data of the evaporation tests by Azenha et al (2007) is compared with the data on usual psychrometric charts [186] and as a conclusion a wetbulb temperature of 16oC is obtained. more experimental data are required for further verification of the model in the evaporation rates higher than 0. (2007) have been used to verify the effect of wind velocity.
6 4.051x0.051x0.0688 20 0. u∞ (m/s) m′′ (kg/m2/h) m′′ m′′ (kg/m2/h) (kg/m2/h) Case1: 0.6 1.051x0.17 1.1.0 Azenha et 75 al (2007) Case 4: 0.61 0.26 0.1128 24 0.0688 20 0.28 0.5.5 0.20 154 .40 0.3x0.29 0.21 0.34 0. (2008) Case 8: Hammer (2001) Case 9: 0.43 0.1 0.24 0.φ () Wind Velocit y.3 d=0.19 0. Evaporation Rate assuming surface cooling.40 0.1 30 0.Table 5.0 Azenha et 74 al (2007) Case 3: 0. (2008) Case 7: Slowik et al.25 0.31 0.44 0.17 0.1128 20 0. (m) (m) Characteristic Length.3x0.5 0 0.34 0.08 0.3385 20 0.07 0.1 0.58 0.45 4..1x0.8 0. Rate assuming Tc=Ta.3 0.6 3.1 Authors 0.3385 22 0.1x0. Exp.2 0.50 6.8 0.23 0.8 1.32 0.3 4. Ta (oC) Relative Humidity .0688 20 0.54 0.0 Azenha et 76 al (2007) Case 5: Lura et al (2007) Case 6: Slowik et al.01 0.0 Azenha et 73 al (2007) Case 2: 0. Results of the evaporation rate in comparison with the experiments Evaporation Surface Dimensio Reference ns.8 0.3 0. L Air Temp.30 0.65 1. Rate of Evaporatio n.6 0 0.051x0.12 0.0688 20 0.
202 1.49E05 2.7315 3777 36 6.s1] saturated water vapor pressure.0051 0.7315 2304 28 8.7315 8293 54 1.s1] diffusivity.59E2 1825 1408 Case 3 291 1.04E05 0.0145 0. Sc Reynolds number.38E3 1825 1408 Case 2 291 1.0104 0.49E05 2.m3] mass transfer driving force.0137 0.212 1.s1] density of the air.m3] mass fraction of the vapor at the interface.196 1.0107 0.0113 0.18E2 2206 796 Case 8 291 1.0113 0. Tfilm [K] dynamic viscosity.212 1. Parameters used for calculation of evaporation rates Parameters boundary layer temperature.60E05 2.997 0.04E05 0.54E3 1825 1174 Case 9 295 1. psat [Pa] partial pressure of the vapor in air.997 0.0078 0.0033 0.87E05 1.995 0. D [m2.81E05 1.m3] kinematic viscosity. Bm [kg.0115 0.s1] Schmidt number.7E2 2 1825 1408 3796 2256 2 1825 1056 0.49E05 2.0137 0.166 1. ce [kg. BF Case1 291 1.0033 0.7315 19350 82 2.43E Case 5 303 1.0137 0.0087 0.5.81E05 1.10E05 0.0234 0.0161 0.49E05 2.81E05 1.m3] blowing factor.81E05 1. ms vapor concentration in the air stream.212 1.0104 0.0104 0. cs [kg. ReL avergae Sherwood number.Table 5.0136 0.7262 43762 123 Case 6 291 1.17E Case 7 294 1.0101 0.52E05 2.04E05 0. ρ [kg.7315 15204 73 2.212 1.m1.04E05 0.0074 0.0174 0.0273 0.15E2 1825 1408 Case 4 291 1.20E05 0.0113 0.0033 0.0113 0.0104 0.0137 0.53E05 2. Sh L average lowrate mass transfer coefficient. hm [m.998 0.0058 0.994 0.7303 107166 192 1.212 1. pv [Pa] vapor concentration at the interface.04E05 0.995 155 .212 1.7296 3686 36 6.04E05 0.998 0.0113 0.08E05 0. ν [m2.0033 0.7315 108803 194 1.81E05 1.0113 0.998 0.81E05 1.2. μ [kg.82E05 1.998 0.0137 0.0137 0.49E05 2.83E05 1.0060 0.0164 0.49E05 2.63E3 2347 1019 2.
2 Ta = 20oC Tc = 16oC L = 0.8 1.0.2. (2007a.6 Theoretical Rate of Evaporation.6 Azenha et al.4 Azenha et al. 2007a. 2007 Hammer 2001 Authors 1. kg/(m2h) Boundary Layer Model 0.5 Rate of Evaporation.8 0. The comparison between the experimental data and the theoretical modeling of initial rate of evaporation based on a boundary layer mass transfer theory 156 . 2007a. Prediction of experimental data of Azenha et al.b Slowik et al.1. 2008 Lura et al.1 ϕ = 60% 0 1 2 0 3 4 5 Wind Velocity (m/s) Figure 5.069 m 0.b) by a method based on a boundary layer mass transfer concept Experimental Rate of Evaporation. kg/(m2h) 1.4 0 0 0.5.2 0.4 0.5.b 0. kg/(m2h) Figure 5.3 0.2 1.
Conclusion Plastic shrinkage crack. the model is capable of predicting the evaporation rates with a good accuracy. mass transfer.6. The theory of water evaporation based on the boundary layer theory. 157 . which more significantly increases by increasing wind speed and lowering ambient relative humidity. are often observed in the first few hours after placing which is a time period well within the initial stage of drying. Results of parametric studies also show that the size of sample in the direction of wind flow (characteristic length) plays a significant role in the change of evaporation rates. During stage I drying. Results of parametric studies show that increasing ambient temperature leads to an increase in the rate of evaporation. diffusion. The results show that given the environmental parameters as inputs. The proposed method is verified by the recent experimental data from several researchers on cementbased materials. rate of evaporation from concrete surfaces is roughly constant and similar to the external potential evaporation rate from water surfaces which controls the drying process. and convection is described and a driving force of concentration gradient is introduced as the main cause of mass transfer. Mass transfer at low and high rates is discussed separately along with other related parameters such as mass transfer coefficient and a method for determination of evaporation rate is suggested.5. a threat to structural integrity of concrete structures.
APPENDIX A FRee Shrinkage Test Results
158
An experimental study was conducted to evaluate the effects of adding low content of ARglass fibers (5, and 10 lb/yd3) on the free shrinkage of cement pastes. The free shrinkage tests were performed according to ASTM standard C 15704. This test method covers the determination of the length changes that are produced by causes other than externally applied forces and temperature changes in hardened hydrauliccement mortar and concrete specimens made in the laboratory and exposed to controlled conditions of temperature and moisture. Measurement of length change permits assessment of the potential for volumetric contraction (shrinkage) of concrete due to drying. This test method is particularly useful for comparative evaluation of this potential in different hydrauliccement mortar or concrete mixtures. In this test, specimens of 1” [25mm] square crosssection and approximately 111⁄4” [285 mm] in length are casted for each mixture. The test specimens are cured in the molds covered with a plastic sheet for 24 hours while water dripping from specimen is protected. Upon removal of the specimens from the molds, the initial Length reading is taken using a digital comparator dial. After the initial comparator reading, the specimens stored in the drying room, so that the specimens have a clearance of at least 1 in. [25 mm] on all sides. Comparator readings of each specimen are taken every one or two days after initial reading. These readings were done in a room maintained at a relative humidity of 30% while the specimens are at a temperature of 73°F [23°C]. The comparator dial were read with both test specimen and the reference bar in the comparator and the difference between the two readings and initial length of specimens were used to calculate shrinkage strain. The effect of ARglass fibers on free shrinkage was investigated by ASTM standard C 15704. Three different cement paste mixtures were designed including 2 mixtures with ARglass fibers and one control mixture with only plain cement paste and no fiber. All mixtures were designed with the watercement ratio of 0.45 and a high slump (6 in.) such as slurry. The cement content was 2080 lb/yd3 (1450 Kg/m3) and only Portland cement, type I/II and water were used to cast the paste. The only difference between these mixtures, Control, ARG5.0, and ARG10 was due to adding 0, 5, and 10 lb/yd3 fibers respectively. In all ARG mixes, AntiCrak® HPglass fibers with 24 mm length were used. The mix proportions of sample are resented in Table A.1.
Table A.1.
Mixture proportions of the ARGs and Control samples (lb/ft3) Mix ID Portland Cement Water Glass Fiber (HPGlass) w/c Control 90 41 0.0 0.45 ARG5.0 90 41 0.19* 0.45 ARG10 90 41 0.37* 0.45
* The fiber content of 0.19, and 0.37 lb/ft3 is equivalent to 5, and 10 lb/yd3
159
Tow prism specimens of 1” [25mm] square crosssection and approximately 111⁄4” [285 mm] in length are cast for each mixture. For making each sample, Portland cement were added to the mixer and blended for 60 seconds. Then, water poured gradually to the mixer and blended for 5 minutes before adding the glass fibers only to the ARG mixtures and mixing for 5 additional minutes to thoroughly mix all the ingredients. After mixing concrete in a suitable laboratory mixer, the concrete is placed in the mold in two approximately equal layers and each layer compacted with the tamper. We worked with the concrete until a homogeneous specimen is obtained. After the top layer has been compacted, we strike off the mortar flush with the top of the mold, and smooth the surface with a few strokes of a trowel. The test specimens are cured in the molds covered with plastic sheet and then removed from the molds at an age of 24 hours after mixing. After length measurement at the end of the curing period using a digital comparator dial, the samples are stored in a drying room until next reading. The lengths of samples are read in 1 or 2 days intervals up to 28 days and the difference between the reading and initial reading (at 24h) used to calculate free shrinkage. The results of experimental tests are shown in Figures A.1. The results show that adding low content ARglass fibers (5 to 10lb/yd3) to the Portland cement paste reduces the free shrinkage by about 5% and 10%, respectively. Experimental data were analyzed using ACI 209R92 shrinkage model. This model is an empirical model developed by Branson and Christiason [187], with minor modifications introduced in ACI 209R82 [188]. ACI Committee 209 incorporated the developed model in ACI 209R92 [87]. The model for predicting shrinkage strains as a function of time have a simple principle: a hyperbolic curve that tends to an asymptotic value called the ultimate value. The shrinkage strain εsh(t,tc) at age of concrete t (d), measured from the start of drying at tc (d), is calculated by equation (A.1). (t − tc )α (A.1) ⋅ ε shu ε sh (t , tc ) = f + (t − tc )α where, f (in days) and α are considered constants for a given member shape and size that define the timeratio part, ε shu is the ultimate shrinkage strain, and (t – tc) is the time from the end of the initial curing. The shape and size effect can be totally considered on the timeratio by replacing α = 1.0, and f as given by equation (A.2) in inlbs units. 0.36 (V / S ) ] f = 26.0e [ (A.2) where, V/S is the volumesurface ratio in inches.
160
2500 2000 1500 1000 500 0
Control : Vf = 0 lb/yd w/c = 0.45
3
Free Shrinkage Strain (micro strain)
Free Shrinkage Strain (micro strain)
3000
3000 2500 2000 1500 1000 500 0
Sample 1 Sample 2 ARG5.0 : Vf = 5 lb/yd3 w/c = 0.45
Sample 1 Sample 2
0
10
20
30
0
10
20
30
Time (Days) Free Shrinkage Strain (micro strain) Free Shrinkage Strain (micro strain)
3000 2500 2000 1500 1000 500 0
Sample 1 Sample 2 ARG10 : Vf = 10 lb/yd3 w/c = 0.45
Time (Days)
3000 2500 2000 1500 1000 500 0
Control (1) Control (2) ARG5.0 (1) ARG5.0 (2) ARG10 (1) ARG10 (2)
w/c = 0.45
0
10
20
30
0
10
20
30
Time (Days)
Time (Days)
Figure A.1. Free shrinkage results of Control and ARG samples and their comparison
Hence, equation (A.1) is transformed to: (t − 1)1.5 (A.3) ε sh (t , tc ) = ⋅ ε shu 28 + (t − 1)1.5 ACI 209R92 suggests multiplying ε shu by seven factors, depending on particular conditions (A.4) ε shu = 780 γ sh × 10 6 in./in. with
γ sh = γ sh,tc γ sh,RH γ sh,vs γ sh,s γ sh,ψ γ sh,c γ sh,α
(A.5)
where, γ sh represents the cumulative product of the applicable correction factors. γ sh ,tc represents the initial moist curing coefficient for curing times different from 161
7 days for moistcured concrete. γ sh , RH is the ambient relative humidity coefficient while Coefficient γ sh, vs allows adjustment for the size of the member in terms of the volumesurface ratio. γ sh , s and γ sh , ψ are the slump the fine aggregate factors respectively. γ sh , c represents effect of cement content and γ sh , α is air content factor. In order to use ACI 209R92 shrinkage model to analyze our free shrinkage data, α is chosen as 1.5, tc equals to 1 day, and f is chosen as 28 days since V/S ratio is equal to 0.25 in. The values of γ sh ,tc , γ sh , RH , γ sh, vs , γ sh , s , γ sh, c ,
γ sh , α are considered as 1.2, 1.094, 1.164, 1.136, 1.498 and 0.99 respectively since
tc equals to 1 day, RH is 30%, V/S ratio is equal to 0.25 in, slump is 6 in. and air content is about 5%. γ sh,ψ referring to fine aggregate content factor is not applicable here since no aggregate was used in cement paste samples. With a good adjustment, the value of 1.5 is replaced γ sh , ψ factor and as shown in Figure A.2, the data show a good correlation with ACI model. ε shu for plain cement paste or Control samples was calculated as 3012 x 106 in/in.
Free Shrinkage Strain (micro strain)
3000 2500 2000 1500 1000 500 0
Control : Vf = 0 lb/yd3 w/c = 0.45
εshu = 3012 x 106 in./in.
Sample 1 Sample 2 ACI 209R92 Model
0
10
20
30
Time (Days)
Figure A.2. Modeling free shrinkage data of Control sample using ACI 209R92 model
To model ARG samples, a factor as γ f is presented to include in cumulated correction factors, γ sh in order to reduce ultimate shrinkage, ε shu and modify ACI model for fiber effect. Then the shrinkage formula turned out as;
ε shu = 780 (γ sh,tc γ sh,RH γ sh,vs γ sh,s γ sh,ψ γ sh,c γ sh,α γ f ) × 10 6 in./in.
(A.6)
162
The results show that ultimate free shrinkage of cement pastes can reduce by almost 4% and less than 10% when adding 5 and 10 lb/yrd3 ARglass fibers to the batch./in.45 3000 2500 2000 1500 1000 500 0 ARG10 : Vf = 10 lb/yd3 w/c = 0.3. Using these fiber correction factors ε shu is calculated as 2890 x 106 in/in for ARG5. our modified ACI model fitted experimental data appropriately.9 for paste samples with 5 lb/yrd3 and 10 lb/yrd3 respectively.96 and 0. Fitting experimental free shrinkage data of ARG samples with modified ACI 209R92 model Experimental and modeling free shrinkage data indicate that adding low dosage of ARglass fibers to concrete samples reduced the free shrinkage strains so slightly. Sample 1 Sample 2 ACI 209R92 Model εshu = 2710 x 106 in./in. 163 . The final conclusion is that free shrinkage parameters do not significantly differentiate the contribution of fiber. Sample 1 Sample 2 ACI 209R92 Model 0 10 20 30 0 10 20 30 Time (Days) Time (Days) Figure A.45 εshu = 2890 x 106 in. Free Shrinkage Strain (micro strain) Free Shrinkage Strain (micro strain) 3000 2500 2000 1500 1000 500 0 Vf = 5 lb/yd3 w/c = 0.Using γ f as 0. The results of fitting data with modified ACI model are shown in Figure A.3.0 and 2710 x 106 in/in for ARG10 samples.
APPENDIX B Vacuum Evaporation Test Method and Data Transfer Description 164 .
The author developed a new test method which is capable of characterizing evaporation parameters and simulating severe conditions to create 2D shrinkage cracks. Oil level and quality (oil has to be changed after 45 complete tests) 2. A schematic of test setup is shown in Figure B. This test method is based on the vacuum drying technique and test results can be used to analyze evaporation and cracking.1. 1. Initial pressure inside the desiccator (adjust it at 20 in Hg before starting the test) 5. Output file header (to be filled up in details and completed in GCTS software) Testing equipments Various components of the setup are described. Figure B. the following items have to be checked.1. 2D shrinkage cracking test setup based on vacuum technology Note that before running an evaporation test. Dry ice (1015 lbs of dry ice is required for a single 48 h test) 3. Equipped Vacuum Vessel 165 . Load cell (Offset the load cell before starting a test) 6. Mold (to make sure is ready and cleanedup) 4.
Regulator and Hose TBore stopcock was used to split and conduct vacuum flow of the vessel between condenser and pressure gage. As shown in the Figure B. Two Plexiglas plates were attached to both end of the load cell to make weight measurement as convenient as possible. The dry ice was fallen first inside the trap up to 1/3 of the height.2×30. TBore Stopcock. The load cell was calibrated by a precise digital scale (0. 166 . The calibration formula was derived and the obtained voltage gain was used for recording moisture loss. Then alcohol was added to the trap to decrease the atmosphere temperature to 70oF to condense the moisture. Plexiglas parts were used to make a 10×10×1.7×25.1b. A 1000mL glass flask was acquired as a trap to collect condensed moisture and prevent water vapor to flow through the pump. the lid has a valve to induce vacuum pressure. The vessel was equipped with a scale system. TBore stopcock. Pressure Gage.05g (0.5”) lowweight mold.0001 lb) was hired. 1700 Pa (0. Mold As shown in Figure 3. pressure gage and vacuum vessel are connected by high pressure hoses. an accurate load cell with capacity of 5440 g (12 lb) and accuracy of 0. Condenser vessel is a 15.A Pyrex vacuum vessel covered with a transparent Acrylic lid was used to maintain vacuum pressure.01g accuracy) using more than 30 different weights up to 1500 g. condenser system. All the components of the vacuum drying system including vacuum pump. There is another hole on the other side of lid used to pass load cell wires from inside of the vessel to the outside.5 cm (6”×12”) cylinder made up of double layer glass materials with a cylindrical hole of 12.4 cm (5”×10”) inside it. The pump is designed to remove air from a vacuum system at atmospheric pressure to a deep vacuum. The pressure gage was required to control the negative pressure inside the vessel and insure of ultimate pressure.3 cm (4”×4”×0. Double layer glass vessel used as heat insulator to prevent fast evaporation of the dry ice.017 atm) for a long period of time. This thin mold is proper to make samples with high specific surface area exposed to the external evaporation condition. two pairs of holes were drilled in each side of the mold and paper clips were passed through them and bended at the outer side by the right angle to stay fixed during the test.2. Vacuum Pump A beltdrive twostage pump was used as the vacuum equipment operated at its ultimate blankoff pressure. alcohol. dry ice and flask was used in this test method. In order to weigh samples continuously.1. In order to restrain the sample and cause earlyage 2D cracks. Condenser System A condenser system including a vessel.
In such a case. TBore stopcock was opened to let the negative pressure flow from vacuum valve at top of the vessel when the pump is on. a vacuum reaching 34 kPa (10” Hg) absolute was applied to each specimen for a short period of time.5” Hg) absolute. 1012 pounds of dry ice was hooked up and crushed into small chunks before placing inside the condenser vessel.7 kPa (0. Testing procedure After a number of preliminary trial tests. the first 30 minutes to one hour was assumed as an offset time supported 167 . Camera In order to analyze several characteristics of earlyage cracks developed by drying of the cement paste and clay samples. At the beginning of drying tests. Note that the recorded weight was the same as or within the accuracy range of the weighing system inside the vessel. The condenser vessel was filled up to 1/3 of the height with alcohol. the initial added dry ice may evaporate being exposed to the air in room temperature. Since the fresh paste is so fluid at the beginning of the test. the mold was prepared and weighed using an accurate digital scale (0. 30 min . a digital camera (Pulnix TM1325CL) was provided. After some hours. The weight was measured by continuous recording of the load cell and the camera was powered on to capture images at 15 minutes interval during the test. Hence. The cement paste was then made according to the mix design table and poured into the mold. A short period of time (i. The software is capable of getting calibration formula and read the data as weight in grams.e.Data Collection System The load cell was connected to a computer using a strain gage transducer amplifier and a simple computer interface unit. Total weight of the sample including weight of mold and cement paste was measured using the digital scale.1 h) is required to get stable data from load cell. Computer has an acquisition card and equipped with a testing software to read and collect test data. more dry ice is needed during the test to maintain its equilibrium with alcohol. Then the pressure decreased in the vessel to 1. The procedure was set to collect the data every single minute and stop collecting the data at the specified 24 or 48hrs.01g accuracy). Also the load cell placed inside the vacuum vessel shows erratic results at the very beginning of the test due to effect of vacuum pressure. Voids affect on crack characteristics of the sample as obstacles against crack propagation. First. The camera was set automatically to take pictures of the top surface of the sample every 15 min through end of the test. reaching ultimate blankoff pressure of the vacuum pump leads to creation of many permanent voids on the surface of the sample. the following procedure was established to run evaporation test using vacuum technique. The camera mounted 10 cm away from the top surface of vessel lid facing the center of the sample inside the vessel. The cement paste sample was then placed inside the vacuum vessel attached to the load cell.
by the fact that only a mass loss of about 1g was recorded during this period due to very low vacuum pressure.Open the GCTS package and click on the ‘Tests’. Data Collection and Transfer Procedure Step 1. 168 .Click on the ‘Universal’. Step 2.
and press Enter.Step 3. Step 4.Type the Project info and Sample description inside ‘TestID’.Click on the ‘TestID’ when in the Transducer column ‘Scale’ is chosen. 169 .
Then click on ‘F3’ to save the command.In case a new procedure with longer duration or higher frequency is required to be run. Step 6.Step 5.Click on the ‘Execute’ to start the data collection procedure. a command procedure has to be defined. You can load ‘MEHDI_1.CMD’ for test duration of 1440 min (24 h). Step 7. 170 .Click on the ‘Command’ and press on ‘F2’ to load a predefined command procedure. Click on the ‘Servo1’ and change the parameters to the desired values.
Click on ‘F5’ to start the data collection. 171 .Step 8. A name has to be typed for the name of the generated DAT file.
Step 9. Step 10.The DAT file has to be loaded by clicking on or typing the DAT file name and then pressing Enter.After the test is complete. 172 . click on the ‘Report’ to convert the format of the output file from DAT (only readable inside the GCTS package) to Text (ASCII) format.
Type in the file of the ASCII file and press ‘Enter’. Step 12.Choose ‘File’ and press Enter.Step 11. 173 .
Step 13.The data transfer is complete and you can find the ASCII file on: C:\gcts 174 .
APPENDIX C Finite Element Analysis of Evaporation Test Data 175 .
The top face is exposed to a lowpressure condition using a vacuum pump simulating two different stages of drying.. The problem is a 1D moisture transport from bottom end to the top face and therefore.0125 m.Evaporation Problem A thin concrete slab with the lateral surfaces insulated from moisture removal. The results are compared with the experimental data and the results of Abaqus analysis.2. and the initial moisture content. The bottom end of slab is impervious and insulated from moisture removal as well. τ = cn τ = gy + c = 0 τ = hy + c = hn yn τ = gy + c = 0 Figure C. Governing differential equations and boundary conditions of the problem 176 . During stage I. is initially saturated with the moisture concentration of 520 kg/m3. during stage II drying which starts after 11. the test simulates a roughly constant evaporation rate of 0. The forward difference method is applied to calculate several steps of the transient solution to the moisture diffusion problem. A code is written to find the critical step for stable solutions. Moisture loss at each time step is calculated as the difference between the integration of moisture concentration inside the domain.42 kg/(m2h) from the top face.1. L=0. However. D(t) and h2(t) are presented as a function of time according to optimization works on analytical solutions as shown in Figure C.5 hours..2. linear elements is used for this problem. a uniform mesh of n=1. evaporation can be observed as a convectiontype process with zero ambient moisture concentration.
s: mixed ⇒ τ = gy + c I. t ) ∂u ( x. t ) ⎤ − ⎢D =0 ∂t ∂x ⎣ ∂x ⎥ ⎦ ⇒ ρ ( x) = 0. Input diffusivity and moisture transfer coefficient as a function of time Solution General 1D timedependent problem: D. f ( x.E. t ) ∂ ⎡ ∂C ( x. h 40 Figure C. t ) ⎤ ρ ( x) + μ ( x) − ⎢α ( x) + β ( x)u ( x. t0 ) = u0 ( x) Moisture transport problem: D.0015 Moisture Diffusivity. t ) = f ( x.: ∂ 2 u ( x.C.001 4x107 D' 2x107 6x109 0 10 20 30 Drying Time. α ( x) = D Galerkin’s method of minimizing residuals: 177 Moisture Transfer Coefficient. m/h .002 K 0.s: u ( x. m2/s 6x10 7 0.E.C.: ∂C ( x. μ ( x) = 1. t ) = 0. t ) ∂ ⎡ ∂u ( x.2. t ) 2 ∂t ∂t ∂x ⎣ ∂x ⎥ ⎦ time − dependent terms 1D BVP B.0. β ( x) = 0.
Trial solution D. a)φi ( x) ]xn ⎜ ∫ φi ( x)φ j ( x)dx ⎟ ∑ 1 ∂x dx j =1 ⎝ Ω j =1 ⎝ Ω ⎠ dt ⎠ f n c k Forward difference method In forward difference method all quantities are evaluated at backward end of time. t ) ∂ ⎛ ∂C ( x. t ) ⎞ ⎤ − ⎜D ⎟ φi ( x)dx = 0 ∂t ∂x ⎝ ∂x ⎠ ⎥ ⎦ Ω ∫⎢ ⎣ i =1.. 2.. CLii = ∑C j =1 n ij .... n.: C ( x. a ) = ∑ a j (t )φ j ( x) j =1 n ⎡ ∂C ( x... 2. tn1 C{a}n −1 + K {a}n −1 = Fn −1 {a}n − {a}n −1 {a}n −1 = Δtn Moisture transport in forward difference method ⎛ 1 ⎞ ⎛ 1 ⎞ C ⎟ an = Fn −1 + ⎜ C − K ⎟ an −1 ⎜ ⎝ Δtn ⎠ ⎝ Δtn ⎠ Capacity matrix C can be diagonalized on the LHS. t .. t .E... n Integration by parts & substituting in residual equation dφ ( x) ∂C ∂C ⎡⎛ ∂C ⎞ ⎤ ∫ φi ( x) ∂t dx + Ω dx D ∂t dx = − ⎢⎜ − D ∂x ⎟ φi ( x) ⎥ ∫ i ⎠ ⎣⎝ ⎦ x1 Ω i =1. n n ∂C dφ ( j ) = ∑ a j (t ) ∂x j =1 ∂x n da ∂C = ∑ j φ ( j) ∂t j =1 dt xn ⎛ ⎞ da j n ⎛ dφi ( x) dφ j ⎞ x + ∑⎜ ∫ D dx ⎟ a j = − [τ ( x.. Cij = 0 when i ≠ j 178 . i = 1.. 2.
φ2 = L L φ1 φ2 kij = ∫ dφi ( x) dφ j D dx ∂x dx Ω ⎡ D L −D L⎤ ⇒K =⎢ ⎥ ⎣− D L D L ⎦ ⎡2 1⎤ 6 ⎡1/ 3 0 ⎤ −1 ⇒ C = L/6⎢ ⎥ ⇒ [CL ] = L ⎢ 0 1/ 3⎥ ⎣1 2⎦ ⎣ ⎦ ⎡0⎤ ⇒ FstageI = ⎢ ⎥ ⎣c2 ⎦ ⎡ 0 ⎤ ⇒ FstageII = ⎢ ⎥ but is modified by addition of h2 ⎣ h2 a2 ⎦ cij = ∫ φi ( x)φ j ( x)dx Ω fi = −[τφi ]x1n x term to [K] 179 .[CL ] 1 0 ⎡1/ CL11 ⎢ 0 1/ CL 22 =⎢ ⎢ 0 0 ⎢ 0 ⎣ 0 1 0 0 0 ⎤ 0 ⎥ ⎥ 0 ⎥ ⎥ 1/ CLnn ⎦ 0 an =an1+Δtn [CL ] ( Fn1 − Kan−1 ) Solution for 1 linear element problem linear interpolation: φ1 = x2 − x x − x1 .
12 E − 5 = ⎢ ⎥ +Δtn ⎜ ⎟ ⎢ ⎜−⎢ ⎥⎢ ⎥ ⎟ ⎢ ⎥ ⎥ ⎝ L ⎠ ⎣ 0 1/ 3⎦ ⎜ ⎣ −4.0125.12 E − 5 4.0125.12 E − 5 −4.12 E − 5 ⎥ ⎢ a ⎥ ⎟ ⎜ ⎟ ⎝ ⎠⎣ ⎦⎝⎣ 2 2⎦ ⎣ ⎦ ⎣ 2 ⎦ n1 ⎠ ⎣ 2 ⎦ n ⎣ 2 ⎦ n1 −4.12 E − 5 −4.12 E − 5 − h2 ⎦ ⎣ a2 ⎦ n1 ⎟ ⎣ a2 ⎦ n ⎣ a2 ⎦ n1 ⎝ ⎠ L = 0.12 E − 5⎤ ⎡ a1 ⎤ ⎞ ⎢ a ⎥ = ⎢ a ⎥ +Δtn ⎜ L ⎟ ⎢ 0 1/ 3⎥ ⎜ ⎢ h a ⎥ − ⎢ −4. h2 is given as a function of time Assembled matrices for n elements analysis 0 ⎡ D L −D L ⎤ ⎢− D L 2D L − D L ⎥ 0 ⎢ ⎥ ⎢ 0 ⎥ 0 − D L 2D L − D L K stageI = ⎢ ⎥ 0 0 ⎥ −D L −D L ⎢ ⎢ 0 − D L 2D L − D L ⎥ ⎢ ⎥ 0 −D L D L ⎦ ⎣ 0 ⎡ D L −D L ⎤ ⎢− D L 2D L − D L ⎥ 0 ⎢ ⎥ ⎢ 0 ⎥ − D L 2D L − D L 0 =⎢ ⎥ −D L −D L 0 0 ⎢ ⎥ ⎢ − D L 2D L −D L ⎥ 0 ⎢ ⎥ − D L D L − h2 ⎥ 0 ⎢ ⎣ ⎦ K stageII 180 .12 E − 5 4.12 E − 5 ⎤ ⎡ a1 ⎤ ⎞ ⎡ a1 ⎤ ⎡ a1 ⎤ ⎛ 6 ⎞ ⎡1/3 0 ⎤ ⎛ ⎡ 4.12 E − 5 4.42 / 3600 = 1. c2 = − 0.12 E − 5⎤ ⎡ a1 ⎤ ⎞ = ⎢ ⎥ +Δtn ⎜ ⎟ ⎢ ⎢a ⎥ ⎥⎜⎢ ⎥ − ⎢ ⎥⎢ ⎥ ⎟ ⎝ L ⎠ ⎣ 0 1/ 3⎦ ⎜ ⎣ c2 ⎦ ⎣ −4.667 E − 4 Matrix equation at stage II for one linear element ⎡ a1 ⎤ ⎡ a1 ⎤ ⎛ 6 ⎞ ⎡1/3 0 ⎤ ⎛ ⎡ 0 ⎤ ⎡ 4.Matrix equation at stage I for 1 linear element ⎡ a1 ⎤ ⎡ a1 ⎤ ⎛ 6 ⎞ ⎡1/3 0 ⎤ ⎛ ⎡ 0 ⎤ ⎡ 4.12 E − 5 ⎦ ⎣ a2 ⎦ n1 ⎟ ⎣ 2 ⎦ n ⎣ a2 ⎦ n1 ⎝ ⎠ L = 0.
Similarly at stage II drying. and for n=4. 17<Δtcrit<18. The critical time step can be found through trial and error starting by large numbers of time steps and then decreasing it to a number which gives the stable results with no oscillation. for n=1. the critical time step for number of element (n=2) was found to be between 37 and 38 s and therefore. 40<Δtcrit<41. the behavior of the solution as time t → ∞ is monitored. Similarly.3 to C. 151<Δtcrit<152. 16<Δtcrit<17. 9<Δtcrit<10. This formula estimates the critical time step for n=1 as: 181 . a forward difference method was used.E. an which are selected as representative values of moisture concentration through the space domain. At stage I for n=3. 9<Δtcrit<10. In this work. As shown in below Figure C. The monitored results are the moisture concentration on the bottom. FstageII ⎢ ⎥ ⎢ ⎥ ⎣c2 ⎦ the [K] Critical Time Steps 0 1/ 6 0 0 0 ⎤ ⎥ ⎥ ⎥ ⎥ 0 ⎥ 1/ 6 0 ⎥ ⎥ 0 1/ 3⎦ ⎡0 ⎤ ⎢ ⎥ = ⎢ ⎥ since convection terms have been already added to ⎢0 ⎥ ⎣ ⎦ To find out whether the solution technique is stable or not. If the behavior is oscillatory divergence. 35 s was chosen for the rest of analysis. This method is conditionally stable and the critical time step need to be found.3. The results are in accordance with the formula for critical time step in the forward difference method which is: Δtcrit ≅ 2 dπ 2 ( (μ / α )δ ) 2 which can be reduced to Δtcrit ≅ 2 ⎛ L2 ⎞ ⎜ ⎟ π2 ⎝ D ⎠ where. the critical time steps for stage I drying is 152<Δtcrit<153. for n=2. and δ is the distance between two adjacent nodes. Therefore Δt = 150 s was chosen for the rest of analysis. the solution technique is unstable. these critical time steps decrease by descretizing the space domain to more but smaller elements. d is the spatial dimension of the problem. and for n=4.6. As shown in Figures C.⎡1/ 3 0 ⎢ 0 1/ 6 ⎢ 0 6⎢ −1 [CL ] = ⎢ L⎢ ⎢ ⎢ ⎣ ⎡0⎤ ⎢0⎥ FstageI = ⎢ ⎥ . a1 and moisture concentration on top. for n=3. μ and α are from D.
h 8 12 100 0 4 Time.21s < 37 s π ⎝ 2 ⎠ 600 Moisture Concentration.12 E − 7 ⎠ and for number of elements n = 2. Finding the critical time step (152<Δtcrit<153) by trial and error for FE analysis with 1 element 600 Moisture Concentration.85s <152s π 2 ⎝ 4.4.12 E − 7 = 19. kg/m3 Stage I Drying Number of elements: n=1 Δt = 153 s 600 Stage I Drying Number of elements: n=1 Δt = 152 s 500 500 400 400 a1 an 300 a1 an 300 200 200 100 0 4 Time. h 8 12 100 0 4 Time. kg/m3 Stage I Drying Number of elements: n=2 Δt = 38 s 600 Stage I Drying Number of elements: n=2 Δt = 37 s a1 an 500 500 400 400 300 a1 an 300 200 200 100 0 4 Time.0125 ⎞ n = 2 ⇒ ( Δtcrit ) stageI ≅ 2 ⎜ ⎟ 4. kg/m3 Moisture Concentration. estimates the critical time step as: 2 2 ⎛ 0.n = 1 ⇒ ( Δtcrit ) stageI ≅ 2 ⎛ 0.01252 ⎞ ⎜ ⎟ = 76. kg/m3 Moisture Concentration. h 8 12 Figure C. Finding the critical time step (37<Δtcrit<38) by trial and error for FE analysis with 2 elements 182 .3. h 8 12 Figure C.
h 8 12 Figure C. kg/m3 Stage I Drying Number of elements: n=3 Δt = 17 s Stage I Drying Number of elements: n=3 Δt = 16 s a1 an 500 400 300 200 0 4 Time. h 8 12 100 0 4 Time. h 8 12 Figure C. Finding the critical time step (16<Δtcrit<17) by trial and error for FE analysis with 3 elements 2e+169 Moisture Concentration. kg/m3 Stage I Drying Number of elements: n=4 Δt = 9 s 1e+169 500 400 a1 an 0 a1 an 300 1e+169 200 2e+169 0 4 Time.6. h 8 12 100 0 4 Time.5.3E+015 Moisture Concentration. Finding the critical time step (9<Δtcrit<10) by trial and error for FE analysis with 4 elements 183 . kg/m3 Stage I Drying Number of elements: n=4 Δt = 10 s 600 Moisture Concentration. kg/m3 2E+015 1E+015 0 1E+015 2E+015 3E+015 a1 an 600 Moisture Concentration.
the final cumulative moisture loss are approaching to the asymptotic results from the upper side. To verify the results with the experiments.Sensitivity of the 1D Analysis to the Number of Elements The results of this FE analysis are intended to compare with the experimental data. 184 . the integration of moisture concentration over the space domain at each time step is compared with the initial moisture content to find the cumulative moisture loss. but the difference is so insignificant that it may attributed to the noise of the system. The results show that the prediction is within 2% of the experimental results.7. g 64.105 64. For the purpose of analyzing the sensitivity of the results to the number of elements. the moisture concentration through the thickness of the sample can be predicted by this method.115 Final Cumulative Moisture Loss.1 64.09 2 3 4 5 Number of Nodes 6 Figure C. 64.11 Linear 1D elements 64. The comparison between results of 1D FE analysis by the MATLAB code and the analytical results are shown in Figure C. As shown in Figure C. The FE analysis results show that there is not any significant difference between results of modeling with 1 or 2 elements. However. the cumulative moisture loss at the end of the test is used.9.8. the experimental moisture loss data are compared with the analytical results calculated as the difference between initial total moisture content and the integration of moisture concentration in the domain. The number of elements increases from n=1 to n=5 which translates into increase in the number of nodes from 2 to 6. Tracking the values of cumulative moisture loss at end of the test by increasing number of elements Results of 1D FE analysis with the MATLAB code As shown in Figure C. Slight difference between the results of analysis and experiments is observed at the beginning of the stage II drying which can be due to the transition from stage I to stage II and the nature of proposed different boundary conditions. Also analyzing with smaller time steps does not have any effect on the predicted results.095 64. the experimental data are measured as the cumulative moisture loss and not the direct measurement of moisture concentration. Therefore.7.
45 No curing T = 24oC 60 40 Experiment MATLAB 1D Analysis Analytical 20 0 0 10 20 30 Drying Time. The technique utilizes the analogy between the heat transfer and the moisture diffusion.45 No curing T = 24oC 400 n=1 Δt = 150 s n=1 Δt = 5 s n=2 Δt = 35 s 60 40 200 20 Experiment FE Analysis (n=1) Δt = 150 s FE Analysis (n=1) Δt = 5 s FE Analysis (n=2) Δt = 35 s 0 0 10 20 Time.6. h 40 Figure C. (a) Results of FE analysis with different elements and time steps. The governing nonlinear differential equation for the diffusional moisture transport in the domain of the problem can be written as: ⎡ ∂ 2C ∂ 2 C ⎤ ∂C = D(C ) ⎢ 2 + 2 ⎥ ∂t ∂y ⎦ ⎣ ∂x Moisture Concentration on Top Face. (b) Comparing the results with experimental data 80 Cumulative Moisture Loss. g w/c = 0.FE analysis by ABAQUS The finite element analysis is carried out using ABAQUS v. h 40 (a) (b) Figure C.101 [70].9. kg/m3 600 80 Cumulative Moisture Loss. g w/c = 0. h 30 40 0 0 10 20 30 Drying Time. Comparing results of 1D FE analysis by the MATLAB code with the analytical results 185 .8.
y.where C(x. While corresponding boundary conditions for heat transfer during stage I can be in the same form as diffusion problem (J=F0).y. Moisture diffusivity at the boundary/surface of the sample during stage I drying is in the form of J=F0 and during stage II drying is in the form of a convective boundary condition and can be written as: J = − D(C ) where ∂C is the moisture gradient at the drying surface with a unit normal “n”. and k is the isotropic thermal conductivity which may be a function of T. ∂n k is the convective moisture transfer coefficient.t) is the moisture concentration varying in the domain with time. the corresponding equations are given by: ∂T ⎛ k ⎞ ⎡ ∂ 2T ∂ 2T ⎤ = ⎜ ⎟⎢ + ⎥ ∂t ⎝ ρ c ⎠ ⎣ ∂x 2 ∂y 2 ⎦ whrere. Cement paste slab subjected to different boundary conditions during stag I and II drying In equivalent transient thermal problem in 2D space.10. T(x. ρ is the density and c is the specific heat. ∂C = − k (Cs − C* ) ∂n Figure C. the heat convection at the boundary for the stage II drying can be written as: 186 .t) is the temperature varying in the domain with time. and D(C) is the isotropic moisture diffusivity coefficient which is a function of C itself. Cs is the moisture concentration at the surface and C* is the ambient moisture concentration.
Figure C.∂T = h fT (T* − Ts ) ∂n where ∂T is the thermal gradient with a unit normal “n”. Comparing governing and boundary equations of heat transfer and diffusion problems lead us to the conclusion of onetoone analogy between these two boundary value problems under following correspondences: Table C. Analogy and corresponding terms between two boundary value problems Diffusion Problem Heat Transfer Problem C(x.12. a thin slices like the one shown in the Figure C. T* is the ambient temperature. The finite element domain is shown in Figure C. A representative thin slice cut perpendicular to top surface for modeling 1D moisture transport An implicit heat transfer finite element analysis is carried out. the concentration and flux will only changes according to the distance from the bottom/surface of sample.1. ∂n h fT is the convective transfer coefficient for thermal problem.11. In that slice.t) k/ρc hfT T* Ts Descriptions and details of the modeling Due to 1D flux.t) D(C) k C* Cs T(x. 187 .y.11 can be a representative of all other slices perpendicular to top surface. Ts is the temperature at the boundary.y.
This relationship is shown in Figure C.03. n = 18 C0 = 630 kg/m3 Cc = 120 kg/m3 Calculated from Experiment CEBFIP Model 0 200 400 600 Moisture Concentration.Figure C.1E007 4. Finite element domain Material Properties The material values used in the analysis are as follows: (a) The moisture diffusivity of the slab as a function of concentration. D(C) corresponds to k/ρc and the values of moisture diffusivity are introduced for conductivity. 188 .1E007 6. However.13 which is wellfitted by the CEBFIP model [154]. the values of density (ρ) and specific heat (c) should be taken as 1. the values of diffusivity are shown as a function of time. the values of diffusivity can be plotted versus moisture concentration.1E007 Moisture Diffusivity. These values can be introduced to Abaqus as isotropic concentrationdependent data in a tabular format. Steps Different steps defined for this specific problem are as follows. m2/s 5.2.13.86 x 108 m2/s α = D0/D1 = 0. in that figure. 7.1E007 1.0E008 D1 = 5. Considering an average moisture concentration at each time step. Since in the mentioned analogy. The input values come from the analytical works shown in Figure C.12.1E007 3. kg/m3 800 Figure C.1E007 1.1E007 2. Input values of diffusivity as a function of moisture concentration (b) The density and specific heat are 1.15 x 107 m2/s D0 = 1.
This load is assigned to step1 as the boundary condition for stage I. Convergence Studies on Different FEM Meshes 189 . (c) Step2: This step is defined to simulate stage II drying. the boundary condition of stage I drying (constant flux: J=F0) is defined as a load with the magnitude of 0.14. In this problem. a predefined field is created to be used as initial condition. The initial value of Ci = 520 kg/m3 is applied to the whole body as the initial condition.42 kg/(m2h).2 is inputted in the table for convective transfer coefficient and the sink temperature is taken 0 corresponding to ambient moisture concentration in the vacuum desiccator. The boundary condition of stage I drying defined as a load Interaction A single interaction is created for the purpose of imposing a convective boundary condition for the step2. The other boundary conditions which are impermeable surfaces on the sides and bottom of the sample are ignored.14. The interaction then assigned to step2 and the load boundary condition (constant flux for stage I) is inactivated for the step2. the moisture transfer coefficient shown in Figure C. Convective transfer coefficient and ambient temperature are introduced into ABAQUS through defining a “surface film condition” type interaction with timevariable film coefficient presented in a tabular format and ambient temperature as a “sink temperature”. (b) Step1: This step is defined to simulate stage I drying.(a) Initial step: In the initial step. since zero flux corresponds to zero load. these two steps have to be created separately. Loadings As shown in Figure C. The initial condition in this step is selected as “computed” which refers to the results computed in the previous step. This value comes from the experimental analysis. Since the top boundary conditions are different in the stage I from the stage II. Figure C.
Mesh4: 16 layers Figure C.15. Mesh5: 32 layers Figure C.15c.Figure C. Mesh6: 64 layers Figure C. Mesh3: 8 layers Figure C. Mesh2: 4 layers Figure C.15e.15d.15b.15f. Different mesh sizes used for convergence studies 190 . Mesh1: 2 layers Figure C.15a.
30401 0. g 0.1 1 10 100 0.4345 64. h 0. h 40 0 Evaporation Rate. Results of FE analysis with ABAQUS using Q4 elements comparing to the experimental results Model Q4Slab1 Q4Slab2 Q4Slab3 Q4Slab4 Q4Slab5 Q4Slab6 Experiment Element Number type of nodes DC2D4 DC2D4 DC2D4 DC2D4 DC2D4 DC2D4 63 105 189 357 693 1365  Final Number Initial flux Cumulative of (kg/(m2.2 Cumulative Moisture Loss.h) Experiment Mesh1 Mesh2 Mesh3 Mesh4 Mesh5 Mesh6 Figure C.3 0.1 80 60 40 20 0 0 10 20 30 Drying Time. kg/(m2.2. Convergence studies on the results of evaporation tests with Q4 elements Table C.4 0.39079 0.4349 63.Drying Time.40539 0.4335 64.4078 61.5 0.41269 0.9328 65 191 .h)) moisture elements loss (g) 40 80 160 320 640 1280 0.41635 0.2899 57.16.36166 0.42 49.
7196 63.17.1 80 60 40 20 0 0 10 20 30 Drying Time. Results of FE analysis with ABAQUS using Q8 elements comparing to the experimental results Element type Number of nodes Number of elements Initial flux (kg/(m2.41335 0.5142 64.41918 0.5 0.3 0. Convergence studies on the results of evaporation tests with Q8 elements (DC2D8: An 8node quadratic heat transfer quadrilateral) Table C.4 0.2 Cumulative Moisture Loss.5864 64.39298 0.h) Experiment Mesh1 Mesh2 Mesh3 Mesh4 Mesh5 Mesh6 Figure C.3162 65 192 .40662 0.2012 65.h)) Final Cumulative moisture loss (g) Model Q8Slab1 Q8Slab2 Q8Slab3 Q8Slab4 Q8Slab5 Q8Slab6 Experiment DC2D8 165 DC2D8 289 DC2D8 537 DC2D8 1033 DC2D8 2025 DC2D8 4009  40 80 160 320 640 1280  0.41669 0.9723 65.41835 0.1 1 10 100 0. kg/(m2.3. h 0. g 0.Drying Time.42 61. h 40 0 Evaporation Rate.
0.44 0. this number of element ensures the convergence of the results as well.h) . As shown in previous section. the domain is divided by 20 layers.36 Final Cumulative Moisture Loss. g 0. Since there is no moisture transfer along the xdirection (perpendicular to the direction of external flux). Since the shape of domain is very simple and basic. kg/(m2.18.4 0. a 4node linear heat transfer quadrilateral element (DC2D4) is used in this analysis.19 in different times ranging from 2 to 40 h. The type of analysis is an implicit heat transfer analysis.28 Figure C.32 68 64 60 56 52 48 0 500 1000 1500 2000 Number of Nodes 2500 Q4 Q8 0. 193 Initial Evaporation Rate. Final cumulative moisture loss and initial evaporation rates versus number of nodes Comparing FE Analysis and the Experimental Results Mesh and Element types For the purpose of comparison with the analytical analysis which was done in 100 layers. the same number of layers is used for meshing along ydirection (along the direction of external flux). (a) The results of FE analysis regarding moisture concentration at nodes are shown in Figure C.
e) t=40 h 194 . c) t=14 h. d) t=20 h.19. b) t=10 h.(a) t = 2 h (b) t = 10 h (c) t = 14 h (d) t = 20 h (e) t = 40 h Figure C. Distribution of moisture concentration (kg/m3) at: a) t=2 h.
e) t=40 h (c) The results of FE analysis regarding normalized moisture loss can be obtained by the integration of flux at the surface over the time. b) t=10 h. The total moisture loss 195 .20. (a) t = 2 h (b) t = 10 h (c) t = 14 h (d) t = 20 h (e) t = 40 h Figure C. d) t=20 h.(b) The results of FE analysis regarding the flux at integration points are shown in Figure C. Distribution of moisture flux (kg/(m2h)) at: a) t=2 h.20 at different times ranging from 2 to 40 h. c) t=14 h.
9 and C. The examples are: • Analysis of drying of internally cured concrete materials by presoaked lightweight aggregate. This result along with the results of experimental analysis. such as a concrete samples with oblique surfaces. some amounts of moisture is consumed by the hydration process which can be modeled as an internal volume moisture sink. Analysis of drying of concrete element with complex shape. In this case..21 and C.21).22. It seems that the proposed 1D FE method and Abaqus analysis can be used to analyze the drying of cementbased materials in conditions where no analytical solution is available. the results of 1D FE analysis are very close to the ABAQUS results.however. • Analysis of the slow process of drying in cementbased materials consideration the aging factor. analytical works and 1D FE analysis (MATLAB code) is shown in Figure C. Comparing results of FE analysis by ABAQUS with the analytical results As shown in the Figure C..45 No curing T = 24oC 60 40 20 Experiment Analytical Abaqus Results 0 0 10 20 30 Drying Time.22. 80 Cumulative Moisture Loss. g w/c = 0. can be calculated as the product of normalized moisture loss and the surface area. h 40 Figure C. In this case the lightweight aggregate acts as an internal volume moisture source. • • 196 . etc.21. Analysis of 2D/3D moisture transport through concrete elements. Both results along with analytical results are within 2% of the experimental data (Figure C.
5.clc. % surface are (m2) F0 = 0. % time of end of the test ini_content_I = 530.inputs % file and version fname = 'n1_delta_t150s. % element numbers and size n =1.80 Cumulative Moisture Loss. h 40 Figure C.clear all. L = 0. digit = 6. % transition time from stage I to II t_final = 40. g w/c = 0. colWidth = 14.15e7. % test constants Area = 0. Comparing results of analysis by ABAQUS with the experimental results and results of 1D analysis by the MATLAB code Matlab Code close all.0125/n.01. % in kg/m3 % input D_II and h2 as a function of time (from analytical works)  % result file % format column width % format digit used % program version 197 . % evaporation potential at stage I (kg/m2/s) D_I = 5.22.45 No curing T = 24oC 60 40 20 Experiment MATLAB 1D Analysis Abaqus Results 0 0 10 20 30 Drying Time.dat'.42/3600. %. progVer = 'Forward_8 04/10/11'. % diffusion coefficient (m2/s) t_trn = 10.
18528E08].79444E08.0. 1.001137.35556E08.4.46389E07.001137. % time steps for stageI (s) delta_t_I=150.0.0.001137.001215..1..47222E08. 2. 1.001137.8. 0..001727.92222E07.001137.4.001727.001137.0.001137.93056E08.001137...0..001195.3.0.001727.0.001137.5.76139E08..1.3.001166.1.0... . %n=1:152 %n=2:37 %n=3:16 %n=4:9 %n=5:6 %n=6:4 %n=8:2 198 .0.0.. D_II_time =[5.000000031.((t_finalt_trn)*3600)..0.47306E08.11417E07.7.001178.001648.2.001727.001137]/3600.46639E07. h2_time=[0.99778E07. 0.875E08.86389E08.1.0.60).84722E08.001339.001142..86389E07.0.001137.001137.001727..001727.001137.0.001137.40472E08.4.001238.5.001137.0.0.0.001727. % Analysis of stage I % initial condition for stageI a_I = ones([(n+1) 1])*ini_content_I.0.001137.0. 1.0.4. ..001137..20861E08.001162.001727.27222E07.85278E08.11389E08....0..7.17917E07..1. 3. .001137. 3.0..001727.0.08694E08..0.3.1.0.0...30556E08.. 4. 0.2.0..0..001137.9.001727.0.0.0.77778E08. 0...0.57278E08.001137.00146.37528E07.001137..33694E08.1..001137.2.0.0.1.34167E07.001162.61694E08.001392.001264.01389E08.1.46111E08.0.001137.001137..0.0.7.675E08..001137.87778E07.15833E08. 0.001137. 6.26389E08. 2.001161.97222E08.66444E08..29972E07.4.66389E07. 0.54083E07.001137.0.2.0.2.0.32722E08.2. 1.3.0.0.2.1.3.28056E08.0..42222E08.76167E07.t_DII = linspace(0... 8.76944E08.1.001727. 0.001297..88889E08.09167E07.2..0.000000261.001162..0. .. . .0.4.001727.74417E07.0.001137.3..1.9.48639E08.60833E08.25444E08.6.0..2.31111E08.14722E07.5..0.4.001727.07611E07..0...001727.1.001545..10278E07.59444E07.0.0.3...0.62778E08. 0..001152.
kg/m3') 199 . % store round numbers in an array a_I = a_I+delta_t_I*CLm1*(F_IK_I*a_I). end % plot the results of {a}n vs.i1)=D_I/L.a_I_store(n+1.plot(T_M. % convert rounds to actual time figure(1).. end end K_I(1.. % K matrix includes the convection coefficient in addition to alpha terms % at stage I there is no convection and therefore it just has alpha terms K_I = zeros([(n+1) (n+1)]). % number of rounds for time domain of stageI rounds_I = floor(t_trn*3600/delta_t_I).M). h2_I_t(m)=h2_time(1).n+1)=D_I/L.'r'.'a_1').i) = 6/L*1/6. if i1>0 K_I(i..e. tau1=g1y1+c1 & tau2=h2y2+c2). F_I(n+1) = F0. % internal force is zero. for m = 1:rounds_I M(m)=m.n+1)=6/L*1/3. CLm1(n+1. F_i refers to [c1. for i = 1:n+1 K_I(i. h') ylabel('moisture content. end CLm1(1. however because of the mixed boundary condition % (i.% extend the {a}n matrix D_I_t(m)=D_I.cn] % at stage I [c1. a_I_store = [a_I_store a_I].i) =2*D_I/L.1)=D_I/L.. time T_M=M*delta_t_I/3600. title('StageI: Moisture content on top and bottom of sample'). xlabel('Time. end if i+1<n+2 K_I(i.. for i = 1:n+1 CLm1(i.cn1] are zeros but cn is equal to outside flux F0 F_I =zeros ([(n+1) 1])..1)=6/L*1/3. K_I(n+1.M).% inverse of lumping capacity matrix (CL1) CLm1 = zeros([(n+1) (n+1)]).i+1)=D_I/L.T_M.'b') legend('a_n'. % calculate {a}n based on {a}n1 according to forward difference a_I_store = []..a_I_store(1.
(q*delta_t_II).plot(T_M. M_loss_I(end)).090 %n=5:44.4f\n'.03)/(1+((1mean(a_II)/ini_content_I)/(1120/ini_content_I))^18))..093 %n=6:44.098 % Analysis of stage II % initial condition for stageI a_II = a_I_store(:.'r') title('stageI: Cumulative Moisture Loss'). % diffusion m2/s h2 = interp1(t_DII.03+(1. 'moisture loss at the end of stage I (g): %5. %n=2:44.(q*delta_t_II). xlabel('rounds') ylabel('Cumulative Moisture Loss. g') fprintf(1..15e7*(0.end).0953 %%n=8:44. rounds figure(2). %n=1:43.'spline'). for q = 1:rounds_II D_II = interp1(t_DII. %n=3:44. legend('Stage I:CML'). % time steps for stageII (s) delta_t_II=150.D_II_time.h2_time.cn] are also zeros F_II =zeros ([(n+1) 1]).i))*L/2*Area*1000. % convection coefficient m/s % K matrix includes the convection coefficient in addition to alpha terms % at stage II there is a convection term as "h2" which is added to the last 200 .Integ_a_I(i)..i)a_I_store(end.030. end % plot the CML vs. % number of rounds for time domain of stageII rounds_II = floor((t_finalt_trn)*3600/delta_t_II). M_loss_I (i) = ini_content_I*(n*L)*Area*1000 .% cumulative moisture loss sum_step_I = sum(a_I_store).M_loss_I(M)..072 %n=4:44. and [c1.801. %n=1:160 %n=2:50 %n=3:18 %n=4:9 %n=5:6 %n=6:4 %n=8:2 % internal force is zero. for i = 1:rounds_I Integ_a_I(i) = (sum_step_I(i)*2a_I_store(1.'spline'). % calculate {a}n based on {a}n1 according to forward difference a_II_store = []. % diffusivity m2/s %D_II = 5.
xlabel('Time. kg/m3') % cumulative moisture loss sum_step_II = sum(a_II_store). % combine two stages  201 .'r'.l))*L/2*Area*1000.'b') title('StageII: Moisture content on top and bottom of sample'). h') ylabel('moisture content.j1)=D_II/L. end % plot the CML vs. end if j+1<n+2 K_II(j.plot(T_Q.j+1)=D_II/L. g') fprintf(1.plot(T_Q.4f\n'.Q).'r') title('satge II: Cumulative Moisture Loss'). for l = 1:rounds_II Integ_a_II(l) = (sum_step_II(l)*2a_II_store(1. time T_Q=Q*delta_t_II/3600+T_M(end). % store round numbers in an array a_II = a_II+delta_t_II*CLm1*(F_IIK_II*a_II).l)a_II_store(end. % convert rounds to actual time figure(3). 'moisture loss at the end of stage II (g): M_loss_II(end)) %n=1: %5. xlabel('Time. legend('a_n'.T_Q. rounds figure(4).M_loss_II(Q).% term in the K matrix K_II = zeros([(n+1) (n+1)]).j) =2*D_II/L. end end K_II(1. end % plot the results of {a}n vs.1)=D_II/L.Q). for j = 1:n+1 K_II(j. h2_II_t(q)=h2. if j1>0 K_II(j. Q(q)=q. legend('Stage II: CML').a_II_store(n+1. h') ylabel('Cumulative Moisture Loss. M_loss_II (l) = ini_content_I*(n*L)*Area*1000 Integ_a_II(l).n+1)=D_II/Lh2.a_II_store(1. a_II_store = [a_II_store a_II]. D_II_t(q)=D_II. K_II(n+1.'a_1').
MQ = [M Q+M(end)]. % (1. % plot the input D vs.2) print heading of result file textHeading = {'n'.plot output results% (1. M_loss = [M_loss_I M_loss_II]. h') ylabel('moisture content.L.F0. g') % .'F0'.h2_II_t(Q). xlabel('Time.' program version : %30s\n\n'. time figure(6).'r') title('Input moisture diffusivity as a function of time'). xlabel('Time. h') ylabel('convective moisture transfer coefficinet.M_loss(MQ).'a_1').'b') title('Input convective moisture transfer coefficient as a function of time'). legend('Cumulative Moisture Loss').a_store(1.'L'. rounds figure(8). D_t = [D_I_t D_II_t].1) print input and output response fid1 = fopen(fname. plot(T_Q. h') ylabel('moisture diffusivity.'r'. time figure(5).MQ). % (1. fprintf (fid1.'t_trn'.'Area'. kg/m3') % plot the CML vs. time figure(7).'b') title('Cumulative Moisture Loss During the Test'). fprintf (fid1. h2_total = [h2_I_t h2_II_t].a_store =[a_I_store a_II_store].Area.valHeading.T_MQ.colWidth. xlabel('Time. legend('a_n'. legend('h2(t)').3) print table of result file 202 . valHeading = [n.'w'). m2/s') % plot the input h2 vs.MQ).'********************************************************** ********************************************\n\n'). T_MQ = [T_M T_Q]. plot(T_MQ.textHeading.a_store(n+1.t_trn. m/s') % plot the results of {a}n vs.progVer). plot(T_MQ. legend('D(t)').plot(T_MQ.D_t(MQ). PrintHeading(fid1.'b') title('Moisture content on top and bottom of sample').digit). xlabel('Time.t_final].'tfinal'}. h') ylabel('Cumulative Moisture Loss.
a_store(1. PrintTable(fid1.textData = {'Time'.NJ.M_loss}.'moist_loss'} .tableResult.'D_t'.digit).:).'top_moisture'. NJ = NJtableResult{1}. data = {T_MQ. fclose all.'h2_t'.h2_total.textData.colWidth.D_t. tableResult = NJtableResult{2}. NJtableResult = StoreTable(data). 203 .'bott_moisture'.a_store(n+1.:).
APPENDIX D Drying Tests on Clays: An Application of the Developed Test Method 204 .
the clay used in this study has Liquid Limit (LL) of 38. 205 .60 Portland cement 1450 1450 1450 Water 650 650 650 watertosolid ratio 0.60 samples were 74 and 77 g.Results of LowPressure Drying Experiments on Clay Wet clay has different microstructure than Portland cement paste including different particle size and distribution.45 C0. two oversaturated clay samples were prepared and subjected to the vacuum drying. The clay was then mixed with water with watertosolid ratios of 0.73.1.6 and labeled by C0. Casagrade's plasticity chart distinguishes the mineral as Montmorillonite clay.60 As shown in Figure D. To evaluate the efficiency and applicability of the test method in capturing transport properties and evaporation characteristics of geological materials.45 and 0. and has no hydration characteristics to change the microstructure by the time. The mix design is shown in Table D.60 samples compared to Portland cement paste P0.45 and C0. Mix proportions of the test series (kg/m3) Test series P0.45 0.1.45 0. This shows that 95% and 82% of the moisture initially added to the clay samples was evaporated at the end of the tests which confirms the capability of the test method in simulating the drying process. Table D. and Plastic Index (PI) of 16. 30 before placing in an oven maintaining a uniform temperature of 110°C for 24 h.45 was 66 g while the cumulative moisture loss of C0.1.45 C0. The clay was removed after the material has been dried to a constant mass according to ASTM D 221605.43. after 48h drying the cumulative moisture loss of cement paste sample P0. According to Attenberg test method.16. The clay was passed through sieve No.45. Plastic Limit (PL) of 21.45 and C0.
h Drying Time. Results of lowpressure evaporation tests on clay samples and their comparison with plain cement paste sample: (a) cumulative moisture loss curves. (b) drying rate curves The calculated evaporation rates during drying test are shown in Figure D.45 Drying Rate.42 kg/(m2h) for the plain cement paste sample (P0.80 70 0. It may be related to the mass transfer at the boundary of the material which appears to be taken place in a thin film inside porous medium under the vacuum condition.h) 0. This observation indicates that similar to cement pastes the drying of clay under constant lowpressure conditions occurs in two distinct stages of a constant drying rate period (stage I). During stage I drying. the thickness of this transition layer which controls the intensity of the mass flux is regulated by the physical properties and microstructure of the material [189]. respectively.60 C0. During the drying process. This transition layer is defined as the location where the pressure drops from the saturated water vapor pressure to the pressure of the vacuum chamber [189].1b.28 kg/(m2h).8 P0. Comparing these rates with 0.45) indicates that clay decreases the initial evaporation rates during early stage of drying. the material is unable to support the necessary flux to the surface and rate of evaporation starts to fall. and determines the mass transfer at the boundary of the material.6 0. The average calculated drying rate at initial stages of drying for clay samples with w/s ratio of 0. h (a) (b) Figure D.2 0 0. g 60 50 40 30 20 10 0 0 10 20 30 40 50 P0. the removal of moisture is mainly controlled by diffusion which dominates the capillary forces [106]. During stage II drying.45 and 0. Kg/(m2. and a falling drying rate period (stage II).45 Moisture Loss. the difference in the physical properties and microstructure of clay and Portland cement could be the reason of different initial evaporation rates. This pressure drop induces a gas flux between porous medium and the vacuum chamber.1.6 were 0. Reaching a critical value of moisture content at the surface.45 C0.4 0.1 1 10 100 Drying Time. The transition time from stage I to stage II 206 .25 kg/(m2h) and 0. the main phase transitions occur at the boundary surface and vapor phase diffusion into the air is determining rate of evaporation. Therefore.45 C0.60 C0.
a boundary condition of constant flux is applied followed by the boundary condition of constant concentration.60 respectively.7 h for cement paste to 26.45 and 0.2. assuming a constant diffusivity is given by Eq. 108].drying was significantly altered from 9. first. As shown in Figure D. two different boundary conditions at the top surface are required. Figure D. (1) [109]. Boundary conditions and Fick’s 2nd law of diffusion for the presented drying problem The analytical solution for the initial and boundary conditions of stage I drying. that is set to ambient moisture concentration (i. e.1 h for clays with w/s ratios of 0. The moisture transfer inside the matrix is simulated by Fick’s second law of diffusion which governs a majority of moisture transfer problems in porous media [58.8 h and 24. zero in this case). Simplified analysis of drying experiments on clay In order to simulate the results of drying tests using a dualstage drying concept. 207 .1.
However. Ci is the initial moisture concentration (kg/m3). Using this solution. the diffusivities of clay samples C0. moisture loss during the advanced phase of drying is given by [109]: ∞ Mt 8 exp{ − DII (2n + 1) 2 π 2t / 4 L2 } = 1− ∑ M∞ (2n + 1) 2 π 2 n =0 (D. Therefore. diffusion coefficients in Stage II drying depend on moisture concentration [110].3. D is the moisture (liquid and vapor) diffusion coefficient (m2/s) and t is the time (s). Since stage I drying ends when the surface reaches equilibrium moisture concentration (i.60 samples were higher by 20% and 140% compared to cement paste. moisture diffusivity at stage I drying is calculated as long as the time of transition from stage I to stage II has not lapsed.C ( t . analytical solution can be integrated over the thickness of sample. Unlike the stage I drying. which is averaged over space and time.1x107 m2/s which are 4 times smaller than the initial diffusivity of cement paste.3) As shown in Table D.45 and C0.45 and C0. x ) = Ci + F0 L ⎧ DI t 3 x 2 − L2 2 ∞ (−1) n nπ x ⎫ − D n 2π 2t ⎪ )cos − 2 ∑ 2 exp( I 2 ⎨ 2 + ⎬ 2 DI ⎩ L L ⎪ 6L π n =1 n L ⎭ (D. the diffusivities at stage II drying for C0. e. DII (m2/s). can be derived by DII = −4 L2 π 2 ⋅ aln when ⎛ M ⎞ 8 ln ⎜ 1 − t ⎟ − ln 2 = aln t π ⎝ M∞ ⎠ (D. 208 . M ∞ (kg). however.2) Considering only first term of the series.60 at the stage I drying were 1. can be obtained from the corresponding quantity after infinite time. While the difference between the results of clays and cement pastes may be related to their physical and microstructural differences. the difference between the results of two clay samples may be discussed by different amount of moisture concentrations. and L is the thickness of sample (m). For stage II drying. moisture diffusivity.23 and Figure D. Also. moisture concentration at the top surface is obtained providing constant F0 and DI values. Mt (kg). The total amount of diffusing moisture in stage II drying.1) where C is the moisture concentration (kg/m3). in this analysis diffusivity in stage II is taken constant resulting in calculation of an average diffusivity.4x107 and 1. zero). F0 is the constant flux in stage I drying (kg/(m2s)).
4E07 1.3.28 0.h)) Cumulative moisture loss at 48 h (g) Moisture diffusivity at stage I (m2/s) Moisture diffusivity at stage II (m2/s) P0.8 24. which is accounted for all processes namely. Darcian transport of liquid and vapor water.60 C0.176 Evaporation rate at 24 h (kg/(m2.h)) Initial Test series evaporation rate (kg/(m2. cumulative moisture loss. transition time.45 1E008 6E007 8E009 Stage I 4E007 Stage II 6E009 4E009 2E007 2E009 0 0 Figure D. The main mechanisms of drying are identified accordingly in two separate stages of drying based on 209 Diffusivity.25 66 74 77 5.60 0. m2/s .1E07 1.7 26.2.024 0. and diffusivities of tested samples Transitio n time of drying stages (h) 9.3E09 8. Results of diffusion analysis on wet clay and cement paste samples Advanced Moisture Transport Analysis on the Drying Clay The advanced modeling of drying of cementitious materials as porous media is presented in the Appendix.45 C0. Initial evaporation rates. m2/s Diffusivity.167 0.Table D.0E09 8E007 P0.45 C0.42 0.0E09 4.45 C0.1 0. evaporation rate at 24 h.1E07 3. It first provides an overview of existing theories for the drying of porous media using a single nonlinear diffusion equation. and diffusion of vapor in dry air.
g 0. Results of analysis are shown in Figure D. Then a quantitative model of drying is proposed.6 and Figure D. The cracking area versus time is shown in Figure D.1 80 60 40 20 0 0 10 Drying of Clay w/s = 0. and moisture distribution as well as moisture diffusivity and its variations with time. These results are in good agreement with the results of Ketelaars [ 190] who reported diffusivity values in the range of 1x107 to 1x109 m2/s for different clay samples. h 0. h 40 (Figure D.34x108 m2/s.4 0.connectivity of liquid and vapor phases inside the pores. It is shown that the developed model can well predict the experimental data of the drying of cementbased materials. Results of analysis show that the diffusion coefficients vary in the range of 2. in which each mechanism is accounted for by appropriate equations. The same analytical method is applied to clay samples.h) . Image Analysis on EarlyAge Cracked Samples Timelapse photography shows that visible earlyage cracks were detected in the clay samples sample as early as 6 h and it developed throughout the test. kg/(m2.7 shows a cracked clay sample after 48 h of vacuum drying.5 0.4 which presents a fairly well prediction of the drying data by the variation of diffusivity and slight variation of moisture transfer coefficient.2x107 to 1.2 Cumulative Moisture Loss.45 0 20 30 Drying Time.4(a)) 210 Evaporation Rate.3 0. observed qualitatively through experiments.1 1 10 Experiment Simulation 100 0. Drying Time.
4.5. h D' 0. m/h 13 h . (b) Backcalculated D and k for the best fit of data drying time: 6 h 7h 8h 9h drying time: 10 h 11 h 12 h Figure D.0. (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the experimental data of an evaporation test. 2D Crack development during drying of a clay sample 211 Moisture Transfer Coefficient.002 K 0.4(b)) Figure D. m2/s 3x107 2x107 1x107 1x108 0 10 20 30 Drying Time.0015 Moisture Diffusivity.001 40 (Figure D.
7.6. Image of a 2D crack pattern in a clay sample after vacuum drying for 48 h 212 . in2 2 1 0 0 4 8 Drying Time.45) Crack Area.45) Cement Paste (w/c=0. h 12 16 Figure D. time with a Portland cement paste Figure D.3 Clay (w/s=0. Comparing cracking area of a clay sample vs.
APPENDIX E Image Analysis on Earlyage Cracks: Quantification and Orientation 213 .
Total number of cracks. NA (1/mm2): NA is the total number of features (cracks) divided by the area of image. 2. these parameters are introduced and then their values for different samples using an automatic procedure developed by a computer code are compared. Ntotal (): Ntotal is the total number of objects in the binary image of cracks. A (mm2): A is the total area of cracks on the image. Total number of crack intersections. Average nearest neighbor distance in plane. LA (mm/mm2): LA is the total dendritic length of cracks per image area. The parameters are used to compare shrinkage cracks in different samples. Beside above parameters. 9. The graphs of normal probability density and normal cumulative density present more information of the distribution of crack length and crack width in the image. Δ2 (1/mm): Δ2 presents a measure for the average centretocentre distance of randomly dispersed nearest neighbors in a plane which is closely related to the configuration and can be demonstrated as Δ 2 = 0. 6.Quantification of Crack Parameters Several parameters for the quantification of crack patterns are presented. Average crack length. First. 5. Number of cracks per unit area. 8. 3. NInt (): NInt is the count of intersection points in the image of crack. Total length of cracks. 10. L (mm): L is the total dendritic length of all the skeletonized cracks in the image. 7. AA (mm2/mm2): AA is the ratio between the areas of the counted cracks to the entire area of the image.5 N A1/ 2 . Lmean (mm): Lmean is the mean value of the dendritic length of all cracks in the image. length and width of all detected cracks in the image are calculated. W (mm): W is the total area of cracks (A) divided by total dendritic length (L). 4. Total area of cracks. The crack density. Average width of cracks. Areal fraction. 214 . Interesting parameters for quantification of cracks in concrete samples 1.
1.Automatic procedures for the quantification of crack parameters Step 1: make a binary image of crack pattern Step2: Detect cracks intersection points from skeletonized image of crack Step 3: Dilate the intersection points Step 4: Subtract dilated intersection points from the initial binary image Figure E. The process of image analysis on a cracked paste sample 215 .
Normal probability and cumulative density curves of crack width and crack length 216 .6 0.normal probability density .crack width (mm) 1 0.8 1 1.4 0.4 0.8 0.4 0.2 normal probability density .6 0.2 0 0.2.15 0.crack length (mm) 1 0.8 0.25 0.2 0 0.6 0.1 0.2 0.6 0.crack length (mm) 0.2 0 0.8 1 1.4 0.05 0 10 5 0 5 10 15 normal cumulative density .2 0 10 5 0 5 10 15 Figure E.crack width (mm) 3 2.4 0.8 0.6 0.4 0.5 1 0.8 0.5 2 1.6 0.2 0.5 0 0.2 0.2 normal cumulative density .
13 0.020 0.32 0.7 0.3 81.4 0.14 0.05 2.084 0. Results of crack analysis on a plain and FRC samples Parameters obtained from crack analysis Total length of cracks (mm) Total cracking area of image (mm2) Density of cracks in image (1/mm) Mean crack length (mm) Total areal fraction of cracks () Average width of cracks (mm) Number of cracks per unit area (mm2) Average nearest neighbor distance (mm) Total number of crack intersections () Total number of cracks in image () Degree of orientation of cracks (01) Average specific number of intersections (1/mm) Plain Sample 621.058 0.008 0.13 77 551 0.61 55 969 0.60 0.0 200.053 217 .099 0.Table E.062 1.066 Glass Fiber Sample 582.1.10 1.
NL(Θi) = specific number of intersections in the orientation Θi NL = 1/n ∑N 1 n L (Θi) . Analysis by counting number of intersections of cracks network with the oriented secants (parallel equidistant lines) In case of an anisotropic cracking. In this technique.Determination of Crack Orientations in EarlyAge Cracked Cement Pastes Concept The degree of crack orientation is determined on the basis of the classical stereological technique of oriented secants on a plane [191. 192. A count of the intersections of these lines and the cracks network is performed. which is characteristic of the specific surface of cracking. The number of intersections is divided by the total length developed by the straight lines in order to get the specific number of intersections. in a figure which is named "rose of the number of intersections". where n = number of orientations Sv = 2NL For the purpose of emphasizing the anisotropy. given as ω= ( N L max − N L min ) N L max + ( − 1) ⋅ N L min 2 218 π . the different values of NL(Θi) are reported in a polar referenced plane. 194]. these characteristics are calculated by averaging the different values obtained by rotating the family of straight lines. 193. The anisotropy is characterized by a scalar.3. called degree of orientation. NL = specific number of intersections Sv = specific surface of cracking = 2NL Figure E. ω. parallel equidistant straight lines cover the view to be analyzed.
ω. the expected results for the orientation degrees are 0 and 1 respectively. a circle. Degree of orientation. NLmax and NLmin are the maximum and minimum values taken by NL respectively. 3 and 6. Since the circle is perfectly isotropic and parallel lines are completely oriented to one direction and have extreme anisotropy. a series of parallel lines.where. • Doing all above steps for the next incremental angles. takes the value 0 for a perfectly isotropic crack pattern and the value 1 when all the cracks have the same orientation (extreme anisotropy). here steps are 15o) in the format of binary images. • Counting the number of intersections of the cracks with the set of lines. The analysis should be capable of capturing the effect of increase in the values of orientation degree by the elongation of rectangle. known as skeletonization. and four different rectangles. Automatic Process of Orientation Analysis The automatic procedure elaborated for the estimation of crack pattern orientation includes. 2. The rectangles tested here have different values of length to width ratio as. • Superimposition of the image of crack network and image of parallel equidistant lines with a given direction using the logical binary operation “AND”. 219 . • Measuring total length of parallel lines covering the image. • Converting grayscale (8bit unit) images of cracked sample into binary images. • Providing families of equidistant parallel lines in different angles (from 0o to 180o. • Calculating average specific number of intersections and degree of orientation for the crack network as given in the previous section. These trials include. Testing and Calibration of Crack Orientation Analysis Process In order to insure the reliability of this image analysis process. • Calculating specific number of intersections defined as the number of intersections divided by the total length of parallel lines. a series of trial shapes and figures are analyzed. 1 (square). • Reducing the cracks to single pixel lines by binary thinning. Here are the results.
0000 3) For a rectangle with length to width ration of 1 (square): Degree of orientation: 0.1) For a circle: Degree of orientation: 0.3328 Rose of the number of intersections: 220 .0519 Rose of the number of intersections: 2) For parallel lines: Rose of the number of intersections Degree of orientation: 1.
7466 Rose of the number of intersections: 221 .6136 Rose of the number of intersections: 6) For a rectangle with length to width ration of 6: Degree of orientation: 0.5078 Rose of the number of intersections: 5) For a rectangle with length to width ration of 3: Degree of orientation: 0.4) For a rectangle with length to width ration of 2: Degree of orientation: 0.
1333 mm1 225 0 0.06 0.04 0.02 0.0998 0.Crack Orientation Analysis on Cement Paste Samples The process of image analysis on an earlyage cracked cement paste sample (plain sample 1) after 24 hours of vacuum drying is shown as below.08 Specific Number of Intersections. original grayscale image (8bit unit) binary image binary image (black and white) opposite of skeletonization of crack network intersection superimposing secants with crack image 90 points of 135 45 Rose of the number of intersections: 180 ω: NL: Sv: 0.1/mm 0 315 270 222 .0666 mm1 0.
02 0.1455 mm1 90 135 45 180 0 0.06 0.08 Specific Number of Intersections.06 0.1376 0.The results of analysis for the other samples are as follow.04 0.1354 mm1 223 .0727 mm1 0.0677 mm1 0.04 0.1/mm 0 225 315 270 ARglass fiber cement paste (sample 1) ARglass content: 3 kg/m3 ω: NL: Sv: Rose of the number of intersections 0.02 0.0922 0. 90 135 45 180 0 0.08 Specific Number of Intersections.1/mm 0 225 315 270 Plain cement paste (sample 2) ω: NL: Sv: Rose of the number of intersections 0.
08 0.02 0.0528 mm1 Sv: 0.04 0.90 135 45 180 0 0.06 0.1070 mm1 0.04 0.1272 0.1/mm 0 225 315 270 ARglass fiber cement paste (sample 1) Rose of the number of intersections ARglass content: 6 kg/m3 ω: 0.2140 mm1 90 135 45 180 0 0.1/mm 0 225 315 270 ARglass fiber cement paste (sample 2) ARglass content: 3 kg/m3 ω: NL: Sv: Rose of the number of intersections 0.1056 mm1 224 .12 Specific Number of Intersections.08 Specific Number of Intersections.0844 NL: 0.
Results of crack analysis and quantifications Sample Degree of Orientation .0874 0.ω 0.0538 Specific Surface of Cracking. NL (mm1) 0.90 135 45 180 0 0.1324 0.1168 mm1 Table E.096 0.02 0.04 0.2.1097 Specific Number of Intersections.0697 0.06 0.1076 Plain cement paste ARglass cement paste (3 kg/m3) ARglass cement paste (6 kg/m3) 225 .1350 0. Sv (mm1) 0.08 Specific Number of Intersections.1/mm 0 225 315 270 ARglass fiber cement paste (sample 2) ARglass content: 6 kg/m3 ω: NL: Sv: Rose of the number of intersections 0.1747 0.0584 mm1 0.1393 0.
APPENDIX F Modeling Earlyage Shrinkage of Concrete Due to High Rate of Evaporation 226 .
qv is the flux of water vapor. except in very early stage of drying. In nonsaturated state.l ∇θl ) − El −v On the other hand. whereas ρl results from the transformation of the mass flux for vapor into an equivalent volumetric liquid water flux [141].Drying of CementBased Materials Governing Differential Equations Drying of cementbased materials can be considered as a moisture flow through porous media. [195]: k (F. R is the gas constant.5) ρl RT v v where. macroscopic volumetric flux through a porous medium with only air and vapor is obtained from equation (F. μl is the dynamic viscosity of liquid water and p is the macroscopic pressure. The liquid water flux for transport through a nonsaturated medium can be obtained by the extended Darcy’ law as presented in equation (F. ρl is the density of liquid water. is a function of moisture content. T is the absolute temperature. ( Dθ .3) Conservation of mass for liquid water expressed in volumetric quantities can be (F.2) In this equation kl (θ ) . qv = − Dθ . the vapor transport under isothermal conditions at the microscopic level is given by Fick's law: 1 M qv = − D ∇p (F.ql − El −v = ∇. The term M/RT results from the ideal gas law. kl is the liquid permeability.2) ql = − l μl K is the hydraulic conductivity defined as K = g ρ kl (θ ) / μl and Ψ is the hydraulic potential presented by Ψ = pc ( θ ) / g ρl . Darcy's law can be used to obtain the flux of liquid water in a saturated porous medium.7) 227 . M is the molecular mass of water. liquid water transport is presented as ql = − Dθ .v ∇θl (F. Dv is the diffusion coefficient of water vapor in air.6).v = ⎛ ∂h ⎞ (n − θl ) M * T (θ l ) Dv Pvs ⎜ ⎟ ρl RT ⎝ ∂θ l ⎠ (F. However. k (θ ) ∇pc (θ ) = − K ∇Ψ (F. the effective permeability of liquid water.6) where Dθ . the concrete is not fully saturated.l ∇θl (F.4) written as: ∂θl / ∂t = − ∇. Defining the liquid diffusivity as Dθ .l = K (∂Ψ / ∂θl ) . the liquid water is transported because of gradient of the capillary pressure [160]. Pel (1995) showed that at the macroscopic level.1) ql = − l ∇p μl where ql is the flux of liquid water. and pv is the vapor pressure.
equation (F. as well as in mortar [149] and lightweight concrete [66]. and different wet capillaryporous materials by Kowalski [101] indicate that evaporative flux at the boundary is roughly constant for a period of time at the beginning of the drying referred to as stage I and then falls significantly by the time in the second period or stage II. ⎛ 7.v . the density of water vapor is 0. T* is the tortuosity accounting for the extra path length resulting from the tortuous pores.where n is the porosity. ( Dθ .4 × 10−4 ⎞ θ = ⎜1 + (F. the total moisture content can be appropriately approximated by the liquid water content. for liquid water content to vapor content ratio greater than about 103. (nθl ) is the voidspace available for vapor diffusion.74 kg/m3. the moisture transport through the sample is assumed to be onedirectional (1D) due to high surface areatodepth ratio. For modeling the internal 228 .13) = ∇. ∂ (θ v + θl ) = ∇. ( Dθ ∇θ ) ∂t Dualstage Drying Methodology Studies of drying behavior of clay brick ceramics by Hall et al.8) are combined. equation (F. To find a single nonlinear diffusion equation for water in both phases. In most of earlyage drying tests. Cooling [100]. Platten [146].4) and equation (F. and exposing only top surface to the drying condition.l + Dθ . Dv is the vapor diffusivity. the mass density of the vapor in the pores and total moisture content are given as: θl θ ⎛ ρ ρ ⎞ = n v ⇒ θ = θ v + θl = ⎜1 + v l ⎟ θl (F. The macroscopic conservation of mass for water vapor expressed in volumetric quantities can be written as: ∂θ v = − ∇.v ) ∇θl (F.qv + El −v (F.8) ∂t where El −v is the rate of evaporation.9) is represented as: ∂ (θ v + θl ) = ∇.10) ∂t Assuming equilibrium between water and vapor. pvs is the saturation pressure of water vapor. ( Dθ ∇θl ) (F. equation (F. The same drying behavior was observed for the Portland cement pastes in recent experimental works by authors [148]. [99].12) ⎟ θl n ⎝ ⎠ As one can see from equation (F.10) can be represented as a nonlinear diffusion equation for total moisture as: ∂θ (F.11) ρl ρv n ⎠ ⎝ At the room temperature.12). and h is the relative humidity. Therefore.l + Dθ .9) ∂t Defining moisture diffusivity as Dθ = Dθ .
15) for stage I as well as each time step for stage II. the initial and boundary conditions are 0≤ x≤ L t = 0. Both chemically and physically bound water may be assumed to be in local equilibrium with free water due to rapid rate of hydration at the early age. Consequently.moisture transport in stage I and stage II drying.18) = = D′ 2 ∂t 1 + R ∂x 2 ∂x Replacing the term D/(1+R) with D’ as the effective moisture diffusion coefficient. equation (F. the equation is seen in the usual form of diffusion equations. θ (). ∂C ∂ 2C =D 2 (F. is the constant diffusivity at stage I. ∂C D ∂ 2C ∂ 2C (F. liquid water will be chemically bounded into hydration products as well as physically bound water to increasing internal solid surfaces. the concentration of bound water. equation (F. To simplify the concept.19) 229 . C. S = RC (F. is directly taken proportional to the concentration of the free water. while the diffusivity is variable at stage II drying. with the dimensional variable C (kg/m3). the timevarying diffusivity values at stage II are only averaged over the space. 2 The cement paste is assumed to be homogeneous at initial conditions with an impermeable bottom surface.13) is used with the replacement of dimensionless moisture concentration variable. 1D form of equation (F. Therefore. S.16) can be presented in a new form by substituting for S from equation (F.14) can be simplified to equation (F. With time. C = Ci (F.16). the substance free to diffuse.15) ∂t ∂x Hydration of cement and its implications on transport properties need to be taken into account during earlyage drying of cementitious materials. D’ presents the potential for conductivity of free water through the fresh cementitious materials.16) During earlyage drying when the moisture diffusion is accompanied by immobilization of bound water. This can be acceptable considering high evaporative surfacetodepth ratio of the samples. Therefore. ∂C ∂ ⎡ ∂C ⎤ dD ⎛ ∂C ⎞ ∂ 2C (F. The assumption is based on experimental observations that concrete is capillary saturated during stage I [105] but unsaturated at later stages of drying and its value will be a strong function of moisture concentration.15) is modified as: ∂C ∂ 2C ∂S =D 2 − (F. the time of drying is divided into several time steps at stage II and constant diffusivity is applied at each time step with respect to the time and the location.17) ∂t ∂x ∂t Equation (F. Using this method.14) = D(C ) = +D 2 ⎜ ⎟ ∂t ∂x ⎢ ∂x ⎥ dC ⎝ ∂x ⎠ ∂x ⎣ ⎦ The only difference in applying this formula in stage I and II. Therefore.
the boundary condition at the top surface is at x = L.24) Where. The analytical solution for equations (F.22) In this model transition time from stage I to stage II.20) =0 ∂x where Ci is the initial moisture concentration (kg/m3). governing equations and boundary conditions during stages I and II drying are presented in Figure (F. Cs is the surface moisture concentration and C* is the ambient moisture concentration (kg/m3).1). F0 is the constant flux in stage I drying (kg/m s). For modeling stage II drying based on the concept of convectivediffusive transfer of vapor at the surface. Prediction of the transition time based on physical grounds requires further studies. k parameter takes into account the wind velocity. x ) = C∗ + ( Ci − C∗ ) + ∑ (F. referring to the loss of continuity in the liquid phase. J = − D′ (F. can be characterized by the convection equation . λn tan(λn ) = 230 . the surface temperature. and the relative humidity for drying to the atmosphere or in the case of lowpressure drying encompasses vacuum condition and temperature. is defined by the user which may be obtained from experimental data. J = F0 (F. and J is the moisture flux (kg/m2).21) 2 where. the exchanged flux of moisture. 2sin(λn ) x cos(λn ) L n =1 λn + sin(λn )cos(λn ) ∞ ⎛ 2 D′t ⎞ ⎜ − λn 2 ⎟ L ⎠ e⎝ kL (F.23) is given by [153] C ( t .23) ∂x where k (m/h) is convective moisture transfer coefficient or surface factor. Thus.25) D′ Geometry. J.1821) during each time step at stage II drying with the convective boundary condition as of equation (F.∂C (F. Solution to moisture concentration values as a function of time and location during stage I drying based on equations (F.1821) is [109] F0 L ⎧ D ′t 3 x 2 − L2 − D ′n 2π 2 t 2 ∞ ( − 1) n nπ x ⎫ ⎪ exp( ) cos C ( t . Moisture transport during stage I drying is considered as an external mass transfer control process with a constant drying rate. x ) = Ci + − 2∑ ⎨ 2 + ⎬ 2 2 2 D′ ⎩ L L ⎪ 6L L π n =1 n ⎭ at x = 0. ∂C J = − D′ = k (Cs − C* ) (F.
Figure F. (b) stage II drying assuming convective moisture flow on the top surface 231 .1b Figure.1a Figure F.1. Geometry. governing equations and boundary conditions during: (a) stage I drying. F.
28). At each time step. D (est ) are estimated by the assumption of proportional decrease of diffusivity with the flux. The objective function for minimization is the weighted sum of two errors of cumulative moisture loss and flux.26) which is derived by rearranging equation (F. k (est) is the estimated surface factor (m/h) and J (exp) (t = ttrans ) is the experimental rate of evaporation at the transition time which is roughly equal to initial evaporation rate. Optimization is implemented to find two best fit parameters. while initial values for k are estimated from equation (F.26) k (est ) = Ccrit − C* where. initial D’ values for optimization process. 232 . In optimization algorithm. Once the optimal solution Dnorm and knorm are found. D’(t) and k(t) to predict the experimental results. diffusion coefficient at the current time step is normalized with its value at previous time step.2) and (F.23). D(est) and k (est) are passed to the optimizer and the optimizer calls the function evaluation to calculate the error and update these two parameters several times until optimal solution for D’ and k are found. Ccrit is the surface moisture concentration at end of stage I or critical moisture concentration (kg/m3) and C* is the ambient moisture concentration.27) D (est ) = Di −1 (exp) (i −1) (t ) J k ( est ) = J (exp) (t (i ) ) C (t (i −1) . Thus. k (est) are estimated by a convection transport formula. J (exp) (t = ttrans ) (F. normalized design variables (D’ and k) are used to avoid numerical problems caused by mixing big and small numbers in numerical operations. so does the mass transfer coefficient knorm= ki /ki1. while initial k values. they are reverted back to D’ and k. J (exp) (t (i ) ) (F. these two initially estimated parameters.27) and (F. Dnorm=D’i /D’i1. t is the current time step. Initial values for D’ are obtained from analysis of stage I drying. Inequality constraints are imposed to the design variables D’ and k in terms of lower and upper bounds.28) In equation (F. Results of back calculations for the reported data by Selih (1996) are shown in Figure (F. while t is the previous time step.3).BackCalculation Procedures A back calculation model is used to predict diffusivity and moisture transfer coefficient at any time step based on given cumulative moisture loss and evaporation rate curves. x = L) − C* (i ) (i −1) (F. For other time steps.
2b Figure F.Drying Time.40 T = 22oC RH = 50% No Initial Curing 100 0.1 0. m/h 0. (b) Backcalculated D’ and k for the best fit of data 233 Evaporation Rate.05 0 Cumulative Moisture Loss. d 30 Figure F.0005 Moisture Diffusivity. kg/(m2.0015 0. m2/s 3x107 2x107 1x10 7 K 0 D' 6x109 0 10 20 Drying Time. g 60 40 20 0 Selih (1996) Simulation 0 10 20 Drying Time. d 0.2.15 0.1 1 10 w/c = 0.2a Moisture Transfer Coefficient. d 30 Figure F.001 0. (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the experiment performed by Šelih (1996).h) .
1 m Time t/h 0.07E007 2.8 t h 0.7 1.4.07E007 BackCalculated Diffusivities CEBFIP: n = 10 CEBFIP: n = 15 CEBFIP: n = 20 Moisture Diffusivity.4 1d 3d 0.07E007 7. (b) moisture concentration distributions through the thickness of the sample in comparison with experimental results 234 .78 x 107 m2/s D0 = 6.4 0. m2/s 3.2 7d 28 d Exp Sim Exp Sim Exp Sim Exp Sim 0 0 40 80 120 160 Moisture Concentration.94 x 109 m2/s α = D0/D1 = 0.00E009 0.6 0.3a 1 0. kg/m3 200 Figure F.07E007 Analysis on Experiments by Selih (1996) D1 = 2. Results of analysis on drying data by Šelih (1996): (a) Moisture diffusivity as a function of normalized moisture concentration expressed by CEBFIP models.025 Cc = 0.6 h = 0.8 Normalized Moisture Concentration 1 Figure F.3b Figure F.3.
210] used the coefficient of hygral contraction as α sh C =1. Wittmann and Roelfstra [ 196 ] described the unrestrained hygral length change. ε sh = α sh C . [203] modeled the developed strain in the drying concrete according to equation (F. 235 . Alvaredo and Wittmann [197] have used this linear relationship with the α sh h ranging from 1. the experimental measurement indicate that the empirical shrinkage coefficient. 209. Having moisture distributions at different time in drying of a concrete sample. Granger et al. by a linear relationship such as: ε sh = α sh h . h.2930) is an easy implementation in drying models to obtain developed strains and related stresses. Benboudjema et al. However. the drawback is the lack of a physical interpretation between nature of the porosity in the material and the developed strains. ε sh .ΔC (F. shrinkage. The major reason for dealing with water content rather than relative humidity is that water content can be more easily measured by weight loss monitoring [55].29).29) where α sh h is an empirical shrinkage coefficient depending on concrete composition and maturity. [205] and Verbeck [206] showed that particularly when h remains above 50 percent. creep and cracking [207].Δh (F.22×105 . as a function of relative humidity. This model has been used in hydromechanical models accounting for the full coupling of drying. Thelandersson et al.99. h. Therefore. 202]. the moisture concentration can be easily translated into an internal stress distribution. and α sh C is the coefficient of hygral contraction which can be obtained from the linear part of unrestrained shrinkage versus weight loss curve [204].52. Similar efforts have been made on linearly relating moisture content to the shrinkage strain assuming a quasilinear relationship between the moisture content and the relative humidity in the range of 50100% in desorption isotherm curves [201. and using a Finite Element Method (FEM). there is no requirement for desorption isotherms of the concrete to relate humidity. Bazant and Xi [198] and Bazant et al. C. [199] also introduced the same concept into their models. this relationship is confirmed fairly well by experiments. α sh h is constant in the range 0.Shrinkage Strains and Drying Simplified Models Some researchers have focused on developing linear relationship between shrinkage strains and the relative humidity.5x103. This stress distribution can be finally used to calculate the total deformation of a drying specimen [196]. to the water content. According to van Zijl [ 200 ].4 < h < 0. The advantage of these simplified models (equations F.30) where C is the moisture content (). [208.
which shows the high magnitude of capillary pressure in the pores. under certain assumptions. This nonlinearity which is more pronounced at the beginning and at the end of the drying process [215] led to suggestions of bilinear and nonlinear relationships [ 216 . pc is the capillary pressure. capillary water is not under any external loads such as gravity. much of the shrinkage that is observed in earlyage concrete can be attributed stresses associated with capillary tension [218]. and h is the fractional relative humidity. Kelvin equation presents the capillary pressure as: ρ RT pl − pg = pc = l ln(h) (F. disjoining pressure. According to Baron [219] these equations can be derived from the equilibrium of chemical potentials between liquid and vapor phases of water. These assumptions include capillary water forms a single continuous phase.5 (RH = 50%) is 96 MPa.32) rm where σlg is surface tension. Capillary tension.31) Mv where pl is the pressure of liquid water. Kelvin and Laplace equations are used to present the magnitude of capillary pressure developed in the concrete pores due to drying phenomena [219]. and solid surface tension (GibbsBangham) are among several mechanisms recognized for shrinkage of cementbased materials [217]. 200]. water vapor is an ideal gas. capillary water is homogenous and incompressible. the gas pressure is equal to the atmospheric pressure and the water is in the chemical equilibrium with ambient atmosphere. According to this formulation. R is the universal gas constant. T is temperature. velocity is zero at any point.01801 kg/mol). This relationship allows the use of internal relative humidity (h) as a parameter for modeling developed stress and strain in concrete due to drying [220]. 213.Capillary PressureBased Models Experimental results [205. pg is the pressure of the gas phase (dry air and water vapor).8 oC) and relative humidity of 0. 211] and models based on the mechanics of porous media [212. 214] reveal the nonlinearity between the shrinkage strain and relative humidity or water content. ρl is the water density (1000 kg/m3). and rm is the radius of assumed circular meniscus. Mv is the molar mass of water (0. Assuming a circular meniscus with the radius rm. the capillary pressure at 300 K (26. 236 . The capillary pressure between humid air (dry air and water vapor) and liquid water should be balanced by the surface tension σ lg at the interface of liquid/gas phase. the first law of Laplace can be derived as: 2σ pl − pg = pc = lg (F. However.
compression develops in the solid microstructure [222]. and therefore. the variation of 237 . pd is the disjoining pressure.36) shows only the microscopic pressure on the solid skeleton due to the capillary pressure. the pressure on solid skeleton can be presented as ps _ c = Sl pc (F.36) While equation (F.33) and rearranging gives α sin γ rm ps _ c = α l ( pl − pg ) + pg + lg pc − patm (F. ρ RT Δpd = a ln(h) (F. αlg is a coefficient incorporating effect of surface tension (σlg).35) ⎟ pc = α sol _ c pc 2 ⎝ ⎠ Gray and Schrefler [223] replaced the coefficient αsolc by the saturation degree of liquid phase. 4. Bazant [ 224 ] presented variation of disjoining pressure in a similar fashion to capillary pressure. psol_c is calculated as (F. b) different pressures applied on solid skeleton inside capillary pores [221] As a reaction to this negative pressure within the pore fluid.33) ps _ c = α l pl + (1 − α l ) pg + α lg sin γ σ lg − patm where.(a) (b) Figure F. respectively. Sl.32) into equation (F. αl is the volumetric fraction of liquid water in the capillary pores. Similar to the capillary pressure. a) Equilibrium between liquid water and humid air at the interface of liquid/gas phase inside a pore in concrete. An average pressure on the solid skeleton due to the capillary pressure.34) 2 According to Baron’s assumption patm = pg and therefore α lg sin γ rm ⎞ ⎛ ps _ c = ⎜ α l + (F. and ρa and Ma are the density and molar mass of adsorbed water. Inserting equation (F.37) Ma where.
E is the young’s modulus. and a homogenization factor. The shrinkage strain can therefore be obtained by combining equation (F.36) and (F.42) To calculate shrinkage strains based on equation (F. αs [215] as: ps = α d Sl pc (F.43) φ= w / c + 0. the total applied pressure on the solid skeleton.the applied pressure on the solid skeleton due to the disjoining pressure can be presented as: Δps _ d = α sol _ d Δpc (F.313 where w/c is the watertocement ratio. σ = (1 − φ )σ − φ ps (F. σ = Eε s (F. As shown in Figure (F. The model proposed by Bentz and Garboczi [226] is used in this study with the following equation: w / c − 0. and can be obtained from shrinkage tests.39) The homogenization factor incorporates effects of both capillary and disjoining pressures as well as the effect of scale change from microscopic to macroscopic levels. 227 . has been predicted with different models [ 225 . On the other hand. εs = φ ps (1 − φ ) E (F.38). capillary porosity. the external stress is zero ( σ = 0 ). φ and modulus of elasticity. E has to be modeled. initially. In unrestrained shrinkage tests. σ can be related to the strains due to elastic behavior of solid skeleton. psol due to the capillary and the disjoining pressures can be obtained using the saturation degree of liquid phase.5). The capillary porosity. concrete can be approximated by a square in which the pores are placed in the center and the rest of area is the solid phase.41α (F. Effective stress σ is defined as the stress acting on the solid skeleton which can be obtained from the equilibrium. φ .38) Considering equation (F.41).41) where. and ε s is the shrinkage stain.42). A unit cell approach can be taken to study this effect on porous media such as concrete materials. 226 . 238 .40) where σ is the external stress (restraining effects) and φ is the porosity.40) and (F. 228 ].39) can be used to determine the strains induced by the pore pressures due to the drying process. and α is the degree of hydration. Sl. Equation (F.
Effective stress on solid skeleton based on unit cell modeling [221] The degree of hydration is modeled as a function of time using a power function. and α f is the degree of hydration at the final a drying time. ⎛t α =⎜ ⎜t ⎝ f ⎞ ⎟ αf ⎟ ⎠ a Figure F.6. ⎛ t ⎞ α =⎜ ⎟ αf (F.5.Figure F. tf is the final drying time.44) ⎜t ⎟ ⎝ f ⎠ where.45 at 24 h is about 0.2 (see Figure F. Results of simulation of degree of hydration are in good agreement with the experimental data reported by Lamond and Peilert [229] showing that the degree of hydration of ordinary Portland concrete with w/c of about 0.6). Simulation of degree of hydration in comparison with experimental data 239 . This model is used for studying earlyage drying shrinkage tests.
45) b ⎛ t − ti ⎞ ⎪ . t > ti ⎪ Ei + ( E f − Ei ) ⎜ ⎜ t −t ⎟ ⎟ f i ⎠ ⎪ ⎝ ⎩ where E. such a model is capable of simulating the development of modulus of elasticity at very early age (<6 h) as well as from very early hours to few days (0150 h). t ≤ ti ⎪ ⎝ ti ⎠ ⎪ E=⎨ (F. and relative humidity of 240 . ti and t f are the time.Elastic modulus development in the early age concrete can be modeled using a twostage methodology. As Shown in Figure (F. initial time of setting. the method can be applied to different sets of data with available experimental moisture loss curves and shrinkage versus time curves. Ei . under temperature of 3540 oC provided by a radiator heater. Modeling development of modulus of elasticity in early ages (0150 h) Modeling EarlyAge Shrinkage from Drying Data In order to confirm the soundness of this method in calculating earlyage shrinkage from the drying data. ⎧ ⎛ t ⎞a ⎪ Ei ⎜ ⎟ . respectively. and E f are the elastic modulus at different times. a should be more than 1 and b less than 1 to conform to the experimental data [Dao et al. and the final time of drying. 2009] and [Van Heyningen and Boon 1973]. t. a and b are constants.7. However. ⎛t⎞ E = Ei ⎜ ⎟ ⎝ ti ⎠ a ⎛t⎞ t ≤ ti : E = Ei ⎜ ⎟ ⎝ ti ⎠ b ⎛ t − ti ⎞ t > ti : E = Ei + ( E f − Ei ) ⎜ ⎟ ⎜ t −t ⎟ ⎝ f i⎠ a Figure F. The experimental data of Wongtanakitcharoen and Naaman [20] were chosen for this purpose which involve onedimensional drying of linear prismatic specimens 1000 mm in length and 100×60 mm in cross sectional area.7).
and the fine aggregate was NS2 sand. Similar to author’s drying results on plain and glass fiber reinforced concrete [148].1.2% volumetric fraction were chosen for analysis. the moisture loss curves for these two samples (plain and 0. Unrestrained earlyage shrinkage test setup (Wongtanakitcharoen and Naaman 2007) Among different earlyage shrinkage tests performed by Wongtanakitcharoen and Naaman. The weight ratio of Portland cement to water. C* was obtained 56 kg/m3.5311 and the properties of PVA fiber is shown in Table F.53. can be obtained from the desorption isotherm curves which relate internal relative humidity to the water content. As shown in Figure (F. The water to cement ratio by weight was 0.9a).5%. sand and coarse aggregate in this mixture were 10. Figure F. C* . the evaporation rates were computed from cumulative moisture loss data using a numerical differentiation procedure.8) rapid rate of evaporation was insured using a high velocity fan provided constant high flow of air over the specimen’s surface. ASTM C150 Type III Portland cement was used in the concrete mix and the coarse aggregate was crushed limestone of 9. Properties of PVA fiber used in this study The ambient moisture concentration.1. The drying results were used to backcalculate diffusion and convective moisture transfer coefficients as shown in Figure (F.22. Table F. Using the BSB model [110]. As shown in Figure (F.9b).525mm maximum size.2% 241 .8. test results on plain concrete and PVA fiber reinforced concrete with 0.
H is relative humidity (fraction) and A.7 Vct = 0. the BSB model [230].3 < w/c < 0. C* is moisture content (kg of water / kg of solid).10) and (F. capillary pores pressure.0. Results of analysis including simulation of profiles moisture concentration. C* = AkVm H (1 − kH )[1 + ( A − 1)kH ] (F.85 + 0.024(0.5 for cement type 1 to 4. for t > 5 days and 0. macroscopic pressure on solid skeleton are shown in Figures (F.60 for cement type 1 to 4.2 w / c) N ct . (a) Prediction of cumulative moisture loss and evaporation rate as a function of time for the reported experimental data.15 and 1.3 < w/c < 0. 1 (1 − ) A − 1 n k= A −1 n = 5. is used to relate internal relative humidity to the water content.9.9. respectively.46) where. (a) (b) Figure F. A = exp( for t< 5 d and 0. 1. relative humidity.85 and 0. Vm are three empirical parameters defined as: A0 ).PVA) are very similar and therefore a single cumulative moisture loss curves is used for the rest of analysis.7 Nct = 1. (b) Backcalculated D’ and k for the best fit of data 242 . A0 = 855 T Vm = 0. 0.45w / c)Vct .11). respectively. A desorption isotherm model. k.0.33 + 2.1. 1. 1.5(0.
(b) simulation of relative humidity profiles from desorption isotherms (BSB model) 243 .2% concrete. (a) Simulation of moisture concentration profiles during drying of plain and PVA0.10.(a) (b) Figure F.
(b) simulation of macroscopic pressure on solid skeleton during earlyage drying of plain and PVA0.6 α d = 1.6 0.(a) 1 t = 24 h 3 t=1.2 40 20 0 Macroscopic Pressure on Solid Skeleton. (a) Simulation of capillary pores pressure profiles. MPa (b) Figure F.2% concrete 0 244 .4 0.8 0.5 0.11.
13). Introduced elastic modulus for plain and PVA 0.14).13.58 0.7 0.12.Simulation results for the development of capillary porosity are shown in Figure (F. Change in Capillary porosity according to porosity model by Bentz and Garboczi [226] ⎛t⎞ t ≤ ti : E = Ei ⎜ ⎟ ⎝ ti ⎠ a ⎛ t − ti ⎞ t > ti : E = Ei + ( E f − Ei ) ⎜ ⎜ t −t ⎟ ⎟ ⎝ f i⎠ b Figure F.66 Simulation 0. 0. h Figure F.2% concrete 245 . the profile of shrinkage strains during the drying process is simulated as shown in Figure (F.62 0.12). Based on these data.54 0.5 0 5 10 15 20 25 Time. Also elastic modulus development curves introduced the model is shown in Figure (F.
14. Simulated earlyage shrinkage strain profiles during drying of: a) plain concrete.2% concrete 246 . b) PVA0.(a) (b) Figure F.
15). Simulation of earlyage shrinkage strain averaged over the thickness of the sample in comparison with the experimental data reported by Wongtanakitcharoen and Naaman (2007) 0 0 5 247 . simulation results indicate that the proposed method is capable of predicting experimental shrinkage strains during earlyage drying process of cementbased materials and presenting the variation of moisture concentration.15. As shown in the figure. 5000 Wongtanakitcharoen and Naaman (2007) 4000 3000 2000 1000 10 15 20 25 Drying Time.14) are averaged over the thickness of samples and shown in Figure (F. capillary pores pressure and applied macroscopic pressure on the solid skeleton within the time and space domain of the test.Since the reported data by Wongtanakitcharoen and Naaman (2007) are the average shrinkage strains. h Figure F. the simulated shrinkage strains in Figure (F. internal relative humidity.
Capillary pressure in fresh cementbased materials and identification of the air entry value. Yang CY. Pease B. Schmidt M. Service life prediction of concrete wharves with earlyaged crack: Probabilistic approach for chloride diffusion. Calif.References [1] [2] [3] [4] Horvath A. Annu Rev Environ Resour 2004. Influence of shrinkagereducing admixtures on the reduction of plastic shrinkage cracking in concrete. Rajabipour F.30:557–565. Cohen MD. The effect of cracks on chloride penetration into concrete. The cement industry and climate change challenges.302303:6306. Na UJ. Mazzotta GB. de Rooij MR. Weiss J. Cem Concr Res 1990. Earlyage concrete cover crack and its effects on concrete cover. Rue d’Arlon 55. DC: CIA. Tietz. van Breugel K. Cem Concr Compos 2008. ACI Mater J 2007. [5] [6] [7] [8] [9] [10] [11] [12] 248 . Washington. Jung SH. MoraRuacho J. Gupta R.104(2):18794. T. Mater Struct 2009. CEMBUREAU (the European Cement Association). Gettu R. Kwon SJ. Dolch WL. US Central Intelligent Agency.39(3):1416. Lura P. 2008. Huang XM. 2003. Aguado A. 29: 181204. Olek J. Schlangen E. Banthia N. Influence of shrinkagereducing admixtures on development of plastic shrinkage cracks. California Nevada Cement Association. Slowik V. Mechanism of plastic cracking in Portland cement and portland cementsilica fume paste and mortar. Fritzsch R.31(1):7583. Plastic shrinkage cracking in cementitious repairs and overlays. BE1040 Brussels.42(5):56779. Belgium. The world factbook. Key Eng Mater 2006. Cem Concr Res 2009. March 30April 3. Los Angeles. Key Eng Mater 2007. Construction materials and the environment.20(1):10319. Struct Saf 2009. Yoon IS.348349:76972. Park SS. A presentation at American Concrete Institute’s Spring Convention.
Cem Concr Res 19766. Shah SP. 2002. Cracking of plastic concrete. On the action of capillary pressure in fresh concrete. Application aux enceintes de confinement de bâtiments réacteurs des centrales nucléaires. Benboudjema F. Cem Concr Compos 2004. Morris PH. Shrinkage Cracking of Fiber Reinforced Concrete. Experimental Investigation into early age shrinkage of cement paste by using fiber Bragg gratings.26(5):473–9. ASTM C15708. Aust J Civ Eng 2003. Mater Struct 2009. Wittmann FH.04. and carbon fibers.[13] Radocea A. Université de MarnelaVallée. Modélisation des déformations différées du béton sous sollicitations biaxiales. Naaman AE. PhD dissertation. ACI Mater J 1990.40:289300. [21] [22] [23] [24] 249 . Klink T. PhD thesis.42:461–468. [19] [20] Hammer TA. [14] [15] [16] [17] [18] Slowik V. ACI Mater J 2009. Chalmers University of Technology. A study on the mechanism of plastic shrinkage of cementbased materials. Esping O. Cem Concr Res 2008. Mater Struct 2007. Dux PF. Tensile properties of earlyage concrete.1992.6(1):49–56. 87(2):138148. Mechanism of early age shrinkage of concretes. Schlattner E. Mater Struct 2001. Dux P.38:938–944.1(1):1721. Göteborg. Standard test method for length change of hardened hydrauliccement mortar and concrete. PVA. Effect of limestone filler BET(H2O)area on the fresh and hardened properties of selfcompacting concrete. Grzybowski M. ASTM Book of Standards 2008.106(6): 483492.34:273278.02. Göteborg. Unrestrained early age shrinkage of concrete with polypropylene. Effect of silica fume on the plastic shrinkage and pore water pressure of highstrength concretes. Sanjayan JG. Morris P. Wongtanakitcharoen T. Dao VTN. Aly T.
K. STF 65 A82036.30(9):775778. Concr Constr 1985. ACI SP206 on Concrete: Material Science to Application.87(4):327–332. 1999:61–72. Earlyage autogenous restrained shrinkage: stress build up and relaxation. Kvam SE.[25] Paillere AM. JG. 2002:189204. Shah.85(6):495504. Proceedings of the Fourth International Research Seminar on ‘SelfDesiccation and its Importance in Concrete Technology’. Shah SP. Carlson RW. Bercelo L. Weiss WJ. Trondheim. Influence of mix proportions and construction operations on plastic shrinkage cracking in thin slabs. A Tribute to Surendra P. Kovler. ACI Mater J 1988. Reading TJ. Lund University. Fagerlund. June 1982. Ed. Report No. Effect of polypropylene fibers on shrinkage and cracking of concretes. Model study of shrinkage cracking in concrete building walls. Testing system for determining the mechanical behavior of early age concrete under restrained and free uniaxial shrinkage. [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] 250 . Hover KC. SINTEF dep. Lange DA. Betong med EEstalfiber (Concrete with FEsteel fibers). Toma G. Shaeles CA. Gripspecimen interaction in uniaxial restrained test. Mater Struct 2008. Padron I. Collins F. 1997:217228. Kraai PP. Persson and G. Effect of synthetic fibers on volume stability and cracking of Portland cement concrete and mortar. Marchand J. Sanjayan. ACI Mater J 1990. Opsahl OA. Cement and Concrete Research Institute. Serrano JJ. ACI Struct J 1988. Effect of fiber addition on the autogenous shrinkage of silica fume concrete. Recent trends to reduce shrinkage cracking in concrete pavements.85(4):395404. B. Bissonnette B. Aircraft/Pavement Technology: In the Midst of Change.41(10):17411753. Altoubat AS. Aly T. Seattle. Buil M. WA. ACI Mater J 1989.27(6):324330. Mater Struct 1994. Zollo RF. Pigeon M. A proposed test to determine the cracking potential due to drying shrinkage of concrete.86(2):139144. Technical University of Norway.
Plastic shrinkage cracking of restrained fiberreinforced concrete. Aldea C.26:531540. A testing method for measuring plastic shrinkage in polypropylene fibre reinforced mortars. A. Shah SP. Standard practice for estimating the crack tendency of concrete. J Mater Civ Eng 2006. France. Mater Lett 1994. Bagneux. Banthia N. Turcry P. ACI Mater J 2005. Karaguler ME. Weiss J. [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] 251 . ACI Mater J 1992. Test method for evaluation of plastic shrinkage cracking in fiberreinforced cementitious materials. Naaman AE. ACI Mater J 2006.102(1):4958. Bentur. AASHTO PP 3499. Transp Res Rec 1997.18(1):46–54. Moragues A. Hauser G. Early age cracking in cementitious systems. ASTM C 158104. Cracking tendency of selfcompacting concrete subjected to restrained shrinkage: experimental study and modeling. Loukili A. Exp Tech 2007. Belarbi A. Ludwig DA. Transp Res Rec 1993. Effects of shrinkage reducing admixture on restrained shrinkage cracking of concrete.103(4):2729.1574:2536. Chapter 6. 2004. Gupta R. Cem Concr Compos 2004. Haidar K. Assessing residual stress development and stress relaxation in restrained concrete ring specimens. Hossain AB.31(6):448.89(3):289295. 2003. Nanni A.1382:6972. RILEM Publications. Loukili A. Evaluation of plastic shrinkage cracking of selfconsolidating concrete. Sanjuan MA. Sarigaphuti M. Influence of different fibers on plastic shrinkage cracking of concrete. 2004.5 – RILEM State of the Art Report TC 181EAS.21:23946.[36] Shah SP. PijaudierCabot G. Marikunte S. ed. Bentur A. Turcry P. Wongtanakitcharoen T. Standard test method for determining age at cracking and induced tensile stress characteristics of mortar and concrete under restrained shrinkage. Mcgillis MT. Yang W. Control of cracking with shrinkagereducing admixtures.
Granger L. Water transport in brick. stone. ACI Mater J 2001.49(181):303322. Modeling concrete shrinkage under variable ambient conditions. Aarseth LI. Zhu B. Tan M. Kim. Cem Concr Res 1971. J. Wu K.98(6):45864.1(5):46173. Mirza WH.13(3):171194. Evaporation of water from fresh mortar and concrete at different environmental conditions. 252 [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] . Hoff WD.93(1):3640. Huldén M. Shimomura T. Hansen CM. Wafa FF. Prediction of differential drying shrinkage in concrete. Bazant ZP. Cem Concr Res 1998. Effect of Ytype polypropylene fiber on plastic shrinkage cracking of cement mortar. Genin P. 1985. Maekawa K. New York: Taylor & Francis.37:925. Plastic shrinkage cracking of normal and highstrength concrete: A comparative study”. 1997.32:3844. Prog. Application of √ttype. Predicting moisture movement during the drying of concrete floors using finite elements.96 (1):3539. Garbalinska H. Plastic shrinkage cracking in concrete materialsInfluence of fly ash and fibers”. Mater Struct 2004. and concrete. Jacobsen S.36 :1294303. Najjar LJ. JK.19:674681. Constr Build Mater 2005. Diruy M. West RP. Drying of concrete as a nonlinear diffusion problem. Analysis of the drying shrinkage behaviour of concrete using a micromechanical model based on micropore structure of concrete. ACI Mat.28 (7):985994. Holmes N. 2002. logarithmic and halftime methods in desorptive measurements of diffusivity in narrow humidity ranges. Hall C. Coatings. Torrenti JM. Phuaksuk P.81(6):5605. Org. Lee CS. Cem Concr Res 2006. ACI J Proc 1984. ACI Mater J 1996.[48] [49] Ma Y. Effect of wind on drying from wet porous building materials surfacesA simple model in steady state. Mag Concr Res 1997. Shah SP. Wang K. Samman TA. Water permeation in coatings. Mater Struct 1999. Berhane Z.
1999. Models for estimation of service life of concrete barriers in lowlevel radioactive waste disposal. Bremner TW. Rep.5(1):320. ACI Mater J 1998.133(3):400410. Drying of saturated lightweight concrete: An experimental investigation. Stavarides H.S. Transp Res Rec 2003.1988:6398. Hossain AB. Wilson GW. SIMULIA.5 (4):295303. Fredlund DG. Mater Struct 1996. Dassault Systemes. Chen D. J Struct Eng 2007. Modelling of moisture loss in cementitiously stabilised pavement materials. Chichester. [65] [66] [67] [68] [69] [70] [71] [72] [73] 253 . Young JF. Can Geotech J 1994. Rhode Island. Nuclear Regulatory Commission. Physical mechanisms and their mathematical descriptions. Mahadevan S.76 (3):443460. Can J Civil Eng 2000. Concr Mater Const 1834:2433. ACI 305.1990.29(7):401405. Weiss WJ. Kodikara JK. Influence of fiber reinforcement on restrained shrinkage cracking.95(4):365375. Swamy RN. Wiley. ACI J 1979. Mater Struct 1972.27 (4):774784. Pease B. USA.31(2):151161. Hot weather concretingACI 305R99. Šelih J. ASCE J Geomech 2005. Steel fiber reinforced concrete and multiple cracking. 2010. Ed. Therrien Y. Cracking analysis of plain concrete under coupled heat transfer and moisture transport processes. Plastic shrinkage cracking and evaporation formulas. Barbour SL. Smith RW. Najjar LJ. Coupled soilatmosphere modelling for soil evaporation. Uno PJ. Farmington Hills. Quantifying restrained shrinkage cracking potential in concrete using the restrained ring test with acoustic emission.[61] [62] [63] [64] Bazant ZP. Walton JC. Bissonnette B. Prepared for U. ZP Bazant. Mathematical modelling of creep and shrinkage of concrete. Nonlinear water diffusion in nonsaturated concrete. NUREG/CR5542 EGG2597. Pleau R. Plansky LE. Chakrabarti S. Michigan: American Concrete Institute.
Fracture of glass fiber reinforced cement. London. Bakhshi M. Quantifying shrinkage cracking in fiber reinforced concrete using the ring test. 109123. Kingsbury D. Chicago . ACI SP20614 2002:219242.14:3134.3 Tuma JJ. Vol. New York. Diamond S. Litherland KL. Mater Struct 2006. Yang W. ACI Mater J 1998. Report on physical properties and durability of fiberreinforced concrete. Weiss WJ. Using acoustic emission to quantify damage in fiber reinforced cement mortars. McGrawHill. Diamond. Cem Concr Res 2003. pp. A method to predict shrinkage cracking of concrete. April 2023. Prague Czech Republic. Marikunte S. Shah HR. Weiss WJ. BecqGiraudon E. Soranakom C. Grzybowski M. Desai TK.39 (9):887899. Oakley DR. MI. S.[74] [75] [76] [77] Mane SA. Oxford BSP Professional Books. Mobasher B.33 (2):207214. American Concrete Institute. Role of alkali resistant glass fibers in suppression of restrained shrinkage cracking of concrete materials. Cem Concr Res 1981. 1986. Bentur A. Proc Durability of glass fiber reinforced concrete symposium organized by PCI. Bentur A. Proc 15th International Glass Fibre Reinforced Concrete Association Congress (GRC 2008). Prague. Laws V.95 (4): 339346. Modeling of restrained shrinkage cracking in concrete materials. Shah SP. 1985. 125135. Glass fibre reinforced cement. 2008. Farmington Hills. November 1215.5R10. 148. ACI Committee 544. Engineering mathematics handbook. No. Cem Concr Res 1984. Kim B. 1970. Mobasher B. Shah SP. pp. A model to predict cracking in fiber reinforced concrete due to restrained shrinkage.1991.11:455466. 41. 254 [78] [79] [80] [81] [82] [83] [84] [85] . Ouyang C. ed. 2010. The use of accelerated aging procedures to predict the long term strength of GRC composites. Mag Concr Research 1989). Majumdar AJ. Mechanisms of potential embrittlement and strength loss of glass fiber reinforced cement composites. Proctor BA. UK.
USA. part2: creep. Prata AT. J Geophys Res 1965. Report TVBM3076. Mater Struct 1978.95(4):42935. 2008. Practical prediction of timedependent deformation of concrete.11 (5): 307328. Introduction to micrometeorology.[86] Bazant ZP. [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] 255 . Brocken H. Boston: McGrawHill. Properties and Materials. J Heat Transfer 1985. Sparrow EM. 2004. 1997.36 (8):10216. Lund Institute of Technology. Brutsaert W. part 1: shrinkage. Carmeliet J. Roels S. Guide for modeling and calculating shrinkage and creep in hardened concrete (ACI 209. Control of plastic shrinkage cracking with specialty cellulose fibers. Ravanbakhsh S. Determination of the Liquid Water Diffusivity from Transient Moisture Transfer Experiments. J Therm Envelope Build Sci. Sultan E. Janz M. ACI Committee 209. Arya SP. ACI Mater J 1998. Evaporation from a plane liquid surface into a turbulent boundary layer. Mehta PK. Soroushian P. Amar MB. San Diego: Academic Press. Pasquill’s influence: On the evaporation from various liquids into the atmosphere. Cerny R. Kumaran K. J Appl Meteorol 1997.543:183202. 2006. Division of Building Materials. Adan O.107(1):23942. Monteiro PJM. 2001. Hens H. University of Lund. Hall C. Evaporation of a thin film: diffusion of the vapour and Marangoni instabilities. Panula L. J Fluid Mech 1985. Concrete: Microstructure. Brighton PWM. Michigan. Pel L. J Fluid Mech 2005. Diffusiondriven nonisothermal evaporation.2R08).159:32345. American Concrete Institute. Huang CH. Methods of measuring the moisture diffusivity at high moisture levels. A model for evaporation as a molecular diffusion process into a turbulent atmosphere”.70:501724.27:277305. Boudaoud AI. Pavlik Z.
[106] Vu TH. [105] Garrabrants AC. [103] Plumb OA. The evaporation of water from bricks. New York: Oxford Science Publications. temperature and concentration of solvent.29:3954. J Am Ceram Soc 1990. US Department of Agrriculture. Hayes JC.21(5):775805. Trans Br Ceram Soc 1930. Drying of porous materials at low moisture content. Drying Technol 2001. David JC. [104] Kowalski SJ. and Multiple Solutes in VariablySaturated Media. Prog Org Coat 1987. Germany. Drying Technol 2003.19(1):1320. and Sejna M.73(1):314. New York: Springer. Determination of nonevaporable water in hardened portlandcement paste. Influence of pore size distribution on drying behaviour of porous media by a continuous model. Drying Technol 1999. The Hydrus1d Software Package for Simulating the OneDimensional Movement of Water. OttovonGuerickeUniversität Magdeburg. Kosson DS. [107] Simunek J. [102] Scherer GW. Vergnaud JM. 2003. 256 . Thermomechanical approach to shrinkage and cracking phenomena in drying. MS Thesis. Contribution to the study of florescence. Hoff WD. 2005.19(5):73165. II. Agriculture Research Service. The mathematics of diffusion.0. Evaporation and drying in brick and block materials. Water movement in porous building materialsVI. 1989.17(10):19992011. Hall C. ASTM Bull 1953. Malizewicz M.194:704. [101] Kowalski SJ. 2006. Modelling of drying of coatings: effect of the thickness. [109] Crank J. Build Environ 1984. Riverside. Version 3. Heat. Theory of drying. Webb SW.15:16372. Van Genuchten M. Gu L. Nixon MR.[98] [99] Copeland LE. Illien JP. Thermomechanics of drying processes. [108] Blandin HP. [100] Cooling LF. Modeling moisture transport from a portland cementbased material during storage in reactive and inert atmospheres. CA.
[121] Brutsaert W. Zerbino. 04. Modern drying technology: Computational tools at different scales. 11(3):181187. Germany: WileyVCH. Turner IW. Mujumdar AS. [113] Tsotsas E. D. Wall FT. history. Padevet P. [117] Aldea CM. Bažant Z. [112] American Society for Testing Materials. Shah SP.30(6):438–44. Cem Concr Res 1982. [114] Sebastian P. Dosoretz G. Effect of cracks on hygric and thermal characteristics of concrete. and applications. Effect of cracking in drying and shrinkage specimens. Standard Specification for FiberReinforced Concrete.Moisture capacity and diffusivity. [119] Vejmelková E. Annual Book of ASTM Standards 2010. [120] Torrijos MC.. Sener S. Effect of cracking on water and chloride permeability of concrete.56 (6):715717. Determination of diffusion coefficients from rates of evaporation. 2009. Reidel Publishing Co. Aizawaa Y. Mater Struct 2010. Heat transfer in vacuum membrane distillation: Effect of velocity slip. Agnon Y. 3rd edition. [122] Hisatake K. Evaporation into the atmosphere: theory. J Appl Phys 1993.02.[110] Xi Y. [118] Bažant ZP. R. 2007. ASTM C1116/C1116M10. Molina L. Kim JK. Mat. and Jennings HM. Drying Technol 1994.73 (11):7395401. Tanaka S. Internal cracking and transport properties in damaged concretes. J.12(4):71760. [116] Bažant ZP. Raftshol WJ. Moisture diffusion in cementitious materials.84(5):351–358.43:109–21. ACI Mater J 1987. in Civ.1:258266. Engrg 1999. Constrcution. Cerny. Darmstadt. An investigation of the boundary conditions for a vacuum drying problemIntroducing the transition layer concept. Adv Cem Bas Mater 1994. [123] Kimpton DD. 1982. Karr A.331:11725. J Phys Chem 1952. Effect of cracking on drying permeability and diffusivity of concrete. Boston. Diffusion: mass transfer in fluid systems. Evaporation rate of water in a vessel. Bauphysik 2008. Giaccio G. Section 4. 257 . J Membr Sci 2009. [111] Cussler EL. [115] Ramon G. New York: Cambridge University Press.12:209226.
[136] Bear J. The ‘Physics’ of Soil Water Physics. Schrefler BA. New York. Interfacial phenomena. [132] Nellis G. New York. 2nd edition. Simmons CM. [135] Sposito G. Klein S. Dordrecht.77 (12): 666474. [127] Prandtl L. Tibere GL. New York. Rideal EK. Simmons CM. CRC Press. Cambridge University Press. [134] Philip JR. Flow in porous media. 1998. [131] Tuve GL. Germany. 22 (9S):83S88S. Heidelberg. 8:667678. Fukuda M. New York. Bachmat Y. 1961. Park Y. [137] Lewis RW. Adv Environ Res 2004. Kimura J. [128] Hisatake K. Marcel Dekker Inc. 38. Wiley.[124] Lugg GA. Studies on the movement of soil moisture. Introduction to modeling of transport phenomena in porous media. J Appl Phys 1995.40 (7):10721077. Experimental and theoretical study of evaporation of water in a vessel.190:144967. Diffusion coefficients of some organic and other vapors in air. Maeda M. Third International Mathematical Congress. Heat Transfer. 1997. [129] Davies JT. Ibrahim TH. 2008. Kluwer. Bolz RE. Washington. 1907. [133] Buckingham E. Handbook of tables for applied engineering science. US Department of Agriculture. [130] Wiederhold PR. Park Y. 1976. Annu Rev Fluid Mech 1970. Fukuda Y. [125] Kwon KC. Chem Eng Commun 2003. Ibrahim TH. (1904) 484491. Academic Press. 1991. On fluid motions with very small friction (in German). 258 . 2:177204. Water vapor measurement: Methods and instrumentation. Cleveland. The finite element method in the static and dynamic deformation and consolidation of porous media. Bulletin No. Water Resour Res 1986. Pseudobinary molecular diffusion of vapors into air. Anal Chem 1968. [126] Kwon KC. Molecular diffusion of volatileliquid vapors into air.
Technische Universiteit Eindhoven.35:141–153. Coussy O. Manchester. Int. [144] Sposito G. Springer. Mat. Scaling approach of the convective drying of a porous medium. PhD thesis. New York. 2007.40 (8) :801811. Physics 1931.35 (2): 469480. Poromechanics. Cem Concr Res 2005. Eindhoven. 2011. [143] Richards LA. New York. Drying induced moisture losses from mortar to the environment. UK. Wiley. Lopez CM. Faria R. Moisture transport in porous building materials. [146] Platten AK. University of Manchester Institute of Science and Technology (UMIST).33 (4): 474484. Role of air pressure in drying of weakly permeable materials. 1985. [140] Idiart AE. Netherlands. Struct. [142] Samson E. 44:415–435.15:557566. J. Cushman (ed. Mobasher B. Modeling ion and fluid transport in unsaturated cement systems in isothermal conditions. Academic Press. [141] Pel L. A study of evaporation and drying in porous building materials. Lie group invariance of the Richard equation. Ishida T. in J. J. Marchand J. [149] Azenha M. Capillary conduction of liquids through porous mediums. BaroghelBouny V. 2004. 259 .H. [145] Coussot P. Carol I. [148] Bakhshi M. Snyder K. Experimental observations of earlyage drying of Portland cement paste under lowpressure conditions. [147] Kowalski SJ. Phys. London. Struct. B 2000. 1990: 9398. Cem Concr Compos 2011. Maekawa K. Thermomechanics of drying processes. 1995. Funicular and evaporativefront regimes in convective drying of granular beds. Beaudoin J. Dynamics of Fluids in Hierarchical Porous Media.[138] Coussy O.127:582–592. Kaviany M. [150] Rogers JA. PhD thesis. Eur. ASCE J Eng Mech 2001. Mater.Part I experimental research. Modeling of drying shrinkage of concrete specimens at the mesolevel.1:318333. 2003. [139] Mainguy M. Heat Mass Transfer 1992.).
PhD Dissertation.49 (179): 129137. [159] Šelih J. 260 . 1987. Army Corps of Engineers. 1998.24 (1):81106. Ishida T. Cem Concr Res 1983.72:364371. The University of New Brunswick. Switzerland. Movement of water during drying of fully saturated concrete. Baluch MH. A practical procedure for the analysis of moisture transfer within concrete due to drying.31 (1C):7791.S. 1995. U. Cem Concr Res 1999. CEBFIP model code 1990. [153] Demirkol E. ACI Mater J 1999. Zafar S. J Food Eng 2006. Soranakom C. Erdoğdu F. [158] Šelih J. Palazoğlu TK. Analysis of mass transfer parameters (changes in mass flux. AlGadhib AH. [154] Comite EuroInternational du Beton. Raza A. Canada. Bremner TW. Crack minimization model for Hot Weather Concreting. EPF Lausanne.[151] Rahman MK. [152] Akita H. Sousa ACM. Maekawa K. [162] TM 58227. USACE. Mobasher B. 2011. Earlyage drying of cementitious materials: theory and modeling. Mater Struct 2007. Drying induced moisture losses from mortar to the environment.96 (5):542550.40:813–825. diffusion coefficient and mass transfer coefficient) during baking of cookies. Transp Porous Media 1996.usace.army. Faria R. Part II: numerical implementation. [161] Azenha M. Fujiwara T. A study on moisture diffusion in drying and drying shrinkage of concrete. Ozaka Y. Lee CS. [160] Bakhshi M.13:216224. http://www. under Review. Moisture diffusion of concrete considering selfdesiccation at early ages. [155] Kim JK. [157] Sakata K. [156] Baluch MH. Arabian J Sci Eng 2006. Modeling of shrinkage and creep stress in concrete repair.29 (12):1921–1927. Standard Practice for Concrete Pavements. Moisture transport in initially fully saturated concrete during drying. Rahman MK. Construct Build Mater. AlGadhib AH. Mag Concr Res 1997.mil/inet/usacedocs/.
Farmington Hills. [165] Hover KC. TJ. Washington.15:17. Oct. Rapid evaporation from freshly cast concrete and the Gulf environment. Evaporation of water from concrete surfaces.. Hover KC. USA.. [174] Churchill SW.1. 2007.95C:540–541. D.5:536602. [170] Kohler. 5th Ed. Water Loss Investigations: Lake Hefner Studies. Experimental essays on evaporation. 261 . [164] Australian Technical Bulletin. Technical Bulletin 95/1. 1995.53. McGrawHill. [168] ACI Committee 308. Department of Commerce. 1954:130136. U. 2001. Research Paper No. Causes and prevention of crack development in plastic concrete.C. Proc Portland Cement Association Annual Meeting. Government Printing Office. Nordenson. Australian PreMixed Concrete Association. D. 1964:111 to 1138.S. Evapotranspiration. Introduction to heat transfer. 1991. ed. Washington. 38. May 1955. Technical Report. Committee Report ACI 308R01.103 (5):384389. Fox WE.. American Concrete Institute. Lake and pan evaporation. 5th Canadian Metric ed. [166] Menzel CA. ed. DR Maidment. Const Build Mate 2001. MA.C. MI. Handbook of Appl Hydrol. Geological Survey Professional Paper 269. Ozoe H. Canadian Portland Cement Association. John Wiley & Sons. ACI Mater J 2006. 1802. Evaporation from pans and lakes.S. 1954:127148. U. ACI Manual of Concrete Practice. Proc Manchester Literary and Philosophical Society. [171] Kohler. DeWitt DP. McGraw Hill. [167] Dalton J.[163] CSA Standard A23. J Heat Transfer 1973. Wiley. 1993:4. Handbook of Hydrology. Evaporation. [172] Shuttleworth WJ. [169] AlFadhala M. VT Chow. Cracks in concrete due to plastic shrinkage and plastic settlement. [173] Veihmeyer FJ. Design and control of mixes. New York.14. A correlation for laminar free convection from a vertical plate. Guide to curing concrete. MA. [175] Incropera FP.
New York. Perry's Chemical Engineers' Handbook. Butterworth & Co Ltd. Comparison of dry deposition predicted from models and measured with a water surface sampler. and solids drying. 2008. MA. 1979:160163. Hestermans P. New York. [179] Reist PC. Inc: New York. John Wiley and Sons.A Tribute to Adrian Pauw. and soil. London. SP27: Creep. [188] ACI Committee 209. 2007. McGrawHill. 5th ed. Chemodynamics: environmental movement of chemical in air. Time dependent concrete properties related to designstrength and elastic properties. Prediction of creep. [182] Touloukian YS. 2000. Guide for curing Portland cement concrete pavements. [187] Branson DE. 1971: 257277. Thermophysical properties of matter. 3rd Edition. England. 1971. [177] Crawford M. McGraw Hill: New York. Aerosol science and technology. 1982:193300. [181] Lienhard IV JH. [183] Poling BE. McLean. Christiason ML. creep and shrinkage. Cambridge. A heat transfer textbook. McGrawHill. [178] Friedlander SK.11: Viscosity. 8th Ed. Prausnitz JM. Psychrometry. Farmington Hills. Phlogiston Press. Drying of solids in the chemical industry. [184] Poole TS. 31:272278. evaporative cooling. Farmington Hills.[176] Thibodeaux LJ. 1977. [185] Nonhebel G. 2nd ed. water. IFI/Plenun. NY. ACI SP76: Designing for Creep and Shrinkage in Concrete Structures . 262 . Saxena SC. Shrinkage and Temperature Effects. VA. 1970. Moss AAH. Holsen T. Perry R. O'Connell J. Noll K. 1993. US. John Wiley and Sons: New York. The properties of gases and liquids. American Concrete Institute. American Concrete Institute.II: FHWAHRT05038. Lienhard V JH. dust. MI. Environ Sci Technol 1997. 1976. 2006. MI. and haze: Fundamentals of aerosol behavior. Smoke. [180] Yi SM. John Wiley and Sons: New York.P. Air pollution control theory. [186] Green D. Federal Highway Administration. shrinkage and temperature effects in concrete structures.
Stevin Laboratory. A review of some of the main results of a research on the aging phenomena of concrete: Effect of moisture conditions on strength.27: 3–14. [190] Ketelaars AAJ. Cem Concr Res 2006. Computational modelling of masonry creep and shrinkage. Eindhoven University of Technology. 1993:103113.[189] Perre P. Total deformation of loaded drying concrete. Exposition et application á suivre et des formules á employer dans les questions de duistribution d’eau. [193] Stroeven P. Drying deformable media: kinetics. [197] Alvaredo AM. Hauggaard AB. Cem Concr Res 1980. 263 . 1856. 1973. Geometric probability approach to the examination of microcracking in plain concrete. [200] van Zijl GPAG. Netherland. Victor Dalamont. [195] Darcy HPG. Barcelona. Spain. Les fontaines publiques de la ville de Dijon. [199] Bazant ZP. Drying with internal vaporisation: Introducing the concept of identity drying card (IDC). Baweja S.10: 601610. Stereometrische metallographie. PhD dissertation. Leipzig. France.36:15081515. shrinkage and creep of mature concrete. PhD dissertation. Technological University of Delft. Identification and quantification of cracks in concrete by optical fluorescent microscopy. Delft University. Cem Concr Res 1974. [198] Bazant ZP. Drying creep of concrete: Constitutive model and new experiments separating its mechanisms. PhD Dissertation.13 (57):10771097.4:761771. Drying Technol 1995. RILEM Proc 5th Int Symp Creep and Shrinkage of ConcreteMathematical Modeling. [194] Litorowicz A. Roelfstra PE. 1974. Xi Y. Wittmann FH. Netherland. [196] Wittmann FH. [192] Stroeven P. Ulm FJ. Mater Struct 1997. J Mater Sci 1979. [191] Salytkov SA.123(11):11881194 . 1992. J Engrg Mech 1997. I: aging and drying effects.14 (5):11411151. shrinkage and stresses. 1999. Shrinkage as influenced by strain softening and crack formation. [201] Pihlajavaara SE. Paris. Some aspects of micromechanics of concrete. VEB Deutscher Verlag für Grundstoffindustrie. Microprestresssolidification theory for concrete creep.
Eng Struct 2005. PijaudierCabot. The Netherlands. Sellier A. Meftah F. and J. Mater Struct 1997. Illston JM. Meftah F. R. Comportement différé du béton dans les enceintes de centrales nucléaires: analyse et modélisation. Tension softening and cracking in drying concrete.[202] BaroghelBouny V. Meftah F. JM. Structural effects of drying shrinkage. shrinkage. Lisse. J. de Borst. Torrenti JM. 5th Int. van Mier. shrinkage and durability mechanics of concrete and other quasibrittle materials 2001:245–250. [204] Granger L. G. Torrenti. Matensson A. Mater Struct 1988. 1968. On the prediction of delayed strains for concrete subjected to drying and loading simultaneously. Mazars.30:588598. Torrenti JM. [207] Benboudjema F. M. Fracture Mechanics of Concrete Structures. 1996. Characterization and identification of equilibrium and transfer moisture properties for ordinary and highperformance cementitious materials. 264 .131 (11):11951199. on Chemistry of Cement. 132. Ph. The effect of drying on the drying creep of hardened cement paste.21:41624. J Eng Mech 2005. Structures and physical properties of cement paste. Cem Concr Res 1999.29(8):12251238. Thoughts about drying shrinkage: Experimental results and quantification of structural drying creep. [210] Benboudjema F. Cuffaro P. Tokyo. [206] Verbeck GJ. Hehnuth RH. Dahlblom O. Cem Concr Res 1984. [208] Benboudjema F. MarneLaVallée. Symp. Ecole Nationale des Ponts et Chaussées. Proc.D. [205] Granger L. F. Torrenti JM. Japan. Torrenti JM. Proc. Meftah F. Drying creep: An elastoplastic damage approach of the structural effect. G. Interaction between drying. 2001. thesis. Mainguya M. [209] Benboudjema. Heinfling G. Acker P. Coussy O. [203] Thelandersson S.27: 239–250. eds. creep and cracking phenomena in concrete.14(3):329–338.1:169–176. Lassabatereb T. France [in French]. [211] Day RL. of CONCREEP 6: Creep.
Modelling of elastoplastic damage in concrete due to desiccation shrinkage. ENPC Press. [213] Gawin D. Mindess. Barcelo L. Pesavento F. [222] Bissonnette B. PhD dissertation. Lange DA. [217] Grasley ZC. 1972:116. Skalny. 265 . [220] Grasley ZC. Mounajed G. Ed. [216] Obeid W. [223] Gray WG. PrenticeHall. 2(1). Thermohygromechanical degradation of concrete: From coupled 3D material modeling to durability oriented multifield structural analyses. 1981.40:163–174. le béton hydraulique. Young JF. Thermal dilation and internal relative humidity of hardened cement paste. Schrefler BA.35:229–236. [214] Grasberger S. Experimental identification of Biot’s hydromechanical coupling coefficient for cement mortar. Mater Struct 2004. Paris. 1982. ASTM J Test Eval 2003. 20:521–538.37:244–256. J. Thermodynamic approach to effective stress in partially saturated porous media. Charron JP. [218] Mindess S. Eur J Mech A/Solids 2001. 2002 [in French]. Modeling drying Shrinkage stress gradients in concrete. Delagrave A. Lange DA. [221] Benboudjema F. A viscoelastic approach for the assessment of the drying shrinkage behaviour of cementitious materials. [215] Benboudjema F. Concrete. Mater Struct 2002. Burlion N.26(2): 115122. Modélisation des déformations différées du béton sous sollicitations biaxiales. Modelling of deformations of high strength concrete at elevated temperature. Mater Struct 2004. Early age behavior of cementbased materials. Meftah F. [224] Bažant ZP. S. Marchand J. Université de MarnelaVallée. 26:759–774.[212] Bourgeois F. Shao JF. Meschke G. Int J Numer Anal Methods Geomech 2002. Torrenti JM. Thermodynamics of hindered adsorption and its implications for hardened cement paste and concrete. Mater Struct 2007. Application aux enceintes de confinement de bâtiments réacteurs des centrales nucléaires. 2001:243326. New Jersey. American Ceramic Society.A. Alliche A. [219] Baron J. Schrefler B. Mater Sci Concr VI. France. Mater Struct 2007. Cement and Concrete Research.40:311–317.37:218–236.
19 (6):423436.[225] Gallé C. 266 . and freezedrying. [229] Lamond J. Peilert J. vacuum. A classical approach. Philadelphia. Garboczi EJ. Bažant ZP. [230] Xi Y. Adv Cem Based Mater 1994. PA. 2006.1:248257. ASTM International.21(23):325344.STP169D. 22(9):971992. Cem Concr Res 2001. Effect of drying on cementbased materials pore structure as identified by mercury intrusion porosimetry: A comparative study between oven. Physical structure of hardened cement paste.31 (10):14671477. [228] Hansen TC. Percolation of phases in a threedimensional cement paste microstructural model. Molina L. Brownyard TL. Significance of tests and properties of concrete and concretemaking materials. J Am Concr Inst 1947. Jennings HM. Cem Concr Res 1991. [226] Bentz DP. [227] Powers TC. Mater Struct 1986. Studies of the physical properties of hardened Portland cement paste. Moisture diffusion in cementitous materialsAdsorption isotherms.
This action might not be possible to undo. Are you sure you want to continue?