MODELLING OF THERMAL STRESSES IN

HYDRATING CEMENT CONCRETE AND

ESTIMATING CRACKING RISK

SRIJEET HALDER

DEPARTMENT OF CIVIL ENGINEERING

INDIAN INSTITUTE OF TECHNOLOGY DELHI
JUNE 2015
 MODELLING OF THERMAL STRESSES IN HYDRATING
CEMENT CONCRETE AND ESTIMATING CRACKING RISK

By

SRIJEET HALDER

Department of Civil Engineering

Submitted in fulfilment of the requirement of the degree of

MASTER OF TECHNOLOGY

to the

INDIAN INSTITUTE OF TECHNOLOGY DELHI

JUNE 2015
CERTIFICATE

This is to certify that the thesis titled as “MODELLING OF THERMAL STRESSES IN

HYDRATING CEMENT CONCRETE AND ESTIMATING CRACKING RISK” being
submitted by Srijeet Halder (Entry No. 2013CET8380) for the award of degree of the Master
of Technology in Construction Engineering and Management from Indian Institute of
Technology, Delhi, is a record of the bonafide work he has carried out under my supervision.

(Supervisor) (Supervisor)

Dr. Shashank Bishnoi Dr. Abhijit Ganguli

Department of Civil Engineering Department of Civil Engineering
Indian Institute of Technology, Delhi Indian Institute of Technology, Delhi
Date: Date:
Place: Place:

i
ACKNOWLEDGEMENT

I take this opportunity to express my deep gratitude to all people who have extended their
cooperation in various ways during the course of the study. It is my pleasure to acknowledge
the help of those individuals.

Firstly, I would like to express my sincere gratitude to my supervisors Dr. Shashank Bishnoi
and Dr. Abhijit Ganguli, Deptt. of Civil Engineering, for their valuable guidance, inspiration
and encouragement during all stages of this project work. They have remained my constant
source of motivation for the period I have worked on this project. Their knowledge and
experience in my field of research have helped me a lot. If not for them, this project would not
have been even started.

I would also like to thank Prof. A. K. Nagpal, Deptt. of Civil Engineering, for helping me in
understanding Finite Element Method. He has showed immense patience and kind attitude
while teaching me some specific topics which I needed exclusively for my project.

I thank IIT Delhi for providing world-class infrastructure and technological backbone for
students like me to work upon. It has provided us the best personnel in the respective fields,
who all have supported me and allowed me to work on my project smoothly.

I would like to thank my parents for providing me wisdom about life and my friends Jatin
Kumar, Ajay Kumar, Rekha Walia and all others for providing me emotional as well as
intellectual support

At last, there is a nameless, faceless supporter of mine, who provided me a vast pool of
knowledge and that is the whole internet community. I cannot thank them individually, but
they all are inevitable part of my knowledge.

SRIJEET HALDER
2013CET8380

ii
ABSTRACT

A new “Finite Element Method”-based computer program called ‘SafeConc’ has been
developed in JAVA programming language. Main purpose of this program is to find
temperature variations in a cement concrete body during its early age and calculate stresses
associated with it. This allows users to find out the temperature distribution, strength
development, thermal stresses and risk of cracking in the body. It is based on an semi-empirical
hydration model developed by Lin and Meyer in 2009, which itself is based on Arrhenius
equation. It also uses finite element method to distribute heat and stresses. Strength is
calculated on the basis of strength-porosity relationships. This program is very flexible and
accounts for chemical composition of cement, water-cement ratio, fineness, size of the body,
etc.

iii
CONTENTS

CERTIFICATE ........................................................................................................................... i

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

ABSTRACT .............................................................................................................................. iii

CONTENTS .............................................................................................................................. iv

LIST OF FIGURES ................................................................................................................. vii

LIST OF TABLES .................................................................................................................... ix

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

1.1 Heat of hydration of concrete ...................................................................................... 1

1.2 Thermal Cracking ........................................................................................................ 2

1.3 Numerical modeling .................................................................................................... 3

1.4 Scope and layout of report .......................................................................................... 4

LITERATURE REVIEW ............................................................................. 5

2.1 Introduction ................................................................................................................. 5

2.2 Research gap ............................................................................................................... 8

CEMENT HYDRATION MODELING ....................................................... 9

3.1 Physical properties of cement ..................................................................................... 9

3.1.1 Fineness................................................................................................................ 9

3.1.2 Particle size distribution....................................................................................... 9

3.2 Cement chemistry...................................................................................................... 10

3.2.1 Phases of cement ................................................................................................ 10

3.2.2 Hydration of cement phases ............................................................................... 11

3.2.3 Heat of hydration ............................................................................................... 12

3.3 Lin and Meyer model ................................................................................................ 13

3.4 Theory of uniform reacted thickness ......................................................................... 16

3.5 Calculation of porosity .............................................................................................. 16

iv
 3.5.1 Microstructure Development ............................................................................. 16

3.5.2 Alternate approach ............................................................................................. 18

3.6 C-S-H density ............................................................................................................ 19

SIMULATION METHODOLOGY ........................................................... 23

4.1 Heat Transfer solution ............................................................................................... 23

4.1.1 Steady-state heat transfer ................................................................................... 23

4.1.2 Transient heat flow ............................................................................................ 27

4.2 Plane Stress/Strain Analysis ...................................................................................... 28

4.2.1 Static Analysis – finite element formulation ..................................................... 28

4.2.2 Dynamic Analysis .............................................................................................. 31

4.3 Numerical Methods and Optimization ...................................................................... 32

4.3.1 Numerical integration – Gauss Quadrature........................................................ 32

4.3.2 Cholesky Factorization ...................................................................................... 34

4.4 STRENGTH ESTIMATION..................................................................................... 36

4.4.1 Strength-Porosity relationship ........................................................................... 36

4.4.2 Extended Zheng’s Model ................................................................................... 37

4.4.3 Elastic modulus .................................................................................................. 38

4.5 CRACKING RISK .................................................................................................... 38

STRUCTURE OF SAFECONC ................................................................. 41

5.1 Program structure ...................................................................................................... 41

5.2 Graphical User Interface (GUI)................................................................................. 42

RESULTS, DISCUSSION AND VALIDATION ...................................... 49

6.1 Degree of hydration ................................................................................................... 49

6.2 Adiabatic temperature rise ........................................................................................ 50

6.3 Heat production rate .................................................................................................. 51

6.4 Temperature rise with convection loss (semi-adiabatic) ........................................... 53

v
 6.5 Strength Development ............................................................................................... 56

6.6 Phases ........................................................................................................................ 57

6.7 Displacements ........................................................................................................... 58

6.8 Stresses ...................................................................................................................... 60

6.9 Cracking Risk ............................................................................................................ 61

6.10 Discussion .............................................................................................................. 62

SUMMARY & CONCLUSION ................................................................. 63

7.1 Summary ................................................................................................................... 63

7.2 Conclusion................................................................................................................. 64

7.3 Potential use of the software ..................................................................................... 64

7.4 Future scope of study ................................................................................................ 64

REFERENCES ........................................................................................................................ 67

FLOWCHARTS ...................................................................................................................... 73

vi
LIST OF FIGURES

Figure 2.1 Effect of temperature on strength of concrete.[19] .................................................. 6

Figure 2.2 Variation of convection coefficient with age of concrete.[9] ................................... 7
Figure 3.1 Effect of fineness on rate of heat evolution [12] ...................................................... 9
Figure 3.2 Reactivity of various phases of cement. [14] ......................................................... 11
Figure 3.3 Rate of heat evolution with time [12] ..................................................................... 12
Figure 3.4 Uniform Reacted Thickness of cement particles irrespective of initial size [15]... 16
Figure 3.5 Flowchart for simulating microstructure development .......................................... 17
Figure 3.6 Illustration for diffusion of hydration products at normal and elevated temperatures
.................................................................................................................................................. 20
Figure 3.7 Calculated C-S-H density with temperature [10] ................................................... 20
Figure 3.8 Strength development at different temperatures with (a) constant C-S-H density, and
(b) temperature-dependent C-S-H density. Source: SafeConc output ..................................... 21
Figure 4.1 Best fit for existing models with experimental results obtained by Chen et al.
[10] ........................................................................................................................................... 37
Figure 5.1 Internal structure of SafeConc – the program ........................................................ 41
Figure 5.2 Flowchart for calculation of degree of hydration ................................................... 42
Figure 5.3 Graphical User Interface of SafeConc – the program ............................................ 43
Figure 5.4 Main Data Input Form ............................................................................................ 43
Figure 5.5 Thermal contour plot .............................................................................................. 45
Figure 5.6 Hydration data input form ...................................................................................... 45
Figure 5.7 Showing deformed state in SafeConc..................................................................... 46
Figure 5.8 Graph of porosity vs time plotted by Safeconc ...................................................... 46
Figure 6.1 Degree of hydration vs time curve for w/c=0.35. Experimental data from [16] .... 49
Figure 6.2 Adiabatic temperature rise curve for w/c=0.3. Experimental data: Bentz (1999) [16]
.................................................................................................................................................. 50
Figure 6.3 Adiabatic temperature rise curve for w/c=0.35. Experimental data: Bentz (2009)
[16] ........................................................................................................................................... 50
Figure 6.4 Adiabatic temperature rise curve for w/c=0.45. Experimental data from Bentz (2009)
[16] ........................................................................................................................................... 51

vii
Figure 6.5 Rate of heat release curve for w/c 0.5 at 200C (isothermal). Experimental data from
Paine (2005) [43] ..................................................................................................................... 52
Figure 6.6 Rate of heat release curve for w/c 0.5 at 400C (isothermal). Experimental data from
Paine (2005) [43] ..................................................................................................................... 52
Figure 6.7 Rate of heat release curve for w/c 0.5 at 200C (isothermal). Experimental data from
Paine (2005) [43] ..................................................................................................................... 53
Figure 6.8 Experimental setup in [22] showing insulating boundaries around the concrete ... 54
Figure 6.9 Specimen diagram showing dimensional layout of the specimen .......................... 54
Figure 6.10 Temperature rise in a concrete prism. Mix 1 in [22] ............................................ 55
Figure 6.11 Temperature rise in a concrete prism. Mix 2 in [22] ............................................ 55
Figure 6.12 Strength development curve simulated by SafeConc ........................................... 56
Figure 6.13 Variation of 28-day strength with w/c ratio (Simulation results) ......................... 56
Figure 6.14 Phase generation and consumption as simulated by SafeConc ............................ 57
Figure 6.15 Deflection of simple cantilever beam (L=10m,B=1m,D=1m,E=30GPa) loaded at
free end. Theoretical=1.333*10-2, Model=1.313*10-2: Error=1.5% ........................................ 58
Figure 6.16 Deflection of centrally loaded simply supported beam (L=10m,
B=1m,D=0.6m,E=30GPa). Theoretical=3.858*10-3, Model=3.833*10-3: Error=0.6%........... 59
Figure 6.17 Sample problem analyzed in (a) SafeConc, and (b) ABAQUS/CAE (parameter –
horizontal displacement) .......................................................................................................... 60
Figure 6.18 Sample problem analyzed in (a) SafeConc, and (b) ABAQUS/CAE (parameter –
horizontal stress) ...................................................................................................................... 60
Figure 6.19 Sample problem analyzed in (a) SafeConc, and (b) ABAQUS/CAE (parameter –
vertical stress) .......................................................................................................................... 61
Figure 6.20 Cracking risk ........................................................................................................ 61
Figure 0.1 Flow Chart for Main Program ................................................................................ 73
Figure 0.2 Flowchart for calculation of degree of hydration ................................................... 74
Figure 0.3 Flowchart for microstructure development ............................................................ 75

viii
LIST OF TABLES

Table 3.1 Phases of ordinary portland cement ......................................................................... 10

Table 3.2 Enthalpy of phases of cement .................................................................................. 12
Table 3.3 Densities of cement phases ...................................................................................... 19
Table 4.1 Gauss points and their weights ................................................................................ 34
Table 6.1 Parameters for modelling of hydration for the material used in [16] ...................... 49
Table 6.2 Parameters for modelling of hydration for the material used in [43] ...................... 51
Table 6.3 Mix design and composition of cement taken for experimental analysis in [22] .... 54

ix
GLOSSARY

Abbreviations Cement Chemistry Notations

FEM – Finite Element Method C – CaO

FDM – Finite difference method S – SiO2

SCM – Scanning electron Microscopy H – H2O

PSD – Particle size distribution A – Al2O3

w/c – water to cement ratio F – Fe2O3

XRD – Xray diffraction $ - SO3

C3S – Tricalcium Silicate

Notations C2S - Dicalcium Silicate

T – Temperature C3A – Tricalcium Aluminate

t – time C4AF - Tetracalcium Aluminoferrite

Ω – Resistance C-S-H - Calcium Silicate Hydrate

σ - Stress CH - Calcium Hydroxide

τ – Shear Stress C6A$3H32 – Ettringite

γ –Shear strain C4A$H12 – Monosulphate

ε - Strain C$H2 - Gypsum

k – Thermal conductivity

ρ - Density

x
 INTRODUCTION

Concrete is the most widely used man-made material. It is relatively cheap, easy to
manufacture, easy to handle and the ingredients of concrete are abundant. Concrete is fluid
when mixed which makes it suitable to form irregular shapes. However, it hardens as hydration
of cement progresses, which gives strength to concrete.

1.1 HEAT OF HYDRATION OF CONCRETE

Hydration of cement in concrete is an exothermic process which liberates heat in the range of
500-1000 joules per gram of cement. Moreover, conductivity of concrete is relatively low
which slows down the dissipation of heat. In large structures such as gravity dams, raft
foundations, deep beams etc., a lot of heat is generated in the body of the structure due to
cement hydration. This causes temperature at the core to rise up to a very high level whereas
temperature at the surface is lower due to dissipation of heat to the environment. Consequently,
a temperature gradient is created across the body which generates thermal stresses. Ideally, no
stress is generated if the temperature is uniform throughout the body and there is no restraint.

Concrete gains its strength gradually as the hydration process continues due to gradual decrease
in porosity; it is very weak at early age. Cracks may appear at any point of time if the stresses
generated due to temperature gradient is higher than the strength developed, which may
increase in width and number because of stress concentration around these cracks. These cracks
may lead to reduced durability of concrete or may result in failure of the structure altogether.

Temperature also effects the ultimate strength of the concrete and it has been seen that at high
temperatures concrete has more porosity, which results in a lower ultimate strength. Delayed
Ettringite Formation (DEF), which is a phenomenon regarding concrete chemistry, takes place
at high temperatures. Ettringite, which is formed by the reaction of aluminate (C3A) and
alumino-ferrite (C4AF) with gypsum and has needle-like microscopic structure. It is the cause
of swelling in concrete. High temperature impedes ettringite formation. Failure of dams have
been reported after few years of use due to DEF. Therefore, it may seem obvious to keep
temperature of concrete within limits to prevent any undesirable events from occurring.

1
1.2 THERMAL CRACKING
As already has been discussed, hydration of concrete produces heat which increases the
temperature at the core of a concrete element while the surface remains relatively cooler. This
difference is more pronounced in concrete elements with large dimensions. Moreover, higher
temperature also increases rate of reaction of cement phases which produces more heat and the
core temperatures increases further. Due to increased rate of reaction core of the concrete
element also gains strength faster and loses its creep capacity at a higher rate than surface of
the element. As the internal portion of the concrete undergoes cooling through transfer of heat
to the surroundings, it is subjected to tensile stress due to restrain provided by the outer layers.
If this tensile stress is greater than the tensile strength acquired by the concrete, cracks appear
and propagate. The outer portion of the concrete also experiences compressive loading and
cracks are formed in this portion as well.

The temperature distribution in the continuum of the body during the early age of concrete
depends upon many factors. Some of these are:

 Cement composition
 Cement fineness
 Enthalpies of various phases of cement
 Specific heat of aggregates and cement
 Thermal conductivity of aggregates and cement
 Ambient temperatures
 Size of concrete element

Some of the measures that can be taken to prevent thermal cracks in concrete are:

 Mixing concrete with ice or chilled water before placing

 Reducing size of the element
 Using low-heat cement
 Providing mechanisms like ducts carrying cold water to dissipate heat from the body faster
 Insulating the surface of concrete to maintain uniform temperatures throughout the body
 Providing shadow to the hydrating concrete to prevent further influx of heat from sun

2
1.3 NUMERICAL MODELING
Heat liberated by hydrating cement is modeled using numerical simulation. In this research no
new simulation methods has been developed, but already available techniques have been used.

The heat generated within the concrete is distributed in the whole body and transferred to the
environment by convection. This is modeled by Fourier’s equation for heat flow.

𝜕 𝜕𝑇 𝜕 𝜕𝑇 𝜕 𝜕𝑇 𝜕𝑇
(𝑘𝑥 𝜕𝑥 ) + 𝜕𝑦 (𝑘𝑦 𝜕𝑦) + 𝜕𝑧 (𝑘𝑧 𝜕𝑧 ) + 𝑄 = 𝑐. 𝜌 𝜕𝑡 (1.1)
𝜕𝑥

Eq. (1) is a partial differential equation on four variables – 3 space and 1 time. This equation
can be solved by employing Finite Element Method (FEM). To solve the equation, the whole
domain of the problem is divided or broken up into finite number of very small elements. The
number of elements is selected by negotiating between accuracy and computational effort.
FEM has many advantages over other methods for solving partial differential equations. One
of them is that, any irregular shape can be modeled using FEM and reasonable accuracy is
expected without any special considerations. However still, FEM is an approximate method
and does not give very precise results, but can be acceptable for most of the practical cases.

The microstructure development of cement is modeled by simulating reaction of cement

particles in a sample domain which gives the porosity of the concrete at all times. To keep the
program executable on general desktop computers or laptops, the sample domain is chosen to
be very small in the range of few hundred micrometers. This captures the hydration of cement
particles within reasonable accuracy level. However, reaction of cement is temperature-
dependent. Therefore, simulation of hydration and transfer of heat requires to be executed
simultaneously. For this, a java-based computer program called ‘SafeConc’ has been
developed. Due to object-oriented nature of java, this program is able to simulate generation of
heat within elements and distribution of this heat to the other elements in parallel to each other.
To properly incorporate the parallel execution of the models, SafeConc is designed in a
modular form with generation and distribution of heat implemented in different inter-linked
modules. Other modules will be developed to calculate stresses and determine strength
development in the later stages of the research. These modules are self-sufficient and complete,
such that each of these can be used independently for different purposes like to take a look into
the microstructure development of cement or transfer heat from sources other than hydration
of cement.

3
1.4 SCOPE AND LAYOUT OF REPORT
This study employs numerical model for cement hydration developed by Lin and Meyer (2009)
[1]. It is an empirical model based on regression analysis of experimental data on various
cements. Although, the authors have factored all the measurable parameters like chemical
composition of cement, fineness, water cement ratio etc., yet some immeasurable
miscellaneous parameters/factors may have been overlooked. Because of this, very high
accuracy cannot be expected from this model as well as SafeConc which is based on it.

This study doesn’t take non-linearity of concrete into account, for efficient use of computing
resource. Since mostly tension cracks are formed due to thermal gradient and concrete behaves
almost linearly in tension region, this assumption doesn’t introduce large error.

Many studies have been conducted in this area. Some of them have developed similar computer
packages. But all of them lacked user-friendly environment for the modeling process or are not
available open-source. SafeConc is a GUI-based computer program which incorporates object-
oriented programming methodology and provides intuitive modeling environment.

Before moving to the discussion on internal working structure of SafeConc, few of the
researches in the area of this study shall be discussed, followed by development of the finite
element formulations regarding heat flow and small discussion on cement hydration kinetics
and microstructure development.

4
 LITERATURE REVIEW

2.1 INTRODUCTION
Thomas et al. (2011) [2] reviewed various cement hydration models such as single particle
models, mathematical nucleation & growth models, and vector & lattice-based approaches to
simulate microstructure development that have been developed in past 40 years. JMAK
nucleation and growth equation, also known as Avrami equation has a good fit with
experimental results for degree of hydration of cement.

As already discussed in previous chapter, Lin and Meyer (2009) [1] proposed a numerical
model for cement hydration. They carried out regression analysis on experimental data on
hydration of various cements. They derived a semi-empirical relationship between rate of
hydration of cement and its different properties, viz., phase compositions, fineness etc.

𝛼̇ = 𝐴𝛼 𝜂𝛼 𝑒𝑥𝑝(− 𝐸𝑎 ⁄𝑅𝑇) 𝑒𝑥𝑝(𝐸𝑎 ⁄𝑅. 293) (2.1)

where, 𝛼̇ is rate of reaction of cement, 𝐴∝ called as chemical affinity incorporates effect of

water-cement ratio and fineness as well as proportion of phases in cement and considers ‘so
called’ water-shortage effect by using concept of ultimate degree of hydration which depends
upon water-cement ratio. 𝜂∝ is permeability of hydrated cement layer and represents
increasing difficulty of water to penetrate layer of hydration products to reach anhydrous
cement. Ea, known as apparent activation energy, depends upon chemical composition. Cement
hydration is a complex process which involves multiple reactions occurring simultaneously,
which makes it difficult to determine activation energy individually for each reaction. Thus,
commonly used value which is measured by adiabatic or isothermal calorimetry of cement
paste is a superposition of activation energies for all these reactions, hence the prefix
‘apparent’.

[3] and [4] independently developed hydration models and coupled them with Finite Element
Method to get the temperature distribution inside a body of hydrating cement concrete.
However, their hydration models significantly differ in approach. The former model is based
on a multiphasic approach and allows for interaction between cement and surrounding
environment. Whereas, the latter applies Artificial Neural Network (ANN) in modeling of

5
cement hydration for determination of adiabatic temperature curve of hydrating cement. Both
have their own advantages and limitations.

In a similar study by Najafi and Ahangari [8], artificial neural network has been used to model
liberation of heat from hydration of cement and programmed using Matlab® programming
environment.

Yazdi (2013) [14] applied Galerkin Finite Volume method for analysis of a typical mass
concrete structure (a dam of 30 m height) using unstructured triangular mesh. A novel method
for application of natural boundary conditions was also developed which gives computationally
more efficient solution, thus reducing the computing time.

Wan (2002) [5] developed a computer program similar to SafeConc in C++ programming
language. The program called as ‘Concrete Cracking Control (CCC)’ uses finite difference
method (FDM) to distribute heat generated from hydration of cement in the body and transfer
to the environment through the exposed surfaces. It then uses FEM to calculate thermal stress
distribution inside the same body. FDM has some inherent limitations, couple of them being
inability to model irregular geometries and requirement for finer mesh as compared to FEM.
Therefore, finite element method is superior to finite difference method in some aspects and it
has been used for both temperature and stress distribution in SafeConc.

Figure 2.1 Effect of temperature on strength of concrete.[19]

[5] also ignored effect of temperature on strength development of concrete. It has been
repeatedly reported that high temperature at early age of concrete has adverse effects on its
ultimate strength (Figure 2.1). This is a major limitation of this study.
6
Nanayakkara [6] discusses the role of delayed ettringite formation in failure of many concrete
structures in U.S. High temperature is the major cause of DEF. Ettringite has needle like
structure and causes expansion in concrete. If ettringite is formed in the plastic state, no stresses
are induced due to this expansion. However, in hardened state this expansion can be disastrous.

Lee et al. [9] studied the variation of coefficient of convection with age of concrete. They
reported a huge variation in coeff. of convection with and without evaporation of water from
hydrating concrete by testing concrete with and without curing blanket.

(a) With evaporation effect (b) Without evaporation effect

Figure 2.2 Variation of convection coefficient with age of concrete.[9]

Chen et al. (2013) [10] reviewed various strength-porosity relationships and found that Schiller
and Ryshkewithch models are numerically indistinguishable and give best fit with experimental
results for general porous materials. They reported that Extended Zheng’s model gives best
representations of experimental data on the strength of cement mortar. It can also be applied
for other cement-based materials.

Kumar and Bhattacharjee [11] performed Mercury Intrusion Porosimetry (MIP) on cores
drilled from concrete beams cast in laboratory. They reviewed the popular strength-porosity
relationships, namely, Balshin’s, Ryskovitch’s, Schiller’s, Hasselman’s, Luping’s and
Atzeni’s. They found poor correlation for all the above relationships, in the range of 50 to 90%.
They proposed a new model for concrete based on Griffith’s fracture theory.

Gallucci, Zhang, and Scrivener [13] in their study has calculated apparent densities of C-S-H
using XRD-Rietveld analysis. The actual densities of solids were not measured in the study but

7
were obtained by dividing the mass of C-S-H measured from XRD with its volume derived
from volume balance, hence the word ‘apparent’.[13] Since it is an approximate approach, they
scaled the values in such a manner that calculated density at 20oC matches the data from most
reliable literatures. This research does not aim to predict the actual density of C-S-H gel but to
study its variability.

2.2 RESEARCH GAP

Although, many theoretical and numerical models have been developed in past few decades
for analysis of cement hydration in concrete. Its practical application using computer programs
has not been done by many. With the increased computing capacities of modern computers,
these models no longer require high-end mainframe computers. It is now possible to run these
models on regular desktop computers and/or laptops.

So far there is no computer program available to apply numerical modeling in cement hydration
and use it for prediction of temperature and stresses in concrete. This study is focused on
developing such a program and release it open-source. The main purpose of the program will
be to aid practicing engineers and contractors in dealing with mass concrete structures where
release of large amount of heat from hydrating cement may cause temperatures inside core of
the structures to rise at devastatingly high levels. Other researchers may also benefit from this
program to analyze the effect of various parameters on the heat of hydration of cement and its
distribution within the concrete mass.

One of the main feature of this program is that it considers effect of temperature on rate of
hydration of cement as well as strength development of concrete, which have been neglected
by many others. CSH density is considered to be temperature-dependent, which is an important
feature in strength prediction of concrete as discussed later.

8
 CEMENT HYDRATION MODELING

3.1 PHYSICAL PROPERTIES OF CEMENT

3.1.1 Fineness
Fineness of cement is measured as cumulative surface area of cement particles per unit volume
of cement. It is measured in Blaine (m2/kg) using Blaine’s air permeability apparatus.

Bentz et al. [10] found that fineness has huge effect on cement hydration. Finer cement reacts
faster with water due to more surface area. Also, there is a limitation on water to penetrate
through hydrated layer of cement. In larger particles, when hydrated layer of cement has
enough thickness, water is no longer able to penetrate through the layer for hydration to
continue. After this, hydration stops for the particle, therefore decreasing the ultimate degree
of hydration for the cement.

Figure 3.1 Effect of fineness on rate of heat evolution [12]

3.1.2 Particle size distribution

Fineness is a measure of mean diameter of cement particles, whereas the particles may hugely
vary in diameter in same cement sample. Finding particle size distribution is not simple for
9
cement, since its particles are generally smaller than 60μm and sieve analysis cannot be
performed on such particles. Only methods available for finding particle size distribution are
X-ray diffraction and Sedimentation.

However, theoretical methods like Rosin-Rammler equation (Eq. 3.1) may also be used to get
a quick estimate of particle size distribution of cement.

𝑑 𝑤𝑑
𝑓 = 1.0 − 𝑒𝑥𝑝(− (𝑑 ) ) (3.2)
𝑚

where, dm is the mean diameter of cement particles, wd is width which ranges from 0.98 to 1.2
and f is fraction finer than the diameter d.

3.2 CEMENT CHEMISTRY

3.2.1 Phases of cement

Ordinary Portland Cement (OPC) has 4 major phases as mentioned in Table 3.1. C3S which is
the largest phase provides most of strength. C3A is responsible for initial setting of cement.
This phase also has highest amount of heat per unit mass, also called as enthalpy. Too much of
C3A can cause flash setting of concrete.

Table 3.1 Phases of ordinary portland cement

Compound Phase Name Abbreviated Name Typical amount
Tricalcium Silicate Alite C3 S 50-70%
Dicalcium Silicate Belite C2 S 15-30%
Tricalcium Aluminate Aluminate C3 A 5-10%
Tetracalcium Aluminoferrite Ferrite C4AF 5-15%
* C – CaO; S – SiO2; A – Al2O3; F – Fe2O3

Chemical composition of cement in terms of proportion of its phases can be determined by

Bogue’s equations.

C3S = 4.071C – 7.6S – 6.718A – 1.43F - 2.53S (3.3)

C2S = 2.687S - 0.7544C3S (3.4)
C3A = 2.65A – 1.692F (3.5)
C4AF = 3.043F (3.6)

10
These are based on the assumption that all the iron oxide goes into the C 4AF phase.
Concentration of the molecules can be measured by use of X-ray florescence.

Figure 3.2 Reactivity of various phases of cement. [14]

3.2.2 Hydration of cement phases

Reaction of Aluminate

C3A + 6H → C3AH6 (3.7)

C3A + 3C$H2 + 26H → C6A$3H32 (3.8)
2C3A + C6A$3H32 + 4H → 3C4A$H12 (3.9)
If gypsum is present, aluminate reacts with gypsum first to produce ettringite (Eq. 3.8). Once
gypsum is consumed, only then aluminate reacts with ettringite to form monosulfate (Eq. 3.9).
When both gypsum and ettringite are finished, aluminate reacts with water and forms
hydrogarnet (Eq. 3.7). Therefore, all these reactions does not occur simultaneously.

Reaction of Ferrite

C4AF + 10H → C3AH6 + CH + FH3 … (11)

C4AF + 3C$H2 + 30H → C6A$3H32 + CH + FH3 … (12)

2C4AF + C6A$3H32 + 12H → 3C4ASH12 + 2CH + 2FH3 … (13)

Reactions with alumino-ferrite also follow similar order as with aluminate.

11
Reaction of Silicates

C3S + 5.3H → C1.7SH4 + 1.3CH … (14)

C2S + 4.3H → C1.7SH4 + 0.3CH … (15)

Although, all the phases react simultaneously, yet each of them proceeds at different rates and
may have little to none offsets in starting. Because of this, curve showing rate of overall heat
evolution from the cement hydration has multiple peaks as shown in Figure 3.3.

Figure 3.3 Rate of heat evolution with time [12]

3.2.3 Heat of hydration

Total heat of hydration is given as sum of heat of hydration from each cement phase.

ΣQfi ∙α𝑖 ∙p𝑖

Q(t) = … (16)
Σp𝑖

Here, Qfi, αi and pi are heat of hydration, degree of hydration and proportion of each phase.
Average values for Qfi, also known as enthalpy, has been measured as given in Table 3.2.

Table 3.2 Enthalpy of phases of cement

𝑄𝑓𝑖 (J/g) 𝑖= 𝐶3𝑆 𝑖= 𝐶2𝑆 𝑖= 𝐶3𝐴 𝑖= 𝐶4𝐴𝐹
517 262 1144 725

12
3.3 LIN AND MEYER MODEL
The numerical model used in this study for simulation of cement hydration is the one developed
by C. Meyer [1] in 2009. It is based on the well-known Arrhenius equation used for
determination of rate of reaction for any chemical reaction. The parameters have been adjusted
for cement hydration by carrying out regression analysis on experimental data on degree of
hydration of different cements of varying chemical and physical properties.

The major assumption of this model is that it considers equal degree of hydration of different
phases. This assumption reduces computational load significantly, while maintaining certain
level of accuracy. This assumption has been adopted by many other researchers, such as, [5],
and [11]. Another approach is to consider degree of hydration of phases independent of each
other. This approach has been adopted by [3], and [12]. It has now been widely accepted that
actual mechanism of hydration of cement is in between these two approaches. However,
modelling such complex interactions is not possible in contemporary computers. Further
advancements of computer technology may make it possible in near future.

The base equation of Meyer model is as given in equation 3.10:

𝛼̇ = 𝐴𝛼 𝜂𝛼 𝑒𝑥𝑝(− 𝐸𝑎 ⁄𝑅𝑇) 𝑒𝑥𝑝(𝐸𝑎 ⁄𝑅. 293) (3.10)

In Eq. 3.10 α̇ is the rate of hydration. Aα is known as chemical affinity of cement and is a
function of degree of hydration. Eq. 3.11 is used to relate it with chemical composition of
cement.

𝐴
𝐴𝛼 = 𝑘 (𝑘.𝛼𝑜 + 𝛼) . (𝛼𝑢 − 𝛼). 𝑠 (3.11)
𝑢

s=1 for hydrostatic pressure equal to atmospheric pressure = 1 atm or 0.1 Mpa
Note: - The effect of pressure incorporated by s is not significant upto 6.8 MPa. Thus, s can be
taken as 1 for all practical purposes.

𝛼𝑢 is the ultimate degree of hydration. Eq. 3.12 is used to calculate 𝛼𝑢 . It relates ultimate degree
of hydration with the fineness of cement and water cement ratio.
𝑤
𝛽1 (𝐵𝑙𝑎𝑖𝑛𝑒)∗
𝑐
𝛼𝑢,293 = 𝑤 ≤ 1.0 (3.12)
𝛽2 (𝐵𝑙𝑎𝑖𝑛𝑒)+
𝑐

The parameters 𝛽1and 𝛽2 in Eq. 3.12 are given by,

13
 1.0
𝛽1 = 𝐵𝑙𝑎𝑖𝑛𝑒 −2.82 (3.13)
9.33( ) +0.38
100

𝐵𝑙𝑎𝑖𝑛𝑒 − 220
𝛽2 = (3.14)
147.78 + 1.656(𝐵𝑙𝑎𝑖𝑛𝑒−220)

However, ultimate degree of hydration has a theoretical maximum limit of:

𝑤
𝑐
𝛼𝑢 ≤ (3.15)
0.4

A0 is the initial chemical affinity, i.e., at 𝛼 = 0. The initial rate of hydration depends upon
available nucleation sites for new nuclei to form and grow. It, thus, mostly depends upon the
total surface area of the cement grains. Although, particle size distribution is also an important
physical characteristic of cement, yet fineness alone is used as a measure of surface area in this
model. Initial chemical affinity is related to fineness by the following expression:

𝐴0 ′ ∗𝐵𝑙𝑎𝑖𝑛𝑒
𝐴0 = (3.16)
350

The second factor in Eq. 3.10, 𝜂α , is the permeability of hydrated cement layer. Both degree of
hydration and fineness influences this factor. Experimental results show that it decreases
exponentially as hydration proceeds.[1]

𝜂α = 𝑒𝑥𝑝(−𝑛. 𝛼) (3.17)

Fineness influences permeability in a way that finer the cement, more is the permeability. This
is due to the fact that, thinner layer of products are formed around finer particles. However,
this effect is diminished as particle grows with the hydration. This can be seen from Eq. 3.19.

𝑛 = 𝑛0 ∗ 𝑓1 (𝐵𝑙𝑎𝑖𝑛𝑒, 𝛼) ∗ 𝑓2 (𝑇, 𝛼) (3.18)

350
𝑓1 = 1 + (1 − 𝛼)2 . 𝑙𝑛 𝐵𝑙𝑎𝑖𝑛𝑒 (3.19)

𝑉 10𝛼4
293
𝑓2 = (𝑉(𝑇) ) = [exp(28 ∗ 10−5 ∗ 𝛼 4 (𝑇 − 293)2 )] (3.20)

Finally, a total of four material constants are required to incorporate the properties of cement
into the mode, namely, k, A0’, n0 and 𝐸𝑎 . These constants are calibrated by using experimental

14
results for different cements. [1] carried out regression analysis on the experimental results
from various researchers and provided expressions for these constants (Eq. 3.21 to 3.24).

k = 0.56 * (pC3S)-0.206 * (pC2S)-0.128 * (pC3A) 0.161 (3.21)

n0 = 10.945pC3S + 11.25pC2S – 4.10pC3A – 0.892 (3.22)

A0’ = -0.0767.pC4AF + 0.0184 (3.23)

Ea = 22100 * (pC3A)0.30 * (pC4AF)0.25 * (Blaine)0.35 J/mol (3.24)

Furthermore, once rate of hydration is calculated, degree of hydration can be calculated using
Avrami’s equation or the JMAK equation given as:
𝑛
𝛼 = (1 − 𝑒 −𝑚𝑡 ) (3.25)

However, Eq. 3.25 cannot be directly used to calculate 𝛼, since m being the rate constant
changes with temperature. Using this equation directly gives oscillations in the results.
Therefore, Eq. 3.25 is written in differential form for n=1 as,

∆𝛼𝑖 = 𝑚. 𝑒 −𝑚𝑡 . ∆𝑡𝑖 (3.26)

The model parameter n can be modified to fit the curves with experimental data for better
degree of prediction. Degree of hydration, α at any time-step is calculated as,

𝛼𝑖+1 = 𝛼𝑖 + ∆𝛼𝑖 (3.27)

Eq. 3.28 is used to determine the amount of heat release since the start of hydration process till
time t. It, basically, states that cumulative amount of heat released is directly proportional to
the degree of hydration. Qmax is total heat stored in all of the cement phases, i.e., the maximum
amount of heat that can be released from the hydration process.

Q(t) = α(t).Qmax (3.28)

Therefore, the primary inputs of this model are proportions of cement phases and water-cement
ratio. The outputs are rate of reaction of cement and degree of hydration. Its corollary is amount
of heat liberated at any time step.

15
3.4 THEORY OF UNIFORM REACTED THICKNESS
Theory of uniform reacted thickness is a simple assumption that thickness of hydrated cement
layer is same for all the particles in a cement mass irrespective of the size of the particle. [13]
first put this theory forward. This assumption effectively captures the influence of fineness on
particle growth of hydrating cement.

reacted initial
particle
thickness
diameter di

unhydrated cement

Figure 3.4 Uniform Reacted Thickness of cement particles irrespective of initial size [15]

3.5 CALCULATION OF POROSITY

3.5.1 Microstructure Development

By applying theory of uniform reacted thickness, it is possible to simulate microstructure
development of cement using simple algorithms. Since, thickness of layer of reacted cement is
same for all the particle irrespective of its size, it can be assumed that 100% hydration happens
when the biggest particle is hydrated completely, i.e., the reacted thickness is equal to half of
the max dia. of particles. This assumption is used to calculate diameter of all the particles, from
which porosity can easily be calculated.

First step for simulation of microstructure development is to generate particles according to

particle size distribution which can either be fed as input by the user from X-ray diffraction
results or can be estimated using Rosin-Rammler equation (Eq. 3). For this maximum,
minimum and mean diameter of the particles should be calculated from sedimentation or other
methods.

16
These particles are then distributed randomly in a closed space whose volume is calculated
from water-cement ratio and a prescribed mass of cement to simulate. This can be taken as
0.00001 g.

The reacted thickness of particles at any time step is calculated from degree of hydration. The
3.15
volume inside the reacted thickness is then increased by a ratio of which is the ratio of
2.0
densities of anhydrous and hydrated cement. Diameter of the anhydrous part is reduced by
twice of the reacted thickness. This gives us the total volume of the partly hydrated cement
particle.

Start

Fix sample mass

(say 0.00001g)

Yes Get degree of hydration

at time step

If D>0
Calculate reacted thickness =
𝛼∗𝐷𝑚𝑎𝑥
2

Calculate the volume

Deduct reacted thickness
inside reacted
from radius
thickness * 3.15/2

Calculate new dia.

from above two

Figure 3.5 Flowchart for simulating microstructure development

This method is only effective if the particles are distributed such that there is no overlap
between two particles. Hydration products should be deposited only in the pore spaces around
the particle, or otherwise the porosity calculated will not be correct. Also, the density of

17
 3.15
hydration products of cement changes with temperature. Thus, a constant ratio of cannot
2.0
be used to calculate the volume of hydration products.

3.5.2 Alternate approach

Another way to calculate porosity of cement paste from degree of hydration of cement is by
determining the volume of cement that has hydrated and the volume of hydration products that
has been produced. This can be done by calculating the volumes of reactants and products of
all the chemical reactions involved in cement hydration.

Porosity is calculated by adding the volume of solids that is still anhydrate to the volume of
products that has formed after hydration and finally dividing it by the total volume that was
present initially.

Example 3.a:

C3S 5.3H C-S-H 1.3CH

Density (g/cc) 3.15 1.00 2.00 2.24
Molar Mass 228.0 18.0 227.2 74.0
Mass reacting (g) 228.00 95.40 227.20 96.20
Volume reacting (cc) 72.38 95.40 113.60 42.95
Solids volume (cc) 72.38 156.55

Example 3.a shows the process of calculation of volumes of various phases before and after
hydration. As we can see, volume of solids increases after hydration. This additional volume
goes into the pores and reduces the porosity. The difference between total volume of reactants
and total volume of products (72.38+95.40-156.55 = 11.23), is chemical shrinkage.

In similar way, increase/decrease in volume of solids can be determined for all the reactions
mentioned in section 3.2.2. At any time step, mass of phases that has hydrated can be
determined from degree of hydration. Thus, overall porosity can be calculated from calculating
amount of reacted and unreacted reactants and that of products.

Initial mass of phases to be considered for calculation comes from the phase composition. 1 m3
of concrete is taken as the sample size. Mass of cement in 1 m3 concrete is called the cement
content of the concrete and generally lies between 300 and 500 kg/m3. Density of cement is
calculated from the density of its phases as,
18
𝜌𝑐𝑒𝑚𝑒𝑛𝑡 = ∑𝑝ℎ𝑎𝑠𝑒𝑠 𝜌𝑖 ∗ 𝑝𝑖 (3.29)

In Eq. 3.30 𝜌𝑖 is the density of ith phase and 𝑝𝑖 is its proportion in the cement.

Accuracy of this method depends upon the size of the time-step. If a large time-step is chosen,
reactants present in small amounts and common in multiple reactions, like gypsum and water,
might be consumed in one or two reactions. Other reactions would not even start. This way,
some of the phases may have hydrated more than 40% in a single time-step and others may
remain at 0%. Therefore, reactions should proceed sequentially and repeatedly for small
intervals.

Table 3.3 shows the densities of all the phases of cement. These are required for volume
calculations. All the phases except C-S-H have constant densities. C-S-H density is highly
influenced by curing temperature. Section 3.6 discusses this property of C-S-H in more detail.
Table 3.3 Densities of cement phases
Phase Density
C3S (Alite) 3.15
C2S (Belite) 3.28
C3A (Aluminate) 3.03
C4AF (Ferrite) 3.73
C-S-H (gel) 1.6-2.1
C$H2 (Gypsum) 2.80
C3AH6 (Hydrogarnet) 2.52
FH3 (Iron hydroxide) 3.00
C4A$H12 (Monosulphate) 2.00
CH (Portlandite) 2.24
H (Water) 1.00

3.6 C-S-H DENSITY

It is now a well-established fact that at elevated temperatures cement concrete exhibits a higher
strength only at initial stages, its ultimate strength is in fact lower than that of concrete at normal
temperature.

Recent studies like [20] and [13], have found that the density of hydrates, especially calcium-
silicate-hydrates (C-S-H), does not remain same for all temperatures. This is due to the fact

19
that at elevated temperatures, there is less time for the hydrates that are formed inside the
cement grain to diffuse out. Thus, more hydrates will be formed inside than outside of the grain
as illustrated in Figure 3.6. Due to this, there will be less products to fill the pores between the
grains and the porosity will be higher even at higher degree of hydration.

Hydration product Original cement

grain

Unhydrated cement
(a) Normal temperature (b) Elevated temperature

Figure 3.6 Illustration for diffusion of hydration products at normal and elevated
temperatures

Galluci et al. [13] have quantified the dependence of C-S-H density on temperature. Using X-
ray diffraction, they have calculated ‘apparent’ density of C-S-H at different temperatures.
These results are shown in Figure 3.7.

Figure 3.7 Calculated C-S-H density with temperature [10]

An approximate relationship has been derived by performing linear regression on these data:

ρ = 0.0066 * T + 1.8909 (R2 value = 0.9848) (3.30)

20
This relationship has been used in SafeConc for calculation of porosity and strength. This
captures the temperature dependence of concrete strength more effectively, as can be seen from
Figure 3.8 (a) and (b).

(a)

(b)
Figure 3.8 Strength development at different temperatures with (a) constant C-S-H density,
and (b) temperature-dependent C-S-H density. Source: SafeConc output

21
 SIMULATION METHODOLOGY

4.1 HEAT TRANSFER SOLUTION

4.1.1 Steady-state heat transfer

This subsection is dedicated to derivation of finite element equation for steady-state heat
transfer. Only bilinear four-noded rectangular element (Q4) is covered, since only this element
has been used throughout this study.

Eq. 4.1 is the assumed polynomial function for the temperature field, also known as the basis
function.

T(x,y) = a1 + a2 x + a3 y + a4 xy (4.1)

The nodal temperature T1, T2, T3 and T4 can be written in matrix form as:

𝑇1 1 𝑥1 𝑦1 𝑥1 𝑦1 𝑎1
𝑇2 𝑥2 𝑦2 𝑥2 𝑦2 𝑎2
{ } = [1 ]{ } (4.2)
𝑇3 1 𝑥3 𝑦3 𝑥3 𝑦3 𝑎3
𝑇4 1 𝑥4 𝑦4 𝑥4 𝑦4 𝑎4

Eq. 4.2 can also be written as:

𝑎1 1 𝑥1 𝑦1 𝑥1 𝑦1 −1 𝑇1
𝑎2 𝑥2 𝑦2 𝑥2 𝑦2 𝑇
{𝑎3 } = [1 ] { 2} (4.3)
1 𝑥3 𝑦3 𝑥3 𝑦3 𝑇3
𝑎4 1 𝑥4 𝑦4 𝑥4 𝑦4 𝑇4
Substituting Eq. 4.3 into Eq. 4.2 and writing it in the matrix form, we get,

1 𝑥1 𝑦1 𝑥1 𝑦1 −1 𝑇1
𝑥2 𝑦2 𝑥2 𝑦2 𝑇
T(x,y) = {1 𝑥 𝑦 𝑥𝑦} [1 ] { 2} (4.4)
1 𝑥3 𝑦3 𝑥3 𝑦3 𝑇3
1 𝑥4 𝑦4 𝑥4 𝑦4 𝑇4

Product of the first two matrices in the right hand side of the Eq. 4.4 forms what is called as
shape-function matrix. Shape-function matrix relates the temperature at any point inside the
element to the nodal temperatures of the elements.

23
 𝑇1
𝑇2
T(x,y) = {𝑁1 𝑁2 𝑁3 𝑁4 } { } (4.5)
𝑇3
𝑇4
(𝑏−𝑥)(ℎ−𝑦) (𝑏+𝑥)(ℎ−𝑦)
N1 = ; N2 =
4𝑏ℎ 4𝑏ℎ
Shape functions
(𝑏+𝑥)(ℎ+𝑦) (𝑏−𝑥)(ℎ+𝑦)
N3 = ; N4 =
4𝑏ℎ 4𝑏ℎ

where, h and b are height and width of the rectangular element respectively.

Before formulating the finite element equations, we need to define the thermal gradient matrix,
𝜕𝑇 𝜕𝑁1 𝜕𝑁2 𝜕𝑁3 𝜕𝑁4 𝑇1
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝑇2
{g} = {𝜕𝑇 } = [𝜕𝑁1 𝜕𝑁2 𝜕𝑁3 𝜕𝑁4] {𝑇 } (4.6)
3
𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑦 𝑇4

The matrix in the right hand side of Eq. 4.6 is the [B] matrix. The choice of notation is kept
consistent with the popular literature on this subject.

To calculate temperatures at different points inside the element, it is required to first calculate
the temperatures at the nodes. The nodal temperatures are calculated by solving Eq. 4.7 for {𝑇}.

[𝑘]{𝑇} = {𝑓} (4.7)

Eq. 4.7 is similar to the equation of spring, where k is the stiffness of the spring or spring
constant. That’s why the matrix [𝑘] is generally called as the ‘stiffness matrix’ and {𝑓} is called
the ‘load vector’.

Eq. 4.8 defines the ‘so-called’ stiffness matrix for the element. Derivation of stiffness matrix
is a lengthy process and requires knowledge of various thermodynamic principles. Also, it is
out of scope of this study. Therefore, for sake of simplicity and comprehensiveness of the
thesis, the derivation is skipped. For detailed read on this topic, reader is referred to [26], [44],
[45], and [46].

Stiffness Matrix, [k] = ∭𝑉[𝐵]𝑇 𝐷[𝐵]𝑑𝑉 + ∬𝑆2 ℎ[𝑁]𝑇 [𝑁]𝑑𝑆 (4.8)

24
Load Vectors:
The load vector {𝑓} in Eq. 4.7 is the sum of load vectors due to various sources of heat, namely,
body source, surface influx and convective source (which can be source or sink depending on
the surrounding temperature and element temperature)

{fQ} = ∭𝑉 𝑄[𝑁]𝑇 𝑑𝑉 (Body heat) (4.9)

{fq} = ∬𝑆1 𝑞[𝑁]𝑇 𝑑𝑆 (Surface heat influx) (4.10)
{fq} = ∬𝑆2 ℎ𝑇∞ [𝑁]𝑇 𝑑𝑆 (heat loss due to convection) (4.11)

Note: S1 and S2 different surfaces, since usually heat influx and convection loss do not happen
through same surface.

Evaluation of Stiffness Matrix:

1 −(ℎ − 𝑦) (ℎ − 𝑦) (ℎ + 𝑦) −(ℎ + 𝑦)
𝐵= [ ] (4.12)
4𝑏ℎ −(𝑏 − 𝑥) −(𝑏 + 𝑥) (𝑏 + 𝑥) (𝑏 − 𝑥)

Matrix D in Eq. 4.8 is the conduction matrix for the element. kxx and kyy are equal for isotropic
materials, such as concrete.

𝑘𝑥𝑥 0
𝐷=[ ] (4.13)
0 𝑘𝑦𝑦

Stiffness matrix in Eq. 4.8 is summation of two functions. The first part arrives from the
conductivity of the material and second part is due to the convection of the surface. For non-
convective elements, like those in the middle of the body, the second part becomes zero.

k = kc + kh (4.14)

ℎ 𝑏
kc = ∫−ℎ ∫−𝑏[𝐵]𝑇 𝐷[𝐵]𝑡ℎ 𝑑𝑥𝑑𝑦 (4.15)

1 1
kc = 𝑏 ∗ ℎ ∗ 𝑡ℎ ∫−1 ∫−1[𝐵]𝑇 𝐷[𝐵]𝑡ℎ 𝑑𝜉𝑑𝜂 (4.16)

Here, 𝜉 = 1/𝑏 and 𝜂 = 1/ℎ and th is thickness of element

25
 (1 − 𝜂)2 𝑠𝑦𝑚𝑚.
𝑘𝑥𝑥.𝑡ℎ.ℎ −(1 − 𝜂)2 (1 − 𝜂) 2
...,kc = 16𝑏 ∫
−(1 − 𝜂)(1 + 𝜂) (1 − 𝜂)(1 + 𝜂) (1 + 𝜂)2
[ (1 − 𝜂)(1 + 𝜂) −(1 − 𝜂)(1 + 𝜂) −(1 + 𝜂)2 (1 + 𝜂)2 ]

(1 − 𝜉)2 𝑠𝑦𝑚𝑚.
𝑘𝑦𝑦.𝑡ℎ.𝑏 (1 − 𝜉)(1 + 𝜉) (1 + 𝜉)2
+ 16ℎ ∫ (4.17)
−(1 − 𝜉)(1 + 𝜉) −(1 + 𝜉)2 (1 + 𝜉)2
[ −(1 − 𝜉)2 −(1 − 𝜉)(1 + 𝜉) (1 − 𝜉)(1 + 𝜉) (1 − 𝜉)2 ]

2 −2 −1 1 2 1 −1 −2
...,kc =
𝑘𝑥𝑥.𝑡ℎ.ℎ −2 2 1 −1 ] + 𝑘𝑦𝑦.𝑡ℎ.𝑏 1 2 −2 −1]
∫[ ∫[ (4.18)
6𝑏 −1 1 2 −2 6ℎ −1 −2 2 1
1 −1 −2 2 −2 −1 1 2

kh = ∬𝑆2 ℎ[𝑁]𝑇 [𝑁]𝑑𝑆 (4.19)

2 1 0 0
kh = 6 ∫ [1 2 0 0]
ℎ.𝐿.𝑡ℎ
(4.20)
0 0 0 0
0 0 0 0

Note: kh matrix in Eq. 4.20 is for surface1-2. For surfacei-j apply following matrix operations:

R1 ↔ Ri R2 ↔ Rj

C1 ↔ Ci C2 ↔ Cj

Evaluation of load vectors:

Integration of Eq. 4.9 to 4.11 is done to obtain load vectors. Final results of these integrations
are given in Eq. 4.21 to 4.23

1
{1}
𝑄.𝑏.ℎ.𝑡ℎ
fQ = (4.21)
4 1
1

26
 1
(fq)surf1-2 = 2 {1}
𝑞.𝑡ℎ.𝐿
(4.22)
0
0 For heat influx
0
(Fq)surf2-3 = 2 {1}
𝑞.𝑡ℎ.𝐿
(4.23)
1
0
…and so on

For load vector due to convection, replace q with h.𝑇∞ ; 𝑇∞ being the ambient temperature.

4.1.2 Transient heat flow

Since heat is liberated from cement continuously, but with varying rate, static analysis is not
capable to determine the temperatures at any intermediate stage during the whole hydration
purpose. For this reason, a transient-state analysis is required.
Eq. 4.7 is slightly modified to incorporate thermal mass of the element. It is the amount of heat
that can be stored inside the element without changing the temperature or conversely, it is the
amount of heat that is needed to increase the temperature of the element by unit degree celsius.

{f} = [k]{T} + [M]{𝑇̇} (4.24)

Thermal mass matrix [M] is written as:

[𝑀] = 𝑐𝜌 ∭[𝑁]𝑇 [𝑁]𝑑𝑉 (4.25)

Eq. 4.26 is the explicit equation for nodal temperatures at (i+1)th step in terms of temperatures
at ith step. It is called explicit since temperatures at the previous step only appears in the right
hand side of expression. It is easier to solve explicit equation than it is to solve an implicit one.

𝑀 𝑀
[∆𝑡 + 𝛽. 𝐾] {𝑇𝑖+1 } = [∆𝑡] {𝑇𝑖 } + (1 − 𝛽)𝐹𝑖 + 𝛽𝐹𝑖+1 (4.26)

In the above equation, choice of β depends upon the level of the accuracy required. It generally
lies between 0 and 1. While lower β means higher accuracy, it also means lower stability. For
stable results and to avoid oscillations in the temperature results with respect to time, β should
be greater than ½. However, lower β can also be taken, but the size of time-step needs to be
checked. It should be kept small.

β=0 → Forward difference (Euler); Conditionally stable (∆𝑡 should be small)

27
 =1/2 → Trapezoid rule (Crank-Nicolson); Unconditionally stable (but gives spurious
oscillations)
=2/3 → (Galerkin); Unconditionally stable
=1 → Backward difference; Unconditionally stable

β=2/3 has been used in SafeConc for solving transient heat flow equation.

4.2 PLANE STRESS/STRAIN ANALYSIS

Finite element method for plane stress/strain analysis is based upon minimum potential energy
theorem. Strain energy of each element is formulated in terms of stiffness of its nodes. Then,
the total strain energy of the system is minimized with respect to the displacements of all of its
nodes.

Since, stiffness equations for simple homogenous, isotropic materials are well known, the
major challenge is to solve these equations as simultaneous equations determine the
displacements. For a 100 node mesh, 200 simultaneous equations are formed. Solving this is
beyond human capability. That’s why computer programs are needed. Since equations are
generated dynamically and number of nodes and other parameters are variable, system should
be versatile enough to be able to solve varying types of model.

4.2.1 Static Analysis – finite element formulation

For bilinear four-noded rectangular element (Q4), displacement functions (or basis functions)
are chosen as Eq. 4.27 and Eq.4.28

u(x,y) = a1 + a2 x + a3 y + a4 xy (4.27)
v(x,y) = a5 + a6 x + a7 y + a8 xy (4.28)

Same procedure as done for heat transfer equation, is repeated for plane stress analysis to obtain
shape functions.

(𝑏−𝑥)(ℎ−𝑦) (𝑏+𝑥)(ℎ−𝑦)
N1 = ; N2 =
4𝑏ℎ 4𝑏ℎ
Shape functions
(𝑏+𝑥)(ℎ+𝑦) (𝑏−𝑥)(ℎ+𝑦)
N3 = ; N4 =
4𝑏ℎ 4𝑏ℎ

h and b are height and width of the rectangular element respectively.

28
Since there are two displacement fields (one for each direction in 2-D), and eight degrees of
freedom in plane stress/strain. The displacement equation in matrix form is written as:

𝑢1
𝑣1
𝑢2
𝑢(𝑥, 𝑦) 𝑁 0 𝑁2 0 𝑁3 0 𝑁4 0 𝑣2
{ }=[ 1 ] (4.29)
𝑣(𝑥, 𝑦) 0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 𝑢3
𝑣3
𝑢4
{ 𝑣4 }
Just like thermal gradient vector (Eq. 4.6) in heat transfer, strain vector is defined in terms of
nodal displacements.

𝜕𝑢
𝜕𝑥
𝜕𝑣
Total Strain vector, {εTotal} = 𝜕𝑦
(4.30)
𝜕𝑢 𝜕𝑣
{𝜕𝑦 + 𝜕𝑥}
𝑢1
𝜕𝑁1 𝜕𝑁2 𝜕𝑁3 𝜕𝑁4 𝑣1
0 0 0 0 𝑢2
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝑁1 𝜕𝑁2 𝜕𝑁3 𝜕𝑁4 𝑣2
= 0 𝜕𝑦
0 𝜕𝑦
0 𝜕𝑦
0 𝜕𝑦 𝑢3 (4.31)
𝜕𝑁1 𝜕𝑁2 𝜕𝑁3 𝜕𝑁4
𝜕𝑁1 𝜕𝑁2 𝜕𝑁3 𝜕𝑁4 𝑣3
[ 𝜕𝑦 𝜕𝑥
𝜕𝑦
𝜕𝑥
𝜕𝑦
𝜕𝑥
𝜕𝑦
𝜕𝑥 ] 𝑢4
{ 𝑣4 }

= [B]{d} (4.32)

Again, the first matrix in Eq. 4.31 is called as [B] matrix. It is evaluated by differentiating the
shape functions, as done in Eq. 4.33.

𝐵=
−(ℎ − 𝑦) 0 (ℎ − 𝑦) 0 (ℎ + 𝑦) 0 −(ℎ + 𝑦) 0
1
[ 0 −(𝑏 − 𝑥) 0 −(𝑏 + 𝑥) 0 (𝑏 + 𝑥) 0 (𝑏 − 𝑥) ] (4.3
4𝑏ℎ
−(𝑏 − 𝑥) −(ℎ − 𝑦) −(𝑏 + 𝑥) (ℎ − 𝑦) (𝑏 + 𝑥) (ℎ + 𝑦) (𝑏 − 𝑥) −(ℎ + 𝑦)
3)

Stiffness matrix [k] is evaluated according to Eq. 4.34. Its derivation is again skipped in this
thesis. For further read, refer [44] or [46].

29
Stiffness Matrix, [k] = ∭𝑉[𝐵]𝑇 [𝐷][𝐵]𝑑𝑉 (4.34)

Load vector for thermal load is given by Eq. 4.35. In this study, external loads like those arising
due to applied concentrated forces or uniformly distributed loading is not considered.

Load Vector, {f} = ∭𝑉[𝐵]𝑇 [𝐷]{𝜀 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 }𝑑𝑉 (4.35)

In Eq. 4.35, matrix [𝐃] is called as ‘Constitutive Matrix’. It relates stresses to strains and is
evaluated as follows,

1 𝜈 0
𝐸
Constitutive matrix, [𝑫] = 1−𝜈2 [𝜈 1 0 ] for plane stress condition
1−𝜈
(4.36)
0 0
2

1−𝜈 𝜈 0
𝐸
[ 𝜈 1−𝜈 0 ] for plane strain condition
1−2𝜈
(4.37)
(1+𝜈)1−2𝜈)
0 0 2

In Eq. 4.37, E is the elastic modulus of the material and ν is the poisson’s ratio.

The total strain is the sum of mechanical strain and thermal strain.

{𝜺𝑻𝒐𝒕𝒂𝒍 } = {𝜀 𝑴𝒆𝒄𝒉𝒂𝒏𝒊𝒄𝒂𝒍 } + {𝜀 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 } (4.38)

Work is done by the mechanical strain and not the thermal strain. Since, energy required for
thermal strain comes from the heat energy. But using finite element method, we can calculate
only total strain. Calculating mechanical strain directly from finite element equations would be
a little complex and unnecessary. Therefore, it is calculated in two steps.

Firstly, the total strain is calculated from Eq. 4.30. Thermal strain is defined as,

𝛼. 𝑇
Thermal strain vector, {𝜺𝑻𝒉𝒆𝒓𝒎𝒂𝒍 } = {𝛼. 𝑇} for plane stress condition (4.39)
0
𝛼. 𝑇
(1 + 𝜈) {𝛼. 𝑇} for plane strain condition (4.40)
0
where, α is the coefficient of linear expansion of the material.

Mechanical strain is calculated as the difference of total strain and thermal strain,

30
{𝜺𝑴𝒆𝒄𝒉𝒂𝒏𝒊𝒄𝒂𝒍 } = {𝜀 𝑇𝑜𝑡𝑎𝑙 } − {𝜀 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 } (4.41)
Thermal stresses are given as the product of constitutive matrix and mechanical strain vector.

Stress vector, {𝝈} = [𝐷]{𝜀 𝑀𝑒𝑐ℎ𝑎𝑛𝑖𝑐𝑎𝑙 } = [𝐷][{𝜀 𝑇𝑜𝑡𝑎𝑙 } − {𝜀 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 }] (4.42)

𝜎𝑥
= { 𝜎𝑦 } (4.43)
𝛾𝑥𝑦
In 2-D isotropic, homogenous material, there are 3 stress variables, namely, horizontal normal
stress (σx ), vertical normal stress (σy ) and shear stress (γxy ).

The stresses calculated by this method are discontinuous in nature. Elements sharing a node
will give different stress results for that node. That’s why stresses calculated at the nodes from
all the elements sharing that node are averaged before calculating stresses at intermediate
points. This process is called ‘smoothing’. It gives a visually appealing plot of results. The
stresses at intermediate points are then calculated from the nodal stresses using shape functions.

4.2.2 Dynamic Analysis

Dynamic plane stress/strain analysis is done by considering acceleration component while
formulating the finite element equations. It is then integrated in time using Newmark’s beta
method.

Numerical integration in time – Newmark’s beta method [43]

{𝑑̇𝑖+1 } = {𝑑̇𝑖 } + (∆𝑡)[(1 − 𝛾){𝑑̈𝑖 } + 𝛾{𝑑̈𝑖+1 }] (4.44)
1
{𝑑𝑖+1 } = {𝑑𝑖 } + (∆𝑡){𝑑̇𝑖 } + (∆𝑡)2 [( − 𝛽) {𝑑̈𝑖 } + 𝛽{𝑑̈𝑖+1 }] (4.45)
2

Where,

{𝑑𝑖 } = Displacement vector at ith time step

{𝑑̇𝑖 } = Velocity vector at ith time step
{𝑑̈𝑖 } = Acceleration vector at ith time step

Consistent-Mass matrix for dynamic plane stress analysis of rectangular finite element is given
by,

[M] =∭𝑉 𝜌[𝑁]𝑇 [𝑁]𝑑𝑉 , ρ being the density of the material. (4.46)

Multiplying Eq. 4.45 with [M], we get

31
 1
[𝑀]{𝑑𝑖+1 } = [𝑀]{𝑑𝑖 } + (∆𝑡)[𝑀]{𝑑̇𝑖 } + (∆𝑡)2 [𝑀] ( − 𝛽) {𝑑̈𝑖 } + 𝛽[𝑀]{𝑑̈𝑖+1 } (4.47)
2

1
[𝑀]{𝑑𝑖+1 } = [𝑀]{𝑑𝑖 } + (∆𝑡)[𝑀]{𝑑̇𝑖 } + (∆𝑡)2 [𝑀] ( − 𝛽) {𝑑̈𝑖 } + 𝛽[{𝐹𝑖+1 } −
2

[𝐾]{𝑑𝑖+1 }] (4.48)

Simplifying the Eq. 4.48, we get

[𝐾′]{𝑑𝑖+1 } = {𝐹′𝑖+1 } (4.49)

where,
1
[𝐾′] = [𝐾] + [𝑀] (4.50)
𝛽(∆𝑡)2
[𝑀] 1
{𝐹′𝑖+1 } = {𝐹𝑖+1 } + [{𝑑𝑖 } + (∆𝑡){𝑑̇𝑖 } + (2 − 𝛽)(∆𝑡)2 {𝑑̈𝑖 }] (4.51)
𝛽(∆𝑡)2

Finally, Eq. 4.49 is solved with initial displacement, velocity and acceleration already known.
This is done recursively, each time proceeding to next time step and using data from the
previous time as input for the next. This is called explicit dynamic analysis.

Displacement at all nodes for all time steps are stored in the elements and manipulated to
calculate strains, and stresses whenever needed.

After performing dynamic analysis, it was noted that there is no significant change in results.
But the execution time was increased manifold. This is due to the fact that, heat liberation is
very slow in concrete. Although, there is some acceleration in the body, yet it can be stated as
‘pseudo-static’ and the very small acceleration can be considered as zero. This simplifies the
analysis and reduces the execution time.

4.3 NUMERICAL METHODS AND OPTIMIZATION

4.3.1 Numerical integration – Gauss Quadrature

As it can be seen stiffness and mass matrices, and the load vector require triple integration, or
in case of plane stress-strain analysis double integration over the whole element. Moreover,
this can only be done when all the data regarding the behavior of the finite element is known.
Since data can be altered anytime during the runtime, it is needed that the integration be done
dynamically during the runtime. For this purpose, numerical methods have to be employed.
Methods like trapezoidal method, Simpsons’ one-third and three-eighth rules, Newton-Cotes
32
method and Gauss Quadrature method are all designed to solve numerical integrations in a
known domain.

SafeConc uses Gauss-Quadrature for numerical integration, which is discussed in this sub-
section.

Gauss Quadrature rule in its simplest form can be written as,

+1
∫−1 𝑓(𝜉)𝑑𝜉 = ∑𝑟𝑖=1 𝑊𝑖 𝑓(𝜉𝑖 ) (4.52)

where, 𝜉 i are the roots of the Legendre polynomial given by,

2𝑘−(1 𝑘−1
𝑃𝑘 (𝜉) = 𝜉𝑃𝑘−1 (𝜉) − 𝜉𝑃𝑘−2 (𝜉) (4.53)
𝑘 𝑘

and, Wi are the weights obtained as

2
2(1−𝜉𝑖 )
𝑊𝑖 = (4.54)
(𝑟(𝑃𝑟−1 (𝜉𝑖 )))2

r is the number of gauss points, which is kept at 3 in this study. Higher the r, more accurate is
the integration, however, computational effort increases O(r2).

For any arbitrary interval [a,b], we change the variable and apply a linear transformation
according to the following relation:
𝑏−𝑎 𝑏+𝑎
𝑥= 𝜉+ (4.55)
2 2
𝑏−𝑎
𝑑𝑥 = 𝑑𝜉 (4.56)
2

Therefore, the integration becomes,

𝑏 𝑏−𝑎 +1 𝑏−𝑎
∫𝑎 𝑓(𝑥)𝑑𝑥 = ∫−1 𝑓(𝑥(𝜉))𝑑𝜉 = ∑𝑟𝑖=1 𝑊𝑖 𝑓(𝑥(𝜉𝑖 )) (4.57)
2 2

Extending this method to two variables, a double integration can be solved as,

𝑏2 𝑏1 +1 +1
𝑏1 − 𝑎1 𝑏2 − 𝑎2
∫ ∫ 𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦 = . ∫ ∫ 𝑓(𝑥(𝜉), 𝑦(𝜂))𝑑𝜉𝑑𝜂
2 2
𝑎2 𝑎1 −1 −1

33
 𝑏1 −𝑎1 𝑏2 −𝑎2
= . ∑𝑟𝑖=1 ∑𝑟𝑗=1 𝑊𝑖 𝑓(𝑥(𝜉𝑖 ), 𝑦(𝜂𝑖 )) (4.58)
2 2

For simplification and less runtime computation, 𝜉𝑖 ’s and 𝑊𝑖 ’s are pre-calculated for r=3 and
hard-coded into SafeConc. This saves a lot of time during execution. The values for 𝜉𝑖 and 𝑊𝑖
are shown in Table 4.1.

Table 4.1 Gauss points and their weights

i ξ or η W
1 -0.774596669241483377 0.555555555555555555
2 0.000000000000000000, 0.888888888888888888
3 0.774596669241483377 0.555555555555555555

4.3.2 Cholesky Factorization

Stiffness matrices generated in finite element analysis have some unique properties. Such as
they are always symmetric, sparse (most of the elements are zero), banded (the non-zero
elements lie near the diagonal), and positive definite.

They are symmetric because, according to Maxwell’s reciprocal theorem, displacement at any
point i due to load at any point j is always equal to displacement at j due to same load at i. This
means kij = kji. However, this is only true if the material is homogenous and isotropic which
are also the prerequisite for Maxwell’s theorem.

They are sparse because load within an element is shared only by its own nodes. Therefore,
contribution of other nodes in the stiffness of the element is zero. Due to same reason, the non-
zero elements also lie near diagonal.

A square matrix [M] of order n is said to be positive definite if and only if it is symmetric and
for any column vector {X} of order n, {X}T[M]{X} is always positive. However, no proof is
provided here for positive definite nature of stiffness matrices. For further read on properties
of stiffness matrices, reader can refer to “A first course in Finite Element Method” by Daryl
Logan[43].

This sub-section discusses about the optimization technique used for solution of finite element
equation in SafeConc. When a matrix [M] is symmetric positive definite or SPD, Cholesky
factorization can be applied to find out a lower triangular matrix [L] such that, [M] = [L][L]T.

34
Since, [L] is a lower triangular matrix, it is known that elements above the diagonal are zero.
Therefore, we can store the non-zero elements in 1D array of size (n!), n being the order of the
square matrix [M]. This way the huge matrix can be stored efficiently with a reduction of about
40% in memory space. This whole process has a O(n3) complexity. After the matrix has been
factorized it takes another O(n2) time to solve a linear equation [M]{x} = {D}. This is more
efficient than inverting the matrix. Also, it should be noted that factorization needs to be done
only once and not for every time step, because stiffness matrix doesn’t change.

Now we shall see how Cholesky factorization is done and how it is used to solve linear
equations efficiently.

Suppose, we have a square matrix of size n x n,

𝑎 𝐴21 𝑇 𝐿 𝑂 𝐿11 𝐿21 𝑇

A = [ 11 ] = [ 11 ][ ] (4.59)
𝐴21 𝐴22 𝐿21 𝐿22 𝑂 𝐿22
where,
A21, L21 are sub-matrices of order (n-1) x 1 and A22, L22 are sub-matrices of order (n-1)
x (n-1).

Programming Steps –

0 … 0
Step 1: Initialize a matrix L = [ ⋮ ⋱ ⋮] of same size as A
0 … 0

Step 2: Compute L11 = √𝑎11

Step 3: Populate L21 = A21/L11

Step 4: L22L22T = A22 – L21L21T

Step 5: Create a new matrix A’ = L22L22T

Step 6: Carry out Cholesky Factorization on A’ and continue the process until A’ is 1 x 1.

For solving a system of n linear equations,

Step 7: [A]{X} = {D} (4.60)

or [L]([L] T{X}) = {D} (4.61)
Step 8: Solve for [L]T{X} in equation (4.62) and denote it as {Y},

Step 9: Solve for {X} in [L]T{X} = {Y}

35
Note: - Since [L] is a lower triangular matrix, it doesn’t require inversion. The equation [L]{Y}
Di − ∑ij=1 Kij Yj
= {D} can be solved recursively by putting Yi =
Kii

Code snippet for cholesky factorization

L = new double[A.length][A[0].length];
for (int i = 0; i < A.length; i++) {
L[i] = A[i].clone();
}
int n = L.length;

for (int i = 0; i < n; i++) {

L[i][i] = Math.sqrt(L[i][i]);

for (int j = i + 1; j < n; j++) {

L[j][i] = L[j][i] / L[i][i];
L[i][j] = 0;
}

for (int j = i + 1; j < n; j++) {

for (int k = i + 1; k < j + 1; k++) {
L[j][k] -= L[j][i] * L[k][i];
}
}
}

4.4 STRENGTH ESTIMATION

4.4.1 Strength-Porosity relationship

Various studies have been conducted to ascertain the dependence of concrete strength on
various parameters. Whatever be the approach, one thing most researchers have agreed upon
is that porosity and pore size distribution are the major factors which influences the strength
most. Chen et al. (2013) [10] studied various porosity-strength relationships. In their study,
they performed experimental tests on different cement mortar for porosity, compressive
strength etc. and investigated the efficiency of all these models to predict strength on the basis
of porosity.

Following models were tested by them:

1. Balshin: 𝜎 = 𝜎0 (1 − 𝑝)𝑏
Where, 𝜎0 is strength at zero porosity and b is the empirical constant.

36
 2. Ryshkewitch: 𝜎 = 𝜎0 𝑒 −𝑘𝑝
Again, 𝜎0 is strength at zero porosity and k is the empirical constant.
𝑝
3. Schiller: 𝜎 = 𝑛 ln( 𝑝0 )

Here, n is the empirical constant and p0 is the porosity at zero strength.

4. Hasselman: 𝜎 = 𝜎0 − 𝑐𝑃
c is the empirical constant.

Figure 4.1 Best fit for existing models with experimental results obtained by Chen et al. [10]

4.4.2 Extended Zheng’s Model

Extended Zheng’s model, based on Griffith’s brittle fracture theory, gives a better fitting
strength-porosity relationship, since, it also considers stress concentration around the pores. It
captures the random nature of cement microstructure more accurately. Zheng et al. [12] first
proposed the relationship in 1992 for porous materials. The basic underlying principle of this
relationship is available in [12]. Extended Zheng’s model is mathematically written as:
1/2
𝑝𝑐 −𝑝 1.85
𝜎 = 𝜎0 [( ) . (1 − 𝑝2/3 ] (4.63)
𝑝𝑐

37
σ0 is the strength at zero porosity and pc is the critical porosity beyond which strength becomes
zero.

The model needs to be calibrated to obtain the parameters in Eq. 4.63. The recommended
method of calibration is to measure the porosity of samples taken from the actual concrete
using any available method (eg. Mercury Intrusion Porosimetry) and compare it with its
compressive strength.

4.4.3 Elastic modulus

Elastic modulus of concrete is not constant but depends upon its maturity. IS456:2000 (cl.
6.2.3) recommends a theoretical value for elastic modulus according to the formula: 5000√𝑓𝑐𝑘
where fck is the characteristic cube compressive strength.

4.5 CRACKING RISK

A large number of failure criteria are available with their advantages and limitations. For brittle
materials such as concrete and cement mortar, few of the criteria which are applicable are
Drucker–Prager yield criterion, Bresler Pister yield criterion, and Willam-Warnke yield
criterion.

In this study, a simple approach to failure has been considered. It is assumed that failure doesn’t
occur till at any point the major principal stress (σ1) is less than the tensile strength of the
concrete (σt) and the minor principal stress (σ2) is greater than the compressive strength of the
concrete (σc). Mathematically,

σc < σ2 < σ1 < σt = no failure

Thus, once this inequality is dissatisfied there is a “risk” of cracking in the body.

𝑚𝑎𝑗𝑜𝑟 𝑝𝑟𝑖𝑛𝑐𝑖𝑝𝑎𝑙 𝑠𝑡𝑟𝑒𝑠𝑠 𝑚𝑖𝑛𝑜𝑟 𝑝𝑟𝑖𝑛𝑐𝑖𝑝𝑎𝑙 𝑠𝑡𝑟𝑒𝑠𝑠

Risk of cracking = 𝛾. 𝑀𝑎𝑥( , )
𝑡𝑒𝑛𝑠𝑖𝑙𝑒 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ 𝑐𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑖𝑣𝑒 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ

𝛾 is the factor of safety assigned by the user.

The principal stresses σ1 and σ2 are calculated as the eigen-values of the Cauchy stress tensor,
or mathematically,

38
 𝜎𝑥 − 𝜎1 𝜏𝑥𝑦
[ 𝜏 𝜎𝑦 − 𝜎2 ] = 𝑂 (4.64)
𝑦𝑥

Pseudo code for calculation of cracking risk

For All Elements
For All Discrete Points
majorStress = element.getStressAt(point,time,Major_Principal)
minorStress = element.getStressAt(point,time,Minor_Principal)
compressiveStrength = element.getStrength(Compressive)
tensileStrength = element.getStrength(Tensile)
risk = Max(majorStress/tensileStrength , minorStress/compressiveStrength)
risk = risk * F.O.S.
End For
End For

39
 STRUCTURE OF SAFECONC

5.1 PROGRAM STRUCTURE

In this chapter, the inner working of the program is explained in brief. SafeConc has been
designed in a modular way having four interlinked modules (or packages) for 1) Cement
hydration modelling, 2) Temperature distribution, 3) Stress distribution and 4) Strength
calculation.

Figure 5.1 Internal structure of SafeConc – the program

Figure 5.1 depicts the internal structure or the logical flow of control of SafeConc. The Module
1 (for hydration simulation) and Module 2 (for temperature distribution or heat transfer) are
interlinked with each other. This type of system is called a ‘feedback system’, where output of
one becomes input for the other and both are executed parallel to each other. The output from
Module 1, i.e. heat generated from hydration at a particular time step is passed to Module 2 for
each element of the finite element mesh. Module 2 simulates the transfer of this heat to the

41
surrounding by convection. This process is continued for the prescribed simulation time and
the temperatures are saved in a 2D array of doubles in the memory. The first dimension of the
array corresponds to time step and the second to node number. This can also be printed to a
‘.csv’ file on user’s request.

Output from Module 1 is saved as degree of hydration and porosity for each element in separate
2D arrays. The porosity is then transferred to Module 3 for further manipulation. Module uses
extended Zheng’s model to calculate strength from porosity.

Figure 5.2 Flowchart for calculation of degree of hydration

The temperatures from Module 2 is also passed to Module 3 for stress calculations. Module 3
analyzes the temperature gradient from node to node and calculates stresses using finite
element method.

Cracking risk – At any time step, if stress at any node is greater than the strength developed at
that node, a location for potential cracking is expected at that node.

5.2 GRAPHICAL USER INTERFACE (GUI)

Figure 5.3 shows the basic GUI developed for SafeConc. Since it is in development stage, some
of the items do not perform any task right now and are only kept as placeholders.

42
 Figure 5.3 Graphical User Interface of SafeConc – the program
1. Creating a new model
 Go to Project>>New (or press Ctrl+N)
 Enter the details about the model at corresponding fields (see Figure 10)

Figure 5.4 Main Data Input Form

2. Generating regular Mesh

Follow these steps:
 Go to Mesh>>Generate Mesh (or press F5)

43
 Pseudo code for generating regular structured mesh of a rectangular body
horizontalGridSize = width ÷no of grid columns
verticalGridSize = height÷no of grid rows
For i=0 to no. of grid rows
For j=0 to no. of grid columns
create new node at (j*horizontalGridSize , i*verticalGridSize)
save a reference to the new node
End For
End For

For i=0 to no. of grid rows-1

For j=0 to no. of grid columns-1
{create new element with nodes – (node[i*nCols + j], node[i*nCols + j + 1],
node[(i + 1) * nCols + j + 1], node[(i + 1) * nCols + j]) }
save a reference to the new element
End For
End For

3. Solve
 Go to Run>>Calculate Temperatures (or press F6)
 Go to Run>>Calculate stresses (or press F7)

4. Plotting Contours
Contour maps of temperature, stresses, strength and displacements is plotted directly on the
screen. For this, the selected parameter is calculated at some discrete points (25 in each
direction per element, i.e., 625 in total in a single rectangular element). These values are then
𝑣𝑎𝑙𝑢𝑒−𝑚𝑖𝑛
scaled from 0 to 255 according to the formula: 𝑠𝑐𝑎𝑙𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 = ∗ 255. The scaled
𝑚𝑎𝑥−𝑚𝑖𝑛

values are then converted into RGB values according to following function:

𝑅 = 0, 𝐺 = 4𝑥, 𝐵 = 255 𝑓𝑜𝑟 𝑥 = 0 𝑡𝑜 63

𝑅 = 0, 𝐺 = 255, 𝐵 = 255 − 4(𝑥 − 64) 𝑓𝑜𝑟 𝑥 = 64 𝑡𝑜 127
f(x) = {
𝑅 = 4(𝑥 − 128), 𝐺 = 255, 𝐵 = 0 𝑓𝑜𝑟 𝑥 = 128 𝑡𝑜191
𝑅 = 255, 𝐺 = 255 − 4(𝑥 − 192), 𝐵 = 0 𝑓𝑜𝑟 𝑥 = 192 𝑡𝑜 255

To see the plots, follow these steps:

 Go to Results>>Show contour>>Temperature
 Output will be plotted as thermal contour (see Figure 11)

44
  Other contours available: Stresses (horizontal, vertical, shear, Principal stresses, von
mises), Displacement, Strength, Cracking risk.

Figure 5.5 Thermal contour plot

5. Modifying cement properties
 Go to Hydration>>Modify Data
 Enter cement phase proportions, fineness and water-cement ratio in the input form that
appears (see Figure 12)
 Enter activation energy if it is known, otherwise leave it zero and the apparent activation
energy will be calculated using empirical relationship provided by Lin and Meyer.

Figure 5.6 Hydration data input form

45
6. Showing the deformed shape
 After temperature and stress analysis have been completed, select View>>Show
Deformed shape

Figure 5.7 Showing deformed state in SafeConc

7. Plotting graphs

Figure 5.8 Graph of porosity vs time plotted by Safeconc

 Go to View Results>>Charts
 Go to Y-axis>>Select Porosity for plotting porosity vs time curve
 Go to Element>>Select element number
46
  Other graphs are available such as Strength vs Time, Degree of Hydration vs Time,
Temperature vs Time, Rate of heat release vs time, Cumulative heat released vs Time
and Phase volume vs Time curves.
 Data can be tabulated for export to MS Excel by clicking Table>>Tabulate.

In this section, working of the program ‘SafeConc’ has been discussed. Modular design (Figure
5.1) enables easier debugging and revisions. Extra effort has been made to make the program
intuitive and user-friendly. Users with zero programming knowledge can also use it. However,
to make radical changes like using hydration model other than Lin and Meyer model, some
amount of programming is required to understand and alter the code.

47
 RESULTS, DISCUSSION AND VALIDATION

6.1 DEGREE OF HYDRATION

The output of Module 1 has been taken in terms of change in degree of hydration with time.
Experimental data from [16] has been selected and compared with output from SafeConc
(Figure 6.1). The parameters used for the modelling are as according to Table 3. Temperature
was maintained at 25oC, i.e., isothermal conditions.

Table 6.1 Parameters for modelling of hydration for the material used in [16]
pC3S pC2S pC3A pC4AF Fineness w/c Cement content
57.1% 15.3% 7.7% 8.4% 325 m2/Kg .35 529.5 Kg/m3

0.8

0.7
Degree of hydration

0.6

0.5

0.4 Model
Experimental
0.3

0.2

0.1

0
1 10 100 1000
Time /hr

Figure 6.1 Degree of hydration vs time curve for w/c=0.35. Experimental data from [16]

49
6.2 ADIABATIC TEMPERATURE RISE
Adiabatic tests on cement concrete has been conducted by Bentz [16]. The chemical
composition and fineness of the cement were that in Table 6.1. These properties have been
used as parameters in SafeConc and the hydration of cement has been modelled. The results
from SafeConc are compared with experimental results by [16] in Figure 6.2 through Figure
6.4.

Adiabatic Temperature Rise

100
90
80
Temperature 0C

70
60
50
Model
40
30 Experimental
20
10
0
0 20 40 60
Time (hr)

Figure 6.2 Adiabatic temperature rise curve for w/c=0.3. Experimental data: Bentz (1999)
[16]

Adiabatic Temperature Rise

100
90
80
Temperature 0C

70
60
50
Model
40
30 Experimental
20
10
0
0 50 100 150
Time (hr)

Figure 6.3 Adiabatic temperature rise curve for w/c=0.35. Experimental data: Bentz (2009)
[16]

50
 Adiabatic Temperature Rise
90
80
70

Temperature 0C
60
50
40 Model
30 Experimental
20
10
0
0 50 100 150
Time (hr)

Figure 6.4 Adiabatic temperature rise curve for w/c=0.45. Experimental data from Bentz
(2009) [16]

6.3 HEAT PRODUCTION RATE

Similar type of work has been done by Paine et al. (2005) [43]. In their study, they examined
the rate of heat release with time and maximum heat value. Calorimetric tests were conducted
on 30g cement mortar samples, each having 0.5 w/c ratio. Cements used were Ordinary
Portland Cement (OPC) as well as OPC blended with GGBS or Fly Ash. Results from their
study are compared with those obtained from SafeConc and shown in Figure 6.5 through Figure
6.7. Phase composition of cement used for modeling has been calculated using Bogue’s
relationship from the oxide composition of cement mentioned in the paper. Parameters used
for hydration modelling have been tabulated as follows:

Table 6.2 Parameters for modelling of hydration for the material used in [43]
pC3S pC2S pC3A pC4AF Fineness w/c
57.9084 17.966 9.91 7.91 405 m2/Kg 0.50

51
 Rate of Heat Release @ 200C
3.5

3
Power (W/Kg of cement)

2.5

1.5 Experiment
Model
1

0.5

0
0 10 20 30 40 50 60 70
Time (hr)

Figure 6.5 Rate of heat release curve for w/c 0.5 at 200C (isothermal). Experimental data
from Paine (2005) [43]

Rate of Heat Release @ 400C

12
Power (W/Kg of cement)

10

6
Experiment
4 Model

0
0 10 20 30 40 50 60 70
Time (hr)

Figure 6.6 Rate of heat release curve for w/c 0.5 at 400C (isothermal). Experimental data
from Paine (2005) [43]

52
 Rate of Heat Release @ 600C
35

Power (W/Kg of cement) 30

25

20

15 Experiment
Model
10

0
0 10 20 30 40 50 60 70
Time (hr)

Figure 6.7 Rate of heat release curve for w/c 0.5 at 200C (isothermal). Experimental data
from Paine (2005) [43]

The curves can be said to have reasonable fit in the declining part. However, during the early
stage, SafeConc predicts a lower peak at low temperature and a higher peak at high temperature
as compared to the respective experimental data. Evidently, SafeConc predicts a greater
influence of temperature on rate of hydration.

6.4 TEMPERATURE RISE WITH CONVECTION LOSS (SEMI-ADIABATIC)

A 1m x 1m x 0.3m concrete prism had been cast in IIT Delhi by Rohit Baghla [22], one of the
M.Tech students under Construction Technology & Management program, in the year 2013.
The temperature rise inside the prism was monitored using temperature gauges along with the
ambient temperature in the month of February. The results from the experiment are borrowed
and compared with the results from this study.

53
 Figure 6.8 Experimental setup in [22] showing insulating boundaries around the concrete

The specimen was constructed according to the dimensional layout shown in Figure 6.9. All
faces except the top were insulated by 100 mm thick Styrofoam layer enclosed in plywood.
The thermal conductivity of these two materials were calculated to be 0.035 W/m/K and 0.15
W/m/K respectively. From this, overall U-value can be calculated by

1/U = t1/k1 +t2/k2 + 1/h (6.1)

Figure 6.9 Specimen diagram showing dimensional layout of the specimen

The convection coefficient applied on the top surface is taken from the curve for zero wind
velocity in Figure 2.2. The convection coefficient at the vertical surfaces and the bottom surface
is equal to the U-value of the form material (plywood+Styrofoam). Mix design, chemical
composition of cement and other physical properties of concrete and other materials are as
follows:

Table 6.3 Mix design and composition of cement taken for experimental analysis in [22]
pC3S pC2S pC3A pC4AF Fineness w/c Cement content
Mix 1 41% 31% 8% 11% 225 m2/Kg 0.467 385 Kg/m3
Mix 2 -do- -do- -do- -do- -do- 0.36 500 Kg/m3
Other parameters:-

1. Conductivity of concrete, kxx = kyy = 2.6 W/m/k

2. Initial temperature = 22.5oC
3. Average ambient temperature = 15oC
54
 4. Average convection coefficient at top, h = 11 W/m2/K
5. Overall thermal coefficient for vertical faces, U = 0.3 W/m2/K

Figure 6.10 and Figure 6.11 compares the experimental results with that obtained from the
simulations. Although, the results do not have a very good fit, yet it is acceptable for practical
purposes.

50 45
45 40

Temperature oC
40 35
Temperature oC

35 30
30 25
25 20
20 15 Model
15 Model 10
10 5 Experiment
5 0
0 0 50 100 150
0 50 100 150 Time (hr)
Time (hr)

(a) 400 mm from bottom (b) 800 mm from bottom

Figure 6.10 Temperature rise in a concrete prism. Mix 1 in [22]

60 50
45
50 40
Temperature oC
Temperature oC

35
40
30
30 25
20
Model
20 Model 15
10
10 5
0
0
0 50 100 150
0 50 100 150
Time (hr) Time (hr)

(c) 400 mm from bottom (d) 800 mm from bottom

Figure 6.11 Temperature rise in a concrete prism. Mix 2 in [22]

The slope of the declining limb of the temperature curve is slightly less in the model than the
experimental curve. This is due to the fact that analysis is done in 2-D in SafeConc, heat leakage
through the formwork in z-direction could not be modeled in the simulation. This strengthens
the argument that the accuracy of simulation depends upon the quality of the input parameters.
55
6.5 STRENGTH DEVELOPMENT
Module 4 of SafeConc takes input as porosity from the hydration module, i.e., Module 1 and
uses extended Zheng’s model[10] (strength-porosity relationship) to calculate strength as
discussed in Section 4.4. The output from module 4 is shown in Figure 6.12 for arbitrarily
chosen model constants. These curves do not represent any real cement, and are included here
only to give demonstration of how water-to-cement ratio influences the strength development
of concrete.

30
Compressive Strength (N/mm 2)

25

20
w/c = 0.35
15
w/c = 0.4
10 w/c = 0.45
w/c = 0.5
5

0
10 100 1000
Time (log hr)

Figure 6.12 Strength development curve simulated by SafeConc

30
Compressive Strength (N/mm 2)

25

20

15

10

0
0.35 0.4 0.45 0.5
w/c ratio

Figure 6.13 Variation of 28-day strength with w/c ratio (Simulation results)

56
The cement composition used for simulation is taken as that in Table 6.1. Isothermal condition
were maintained throughout the analysis by providing very high (in the order 107) convection
coefficient. The isothermal temperature used was 25oC. Model constants are 𝜎0 = 69.4 & 𝑝𝑐 =
0.562.

Figure 6.13 shows variation of 28-day compressive strength with water-cement ratio as
predicted by the model.

6.6 PHASES

Figure 6.14 Phase generation and consumption as simulated by SafeConc

Figure 6.14 shows amount of reactants consumed and products produced. In this diagram, only
the solid phases are shown. This diagram points out the order of reactions that are taking place.
It should be noted that ettringite is only produced till gypsum is present. Once gypsum is
consumed completely, C3A starts to react with ettringite (AFt phase) to produce monosulphate
(AFm). Due to this, concentration of ettringite starts to drop, which is evident in the figure.

If ettringite is not produced due to absence of gypsum, or it has been consumed after reacting
with aluminate and ferrite, hydrogarnet is produced by reaction of C3A directly with water.

57
This reaction is very fast and is the cause of flash set in concrete. That’s why gypsum is added
in cement clinker before grounding.

6.7 DISPLACEMENTS
So far we have checked the validity of results from module 1, 2 and 4. Module 3 concerns
about the stresses and deformations associated with temperature change in hydrating concrete.
The prime objective of this study is to analyze the thermal stresses and calculate the risk of
cracking.

However, this part is the most difficult to validate. That is due to the difficulty in measuring
actual stresses inside a concrete body. A lot of instrumentation and monitoring is required to
log these data. During the whole study no reliable data was found in literature through which
this model could be validated.

But some respite can be taken by accepting the fact that finite element method is nothing but
pure math. It’s like solving hundreds of simultaneous equations. Only thing there to be
concerned about is the validity of the equations. Great care has been taken during formulating
these equations. Some classical problems are taken to check their correctness.

Figure 6.15 shows a cantilever beam loaded with 100 KN concentrated load at its free end. The
deflection of the beam is calculated using beam bending theory and compared with that
obtained from the finite element method.

𝑃𝐿3
= = 1.333 ∗ 10−2
3𝐸𝐼

Figure 6.15 Deflection of simple cantilever beam (L=10m,B=1m,D=1m,E=30GPa) loaded at

free end. Theoretical=1.333*10-2, Model=1.313*10-2: Error=1.5%

58
 𝑃𝐿3
= = 3.858 ∗ 10−2
48𝐸𝐼

Figure 6.16 Deflection of centrally loaded simply supported beam (L=10m,

B=1m,D=0.6m,E=30GPa). Theoretical=3.858*10-3, Model=3.833*10-3: Error=0.6%

To verify stress results generated due to thermal loads, sample problems were created and
tested in SafeConc and a commercially available FEM software – ABAQUS/CAE. One of the
problem consists of a cube of side 10 m, which was modelled and a uniform thermal load was
applied. For an arbitrarily chosen cement type, uniform temperature of the body was found to
be 74oC at 100th hour under adiabatic condition, from Module 1. The initial temperature was
given as 20oC and the bottom face was restrained from any type of motion.

Same structure was modelled in ABAQUS/CAE and same temperature rise was provided as
input.

Figure 6.17 compares results from both software. Although, displacement pattern is same, the
magnitudes are slight off with a deviation of about 0.8% in the extremes. This method is not a
reliable way to validate the model, since both SafeConc and ABAQUS uses finite element
method. This is done only to ensure that the implementation of finite element method in
SafeConc has been done correctly and there is no apparent logical error.

(a) (b)

59
Figure 6.17 Sample problem analyzed in (a) SafeConc, and (b) ABAQUS/CAE (parameter –
horizontal displacement)

6.8 STRESSES
Sample problem in section 6.7 is again taken in this section for analysis of stress data. Figure
6.18 shows the plot of horizontal stresses from both software. Again distribution is similar, yet
the magnitudes have a deviation of about 5%. Figure 6.19 compares the results for vertical
stresses. Deviation is found to be about 16%.

(a) (b)
Figure 6.18 Sample problem analyzed in (a) SafeConc, and (b) ABAQUS/CAE (parameter –
horizontal stress)

60
 (a)problem analyzed in (a) SafeConc, and (b) ABAQUS/CAE
Figure 6.19 Sample (b) (parameter –
vertical stress)

6.9 CRACKING RISK

Figure 6.20 is a plot of cracking risk in a sample body. The magnitude of risk doesn’t show the
probability of cracking. It is just the factor by which stress exceeds the strength. Here, the factor
of safety is taken as unity, which can be changed if necessary. Strength is reduced by dividing
with the factor of safety while calculating risk. In general, if the risk is more than 1, it is more
likely that cracks will appear. However, for a risk of less than 1, there are less chances of
cracking, but it is not nil.

Figure 6.20 Cracking risk

In this problem, since the body is restrained from the bottom, maximum probability of cracking
is near bottom only.

61
6.10 DISCUSSION
The model is able to predict the primary variables, such as, degree of hydration, cumulative
heat of hydration and adiabatic temperature better than the secondary variables such as stresses
and strains. The strength estimation part is very basic and needs revision.

After scrutinizing all the results, it has been found that although the simulation methodology
used in SafeConc is approximate and has many assumptions, yet it can have practical
applications. Its accuracy is reasonable and can be improved by calibrating with experimental
tests. Like any other simulation software, its effectiveness depends upon the quality of the input
parameters as well as experience of the user.

The ‘safety factor’ or the ‘factor of safety’ is used only to tackle the uncertainties arising from
these analysis. It should be always kept in mind that simulations are not designed to eliminate
experiments, but to reduce their number and improve their effectiveness. Experiments are time-
consuming, but are more realistic. Simulations, on the other hand, are fast and points out the
locations where more scrutiny is required.

62
 SUMMARY & CONCLUSION

7.1 SUMMARY
This study has been an attempt on modelling and prediction of thermal stresses in concrete
during early stages of hydration.

 Various chemical reactions involved in cement hydration have been studied in detail.
 Heat liberation during hydration of concrete has been studied extensively.
 A user-friendly computer program has been developed which can simulate heat
liberation in concrete.
 The program is titled as SafeConc and incorporates optimized algorithms to simulate
cement hydration and release of heat associated with it.
 It can find out the thermal profile of a concrete body undergoing hydration, at any point
of time.
 Validation of the model with experimental results shows it to be quite satisfactory.
Though, only the rate of hydration, amount of heat release and temperature rise have
been verified.
 Stresses inside the body is calculated with the help of static finite element analysis.
 Due to lack of data, stresses calculated from the model could not be verified. Maximum
care has been taken during implementation of models and since finite element method
is nothing but pure math, it is assumed that no additional error is incurred except those
associated with the method itself. Despite of various limitations of FEM, it is still
widely used by structural engineers.
 Strength of concrete is calculated on the basis of strength-porosity relationships.
 Finally, cracking risk can be predicted which warns a user for potential failure which
can occur if no further care is taken.

63
7.2 CONCLUSION
After analyzing the various results, the following conclusions can be drawn: -

 Lin and Meyer model gives satisfactory results. Experimental validation of the results
shows that this model is able to predict degree of hydration, rate of hydration, and rate
of heat release with reasonable accuracy.
 Finite Element Method, although an approximate method, gives results accurate enough
for practical implications.
 More work is required in this field to develop a wholesome, versatile and accurate
simulation methodology.
 Although, 2-D analysis is fast and simple, yet 3-D modelling is required to incorporate
several types of boundary conditions and modelling constraints.

7.3 POTENTIAL USE OF THE SOFTWARE

 Research – Researchers working in similar fields can benefit from the software by
comparing their data with the results from SafeConc. Also, since this is an open-source
program, they can use their own model for hydration and analyze the results using the
graphical environment.
 Industry – Although, SafeConc is very basic in nature and has some major
shortcomings like types of boundary conditions available to be applied, yet with some
further development it can be used by practicing engineers to check their design for
possible thermal cracks.
 Students – Students studying concrete technology can use this software to understand
how various properties of cement can influence the quality of concrete prepared from
it. Also, they can understand the effect of temperature on the development of strength
of the concrete.

7.4 FUTURE SCOPE OF STUDY

1. 3D extension – This study is focused on plane stress analysis and 2D heat diffusion. This
simplifies the problem for time-efficient calculations. However, more detailed results can
be obtained if analysis is done in 3D. Only slight modification is needed in the algorithms
to convert it into 3D analysis.

64
2. CAD modelling – Although option is provided in the program to import any type of
complex structure from a 2D diagram. It will be better if user is able to directly model the
structure in the program or import it from popular CAD softwares.
3. Experimental validation – Due to lack of data on thermal stresses, there’s no way to check
the accuracy of the model.
4. Inclusion of fly ash and other additives – This study focuses on concrete with Ordinary
Portland Cement (OPC). Mineral additives like fly ash, ground granulated blast furnace
slag, microsilica etc. have been increasingly used in manufacturing concrete. Including
these in analysis would be more useful.

65
REFERENCES

[1] F. Lin, and C. Meyer, “Hydration kinetics modeling of Portland cement considering the effects of
curing temperature and applied pressure,” Cement and Concrete Research, 39 (2009) 255–265,
2009
[2] J. J. Thomas et al., “Modeling and simulation of cement hydration kinetics and
microstructure development,” Cement and Concrete Research 41 (2011) 1257–1278, 2011
[3] L. Buffo-Lacarriere et al., “Multiphasic finite element modeling of concrete hydration,”
Cement and Concrete Research 37 (2007) 131–138, 2007.
[4] A. Ilc, G. Turk, F. Kavcic, G. Trtnik, “New numerical procedure for the prediction of
temperature development in early age concrete structures,” Automation in Construction,
18 (2009) 849–855, 2009.
[5] Y. Yuan and Z. L. Wan, “Prediction of cracking within early-age concrete due to thermal,
drying and creep behavior,” Cement and Concrete Research, 32 (2002) 1053–1059, 2002
[6] Nannayakara, “Importance of controlling temperature rise due to heat of hydration in
massive concrete elements,” Univ. of Moratuwa.
[7] P. F. Siew T. Puapansawat Y. H. Wu, “Temperature and heat stress in a concrete column
at early ages,” ANZIAM J. 44 (E) ppC705–C722, 2003
[8] Z. Najafi and K. Ahangari, “The Prediction of Concrete Temperature during Curing Using
Regression and Artificial Neural Network,” Journal of Engineering Volume 2013, Article
ID 946829, 2013
[9] Y. Lee , M. Choi , S. Yi and J. Kim, “Experimental study on the convective heat transfer
coefficient of early-age concrete,” Cement and Concrete Composites, vol. 31, pp. 60-71,
2009.
[10] X. Chen, S. Wu, J. Zhou, “Influence of porosity on compressive and tensile strength of
cement mortar,” Construction and Building Materials 40 (2013) 869–874, 2013.
[11] R. Kumar, B. Bhattacharjee, “Porosity, pore size distribution and in situ strength of
concrete,” Cement and Concrete Research 33 (2003) 155 – 164, 2013
[12] M. Zheng, X. Zheng and Z.J. Luo, “Fracture strength of brittle porous materials,”
International Journal of Fracture 58:R51-R55, 1992.
[13] E. Gallucci, X. Zhang, K.L. Scrivener, “Effect of temperature on the microstructure of
calcium silicate hydrate (C-S-H),” Cement and Concrete Research 53 (2013) 185–195,
2013
67
[14] S. Sabbagh-Yazdi, T. Amiri-SaadatAbadi, F. M. Wegian, “2D Linear Galerkin finite
volume analysis of thermal stresses during sequential layer settings of mass concrete
considering contact interface and variations of material properties Part 1: Thermal
analysis,” Journal of the south african institution of civil engineering vol 55 no 1, april
2013, pages 94–103, paper 829 part 1, 2013.
[15] D.P. Bentz , E. Garboczi , C. Haecker and O. Jensen, “Effects of particle size distribution
on performance properties of Portland cement-based materials,” Cement and Concrete
Research, vol. 29, pp. 1663-1671, 1999.
[16] D.P. Bentz, V. Waller and F. de Larrard, “Prediction of adiabatic temp rise in conventional
and highperformance concretes using a 3-D microstructural model,” Cement and Concrete
Research, vol. 28, No.2, pp. 285-297, 1998.
[17] J. Zheng, J. Zhang, G. W. Scherer, “Prediction of the degree of hydration at initial setting
time of cement paste with particle agglomeration,” Cement and Concrete Research 42
(2012) 1280–1285, 2012
[18] B. Osbaeck and V. Johansen , “Particle Size distribution and rate of strength development
of Portland cement,” Journal of American Ceramic Society, vol. 72, pp. 197-201, 1989.
[19] Portland Cement Association website, http://www.cement.org.
[20] J.I. Escalante-Garcı´a and J.H. Sharp, “Effect of temperature on the hydration of the main
clinkerphases in portland cements: Part I, neat cements,” Cement and Concrete Research,
Vol. 28, No. 9, pp. 1245–1257, 1998
[21] P. Termkhajornkit and R. Barbarulo, “ Modeling the coupled effects of temperature and
fineness of Portland cement on the hydration kinetics in cement paste,” Cement and
Concrete Research, vol. 42, pp. 526-538, 2012
[22] R. Baghla, “Determination of thermal gradient and Effect on strength due to heat of
Hydration in mass concrete,” M.Tech Thesis, IIT Delhi, 2013.
[23] S. Joseph, “Modeling early age hydration and risk of thermal cracking in mass concrete,”
M.Tech Thesis, IIT Delhi, 2013.
[24] M. Satyaketu, “Modeling the temperature rise in cement pastes in semi adiabatic
conditions,” M.Tech Thesis, IIT Delhi, 2013.
[25] G. De Schutter, “Finite element simulation of thermal cracking in massive hardening
concrete elements using degree of hydration based material laws,” Computers and
Structures 80 (2002) 2035-2042, 2002.

68
[26] G. P. Nikishkov, “Introduction to the finite element method,” 2004 Lecture Notes.
University of Aizu, Aizu-Wakamatsu 965-8580, Japan.
[27] J. Li, "Predicting early-age thermal behavior of mass concrete for bridge foundation."
Graduate Theses and Dissertations Paper 12659, 2012
[28] Lecture Modules [Online]. http://www.cdeep.iitb.ac.in/nptel/Mechanical/Heat
%20and%20Mass%20Transfer/Conduction/Module%202/interactivepages/prob2.1ps.ht
ml, IIT Bombay.
[29] D S Guo, E Y Chen, G L Low, J L Yang, “Experimental modelling of temperature rise of
mass concrete by FDM method,” 26th Conference on Our world in concrete & structures:
27 - 28 August 2001.
[30] A. Patini, “Numerical analysis of temperature development in concrete at an early age”,
M.S. Thesis, University of the Witwatersrand, Johannesburg, 2011.
[31] X. Wang, H. Cho, H. Lee, “Prediction of temperature distribution in concrete
incorporating fly ash or slag using a hydration model,” Composites: Part B 42 (2011) 27–
40, 2011.
[32] M. Cervera, R. Faria, J. Oliver, T. Prato, “Numerical modelling of concrete curing,
regarding hydration and temperature phenomena,” Computers and Structures 80 (2002)
1511–1521, 2002.
[33] L. D’Aloia and G. Chanvillard, “Determining the ‘‘apparent’’ activation energy of
concrete Ea—numerical simulations of the heat of hydration of cement,” Cement and
Concrete Research, vol. 32, pp. 1277-1289, 2002.
[34] O. M. Abdulkareem, A. H. Ahmad, “Effect of high temperature on mechanical properties
of concrete containing admixtures,” Civil Engineering Dept. University of Mosul, 2009.
[35] L. Hwang, S. Rizkalla, “Effective tensile stress-strain characteristics for reinforced
concrete,” Graduate Thesis, Department of Civil Engineering, University of Manitoba,
Winnipeg, Manitoba.
[36] D. Yoon, H. Song2), and C. Min, “Finite Element Analysis on Hydration Heat of Concrete
under the Influence of Reinforcing Steel Bars,” Journal of KCI, Vol.17, No.1, 2005.
[37] H. Ma and Z. Li, “Realistic pore structure of Portland cement paste: experimental study
and numerical simulation,” Computers and Concrete, Vol. 11, No. 4 (2013) 317-336, 2013.
[38] E. Cerro-Prada, M.J. Vázquez-Gallo, J. Alonso-Trigueros, A.L. Romera -Zarza,
“Modelling hydration process of cement nanoparticles by using an agent-based molecular

69
 formation algorithm,” E.U.I.T. Obras Públicas, Universidad Politécnica de Madrid
C/Alfonso XII, 3, 28014 Madrid Spain.
[39] M. Zajac, M. B. Haha , “Experimental investigation and modeling of hydration and
performance evolution of fly ash cement,” Materials and Structures (2014) 47:1259–1269
[40] D. P. Bentz, “Influence of water-to-cement on hydration kinetics: Simple models based
on spatial considerations,” Cement and Concrete Research, Vol. 36, No .2, pp. 238-244,
2006.
[41] J. Escalante-Garcia and J. Sharp , “Effect of temperature on hydration of the main clinker
phases in Portland cements: Part I, Neat Cements,” Cement and Concrete Research, vol.
38, pp. 1245-1257, 1998.
[42] J.I. Escalante-Garcia, “Nonevaporable water from neat OPC and replacement materials in
composite cements hydrated at different temperatures,” Cement and Concrete Research
33 (2003) 1883–1888, 2003.

[43] K. A. Paine, L. Zheng and R. K. Dhir (2005) “Experimental study and modelling of heat
evolution of blended cements.” Advances in Cement Research, 17 (3). pp. 121-132. ISSN
0951-7197.

[44] D. Logan (2012). "A first course in the Finite Element Method".

[45] K. J. Bathe (2006). “Finite Element Procedures.”

[46] Khennane, A. (2013), "Introduction to Finite Element Analysis Using MATLAB and
Abaqus."

[47] Kim, Soo Geun, "Effect of heat generation from cement hydration on mass concrete
placement" (2010). Graduate Theses and Dissertations. Paper 11675, 2010.

[48] W. M. Lira, P. R. Cavalcanti, L. C.G. Coelho, L. F. Martha, “A modeling methodology

for finite element mesh generation ofmulti-region models with parametric surfaces,”
Computers & Graphics 26 (2002) 907–918, 2002.

[49] C. Dunant, “Experimental and Modelling Study of the Alkali-Silica–Reaction in

Concrete,” Thèse no 4510, École Polytechnique Fédérale De Lausanne, 2009.

[50] P.K. Mehta and P.J.M. Monteiro, “Concrete: Microstructure, Properties, and Materials”.

[51] Lecture Modules, Technological Institute of Phillipines.

70
[52] Petr Krysl, “A Pragmatic Introduction to the Finite Element Method for Thermal and
Stress Analysis”.

[53] N. Ukrainczyk, E.A.B. Koenders, K. van Breugel, “Multicomponent Modelling Of

Portland Cement Hydration Reactions,” Second International Conference on
Microstructural-related Durability of Cementitious Composites, 11-13 April 2012,
Amsterdam, The Netherlands, 2012.

[54] A. Caggianoa, M. Pepe, E. A.B. Koenders, E. Martinelli and G. J. Etse, “Numerical

Modeling of Hydration Process and Temperature Evolution in Early Age Concrete,”
Mecánica Computacional Vol XXXI, págs. 1893-1907 (artículo completo), 2012.

[55] E. Martinelli, E. A.B. Koenders, A. Caggiano, “A numerical recipe for modelling

hydration and heat flow in hardening concrete,” Cement & Concrete Composites 40 (2013)
48–58, 2013.

[56] M. Zacak, S. Garrault, J.P. Korb, A. Nonat, “Effect of temperature on the development of
C-S-H during early hydration of C3S,” 12th International Congress on the Chemistry of
Cement, Jul 2007, Montreal, Canada. pp.W1-06.2, 2007.

[57] Online forum – http://www.stackoverflow.com.

71
FLOWCHARTS

Figure 0.1 Flow Chart for Main Program

73
Figure 0.2 Flowchart for calculation of degree of hydration

74
 Start

Fix sample mass

(say 0.00001g)

Get degree of hydration at

Yes
time step

If D>0 Calculate reacted thickness =

𝛼∗𝐷𝑚𝑎𝑥
2

Calculate the volume

Deduct reacted thickness from
inside reacted thickness
radius
* 3.15/2

Calculate new dia.

from above two

Figure 0.3 Flowchart for microstructure development

75