Fracture Mechanics of Concrete PDF

Aalborg Universitet
Fracture Mechanics of Concrete
Ulfkjær, Jens Peder
Publication date:
1992
Document Version
Publisher's PDF, also known as Version of record
Link to publication from Aalborg University
Citation for published version (APA):

Ulfkjær, J. P. (1992). Fracture Mechanics of Concrete. Aalborg: Dept. of Building Technology and Structural
Engineering.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at [email protected] providing details, and we will remove access to
the work immediately and investigate your claim.
Downloaded from vbn.aau.dk on: januar 23, 2019

INSTITUTTET FOR BYGNINGSTEKNIK
DEPT. OF BUILDING TECHNOLOGY AND STRUCTURAL ENGINEERING
AALBORG UNIVERSITETSCENTER • AUC • AALBORG • DANMARK
JENS PEDER ULFKJÆR

FRACTURE MECHANICS OF CONCRETE
JULY 1992 PH.D. THESIS
FRACTURE MECHANICS OF
CONCRETE
Jens Peder t.ntkjær

Department of Building Technology and Structural Engineering
University of Aalborg
Denmark
ACKNOWLEDGEMENTS
The present thesis Fracture Mechanics of Concrete has been prepared in conneelion with my
Ph.D. study in the period of March 1989 to June 1992 at the Department of Building
Technology and Structural Engineering, University of Aalborg, Denmark.
I would like tothank Dr. H.H. Bache of Aalborg Portland, Denmark, for fruitful discussions
and for technical assistance with the produetion of the high-strength concrete specimens.
Also thanks to assistant engineer Henning Andersen and the rest of the staffat the Labaratory
of Structural Engineering, University of Aalborg for their helpfulness and proficiency.
The financial support from the Danish Council for Scientific and Industrial Research is
gratefully acknowledged.
I would also like to thank my friends at the Department and especially Associate Professor
Niels Andreas Harder, for many hectic and very fruitful discussions on size effects and
brittleness.
Finally, special thanks to my supervisor Associate Professor Ph.D. Rune Brincker for
valuable, inspiring and attenlive guidance offered during my study.
The skillful proof reading of the manuscripts has been made by Secretary Solveig
Hesselvang, and the drawings of many figures have been carried out by Draughtsman Poul
Skørbæk. Their careful work is greatly appreciated.
Aalborg, July 1992 Jens Peder Ulfkjær

List of Contents
l IN'TR.ODUC'riON. . . . III •••••• "' ••••••••••••• III •••• "' •••• "' • • 5
1.1 Background and Motives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Scoj)e of Thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Readers Guide. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 FRACTURE AND FRACTURE MODELS OF

CONCRErE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1 Fracture of Concrete and High Strength Concrete. . . . . . . . . . • . . . . . 8
2.1.1 Concrete on the Micro-level. . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 The Fracture Process in Compression. . . . . . . . . . . . . . . . . 15
2.1.3 The Fracture Process in Tension. . . . . . . . . . . . . . . . . . . . 18
2.2 Linear Elastic Fracture Mechanics. . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.1 Basic L'EFM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.2 LEFM and Concrete. . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Nonlinear Elastic Fracture Mechanics. . . . . . . . . . . . . . . . . . . . . . 26
2.3.1 The Fietitlous Crack Model. . . . . . . . . . . . . . . . . . . . . . . 26
2.3.2 The Crack Band Model. . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3.3 The Two Parameter Model. . . . . . . . . . . . . . . . . . . . . . . 29
2.3.4 The Effective Crack Model. . . . . . . . . . . . . . . . . . . . . . . 30
2.3.5 R-Cu"CVes. • •••••.•••••••.••.••.•.....••.••• 30
2.4 Brittleness and Size Effect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.1 Example I. Size effect of plate made of a perfectly
brittle/perfectly duetile materlaL . . . . . . . . . . . . . . . . . . . 35
2.4.2 Example II. Brittleness and stability of concrete rod. . . . . . . . 39
2.4.3 Example m. Stability and brittleness of glued beam. . . . . . . . 40
2.5 References for Chapter 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3 NUMERICAL METHODS IN CONCRErE

FRACTURE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1 The Hillerborg, Modeer and Petersson Methods. . . . . . . . . . . . . . . . 47
3.2 The sub-structure meth.od. • . • . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3 The Direct Sub-Structure Method. . . • . . . . . . . . . . . . . . . . . . . . . 51
3.4 Stability of The Direct Sub-structure Method. . . . . . . . . . . . . . . . . . 56
3.5 Model Evaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.6 References for Chapter 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 ANALYTICAL METHODS IN CONCRETE

FRACTURE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1 Non-Reinforced. Concrete. . . . . • . . . . . . . . . . . . . . . . • . . . . . . . 59
4.1.1 The Ulfkjær, Brincker and Krenk Method. . . . . . . . . . . . . • 59
4.1.2 The Chuang and Mai Method. . . . . . . . . . . . . . . . . . . . . . 71
4.1.3 The Uorca and Elices Method. . . . . . . . . . . . . . . . . . . . . 75
4.1.4 Model Evaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2 Reinforced Concrete. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2.1 The Carpinteri Method. . . . . . . . . . . . . . . . . . . . . . . . .. 82
4.2.2 The Hededal, Kroon, Ulfkjær and Brincker Method. . ... .. . 88
4.2.3 Model Evaluation. . . . . . • • . . . . . . . . . . . . . . . . . . . . . 93
4.3 References for Chapter 4. . • . . . . . . . . . . . . . . . . . . . . • . . . . . • 93
S EXPERIMENTAL DETERMINATION OF
FRACTURE PROPERTIES OF IDGH-STRENGTH CONCRETE. . . . . . . 96
5.1 Size Effect Experiments. • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
S.1.1 Materials. . . . . . . • . • . • . . • . . . . . . . . . . . . . . . • . . . 96
S .l. 2 Testing equipment and procedure. . . . . . . . . . . . . . . . . . . 98
5.2 Fracture Parameter Results. . . . . . . . . . . . . . . . . . . . . . .. ... . . 101
S.2.1 The Modulus of Rupture. . . . . . . . . . . . . . . . . . .. ... . . 102
S.2.2 The Fracture Toughness. . . . . . . . . . . . . . . . . . .. ... . . 103
S.2.3 Fracture Energies According to RILEM. . . . . . . . .. ... . . 104
S.2.4 Fracture Parameters According to the Fictitious
Crack Model and Data fitting. • . . . . . . . . . . . • . . . . . . . 10S
S.2.S Conclusions on Size Effect Experirnents. . . . . . . . . • . . . . . 113
5.3 Dye Experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
S.4 Conclusions of Chapter S. • • • • • • • • • • • • • . • • • • • • • . • • • • • • • 114
5. S References for Chapter S. • • • • • • • • • • . • • • • • • • • • • • . • • . . • • 116
6 CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.1 Summary of Thesis. . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . 118
6.2 General Conclusions. • . . . . . . . . . . . . • • . . . . . . . . . . . . . . . . . 119
6.3 Future Perspective. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
APPENDIXA1
NOTATION. . . . . . . . . . . . . . • . . . • . . . . . . • . . . . . . . . . . . . . . . . . . . 121
APPENDIX A2
LOAD DISPLACEMENT CURVES. . . . • . • • • • • . . . . . . . . • . • • . • . . . . . 123
APPENDIXA3
FRACTURE PARAMETERS. 129
APPENDIXA4
OPTIMIZATION RES.'ULTS . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 135
APPENDIX AS
CRACKIN'G PROFILm. • . . . • . . . . . . • • • • . . . . • . . . . . . . . . . . . . . • • 153
APPENDIXA6
RESUME, IN' DANISH. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . 156
List of Contents
l INTRODUCTION.
1.1 Background and Motives.
In the last two decades the research tool fracture mechanics for concrete has been developed,
as it has become clear that the traditionally applied calculation tools ( elasticity theory and
piasticity theory), not always are applicable of describing certain phenomenons in concrete
fracture. The tool has also been applied by materlal researches developing new ultra strong
duetile materials, e.g. Compact Reinforced Composiles (CRC).
The compressive strength of concrete has in the same period increased dramatically, and
compressive strengthover 100 MPa by using conventional techniques has been obtained. The
increase in strength is followed by an increase in the brittleness of the materials, and it has
been assumed that the conventional techniques do not suffice for these new brittie materials.
One of the most important models which describes the fracture process of concrete in tension
is the fictitious crack model developed by Arne Hillerborg and his co-workers at the
Universily of Lund, Sweden. The fictitious crack model is a nonlinear fracture mechanical
model, based on observations made in a dispiacement controlled tensile test.
One of the phenomenons which can be deseribed by using the fictitious crack model, is the
well-known size effect on the bending tensile strength. It can also be proved that the
brittleness of a structure is dependent on the size of the structure, explaining why large
struerures are cracking more than small structures.
There are two major problems by using the fictitious crack model. Firstiy, that the materlal
parameters which describe the model must be found by performing a stable deformation
controlled tensile test, which is almost impossible for high strength concrete. Secondly, that
it is almost always necessary to use numerical methods, and only in very special cases are
analytical methods developed.
Most of the developed numerical methods are either not stable or are not able to calculate
the entire load-dispiacement curve. The analytical methods are in general very time
consuming, not accurate enough or do not describe size effects.
Therefore, it would be welcome if powerful numerical techniques, simple accurate analytical

models and indirect methods for determining the constitutive parameters were developed.
J.P. Ulfkjær. Fracturc mechanicø of concrete
1.2 Scope of Thesis.
On the basis the previous scope of this thesis is presented, as a listing of topics.
The main scqpe of the thesis is:
Developmenl ofnwnerical and analyticalfracture mechanical techniques, examination ofthe

applicability ofthese techniques on high strength concrete struerures and to develop indirect
methods for determination of the constitutive parameters.
The above scope of the thesis can be divided into the foliowing topics and a list of general
limitations.
To.pics
The development of fast numerical methods which are applicable for brittie
structures and which can calculate the entire load-dispiacement curve.
The development of approximate analytical methods where the calculation time

is smaller than that of the numerical methods, and which enables the
development of explicit analytical results.
To develop experimental techniques which can perform stable three-point

bending tests
Perform experiments with high strength concrete in order to investigate size

effects and to study the brittleness of the strong material.
Develop methods for indirect determination of the constitutive parameters in

the fictitious crack model.
Limitations.
The foliowing limitations are general in this thesis.
The compressive strength is assumed to be so large that compressive faiture

will not occur.
The fictitious crack model will be used.
lt is in general assumed that the softening relation is piece by piece linear.
6
l. Introduetion
The structure outside the crack is assumed to be modelled satisfactory by

linear elasticity theory.
In chapter 3~5 only the three-point bending geometry is considere(i.
1.3 Readers Guide.
In the foliowing it is the intention to give the reader a preliminary overview of the thesis,
which enables more selective reading.
After the introduction, chapter 2 is used to give an overview of the factors that govems
strength and cracking of normal- and high strength concrete. Models for cracking in concrete
are evaluated, and a few examples which illustrates size effects and brittleness in concrete
are presented.
The chapters 3-5 are the mainbodyin the thesis.
In chapter 3, different numerical methods which have been developed for use with the
fictitious crack model are presented and stability problems by using the models are deseribed
and solved.
Chapter 4 is devoted to analytical models which take the softening behaviour of concrete into
consideration. Also two models which are developed for lightly reinforced concrete are
presented.
In chapter S an experimental investigation of 8 different beam geomelries of a high strength

concrete is presented. Fracture parameters by using different fracture models are presented,
and size effects are observed. The numerical method is applied to determine the fracture
parameters by curve fitting.
An overall summary and condusion is given in chapter 6.
At the end of the thesis there are 6 appendixes which contain the list of notation, all the
fracture results and the load dispiacement curves of the performed experiments.
The references in the text can be found in the list of references in the last section of the
individual chapters.
7
J.P. Ultkjær. Practure mecbanics of concrete
2 FRACTURE AND FRACTURE MODELS OF

CONCRETE.
After a description of concrete and tlle fracture process of concrete a review of tlle most
important models which are used to describe concrete fracture will be given. A short
presentation of linear elastic fracture mechanics (LEFM) is given with emphasis on
terminology and definitions. Four different non-linear elastic fracture mechanical (NLFM)
models which are widely accepted are presented. The models are compared and tlle
applicability of each model is evaluated. Finally the terms 'size effect' and 'brittleness' is
discussed and different examples and stability problems is examined by using LEFM.
2.1 Fracture of Concrete and ffigh Strength Concrete.
Concrete is a heterogene()US anisatropie non-linear inelastic composite materlal tllat consists

mainly of cement, aggregates (sand and gravel) and water. In high-strengtil concrete (here
defined as concrete witll a compressive strengtil above 50 MPa) mineral admixtures and
water reducing additives are usually present.
Wittmann (1983) suggests to consider concrete on three different levels: the micro-level
where tlle structure of the hydrated cement is considered, the meso-level where large
inhomogeneities such as larger aggregates and flaws are taken into consideration and the
macro-level where the concrete is assumed to be a homogeneous isotropie continuum.
Accordingly cracks are categorized as follows:
Micro cracks : Cracks that can only be observed by an electron microscope.

Meso cracks : Cracks that can be observed using a conventional microscope.
Macro cracks : Cracks that are visible to the naked eye.
Depending on the purpose of the investigation tlle appropriate level should be chosen.
However, it will often be fruitful to develop a model at one level based on observations at
a lower level. In practice models which should apply for the practical engineer will be on tlle
macro-level.
In this chapter emphasis will be on subjects related to strength and cracking of concrete.
Strength of concrete wi11 mainly depend on the strength and stiffness of the bardened cement
8
2. Fracturc and fracturc models of concretc
paste, on the strength of the aggregates and on the bonding between the cement paste and the
aggregates. The strength of the cement paste and of the aggregates will to a great extent
depend on the porosity. This is explained in two different theories: fracture mechanics by
Griffith (1921) and the weakest link theory by Weibull (1932). These two theories will be
presented in detaUlater.
The difference between normal strength concrete (NSC) and high strength concrete (HSC)
liesin the difference in the microstructure. According to Rice (1977) "depends strength and
frocture in ceramics eritical/y on microstructural extremes rather than on averages•• and
according to Mai (1991) 11 • • but we want to re-emphasize here the need to understand the
' physical mechanisms ofthe bridges and the mechanics ofcrock-bridging. It is such knowledge
thai teaches us to design better ceramics through the control of micro-structures, such as
grain siz.es•.. 11 • Thus, in arder to understand the fracture process, which is necessary for
developing realistic fracture models, and to explain the difference between normal and high
strength concrete this presentalion starts on the micro-level.
2.1.1 Concrete on the Micro-level.
F.ach component of concrete will first bedeseribed with emphasis on NSC, then HSC will
be deseribed under the section mineral admixtures.
Portland Cement.
In the experiments performed in conneelion with this thesis the Danish low alkali sulphate
resistant portland cement certified as PC (A/HS/EA/G)1 was used.
Cement pawder consists of particles with ~ical sizes of 1-50 #'m, see Fig. 2.1:. The
specific surface (the blaine) is typically 350 m /kg. By increasing the blaine it is possible to
obtain high early strength. The long term strength is, however, not affected by the blaine,
Swammy (1986).
The solid phases of portland cement consist of four principal minerals: alite (impure
tricalcium silicate (C3Sf ), belite (impure dicalcium silicate (C2S)), impure tricalcium
aluminate (C3A) and a ferrite solid solution (tetracalcium alumina ferrite (C4AF or Fss).
When these are mixed with water several chemical processes, called the hydration process,
start. The outcome of the hydration process is the hydration products. The calcium silicates
react with water to give calcium silicate hydrates (C-S-H) and calcium hydroxide (C-H). The
aluminate and the ferrite phases react with added gypsum (calcium sulfate (CaSO..>) to give
1 According to the Danish code: DS-SBC 227
lnle customary cement nomenelature is used: Calcium oxide - CaO = C; Silicium oxide
-SiD" = S; Aluminuro oxide- Al20 3 =A; Water- H=H20;
9
J.P. Ulfkjer. FraelUre mecbania of COIICRite
100
90 l / v Sand l'la. if
L
Micro sill~ t Cement J/
80
c:
bO
7 "!J ]
·c;; 70
~ 60 l f l
....c:ø. 50 ~ 'l
::s
o 40 l Fly ash., )l l
e v l 1/ l{
< 30
20 J
l(
v ,J /
/
10
1/ ./
" ~ ~~ /
o
18 32 64 118 1/4 112 1 2 4 8 16 32 64 1/8 114 112 1 2 4
t_ nm t - l'f'1 1.- mm Grain size
Fig. 2.1: Grain curves for micro silica, fly ash, cement and sand, Herholdt et al. (1985).
two types of products, referred to as AFt (Ettringite) and AFm, see Table 2:. If gypsum is
not added the C3A will almost instantly react with water making the setting time very short,
Herholdt et al. (1985).
elinker Con- Hydration prod- Morphology

mineral tents ucts
c3s 58% C-S-H and C-H Of C-S-H: Wide range of morphology;
Early: Fibrillar or honeycomb structure;
Later: More dense
c2s 24% C-S-H and C-H Of C-H: Euhedral hexagonal habit
C3A 4% AFt Hexagonal rods
C4AF 8% AFm Hexagonal piates
Table 2: Constituents of cement.
!Jle C-S-H is themost important component in concrete and together with C-Hit controls
the strength development and most olher macroscopic properties of the bardened cement
paste. The other elinker minerals and gypsum are important during cement elinker
production, in regulation of setting time and important rheological properties, Skalny and
Roberts (1987).
It is in general difficult to follow the hydration process, and especially in the early phases.
However, some results can be given of the mechanical details of the microstructural
development during hydration by using e.g. the scanning electron mieroscape (SEM) or the
high voltage electron mieroscape (HVEM). In the HVEM it is possible to install an environ-
mental cell which makes it possible to study the hydration process as early as after 10 min
lO
2. Fracture and fracture models of concrete
mixing.
The hydration process is an exothermic reaction, and the overall progress in the hydration
process can be studied under adiahatic conditions. The heat development for each elinker
mineral is different and is approximately 500 J/g for the cement taken as a whole. However,
the rate-of..heat evolution curve is, of considerable more interest since it determines the
60
50
!2
,.,
-g...
~ 40
30
~
.c 20
._
o
~
~
~
10
o
'" ........
o 10 20 30 40 50 60 70 BO 90 100
Time [h]
Fig. 2.2:. Rate of heat evolution curve, Herholdt et al. (1985).
temperature and the temperature gradients in the concrete. A typical ratee.()f-heat curve is
seen in Fig. 2.2:.
By referring to the rate-of-heat evolution curve the hydration process can roughly be divided
into three phases: early (Q-3 hrs), middle (3-24 hrs) and late (beyond 24 hrs). An illustration
of the development of the microstructure of a cement grain during hydration is shown in
Fig. 2.3: a-e.
In the early phase, termed the induction period, the cement remains fluid and a large amount
ofheat is developed, see Fig. 2.2:. In Fig. 2.3:a. a typical anhydrous cement com consisting
of c;s, <;A and C4AF is shown. The ~S is omitted since it basically performs the same
processes as (<;S).
Immediately when the cement surface and the water come in contact the hydration starts.
Calcium, aluminates and other ions are released into the solution forming an aluminate and
silica rich gellayer, Skalny and Roberts (1987), Scrivener (1989) and Herholdt et al. (1985).
Asearly as after 10 minutes hydration small rods of AFt which have nucleated in the gel,
can be observed. At the end of the induction period the reaction of the tricalcium silicate
begins which leads to the formation of C-S-H where the rods of AFt have nucleated. That
is outside the original boundary of the cement grain which leaves a space of approximately
l #'m between the anhydrous cement grain and the AFt and the C-S-H. These produets are
11
J.P. Ulfkjær. Fracture mecbanics of concrete
' IOJ!ftl
a) b) c) d) e) f) g)
Fig. 2.3: The hydration process. a) unhydrated. b) -10 min. c) -10 hrs. d) ·18 hrs. e) 1-3
days. f) -16 days. g) years, Senvener (1989).
referred to as the outer product, Skalny and Roberts (1987) and Mehta and Aitchen (1990).
This mechanism continues with the formation of C-S-H at the outer surface increasing the
distance between the core and the shell. After about 16 hrs the AFt rods start to grow again
through the shell of C-S-H. At the end of this phase all grains smaller than 5 pm are
completely hydrated leaving hollow shells of hydration products, Scrivener (1989). As the
shell of hydration produets thickens the shell beoornes Jess permeable and will eventually
prevent the water transport in the system which ends the second phase.
The subsequent hydration which continues infinitely will then be a slow solid state process
decreasing the distance between the core and the shell. These produets are often referred to
as the inner products, Skalny and Roberts (1989), Senvener (1989) and Mehta and Aitchen
(1990). The hydrated cement grain will thus consist of the porous outer produets and the
more dense inner products.
This explains why the fracture path will go through the outer produets rather than the inner
solid produets as revealed by SEM studies of fractured surfaces. Thus, the larger amount of
inner produets that can be achieved the stronger is the cement paste. This is obtained by low
water/cement ratios, mineral admixtures and super plasticizer, Mehta and Aitcin (1990).
Coarse A~~regates.
The strength ofthe coarse aggregate, usually with maximum sizes up to 64 mm, is controlled
12
2. Fraelure and fraelure models of concretc
by the amount and sizes of cracks and by the amount of weak minerals. A high quality aggre-
gate is therefore clean, free of clay and silt, well graded and with a high density. Certain
types of aggregates have a strengths up to 200 MPa, Mehta and Aitcin (1990). It is usually
assumed that the strength of the aggregates does not have a significant influence on the
compressive strength, but as shown by Aitchen (1990) thisis not the case for high-strength
concrete. It is therefore evident to use high quality aggregates in high-strength concrete.
Bondin&
The final important aspects in concrete strength is the role of the bonding between the cement
pastc and the aggregates. The aggregates depending on shape, size and surface texture will
Fig. 2.4: SEM photo ofthe transition zone. a) (l) Silicious aggregates; crack along edge and
(2) oriented C-H crystals. b) (2) Limestone aggregate; crack in cement paste beyond a zone
of calcitereaction (l) and CaS03 crystals, Regourd (1984).
block the water distribution during bleeding. This leads to a locally increased water/cement
ratio at the paste aggregate interface. This increase leads to a change of the microstructure
next to the aggregates, termed the transition zone typical 40-50 pm wide. In addition to
larger porosity due to the bleeding there are generallylarge crystals of C-H in the transition
zone with a referenced orientation, Mindess (1989), see Fig. 2.4: and Fig. 2.5:. The weakest
zone in the transition zone does not Iie at the physical interphase but 5-10 pm away from the
aggregate. The fracture will often run in the oriented C-H crystals. lt is often assumed that
the transition zone only represents a fraction of the entire volume of the concrete.
Microscopical investigation by Diamond, Mindessand Integers (1982), has revealed that the
meandistance between the aggregates is about 75 - 100 pm which means that most of the
bardened cement paste Iies in the transition zone. These observations indicate that the
strength of the transition zone is of outermost importance in the description of concrete
13
J.P. Ulfkjær. Practuro mecbanica of concrete
Trans1tlon Zone Bulk Cement Poste
,
Fig. 2.5: The transition zone between the aggregates and the cement gel, Mehta (1986).
strength.
Mineral Admixtures.
The mineral admixtures are e.g. puzzolans, blast-furnace slag, fly ash and water reducing
additives. Puzzulanes e. g. micro silica are extremely fine particles with a size of 20 nm- 500
nm while fly ash have atmost the same size distribution as cement, see Fig. 2.1:. The micro
silica is added to make the concrete more dense, i.e. it wil1 fi1l out the voids which are too
small for the cement com and the hydration products, see Fig. 2.6:. Micro silica is not
reacting with water but with the hydration produet calcium hydroxide, making the bardened
cement paste more homogeneous and dense. When micro silica is added the transition zone
also changes. The micro silica wil1 react with the oriented C-H which is the weak part of the
transition zone, thus the interface between the cement gel and the aggregate is strengthened,
Sarkar and Aitcin (1987).
Since it is very difficult to disperse the small micro silica particles in the cement pastes it
will always be necessary to add water reducing additives (WRA), which wil1 increase the
workability. The WRA can be categorized into three groups: a) lignosulphonates b)
sulphonatcd melamine-formaldehyde condensates and c) sulphonated naphtalene-formaldehyde
condensates. a) is also having a retarding effect and is often referred to as a plasticizer. The
groups b) and c) are referred to as super plasticizer due to their low retarding effect which
allows an increased dosage, FIP/CEB (1990). A cernbination of the two types of produets
wil1 therefore often be used.
14
Fig. 2.6: Micro silica concrete. a) Silicious aggregates (l) no crack and no oriented crystals
of C-H low porosity and amotphous C-S-H (2). b) Limestone aggregate (3) no crack
amotphous C-S-H and calcitereaction (l) some C-H crystals in the matrix, Regourd (1984).
Sillea fume and water reducing additives wi1l thus change the microstructure of concrete
significantly making the structure more dense, see Fig. 2.6:.
1.1.11be Fraelure Process in Compression.
Depending of the level of the investigation each part of NSC can be thought of as a
composite materlal that consists of stronger particles which are embedded in a weaker matrix,
see Table 3:.
Level Materlal Strong Particles Weak Matrix

Micro Cement paste Ioner Produets Outer Produets
Meso Mortar Sand Cement Faste
Macro Concrete Gravel Mortar
Table 3:. Strong and weak particles classifled depending on the considered level.
Consider the stress- strain curves for aggregates, concrete, and cement paste see. Fig. 2.7:.
It is observed that the curve for the gravet is almost linear until 95% of the failure load and
that the failure is very brittle. The stress-strain curves for the other materiais are non-linear.
This is probably due to cracking at the different levels.
15
J.P. Ulfijær. Fracture mecbanica of concrete
"'
.....
"'
f:
Vl
Fig. 2. 7: Stress-strain curves for aggregates, concrete and cement paste, Mindess (1983).
The fracture process on the micro-level related to the stress-ss wtrain curve, has as far as the
author knows, only been studied for cement paste and mortar, and only by Attiogbe and
Darwin (1987). A very large study (150 specimens) of microcracking in cement paste and
mortar was made in order to understand the non-linear behavior of those two materials. The
examination was performed in a scanning electron mieroscape with a magnification of 1250
x or 2500 x (cracks with widths below 2.5 l'm). They concluded that cracks run through C-
S-H and C-H in the cement paste. In the mortar the cracks are also running at the interface
between the sand and the cement paste. The result were presented as crack densitles at
different stages of loading. For both materiais it was observed that a substantial amount of
cracking was initial cracking. For the cement paste almost 50% of the cracking at failure
were initial cracks whereas it was approximately 30% for the mortar. However, during
loading the crack densitles where growing more rapidly in the mortar than in the cement
paste. Whereby they concluded ''Ihus, sand particles appear to aet as stress raisers that
resull in a greater degree of softening and of lower strain capocity for mortar than for
cement paste'.
Crackin& on the meso-level
Due to the non-linear stress-strain curve for concrete it is beneficial to divide the ascending
branch of the stress-strain curve into four regions and the descending branch as suggesled by
Mindess (1983), see Fig. 2.8:.
This is illustrated by Hsu et al. (1963) who by using microscope and x-ray techniques studied
NSC and obscrved that there initially was a substantial amount of bonding cracks. According
to Stroeven (1975) the increasc of the specific crack area during loading is increascd with
26% which corresponds to 80% of all cracks were initial. These observations are in
16
2. Fraelure and fraelure models of concrete
Rap1d growth
~ matnx cracks
A!ld1lional growth cl bond cracks.

!! plus s Gw growth ot malm cracks
"'
"'
~ ~ r--+----------------------
V'I
Slow grnwth cl Ilona cracks
Pre·exist ng bond " acks

extend on ly slight ly under toaa
0 ~-----------------------.-
STRAIN
Fig. 2.8: Stress-strain curve for concrete divided into four regions, Ziegeldorf (1983).
accordance with the observation made on the micro-level. These initial cracks are properly
due to swelling, shrinkage and bleeding since swelling yields tangential cracks while cracks
due to creep are radial, Ziegeldorf (1983).
For loads below app. 30-50% of the ultimate load the stress-strain curve is approximately
linear and there is no growth of the initial cracks. Due to crack intensities and the difference
in the elasticity modulus between the mortar and the aggregates bonding cracks start to grow
making the curve increasingly non-linear. Beyond 50% of the ultimate load cracks start to
form in the mortar running between the aggregates parallel with the load direction (indicating
tensile cracks). These cracks grow quasi-statically and this type of crack growth is often
termed slow crack arowth. At 75% of the ultimate load a more complex cracking system
starts to develop and the cracks in the mortar coalesce with the bonding cracks and finally
faiture occur.
In HSC the same mechanisms can be observed, Smadi and Slate (1989). lt can, however, be
concluded:
The amount of cracking is significantly less in HSC than in NSC.

The interfacial cracking is insignificant below 60-70% of the ultimate load for HSC.
Mortar cracking is negligible below 90% of the ultimate load.
By the observations made for HSC it can be concluded that the strengthening of the transition
zone limits the amount of bond cracking and the more homogeneous high-strength cement
paste is more brittle.
17
2.1.3 The Fradure Process in Tension.
The tensile strengtit of concrete is a produet of the above observations and the fracture
process wi1l thus be dependent on the strengthof each link in the cracking process (cracking
of the paste, cracking of the aggregates and debonding). In normal strength concrete the
cracking process wi1l start as debonding or as crack growth of existing micro cracks in the
cement paste at approximately 80% of the peak load, then newcracks wi1l form and some
will stop due to stress redistributions and due to crack arrest (cracks stopped by stronger
aggregates). These cracks are evenly distributed throughout the specimen. When the peak
load is reached a large amount of cracking willlocallze in a narrow zone and eventually the
crack that will split the specimen into two will form in that zone. This zone will probably
develop in the weakest part of the specimen.
In high strength concrete the weakest part can both be in the aggregates and in the cement
paste. lf the cement paste and the transition zone are the strongest then crack initiation wi1l
probably not start as debonding but instead the crack growth wi1l initiate in the aggregates.
After the peak load is reached all cracking will be limited to the zone. By using laser
holographic Ansari (1987) confirmed these observations and he further observed that the
dispiacement profile in the narrow zone was varyingin a random way.
In the following the fracture process in tension will be deseribed on the macro-level by using
a concret.e rod as an example. Consider a concrete rod with the length, L subjected to a
deformation controlled tensile load, P, see Fig. 2.9: . The strain is measured by the strain
gauges A, B, C and D. The strain gauges A,B and C having the length L/3 and D has the
Unitormly distribu ted Localized zone

cracking with cracking
P,llL P,llL
A c
~L ~L
1---~)X
Fig. 2.9: Concrete rod subjected to dispiacement controlled loading.
same length as the rod. It is now assumed that the crack that eventually will split the rod into
two will develop under strain gauge B. Thc stress-strain curves for the four strain gauge are
18
2. Fracture md fracture models of coocrete
shown in Fig. 2.10:. The ascending brancb of tbe four stress-strain curves are identical
q
Fig. 2.10: Stress strain curves for tbe strain gauges A,B,C and D.
whereas tbe descending brancb only is identical for gauges A and C. This will be explained
in tbe following.
The elongation of tbe rod will in general consist of three contributions: a linear elastic part
fully deseribed by tbe elastic strain, Ee, a non-linear part whicb is due to uniformly
distributed cracldng deseribed by tbe non-linear strain, E11 , and a part which is due to
increased cracking in tbe localized zone, w. The total elongation, M...., can tben be written
(2.1)
The strain gauge A and C will only measure the two first contributions in eq. (2.1). Strain
gauge B and D will measure all three contribution, but since strain gauge D is three times
longer than B the last term will be divided by L for gauge D and only L/3 for gauges B. Thus,
tbe strain measureel by a strain gauge will be dependent on tbe lengtb of tbe gauge and of tbe
position on the rod. Eacb term in eq. (2.1) is shown in Fig. 2.11:.
During cracking energy will dissipate into tbe specimen. The dissipated energy can be
categorized according to eq. (2.1). A certain amount of energy will dissipate more or less
uniformly through tbe whole specimen and is deseribed by tbe second term in eq. (2.1) the rest
will be limited to tbe narrow zone. The total amount of dissipated energy can tben be written
19
J.P. Ulflgær. Fracture mecbanics of concrete
L fpc.t; L Erea
Edisp = I J Ji(e,JAde,.dx+ J J h.(e,JAde,.dx

a) u o o
we
o Epc.t;
(2.2)
+ J I Jj(w,x)tdwdx
W lOM 0
b) q
u=Sl,(e) e_<e".-
ITu
lt..(e) E-~E".ø
q
c)
d) q
u. E.u.=A,L+ G,
.dL
Fig. 2.11: The total elongation of a rod can be divided into three parts. a) linear elastic part
b) non-linearin-elastic part and c) localized part. d) the total elongation.
20
2. Fracturc and fracturc models of concrete
where ,A, is the cross sectional area.Thus, the dissipated energy is dependent on the length of
the rod, and of the three materlal functions/JfJ,/lJ andJ3f) defined in Fig. 2.11:.
An accurate model which describes the fracture behavior of concrete should, therefore, inelude
description of all three contributions in eq. (2.1).
2.2 Linear Elastic Fracture Mechanics.
The basis of linear elastic mechanics originated with the pioneer work by Griftith (1921).
Griffith considered glass rods and suggested that fracture in brittie materiais arises from small
inhomogeneities such as flaws and cracks. These inhomogeneities yield stress concentrations
as given by lnglis (1913) and he introduced a parameter, the energy release rate, which is a
state parameter related to crack extension. Until a series of catastrophic failures occurred (e.g.
the liberty ships, fatigue in aero planes) researchers did not pay much attention to the theory
developed by Griffith. The theory was further developed by: Irwin (1957) who introduced the
stress intensity factor, Rice (1968) who detined the J-integral, Rutehinson (1968) who took
yielding in front of the crack tip into consideration and Freudenthal (1968) who developed
probabolistic fracture mechanics. In the foliowing the basic ideas of LEFM will be given.
2.2.1 Basic LEFM.
The theories presented in this chapter is basic linear elastic fracture mechanics and can be
found in one of the many text books on this topic e.g. Hellan (1984), Ewalds and Wanhill
(1984), Alibadi and Rooke (1991), Gansted and Sørensen (1991).
21
J.P. Ulfkjer. Fncturc mechanicø of concrete
F, d
Fig. 2.12:. Arbitrary specimen with an initial crack, energy approach.
Consider a plane arbitrary elastic specimen with the thickness, t, and an initial noteh of length,
a, which is subjected to a set of boundary conditions and loaded remotely from the crack tip
with an arbitrary load, F, see Fig. 2.12:. The total potential energy, n, in the system is then
given by
(2.3)
where, ne, is the elastic energy content in the body, nF, is the potential of the extemal forces
(body and surface) and, nx, is the total kinetic energy in the system. It is now possible to
introduce a state parameter termed the energy reJease rate, G, defined as
G=--
an (2.4)
taa
The energy release rate is a pure continuum mechanical parameter and is not necessary related
to crack growth.
However, if crack growth is considered then eq. (2.3) changes to

n ,., ne+nF+Dx•llc (2.5)
where, llc, is the fracture potential, that is the energy that dissipates during crack growth. By
assuming that crack growth is only dependent on one parameter, a, the equilibrium condition
becomes (the variation of the total potential shall equal zero)
an = o (2.6)
taa
The fracture criteria can then be written as
ane (2.7)
G = - =R
taa
22
Where, R, is the resistance towards crack growtb. In general R is a material, geometry and
crack length dependent parameter. In LEFM, R, is assumed to beamaterlal conslant termed
the crack driving force, Gc. The fracture criteria is then written as
(2.8)
The crack growth is stable if the foliowing stability condition is fulfilled (the second variation
of the total potential shall be positive)
aG > aR = 0 (2 •9)
aa aa
For most struerures the energy release rate, G, wi11 be an increasing function as the crack
. aG
growth , 1.e- > 0 .
aa
The above outlined theory, referred to as the energy approach is, however, not adequate as
a design tool and instead Irwin (1957) developed the stress intensity factor concept, which is
based on stresses rather than on energy considerations.
In general fracture can take place in one of three modes or a combination of these. These
modes are opening mode, shear mode and torsion mode. Here only opening mode will be
considered.
F,8
Fig. 2.13:. Arbitrary specimen with an initial crack, stress intensity approach.
Consider the specimen in Fig. 2.13:, according to linear elasticity theory the stress distribution
close to the crack tip is deseribed by
uiJ(8,r) = K /;i(8) + higher order terms (2.10)

J2rr
where, uij, is the stress tensor, 8 and r are the polar coordinates, fiJ, contains trigonometric
23
J.P. Ulfijær. Fracture mecbanics of concrete
functions and, K, is the stress intensity factor. As the coordinate, r, approaches zero the first
term approaches infinity and the extra terms are constant or tends to zero. Consequently the
first term is dominating in the vicinity of the crack tip. Thus, even for infinitely smallloads
the stresses next to the crack tip will approach zero making a critical stress as failure condition
meaningless. lrwin (1957) considered the elastic work necessary to close the crack and derived
a relationship between the stress intensity factor and the energy reJease rate
(2.11)
K= JGE
The relationship retates the crack driving force to a critical stress intensity factor, K c. Tbe
fracture eriterion can thereby be written as
K=Kc (2.12)
The critical stress intensity factor is usually termed the frøeture toughness. The designing
engineer is therefore only limited to determine the stress intensity factor and compares it to
the critical value in order to establish if the crack growth. The stress intensity factor is often
written as
K = uJra g(geo) (2.13)

where, g(geo), is a dimensionless function of the geometry, which can be determined analytical
or numerical e.g. the boundary element method, Alibadi and Rocke (1991).
In order to ascertain if LEFM is applicable for a given materlal it is then necessary to perform
experiments and see if, Kc, is a materlal parameter. This will only be the case for elastic
perfectly brittie materials.
2.2.2 LEFM and Concrete.
Many researchers have studied the applicability of LEFM to concrete and only some of the
most important results will be given here, however, reference should be made to Mindes s
{1983) and Mindess (1986) where an annotated bibliography from 1928-1986 on fracture and
cracking of concrete is presented with 595 abstracts.
The first to use LEFM on concrete was Kaplan {1964). He performed three and four point
bending experiments on notehed beams at different size scales. He found that K c is a geometry
and size dependent parameter. He attributed this 'size effect' to slow crack growth and to
shear.
Walsh (1976) and Higgins and Bailey (1976) made experiments with notehed beams of
concrete and bardened cement paste respectively and they observed that Kc increases with the
beam depth. They further concluded that the fracture toughness tends to a constant limiting
value as the beam size increases. The same trend was observed by Modeer (1979) for concrete
and he suggesled that in order to measure a true materlal parameter the specimen size should
belargerthan
24
d >lO [~:r (2.14)
where, d, is the beam depth. The condition yields that a beam of NSC should have a beam
depth of more than2overc 2-3 m.
Since HSC is considered a more brittie materlal than NSC, Biolzi and Tognon (1987) and
Tognon and Cangiano (1989) made three-point bending experiments with NSC and HSC
(Compressive strength from 50 Mpa to 170 Mpa), and they concluded that Kc is increasing
with the beam size and with the strength of the concrete. For increasing notch length, Kc
increased to a peak value (at notch depth ratio of approximately 0.4) and then decreased.
The reason for the requirement in eq. (2.14) and discrepancies is that a process zone of micro
cracking, debonding, crack arrest etc. wi11 develop in front of the crack tip as deseribed in
section 2.1.3. If LEFM should be applicable this zone, the fracture process zone, should be
small compared to the overall beam dimensions. The relative size of thc process zone is
dependent on the size and geometry of the considered specimen and of the materlal used. It
can therefore bc concluded that LEFM is only applicable for large brittie concrete struerures
such as solid concrete dams or anchor blocs for large span bridges, Planas and Eices (1989).
2S
J.P. Ulfkjær. Fractute mecbanics of concrete
2.3 Nonlinear Elastic Fracture Mechanics.
Since LEFM is not applicable for labaratory sized specimens several non-linear fracture
mechanical models have been developed. The five most accepted models arebriefly presented.
2.3.1 1be Fietitlous Crack Model.
The Fietitlous Crack Model (FCM) was invented by Hillerborg, Modær and Petersson (1978)
and was fonnulated by Hillerborg (1977) and Petersson (1980), on the basis of the
observations made of a dispiacement controlled tensile tests as deseribed in section 2.1.3.
a a
a HHIH
a
I
y
E c
L+ IiL w B
A
a 0010 E w
a) b)
Fig. 2.14: The fictitious crack model. a) simplified rod b) the materlal function.
Consider the rod in Fig. 2.14: subjected to dispiacement controlled tensile loading. The rod
is divided into three regions A, B and C. Part A and C are assumed to be linear elastic and
completely deseribed by the elasticity modulus, E, and Poisons ratio, v. Before the tensile
strength is reached part B is assumed to be infinitely small. When the first principal stress
reaches the tensile strength, u,_, a crack is assumed to develop in part B. The crack is a so-
called cohesive crack, Barenolatt (1962), which is able to transfer stresses according to a
materlal function, f('), only dependent on the distance between the two uncracked parts. The
distance between the two uncracked parts is tenned the crack opening displacement, w. As
the crack opening dispiacement increases the stress wi11 gradually fall and at some critical
opening, wC' the crack is not able of transferring stress and the specimen wi1l split into two.
Since the crack is able to transfer stresses and therefore not a real crack the crack is phrased
a ractitious crack.
For the FCM eq. (2.1) becomes

u
llL = -L+w (2.15)
E
26
2. Practwc mel fracture models of concrete
The dissipated energy can thereby according to eq. (2.2) be written as

wc
Edisp = Jf(w)dw (2.16)
o
Since f(j is a materlal lunetion the area under the fimetion must be a materlal parameter, here
termed thc fracture energy, GF, Hillerborg (1983).
Thc FCM is not well suited for analytical solutions and it is only in simple cases possible to
derivc analytical results. Instead the FCM is used together with numerlcal methods like the
finitc element method or the boundary element method, Hillerborg (1976), Petersson (1981),
Caipinteri (1989) Brincker and Dahl (1989), Harder (1990).
The greatest problem with the PC-model is that it is extremely difficult to perform a stable
dispiacement controlled uniaxial tensile test of a concrete rod and hitherto experiments have
only been made for normal strength concrete.
The first to performed a stable tensile test on NSC, in a very stiff testing machine, where the
descending branch is measured was Evans and Marathe (1968). The methods for obtaining the
descending branch has become more camplex and Gopolaratnam and Shah (1984) performed
a large test series in a servo-controlled testing scheme. The most important work seems to
have bcen performed at Delft University of Technology, where many experiments have been
performed during the last decade, Reinhardt (1985), Wolinski et al. (1987). Among the
problems which has to be overeorne are:
The axial and rotational stiffness of the testing equipment.

The nucleation of crack growth from one side of the specimen.
The preparation of the specimens, in order to minimize the above
problem.
Measuting of strains and dispiacement on the specimens.
If the above mentioned problems is not dealt with the test wi1l either be unstable or an
incorrect stress dispiacement relation is measured, Hillerborg (1989).
27
J.P. Ulfkjer. Fraelure mecbanics of concrete
2.3.2 Tbe Crack Band Model.
The Crack Band Model was invented by BaZant and Oh {1983) and Bafant (1984), and is
similar to the FCM. Instead of assuming that all the deformation in the process zone localize
into a zone of zero width as in the FCM, BaZant assumes that the deformation willlocalize
into a zone with a width, hc, which is assumed to beamaterlal parameter (experimental fit
seems to yield hc= Jdmax where dmax is the maximuro size of the aggregates). The softening
ao
l1'1'1't11 øo , åL
Crack a
Band u
_.....v w
L ~
lx e•
åL
-L
a)
møo,åL b)
ep
Fig. 2.15: The crack band model a) the simplified rod b) the constitutive relation for the band.
relation can then be formulated as a stress-strain relation instead of a stress -crack opening
dispiacement relation, see Fig. 2.15:.
This leads to some advantages: a) the model is well suited for FEM calculations, b) triaxial
effects are easily taken into consideration, c) some analytical solution can be obtained the most
well-known is the Size Effect Law, Bazant (1984). The disadvantages are that an extra materlal
parameter is necessary (the width of the process zone).
The dissipated energy can be written as

hc Ec hc
Edisp = J Ju(e)btdedx = Jbtg1dx = GF (2.17)
o o o
where g1 is the volumetric fracture energy, Elices and Llorca (1989).
The 'size effect law' is usually written as (only the first term of a Taylor series expansion is
considered)
28
f1 Bo
-:z-=---
~
(2.18)
where qfq11 is the normalized failure load, f111 is the uniaxial tensile strength, d" is the
maximum aggregate size and B0 and )\0 are empirical constants usually determined by linear
regression from experimental data, Bal.ant (1983). The size effect law will be considered
further in section 2.4
2.3.3 The Two Parameter Model.
The two parameter model was proposed by Jenq and Shah (1986) and is a non-linear model
which assumes a crack tip singularity in front of the real crack. The model belengs to a family
of models donated cohesive crack models with crack tip singularity, Elfgren (1989).
In this model it is assumed that the peak load is reached when the stress intensity factor of an
effective crack (that is the initial crack plus the cohesive crack) and the crack tip opening
dispiacement at the initial crack tip reaches two critical values, KJ/ and CI'OD c· The two state
parameters Kl and CI'OD are calculated according to LEFM. 1t is, however, not possible to
measure the CI'OD during experiments and instead the crack mouth opening displacement,
CMOD, is measured. By using the inelastic CMOD at peak load the effective crack length can
be found and by using this crack length the two critical state parameters can be obtained.
Several investigations have been made in order to check the applicability of the two parameter
model, Karihaloo and Nallathambi (1991), where it is concluded that the fracture toughness
is essentially independent of specimen depth and that the variation between 0.93 and 1.53
MPa/iil is mainly due to variations in the mix properties. They did not conetude anything
about CI'ODc , but it is observed that the variation of the CI'ODc is much strenger and that
it is not possible to detect a trend in the observed values, however, Briihwiler et al. concluded
from wedge splitting tests that KJ/ is size independent but CI'ODc displays a significant size
effect (CIOODc increases with specimen size). This phenomenon can be explained as follows.
If the considered specimen outside the crack is assumed to be elastic then the crack tip opening
dispiacement can be written according to the superposition principle
CI'OD - CI'ODF+CI'OD CO (2.19)
where CI'ODF is the crack tip opening dispiacement due to the applied load, F, and CI'OD co
is the crack opening dispiacement due to the cohesive stresses. In the two parameter model the
last term is not taken into consideration, Cotterell and Mai (1987) and it is not always a good
approximation to negleet this component. There is also no evidence for the faet that the crack
tip opening dispiacement should takc on its critical value at peak load. Actually more detailed
analysis using the FC model and the boundary element method shows that the CI'OD talces on
its critical value after the peak load, Ulfkjær, Krenk and Brincker (1991).
lt is therefore concluded that the two parameter model is only a one parameter model where
an effective crack length is used as controiling parameter.
29
J.P. Ulfkjer. Fncture mecbanica of concrete
2.3.4 The Effective Crack Model.
As in the two Parameter model the effective crack model assumes a sharp crack in front of the
real crack as suggested by Nallathambi and Karihaloo (1986). In the effective crack model it
is assumed that slow crack growth occurs prior to the peak load and an effective crack length
which takes the slow crack and the reduced stiffness into consideration is introduced. The
effective crack length is calculated by using the peak load and the deflection at peak load. The
effective critical stress intensity factor, K 1/ is then calculated by using the effective crack
length and the peak load, Nallathambi and Karihaloo (1991).
The effective crack model and the two parameter model are thus very similar, it is, however,
considerably easier to calculate the effective critical stress intensity factor as deseribed in
Nallathambi and Karihaloo (1991).
2.3.5 R-Curves.
Instead of assuming that the resistance, R, is a materlal property a more general theory which
assumes that R is a function of the crack length has been developed. The fraelure eriterion is
then written as
G= R(a) (2.20)
and the stability condition beoornes
aG < aR(a) (2.21)
aa aa
The basic assumption is then that R(a) is a materlal function independent of size and geometry.
G, R ,G(F))
,,
R(a)
G(~)
, ,
, ,,
----.,.--,
, ,
, ,,
~
,, , , ,
..
~: ", _.
ao a
Fig. 2.16: R-curve for a specific materlal and initial crack length.
A correlation between the FC-model and the R-cutve concept can be made. Consider an
30
arbitrary specimen without an initial notch. According to the FC-model a fictitious crack will
develop when the first principal stress reaches the tensile strength. By increasing the loading
the fictitious crack wil1 grow and eventually the real crack wil1 start to grow, see Fig. 2.17:.
If the R-curve concept should be applied it is necessary to introduce an effective crack length,
aeft generally defined as, Hellan (1984)
Fig. 2.17: Fietitlous crack with an effective crack length.
aef • a+aaf a E ]0; l] (2.22)

the dissipated energy during crack growth can then be written as
l
-(a-aer>
a
Edisp = GFt(aeraaf)+ J rp(x,aef)dx
(2.23)
o
where rp{x,ae} is the energy density in the fictitious crack, Brincker (1991), determined by
~.aet>
rp(x,aef) • J u(u)du (2.24)
o
where w(x,aJ is the crack opening profile which is dependent on the size and the geometry
of the consioered structure. According to eq. (2.7) the resistance towards crack growth
becomes
31
J.P. UJDgær. Fraelure mechanics of concrete
aEdisp aa1 1 1
R = • GFt(l-a-)+cp(-(aq-a),aq)-
aaef aaef a a
(2.25)
.!cae~-a)
a J acp(x,aet> dx
aaef
o
since the second term is always zero eq. (2.25) becomes
l
-(ae~-a)
R ., ..!_GF+ a J acp(x,aJ dx (2.26)

a aaef
o
Thus, a complicated relation exists between the FC-model and the R-curve concept revealing
the fracture resistance is dependent on the size and geometry of the considered structure. It is
also obvious from eq. (2.26) that the R-curve is strongly dependent on how the parameter, a,
is chosen. It was shown by Brincker (1990) that if the crack profile and the stress-crack
opening dispiacement are chosen to belinearthen the R-curve will be a parable.
Similar considerations to theabove have also been made by Mai (1991) but his approach seems
to fail since he does not take crack growth and the fictitious crack length into consideration.
Several researchers have made experiments in order to detect if concrete has an R-curve
behavior, Wecharatana and Shah (1982), Wecharatana and Shah (1983), Bal.ant and Cedolin
(1984) and Ba7.ant et al. (1986).
Wecharatana and Shah (1982) concluded that the R-curves seem to depend on the size and
geometry of the concrete specimens considered and attributed this effect to the large process
zone. BaZant concluded the same by using a linear crack band model. It seems, however, that
by introducing an equivalent crack length this dependency is not very strong and he concluded
that if a rough estimation of the load carrying capacity is needed R-curves seem to be suitable.
2.4 Brittleness and Size Effect.
Size effect is a term which is used more or less stringent by different researchers. The
common interpretation of the word is that an assumed materlal parameter is not constant but
is varying with the size of the structure. The author feels, however, that the word is rather
misleading, since the word does not describe the problem which in reality is that for some
reason the model used is not accurate enough. Therefore size effects are always related to
models. Thus, an observed size effect in one model can completely disappear in another more
accurate model. The problem will be further illustrated by considering the three-point bending
geometry.
32
The usual assumption is that the beam is a Bernoulli beam with the assumption that the tensile
strength (or modulus of rupture) is a materlal property. This is, however, a poor model since
experiments show that the bending tensile strength is decreasing as the size of the beam
increases, Reagel and Willis (1931).
The results of a large test series performed in USA, by the Arnerlean Association of State
Highway Officials, Reagal and Willis (1931), are here briefly reported for the dimensions
which are similar to those used in the present investigation, see chapter 5.
The experiments were performed at four different laboratorles in USA, and consisted of 64
different beam geometries, and a total of 768 beams of plain NSC were tested.
6
t • 203.2 mm
!5.9
t • 254.0 mm
t • 152.4 mm
!5.2 t • 52.12 mm
5.1 100 120 140 160 180 200 220 240 260
beam depth. d
Fig. 2.18: Modulus of rupturefor varying values of the beam depth, d, and thickness,
t, and span equal to 762 mm, Willis andReagel (1931).
The modulus of rupture, um, was calculated from

l(Fmax + 3/4W)
C1 = ----=-- (2.27)
m td2
where, W, is the weight of the beam and , FmtDC is the peak load. The factor 3/4 on the beam
weight is a mystery to the author, who expected it to be 1/2. The factor 3/4 wil1 yield a
smaller size effect than the factor 112. This discrepancy wil1 in the foliowing be ignored. The
experiments performed were under high control and the coefficent of variation of the individual
33
J.P. Ultlcjær. Fncture mechanics of concrete
test varled between 2% and 5.% for the considered beam dimension, and the variation between
each laboratory was between 2% and 7%.
The test results for the beam with the length 762 mm are presented in Fig. 2.18:. It is seen
that the modulus of rupture is decreasing with the beam depth for all beam thicknesses, and
the decreasing by doubling the beam depth is 9% for the thielmesses 101.6 mm and 6% for
the other thicknesses. lt is also seen that the modulus of rupture is almost independent of the
beam thickness.
This suggests that the Weibull size effect importance or that another cause for size effect is
interacting when the beam thickness is increased, e. g. the crack profile is changing.
This size effect is known for many brittie materiais and is usually explained by the Weibull
effect, Weibull (1939) and Weibull (1951). The Weibull effect is based on the weakest link
concept and on the assumption that the materlal is perfectly brittie. The mean value of the
strength, R, can be expressed for homogeneous stress distributions as:
EIRJ=x0 +x_r [l+!) [:.)-j (2.28)
where x0 is the lower strengtit limit, V is the volume of the structure V0 is a reference volume,
and X0 and k are positive constants. For concrete Zech and Wittman (1977) found that k= 12.
For the considered beam (2.28) beoornes for x0 =0:
E[R]=-x~ [1+.!.)
k
[ vV 2(k+1)
0
l 2 l-ll
(2.29)
It is seen that it is only the volume and not the geometry of the beam which determine the size
effect.
By using linear elastic fracture mechanics, which is a better model than the Bernoulli model,
it is possible to predict a strong size effect of the form:
ae·-
112 d
Kc~
-g(geo)
Ta
(2.30)
wherc ae is the faiture stress, K0 is the fracture toughness, a is the normalized crack length,
and d is a characteristic dimension of the structure e.g. the beam height. Yet, numerous
experiments have shown that the fracture toughness is not conslant when the size of the
structure is changed, still a size effect, and LEFM should therefore be used with care on
concrete.
In 1977 Hillerborg and co-workers showed that concrete is not a perfectly brittie materlaL This
'non-brittle' behaviour is due to the existence of a cohesive zone in the front of the crack tip,
and can bedeseribed with a model called the fictitious crack model (PC-model). The PC-
model gives a better description of the physical reality of crack growth, and by using the PC-
model the modulus of rupture is size dependcnt.
34
2. Fraelure and fraelure models of concrete
One of the basic assumptions regarding the FC-model and the crack band model is that the
fracture energy Gp is a materlal property. However, the fracture energy measured using the
RILEM method displays a significant increase with beam size (the third type of size effect),
Hillerborg (1985). This size effect is usually attributed to the faet that in the FC-model the
energy dissipation in the bulk region has not been taken into consideration, Elices and Pianas
(1988).
By using dye technic Swartz (1989) showed that the crack profile is not a straight line, but is
rather U-shaped. This could indicate that the strength of the concrete is lower near the surface
of the specimen. If the size of this weak zone is independent of the specimen size then another
cause for size effects is found.
In the foliowing a few examples will be given which further illustrates the terms tbrittleness'
and tsize effectst.
2.4.1 Example I. Size effect or plate made or a perfectly brittle/perfectly duetile material.
Consider a larae plate o f width 2b with an initial crack o f length 2a. The plate is subjected to the
load ø0 at the edges, see Fig. 2.19:. The plate is made of a materlal with the yield strength, O'u•
L
- 2a
Fig. 2.19: Plate with center crack.
and with the fraelure toughness, Kc,. The purpose of this example is to investigate wbich of the
two faiture possibilities will occur depending on the Bia and geometry of the beam.
Yielding will occur in the two parts next to the crack when
CJo a
(2.31)
- •1--
Ø'y d
If tbe width of tbe beam is assumed to be much larger than tbe initial crack length, tben the stress
intemrity factor can be wriUen as
35
J.P. Ulflcjer. Fracture mecb•nics of concrete
K•ø0 r;;; (2.32)
Crack propagation occura when K•Ke, viz
øo Kc l l
ø., • ø., -r.:=:= {ti (2.33)
JTd
When similar piates are considered (e. g. ølb•kst., Kc""'krt. and ø1 -kst) at different size scales
u shown in Fig. 2.20:., then it is seea tbat the yielding eriterion eq.(2.32) is independent of tbe
2
~ 1.8 2
Klc • 1.5 N/mJ/
o 1.6 2
au • 6.0N/mm
b"
1.4 a/d -0.01
1 1.2 dtr • 2.03 m
i
~
l
i
~ 0.4
0.8
0.6
Yielding
0.2
o
o l 2 3 4 s 6 7 8 9
plate size, d [m]
Fig. 2.20: The yielding criteria, the fracture criteria and the siz.e effect law for the considered
plate.
siz.e scale, d, whereas the crack propagation eriterion eq. (2.33) decreases npidly with the siz.e
øcalo. Thua, for small sizea tbe plate will fail due to yielding and for large piates it wiU fait due
to cøck propagation. The tnnsition, 4, between tbe two types o f faiture is determined by setting
eq. (2.32) equal to eq. (2.33)
4•[Kø., ]l T~ t-i
c l 1 (2.34)
( ]
lf tbe teasilo atrength is determined with such a plate then the tensile strength wiU be dependent
OD tbe size of tbe plate if the size of the beam is larger than 4 and tho tensile strength is
calculated according to the yielding criterion. On the other band if fractute toughness is deter-
mined for plate siz.es smaller than 4 tben Ke is also dependent on the beam size. These
36
2. F neture and fracture models of concrete
phenomenoos are often termed 'size effect' since an assumed materlal parameter is dependent on
the size of the specimen. Since concrete is not a perfectly brittie materlal it is necessary to use
buge specimens if the fracture toughness should be determined as a materlal parameter.
Consider now the size effect )aw in cq. (2. 18). AB the size, d, approaches zero the denaminator
approaches one and the normalind faiture load approaches, B. As the size approach infinity the
one in the square root becomes insipificant and the normalind faiture becomes a function of
b -l/1, Thus, by makin& proper adjustment of the two constants B0 and ~ the size effect law
describes the two faiture possibilities as Iimitina cases. This is done by setting
B • 1-~ (2.35)
d
and
{2.36)
whereby the sizo effect law becomes
..!,.Rd
fly
t-!!
l+.!!.
(2.37)
d!r
Yielding
-1
10 1~o~~~~~1~o~o~~~~~1~o~1~~~~~1~o2
Plate Size [m]
Fig. 2.21: Load carrying capacity of plate in log-log scale.
37
1.P. Ulfkjær. Fraelure mecbanics of concrete
The size effect law together with tbe yielding eriterion and the fraelure eriterion are shown in
Fig. 2.20:. Since tbe crack propagation eriterion in loglog scale becomes a straight line with a
slope of O.S it is common to display the faiture criteria in loglog scale, see Fig. 2.21:, Leicester
(1973), which makes it obvious that tbe Size Effect Law describes a gradual transition from the
yielding eriteria IO the crack propagation criteria. lt should, however, be empbasized tbat the
Craclc Band Model and tbe Fictitious Craclc model would not predict any size effect in this
example and the Size Effect Law is tberefore only used as mathematical expression which malces
the above mentioned transition.
In the example with a plate it was seen that the transition size of the plate, 4, plays a
dominant role. In general plastic, elastic or visco elastic faiture criteria can be written as a
dimensionless function of the geometry
.!!.... = gl(geo) (2.38)

fly
and the faiture eriterion for crack propagation, eq (2.8), can be rewritten as
u ... Kc g(geo) l
a, J": {ti (2.39)
whereby the transition size becomes

2
d". = [ Kc] _1 g(geo) ]2 (2.40)
u., a '1"-
[ gl(geo)
d
and the Size Effect Law can be written as
(f gl(geo)
=
R
(2.41)
11tus, for small ratios of {J1 =dldtr, the behavior of the structure is ductile, for large {J 1 the
behavior becomes more brittle, mak.ing {31 a brittleness number. The most commonly used
brittleness number is defined as the ratio between the elastic energy and the fracrure energy
B = elastic energy E _
L 2 L3 u,
2
u, (2.42)
fracrure energy - L 2GF - E Gp
where L is the characteristic dimension of the struerure and uu2/E is the stored elastic energy
at faiture for a unit volume and Gp is the fracrure energy, Elfgren (1989). If the relation
J
K c ,., EGp is used i t is seen that B is a special case of {J 1 where the geometry o f the struerure
is not taken into consideration, but only the size deseribed by, L.
As it is seen from the previous examples the term brittleness is a word which is easy to
38
2. Fracturo and frøeturo models of concrete
misinterpret. 1t has been common to understand brittleness as related to a materlal property,

or to the shape of the load-dispiacement curve, and as was seen, is brittleness affected by the
size and geometry of the considered structure, which again was reflected in the two introduced
brittleness numbers. In the foliowing brittleness will be considered on the basis of a concrete
rod. A material brittleness is introduced and structural brittleness based on observations of the
descending branch of the load dispiacement curve is defined, finally a third kind of size effect
is introduced.
2.4.2 Example II. Brittlenessand stability or concrete rod.
Durlog crack growth energy will dissipate as previously described. The dissipated energy can be
divided into two contributions
(2.43)
whero, Gv, is thc volumc diasipation and G1 is thc surface dissipation durlog crack: growth. A
materlal brittleness can then be defiDed as, Brinck:er (1990),
(2.44)
Structural brittleness is defined by considering the load dispiacement curve of a rod made of a
materlal which bebaves according to the fictitious crack: model. Consider an initial uncracked
concrete rod with the length, L, and tbe cross section, A, subjected to a dispiacement controlled
load, F, and the corresponding displacement, u. The elastic properties are deseribed by the
elasticity modulus, E, and thc softeniøg relation is deseribed by,f(w), see Fig. 2.14:. The total
potential energy in tbe system is given by
(2.45)
Equilibrium is obtained for
.;g •~ (w-u)+Af(w) (2.46)
And øtability is obtained through thc foliowing condition
a2 n •
awi
clf{w) +AE >O
clw L
8./{w)<AE
- (2.47)
8W T
which correspondø to infinitely slope on the descending branch of the load dispiacement curve.
A brittleness number which ineludes thc slope of the descending branch seems, thus, appropriate.
The load dispiacement for the considered rod is as sketehed in Fig. 2.11:. The slope on the
descending branch is given by
39
J.P. Ulflgær. Fraelure mecbanics of concrete
au l
-,;r •
oo
-:;-a---::a-
"+ w
au au (2.48)
l
which becomes for a linear softening relation
au u~-Ewc
-=-=-- (2.49)
a~ Eau
The advantages with this brittleness number is that it is related to a specific bebavior of the load
dispiacement curve. The disadvantages are that it is difficult to calculate, it is not dimensionless
and it is not defined for all structures.
2.4.3 Example m. Stability and briUieness of glued beam.
v- u+ CF
Fig. 2.22: Beam glued to an infinitely rigid foundation, subjected to deformation controlled load.
The purpose of this example is to illustrate how the stiffness of thc testing machine affects the
stability of the fraelure process in a dispiacement controlled experiment. Thc example Curther
illustrates the necessity in measurlog the true dispiacement of the structure.
Consider a beam glued to an infinitely rigid support as shown in Fig. 2.22:. An initial crack with
the leogth, ø, much larger than the beam depth, b, is assumed to exist. The beam is subjected to
a dispiacement controlled load, v. Between the load point and the beam a spring with the
compliance, C, is inserted, modelling the flexibility of the testing machine. During testing the
opening of the crack, u, is measured. The purpose of this example is to investigate how the
flexibility of the testing machine influences the stability of the test.
Since the test is deformation controlled the potential of the load is zero and the total potential of
the system becomes
1
n - .!.uF+.!.CF
l l
(2.50)
40
2. Fracture and frac:ture models of concrete
lf tbe beam is assumed to be a Bernoulli beam the dispiacement becomes
__
.,. 1Fa3
3 El
(2.51)
and the energy release rate becomes
an 1 a 1F 1 (2.52)
G • - - .. - - -
taa 2 Elt
Tbe craclc wi11 extend wben G•Gc and tbe corresponding criticalload becomes
J2GctEl (2.53)
Fc • ..:.---
a
tbe comsponding dispiacement is
.. [~i·+· (2.54)
l f tbe fracture proces& shall be stable avlaa >o yielding the foliowing condition to tbe compliance
of the teating machiDe
C<.,! a3 (2.55)
3El
A set of load dispiacement curveø for different complianceø is sbown in Fig. 2.23: where tbe
controlling parameter is u (virtual Cl'ICk propagation). It is seen tbat even in dispiacement
controlled testing the fracture procesa becomes unatabte for too flexible teating macbines. Instead
it is neceøsary to perform experimeotø wbere thc crack opening is tbe controiling parameter, if
the descending branch of tbc load wsplacement curvc is requested. Tbis can only be done in a
closed loop scrvo controlled testing system, Brown and Hudson (1972).
2.5 References for Chapter 2.
Alibadi, M.H. and Rooke, D.P. (1991), Numerical Fracture Mechanics, Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Ansari, Farhad (1987), Stress-Strain Response af Micro cracked Concrete in Direct Tension,
ACI MaJerials Journal, Nav-Dee 1987, pp. 481-490.
Aitcin, P.C. (1990), Effect of Coarse-Aggregate Characteristics on Mechanical Properties of

High-Strength Concrete, ACI MaJerials Journal, V. 87, No. 2, pp. 103-107.
Attiogbe, K.E. and Darwin D. (1987), Submicrocracking in Cement Paste and Mortar, ACI
MaJerials Journal, Nov-Dec 1987, pp. 491-500.
Barenblatt, G.I. {1962), The rnathematical theory of equilibrium cracks in brittie fracture,
41
J.P. Ulfkjer. Fnctwe mecbanics of concrete
6
F [N] 10
14 Kc •1.5 :MN/mm
C-O 0.5Cc Cc 1.5Cc 2Cc E • 20GN/mm
12
a • 500mm
b •100mm
10 t • 100 mm
.cc - 1.5 10 mm/MN
8
oo 0.005 0.01 0.015 0.02 0.025 0.03 0.035

Displacem~ v [mm]
Fia. 2.23: Load dispiacement curves for lhe glued beam when lhe compliance of the spring is
varied.
Advances in Applied Mechanics, 7, pp. 55-129.
BaZant, Z.P. and Oh, B.H. (1983), Crack Band Theory for Fracture of Concrete, Materials
and Structures, Vol. 16, No. 93, pp. 155-177.
Bafant, Z.P. (1984), Size Effect in Blunt Fracture: Concrete, Rock, Metal, Journal of
Engineering Mechanics, ASCE, Vol. 110, pp. 518-535.
Bafant, Z.P. and Cedolin, L. (1984), Approximate Linear Analysis of Concrete Fracture By
R-Curves, Journal of Structural Engineering, ASCE, Vol110, pp. 1336-1355
BaZant, Z,P, Kim J.K. and Pfeiffer, P.A. (1986), Non-linear Fracture Properties from Size
Effect Tests, Journal of Structural Engineering, ASCE,Vol. 112, No. 2, pp. 289-307.
Biolzi, L. and Tognon,G. (1987), On the Fracture Mechanics ofHigh and Very High Strength
Concrete, Il Cemento, 3, pp. 251-259.
Brincker, R. and Dahl, H. (1989), Fietitlous Crack Modelof Concrete Fracture, Magaz.ine of
Concrete Research, 41, No. 147, pp. 79-86.
42
2. Frøeture md fracture models of concrete
Brincker, R. (1990), General approach to fracture &. stability, personal communication.
Brown, E. T. and Hudson, J.A. (1972), Discussions of the paper Griffith Fracture Criterion
and Concrete, by S.P. Shah and F.J. McGarry in Journal ofEngineering Mechanics Proc.
Paper 8597, Journal of Engineering Mechanics, ACSE, 98 (EMS), pp.l310-1313
Briihwiler, E., Broz, J.J. and Saouma, V.E. (1991), Fracture Model Evaluation of Dam
Concrete, Journal of MaJerials in Civil Engineering, ASCE, Vol. 3, No. 4, pp. 235-250.
Carpintcri, A. (1989), Size Effects on Strength, Toughness, and Ductility. Journal of

Engineering Mechanics, ASCE, Vol. 115, No. 7, pp. 1375-1392.
Cotterell, B. and Mai, Y.-W. (1987), Crack Growth Resistance Curves and Size Effect in the
Fracture of Cement Paste, Journal of MaJerials Science, 22, pp. 2734-2738.
Elfgren, L (1989), Fraelure Mechanics of Concrete Struerures-From theory to application,

Clulpman and Hall, pp.1-407.
Elices, M. and Planas, J. (1988), Materlal Models, in Fracture Mechanics of Concrete

Structures: From theory to applications. Edited by L.Elfgren, pp. 16-66.
Evans, R. H. and Marathe, M. S. (1968), Microcracking and Stress-Strain Curves for Concrete
in Tension, Materials and Structures, l. pp. 61-64.
Ewalds, H.L. and Wanhill, R.J.H (1984), Fracture Mechanics, Edward Arnold, London.
FIP/CEB (1990), High Strengtil Concrete- State of the art, FIP/CEB, Bulletin d'information
No. 197, pp 1-61.
Freudenthal, A.M. (1968), Statistical Approach to Brittie Fracture, in Fracture, an Advanced

Treatise, Volume II, pp. 592-619.
Gansted, L. and Sørensen N.B. (1991), Introduetion to Fatigue and Fraelure Mechanics,
Universiry ofAalborg, Aalborg, Denmark.
Griffith, A. A. (1921), The Phenomena of Rupture and Flow in Solids, Royal Society ofLon-
don, Philosophical Transaction. A, Vol 22, pp 163-198.
Gopolaratnam, V.S. and Shah, S.P. (1984), Softening Response of Plain Concrete in Direct
Tension, Technological Institute Nortilwestern University, Evanston, Dlinois, pp. 1-32.
Harder, N.A. (1990), Computer Simulated Crack Propagation in Concrete, in Fracture

Belulvior and Design of MaJerials and Structures, Proceedings of the 8th European
Coriference on Fracture-ECFB (Editor D. Firrao), Engineering Materials Advisory Services
LTD, Vol. ll pp. 706-712.
43
J.P. Ulflger. Fracture mecbanicø of concrete
Higgins,D.D. and Bailey, I.E. (1976), Fracture Measurements on Cement Paste, Journal of
Materials Science, Vol. 11, pp.773-782.
Hillerborg, A.,Modæt', M. and Petersson, P-E. (1976), Analysis of Crack Formation and
Crack Growth in Concrete by means of Fracture Mechanics and Finite Elements, Cement
and Concrete Research, Vol.6, Number 6, pp. 773-782.
Hillerborg, A. (1977), Materialbrott, Avdelingen Ør bygg1Uldsmateriall1Jre, rappon 1VBM

3004, Lund, Sweden, pp. 1-48.
Hillerborg, A. (1983), Analysis of One Single Crack, in Fracture Mechanics of

Concre~,edited by F.H. Wittmann, Elsevier, pp. 223-249.
Hillerborg, A.(1985), Influence of Beam Size on Concrete Fracture Energy Determined

According to a Draft RILEM Recommendation; Report to RILEM TC50-FMC, Rappon
1VBN-3021, pp. 1-15.
Hillerborg (1989), Stability Problems in Fracture Mechanics Testing, in Fracture of Concrete

and Rock- Recent Developments, Editors S.P.Shah, S.E. Swartz and B. Ba", Cardiff,
Wales, Elsevier Applied Science, pp. 369-378.
Hellan, K. (1984), Introduetion to Fracture Mechanics, McGraw-Hill, New York.
Herholdt, A.D., Justesen C.F.P., Nepper-Christensen P. and Nielsen A. (1985), Beton-Bogen,

00 Aalborg Ponland, 2. Udgave, (in Danish).
Hutchinson, I.W. (1968), Plastic Stress and Strain Fields at a Crack Tip, Journal of
Mechanics, Physics and Solids, 16, pp. 337-347.
Hsu, T.T.C., Slate, F.O., Sturman, G.M. and Winter, G. (1962), Microcracking of Plain
Concrete and the Shape of the Stress-Strain Curve, loumal ofthe American Concrete Insti-
tute, Proceedings V. 60, No. 2, pp.209-223.
Inglis, C.E. (1913), Stresses in a Plate due to the Presence of Cracks and Sharps Corners.
Trans. Inst. Naval Architects, pp. 219-241.
Irwin, G.R (1957), Analysis of stresses and Strains Near the End of a Crack Traversing a
Plate, Journal of Applied Mechanics, Vol. 24, pp. 361-364.
Jenq, Y. and Shah, S.P. (1985), Two Parameter Fracture Model for Concrete, Journal of
Kaplan, M.F. (1961), Crack Propagation and the Fracture of Concrete, Journal of the
American Concrete lnstitute, Proceedings V. 58, No. S, pp 591-610.
Karihaloo, B.L. and Nallathambi, P. (1991), Notehed Beam test: Mode I Fracture Toughness,
in Fracture Mechanics Test Methods for Concrete, Edited by S. P. Shah and A. Carpinteri,
44
2. Føcture amd fracture models of concrete
Chapmann and Hall, pp. 1-86.
Leicester, R.H. (1973), Effect of Size on the Strength of Structures, Forest Produets
Laboratory, Division of Building Research Technological Paper No. 71, pp.1-13.
Mai, Y.-W. (1991), Fracture and Fatigue ofNon-Transformable Ceramics: the role of Crack-
Interface Bridging, in Fracture Process in Concrete, Rock and Ceramics Edited by J. G. M.
van Mier et al., Chapmann and Hall, pp. 3-26.
Mehta, P.K. (1986), Concrete: Structure, Properties, and Materials, Prentice-Hall, Ine., 450
pages.
Mehta, P.K. and Aitcin, P.C. (1990) Microstructural Basis of Selection ofMaterials and Mix
Proportions for High-Strength Concrete, in High-Strength Concrete ,Second IntematioTUJl
Symposium, Arnerlean Concrete Institute, pp. 265-286.
Mindess, S. (1983), The Cracking and Fracture of Concrete: An Annotated Bibliography 1928-
1981, in Fracture Mechanics ofConcrete, edited by F.H. Wittmann, Elsevier, pp. 542-671.
Mindess, S. (1986), The Cracking and Fracture of Concrete: An Annotated Bibliography 1982-
1986, in Fracture Toughness and Fracture Energy of Concrete, edited by F. H. Wittmann,
Elsevier, pp. 629-699.
Mindess, S. (1989), Interfacesin Concrete, in Materials Science of Concrete I Edited by Jan

Slallny, The Arnerlean Ceramic Society, pp. 163-190
Nallathambi, P. and Karihaloo, B.L. (1986), Determination of Specimen-Size Independent

Fracture Toughness ofPlain Concrete, Magazine ofConcrete Research, Vol. 38, No. 135,
pp. 67-76.
Regourd, M. (1984), Microstruerure of High Strength Cement Paste Systems, in Very High
Strength Cement-Based Materials editor J. F. Young, Materials Research Society Symposia
Proceedings Vol. 42, pp. 3-18.
Petersson, P.-E. (1981), Crack Growth and Development ofFracture Zones in Plain Concrete
and Similar Materials, Division ofBuilding Materials, Lund Institute ofTechnology, Report
7VBM-1006, Lund sweden, pp. 1-174.
Planas, J. and Elices, M. (1989), Asymptotic Analysis of the Development of a Cohesive

Crack Zone in Mode I Loading for Arbitrary Softening Curves, Proceedings of Fracture
of Concrete and Rock (S.P. Shah and S.E. Swartz, Editors), SEM-RILEM Conference.
Houston p. 384.
Reagel, F. V. and Willis, T.F. (1931), The Effect of the DimensionsofTest Specimens on
the Flexural Strengthof Concrete, Public Roads, Vol. 12, No. 2, pp.37-46.
Reinhardt, H. W. (1985), Crack Softening Zone in Plain Concrete under Static Loading,
45
I.P. Ulfkjær. Fracture mecbanicø of concrete
Cement and Concrete Research, Vol. 15, pp. 42-52.
Rice, J.R. (1968), A Path-lndependent Integral and The Approxirnate Analysis of Strain
Concentration of Notches and Cracks" Journal of Applied Mechanics, 35, pp. 379-386.
Rice, R. (1977) Microstructure Dependence of Mechanical Behavior Ceramics, Trealise on

Materlal Science and Technology, Vol II, Academie Press Ine., pp. 199-381.
Ziegeldorf, S. (1983), Phenornenological Aspects of The Fracture of Conerete, in Fracture

Toughness and Fracture Energy ofConcrete, edited l1y F.H. Wittmann, Elsevier, pp. 31-41.
Sarkar, L.S. and Aitcen, P.C (1987), Cornparative Study of the Microstruerures of Normal
and Very High-Strengtit Concretes, Cement, Concrete and Aggregates, The Arnerlean
Society for Testing and Materials, pp. 57-64.
Scrivener, K.L. {1989), The Mierostrueture of Conerete, in Materials Science ofConcrete l,

Edited l1y Jan Skalny, The Arnerlean Ceramie Society, pp. 127-161.
Skalny, J. and Roberts, R.L. (1987), High-Strengtit Conerete, Ann. Rev. Mater. Sci., 17, pp.
35-56.
Srnadi, M.M. and Slate F.O. (1989), Mierocraking of High and Normal Strength Coneretes
underShortand Long-Term Loadings, ACI Materials Journal, V. 86, No.2, pp. 117-127.
Stroeven, P (1975), The Influence of Volurne Change on tlte Design and Technology of
Modem Building Struetures, Karlovy Vary, CSSR.
Swamrny, R.N. (1986), Properties ofHigh-Strengtlt Conerete, Cement, Concrete and Concrete
and aggregates, The Arnerlean Society for Testing and Materials, pp. 33-41.
Ulfkjær, J.P., Krenk, S. and Brineker, R. (1992), Analytical Model for Fietitlous Crack
Propagation in Concrete Beams, submitted for publication in Journal of Engineering
Mechanics.
Weibull, W. (1939), A Statistical Theory oftlte Strengtit ofMaterials, Royal SwedishAcademy

o/Eng. Sci. Proc., 151, 1-45.
Weibull, W. (1951), A Statistical Distribution Funetion of Wide Applicability, Journal of

Applied Mechanics, 18, 293-297.
Wolinski, S., Hordijk, D.A., Reinhardt, H.W. and Cornelissen (1987), Influenceof Aggregate
Size on fracture Mechanies Parameters of Concrete, The international Journal of Cement,
Composiles and Lightweight Concrete , Vol. 9, No. 2, pp. 95-103.
46
2. Føcture aod fncture models of concrete
3 NUMERICAL METHODS IN CONCRETE

FRACTURE.
It is only in a few simple cases possible to apply the fictitious crack model in an analytical
manner and instead it is necessary to use numerical methods like the finite element method or
the boundary element method. In the foliowing three different methods based on the finite
element method and the boundary element method are presented. The presentation of the
methods is limited to opening mode, to the threc-point bending gcometry and the softening
relation is assumed to be piece by piece linear. Non-linear softening relations are in general
applicable for all the presented methods, but then iterative procedures are necessary, Elfgren
(1989). The two first models have not been implemented by the author and arc only roughly
presented. The last method is deseribed in detail, implemented in computer and a serions
stability problem has been solved, and this method willlater be used extensively.
3.1 The ffillerborg, Modeer and Petersson Methods.
Fig. 2.2 Beamand element mesh used in the calculations by Hillerborg et al. (1976).
The fint method by means of the finite element method and the fictitious crack model was
performed by this method, Hillerborg et al. (1976). The method is briefly outlined in the
47
J.P. Ulfkjer. Fracture mechanicø of concrete
following.
Consider the beamand the element mesh in Fig. 2.2. First the moment which yields the tensile
strengtit in node l is calculated. A new special very thin element with properties according to
the chosen softening relation is now inserted. As the moment increases a fictitious crack will
develop in node l, and the load which corresponds to the tensile strength in node 2 or the
critical opening in node l is calculated. If the critical opening is reached the element at node
l is removed and a stress free surface is achieved. If the tensile strength is reached in node
2 a new spccial element is inserted at node 2 and the previous step is repeated. In this way the
force-dispiacement curve can be calculated to tbe peak moment and a littie further.
The advantage with this method is that it is directly applicable for all geomelries and loading
configuration and capable of predicting multiple cracking, as long as the struerure can be
satisfactory modelled by the finite element method. The method is therefore also valid for
reinforced concrete. The disadvantage of this method is that it is very time consuming and
many different element meshes are needed, as the fictitious crack develops and advances.
Instead Modeer (1979) developed the force method, a more rapid special geometry method,
which later was refined by Petersson (1981) to the sub-structure method.
3.2 The sub-structure method.
The sub-strucrure melbod was introduced by Petersson (1981) and later refined by Carpinteri
(1989), who made the fracture process, fictitious crack length controlled. The method is more
or less a melbod based on the boundary element method, since only loads at the boundary of
the struerure are considered. The method is superior to the melbod developed by Hillerborg
et al. (1976) since the calculation speed is considerably higher, up to a factor 100, Brincker
and Dahl (1989).
Consider an arbitrary struerure subjectcd to the dispiacement controlled load, F, and the
l
l
.l
ou l au
(
\
' r7
J
./
)
" wc
Fig. 2.3 Structure with virtual crack path, as used in the substructure method.
48
3. Numerical Methods in Concrete Fracture
corresponding displacement, ~' see . The virtuai crack path, 'Y, is assumed to be known on
beforehand. The distance along the crackpathis given by the coordinate y, and the virtual
crack path ends at d . Along the crack path closing stresses are acting according to the
fictitious crack mod~l The crack is assumed to be long, but is not allowed to split the
structure into two, as it is assumed that a certain amount of nodes are necessary if the stress
and strains should be modelled in arealistic way, Petersson (1981).
First the governing equation will be derived. Then the fracture process is divided into three
and the equations are discretizised.
The crackpathis now assumed to consist of three parts, I) for y=O to y=a where the crack
surfaces are stress free (the real crack), m for y=a to y='!! where the crack is loaded by
cohesive stresses, a(y), according to the fictitious crack model, and deseribed by the softening
relation,f(w), and part III) the linear elastic part where the crack opening dispiacement is zero.
The opening between the two virtuai crack surfaces, w, can be determined by the compliances
d'Y
w(y) = JC (y,y
1
1
) a(y 1)dy 1 + ~{y)F (3.56)
o
where C1{y,y 1) is the dispiacement in x from a unit load in y 1 , and C2 (y) is the dispiacement
in y from a external load. Similarly the load point dispiacement can be calculated by
d'Y
~ = J~(y)a (y)dy + CJ(y)F (3.57)
o
where C3 is the load point detleetion from a unit load.
Three set of conditions must be fulfilled. At part I the virtuai crack surfaces are stress free
a(y) =o for Os y s a (3.58)
at the fictitious crack the stresses is deseribed by the softening relation
a(y) =.f{w(y)) for as y
a1 s (3.59)
at the uncracked ligament the crack opening dispiacement is zero

w(y) =O for a1 s y s d'Y (3.60)
which constitutes the governing equations for the sub-structure method.
The structure is now discretizised along the crackpathin n-nodes, with equidistant spaces
between the nodes as proposed by Petersson (1981) and Catpinteri (1989). The stresses are
expressed by the nodal forces, si, and the campliances are given by the appropriate influence
coefficients
The calculations are divided into three phases: Phase I where the entire structure is assumed
to be linear elastic, Phase ll where the fictitious crack is developing and Phase m where the
49
J.P. Ulfkjær. Fraelure mechanics of concrete
real crack is extending.
Phase I
In all three phases the widening of each node can be calculated by using the discretiziced
version of (3.56)
w; = En l
cus1 + ci F
Il
(3.61)
j 3 l
where, w1, is the crack opening dispiacement of node i, s1, is the elosure load in node j, c!1,
is the opening in node i from a unit load in node j, and cff
is the opening of node i due to a
unit externalload. The boundary campliances can be calculated by using the finite element
method on the discretizised structure. By applying Maxwells-Bettis theorem it is seen that
c!u=c!ii, whereby the campliances reduces to n(n + 3
2
>.
Since the opening of each node is zero, the foliowing condition must be fulfilled
W·= 0 for i = l, ..., n (3.62)
Eqs. (3.61) and (3.62)
'
compose a linear algebraical system of n equations with n
unknowns, that is the node forces, s1• When the load, F, and the node forces, s;, are known
it is possible to calculate the load point displacement, ~. as
Il (3.63)
~ =c.J SJ· + eJ'
r
where, cF, is the dispiacement for a unit load. At this first step the load, F 1, which produces
the ultimate nodal force, u", at node l and the corresponding displacement, ~1 , is calculated
by eq. (3.63), (4.95). The load-dispiacement curve in phase I is then the linear curve between
(0,0) and (~1 ,F1 ). Thus, the processis performed in load controlin phase I.
Phase Il
In phase n a cohesive crack develops in front of the real crack tip. The fictitious crack is
assumed to form between nodes l and m whereby eq. (3.62) is replaced by
s, = s. [l -::] for i = o,...,m - l (3.64)
and at the remaining ligament

w1 = O for i = m, m + 1, ..., n (3.65)
Eqs. (3.61), (3.64) and (3.65) constitutes a linear algebraical system of2n+ l equations with
2n+ l unknowns, i. e, the crack opening displacements, the nodal forces and the externalload,
F. At the second point of the load-dispiacement curve the fictitious crack is between node l
and 2 and the nodal force in node 2 is equal to the ultimate nodal force and the load, F2 that
produces that force is calculated. The dispiacement that corresponds to, F2 , is then calculated
by using eq.(3.63), (4.95). The calculations proceed in this manneruntil the critical opening
is reached in node l, which ends phase n. In this way the process is fictitious crack length
50
3. Numerical Mcthods in Concrete Fracture
controlled in phase n.
Phasem
In phase m the real crack starts to grow. The real crack is between node O and k-1 and the
fictitious crack is between node k and m-l whereby eqs. (3.64) and (3.65) are replaced
s1 • O for i • 1,2...,k - l (3.66)
in the fictitious crack
S·
•
=s
"
[1 - wi]
wc
for i = k,k + l...,m - l <3· 67)
and at the remaining ligament

wi =O for i = m, m + l, ... , n (3.68)
Eqs. (3.61), (3.66), (3.67) and (3.68) constitute a linear algebraical system of 2n+ l
equations with 2n +l unknowns.
In phase m the controiling parameter is also the fictitious crack length, it is thcrefore not
possible to directly calculate the real crack length, the externalload and the dispiacement and
instead an iterative procedure must be applied, Carpinteri (1989).
The sub-structure method is a specially designed method, for plain struerures with one crack.
The method is easy to use, and when the campliances are determined the method is applicable
for all sizes of the considered struerure by scaling the compliances appropriately. The
calculation speed is increased tremendously compared with the Hillerborg et al. method, and
by making the fictitious crack length the controiling parameter the method is applicable of
predicting snap-back. If the method was not fictitious crack length controlled, but instead
deflection controlled, the method beoornes unstable at the point where the slope becomes
infinitely. The method is, however, truncated since the crack is not allowed to divide the
structure into two, but only as long as the crack is allowed to be as deseribed previously. The
problem is especially serious for duetile beams where the fictitious crack length is large
compared to the beam height, whereby only a small part of the descending branch is
calculated.
3.3 The Direct Sub-Structure Method.
The direct sub-structure was invented by Brincker and Dahl (1989) in order to be able to
calculate the entire descending branch of the load dispiacement curve. Instead of making a cut
in the structure as in the sub-structure method, the structure is actually divided into two in the
direct sub-structure method. The method is applicable for multiple cracking, but in order of
simplicity the method is only demonstrated on the three-point bending geometry. The method
is deseribed foliowing the procedure of Brincker and Dahl (1989) and an effective method
which solves the stability problem deseribed by Brincker and Dahl (1989) is presented and
demonstrated.
51
I.P. Ulflger. Pmcture mcchanica of concrcte
~6
l F
l ~l·ID
n:
· A' t J:
y
lF/2 lF/2
l& y ;; d ls
Elaslie zone
y ::a+at B
v Fraelure zone
y:. a
Crac k
x y: O
Fig. 2.4 The considered beam divided into two substructures, Brincker and Dahl (1989).
First the general equations will be derived, which are valid in all three phases, then the
calculation technique for deformation control will be described.
Consider the beamin Fig. 2.4, which is split into two parts (a) and (b) by a virtual cut along
what is assumed to be the crack path, 'Y. In general the dispiacements will consist of the sum
of two terms: rigid body displacements, vør, and elastic displacements, vbu, of the virtual
surfaces.
By assuming small dispiacements the rigid body dispiacements are calculated as

r 2o
va(y) = ~y (3.69)
and for beam part (b)
vb(y)
r ~
= uh -
2o
-Y (3.70)
l
where, vø', and, v{, are the dispiacements ofthe virtual crack surfacesand oh is the horizontal
dispiacement of the support. The dispiacements caused by the stresses acting on the virtual
surfaces are calculated by
d
v:(y) • Jq(y)c(y,y )dy' 1 (3.71)
o
and for beam part (b)
d
v;(y) = - Ju(y)c(y,y )dy' 1 (3.72)
o
52
3. Numerical Melbods in Concrete Fracture
where c(y,y 1) is the compliances of the considered dispiacement The opening between the
virtual crack surface is then given by the difference between the total dispiacements fields of
the two bodies (a) and (b)
v(y) = v11(y) - va <Y>
h
4 (3.73)
= - 2 Ju(y 1)c(y,y 1)dy 1 + l)h - /)y
1
o
Depending on which phase is considered one or more of the foliowing conditions must be
fulfilled. The equilibrium condition
d
Ju(y)dy =O (3.74)
a
the constitutive condition
f(w(y)) '"' u(y) for a < y s a + af (3.75)
and the compatibility condition
w(y) =O for a + af <y < d (3. 76)
The extemalload is determined by the equilibrium condition i.e. the equivalent moment at the
virtual crack faces shall outbalance the external load
d
F • - ~J u(y)ydy (3.77)
a
As in the sub-structure method, the system is discretizised, in n nodes along the virtual crack
surfaces with equidistance, 3o- The stresses are expressed by the nodal forces s1 and the
compliances are expressed by the appropriate influence coefficients, cij.
The crack opening dispiacement is then expressed as

n /)
Wi =- 2 L C;fj + Oh - 4]yi for ks i < m (3.78)
)"'k
The equilibrium condition
11
:E s1 =O (3.79)
)•k
the constituti.ve condition is for simplicity assumed to be linear
W·
h• "'s Il(l + w -'>c (3.80)
if the constituti.ve relation is assumed to be piece by piece linear the two constants s" and wc
are simply modified for the appropriate nodes. The only problem is then to establish on which
53
J.P. Ulfijær. Fracture mechanics of concrete
line on the softening relation each node is, Brincker and Dahl (1989). By inserting (3.78) the
constitutive condition (3.80), (3.81) beoornes
l ,. 4 o oh si (3.81)
l - 2- L
wc j • k
Cifi - --Y;
wc l
+ -
wc
- -
s"
=O for ks i < m - l
The compatibility condition is discretizised and combined with (3. 78)

n oh o
E
j"' k
Cifj - - + 2-Yi
2 l
=o for ms i< n (3.82)
The equations (3. 79), (3.80), (3.81) and (3. 82) constitute a linear algebraical system or 2n+ l
equations with 2n+ l unknowns. It is seen that only n(n + l) influence coefficients are
2
needed. The system of linear equations can then be written
(3.83)
where the coefficient matrix is given by
s"
2-Ckk +l
s"
2-ckn - -s"
w , w , Wc
c c
w
s"
2-ck
c
+ 1k
, 2-ck
w
s"
c
+l ,n --wc
Sil
·•. (3.84)
A= Sil Sil
2-cm
w -ln
•
--wc
c
-cm,k - cm,n 0.5
··.
- cn,k - ln,n 0.5
l l o
and the right band side is given by
54
b• (3.85)
and the solution to the system of equations is
X= (3.86)
s"
~h
The non-lineacity of the problem is introduced by updating the matrix A and the right band
sides b. When the nodal forces arc determined the crack openings are determined by
(3. 80), (3. 81), and the extemal load is determined by
4 n
F ... - -l}.
.Ek SjYj (3.87)
The problem then consists in determining which node is the next to changc state (from elastic
to fictitious state or from fictitious state to complete fracture).
Two nodesarepotential of changing state: node k and node m. By applying a small test load,
d~, the opening in node k will change dwk, and the stress in node m will change dsm, whereby
thc foliowing sensitivities can be calculated
SS
J.P. Ulfkjær. Fracture mechanics of concrete
(3.88)
dsm
tm • S
.,
-S
m
The next node to change state will be the one with the large sensitivity. Since the system is
linear the dispiacement increment, L\~ , necessary for changing state is calculated by
L\lj = __!!!.. (3.89)

t max
It is seen that when a node is completely fractured it will simply be removed from the system
of cquations, reducing the calculation time as the real crack advances. In this way the entire
load dispiacement can be calculated, in load point dispiacement control.
3.4 Stability of The Direct Sub-structure Method.
As was concluded by Brincker and Dahl (1989) a serions stability problem arises when the
beam brittleness is increased, and an exorbitant amount of time has been used to solve this
problem.
The problem can be observed in the outlined method in the way that the crack opening
dispiacement of the nodes in the fracture zone increases, which corresponds to that the strain
softening relation becomes a strain hardening relation. Therefore, it was thought that an
unloading branch was necessary in order to overeorne the problem. This was not the case. A
method wherc the softening relation was a continuous fimetion was implemented. In this case
an iterative procedure is necessary since the system is non-linear. The stability problem was,
however, not solved.
It was then realiud that the stability problem is due to the snap-back effect. The load-
dispiacement curve constitutes the set of points where the system is in equilibrium. If the beam
is so brittie that snap-back is introduced, that means that the dispiacement at some point on the
descending branch is decreasing, yielding a negative infinitely slope. However, the crack
opening dispiacement will always be an increasing ftmction.
The problem can therefore be solved by making the fracture process crack mouth opening
controlled, or fietitions crack length controlled ( as by Carpinteri), or crack tip opening
controlled. Here crack tip opening control is presented. As is seen from the previous equations
it is not possible to derive the governing equations with the crack opening as the controiling
parameter. Inst:ead a very simple and effective procedure was developed. The process is made
indirectly crack tip opening dispiacement controlled.
The calculation are performed completely similarly to the just deseribed method. However,
56
if the sensitivity of node k is then the dispiacement increment is calculated by
.åcS - deS (3.35)

lk
Fig. 3.5 Load dispiacement curve where the beamis to brittie compared to the number of
nodes n=21.
Fig. 3.4 Examples of load dispiacement curves, calculated by using the Direct substructure
method, at different size scales
Since, .rk < O the dispiacement will decrease and snap back is introduced. This method is
very stable, and a load-dispiacement curve is always calculated even for extremely brittie
beams. The results are though not reliable if the number of nodes are too small compared
to brittieness of the structure. This method will be used extensively in the rest of the thesis
were many examples of load dispiacement curves using this method are shown.
3.5 Model Evaluation.
Three numerical methods have been presented. The Hillerborg et al. method was the first
method to be used with thc fictitious crack model and is a multi purpose method applicable
for all geometries and loading conditions. The method is, however, very slow and many
element meshes are necessary as the fictitious crack develops and as the real crack extends.
Instead the sub-structure method was developed by Petersson (1981) and Caepinteri (1989).
The sub-structure is only applicable for a certain type of structures but instead it is very fast
compared to the Hillerborg method. The method is also made fictitious crack length
controlled, whereby very brittie struerures can be analyzed. The method is, however,
truncated by that the crack is not allowed to split the structure into two, and is therefore not
well suited for duetile structures.
To overeorne the problem of splitting the structure into two Brincker and Dahl (1989)
developed the Direct sub-structure method, where the crack is extending through the entire
structurc and actually splits the structure into two. The method is dispiacement controlled and
is therefore not applicable for brittie structures. The calculation speed is of the same order
as in the sub-structure method.
To overeorne the problem of snap-back the author has extended the direct sub-structure
method, to be crack tip opening dispiacement controlled. The only limit to how the brittie
57
J.P. mtJgær. Fracturo mech•nica of concrete
struerures the method can analyze is thereby only Jimited to the amount of nodes.
The direct sub-strueture is therefore applied in the restofthis thesis, since it is superior to
the sub-strueture method.
Brineker, R. and Dahl, H. (1989), Fictitious Crack Modelof Concrete Fraeture, Magaline
of Concrete Research, 41, No. 147, pp. 79-86.
Carpinteri, A. (1989), Size Effects on Strength, Toughness, and Duetility. Journal of

Elfgren, L (1989), Fracture Mechanies of Conerete Struerures-From Theory to Application,

Chapman and Hall, RILEM Report.
Modeer, M. (1979), A Fracture Mechanies Approach to Pailure Analysis of Conerete

Materials, Division of Building Materials, Lund lnstitute of Technology, Report
1VBM-1001, Lund Sweden, pp. 1-102.
Hillerborg, A., Modeer, M and Petersson, P.-E. (1976), Analysis of Crack Formation and
Crack Growth in Conerete by means of Fracture Mechanies and Finite Elements,
Cement and Concrete Research, Vol. 6, pp. 773-782.
Petersson, P.-E. (1981), Crack Growth and Development of Fraeture Zones in Plain
Concrete and Similar Materials, Division of Building Materials, Lund lnstitute of
Technology, Report 1VBM-1006, Lund Sweden, pp. 1-174.
58
4 ANALYTICAL METHODS IN CONCRETE

FRACTURE.
As deseribed in the previous section a variety of numerical methods exists which can be used
to predict the load carrying capacity of plain and reinforced concrete struerures in which a
fictitious crack develops. It is, however, often convenient to have more simple methods which
can be used to predict approximate results and more appropriate for the practical engineer. The
numerical methods are in general very time consuming, and if an indirect method is used to
determine the constitutive relations, as will be deseribed in chapter 5, the numerical effort
becomes enormous. In order of convenience the various models arenarned after the authors,
who published the deseribed methods.
4.1 Non-Reinforced Concrete.
In the following, three analytical models for plain concrete will be presented and some
improvements will be suggested. The presentalion is limited to linear softening relations and
the notation used in each model corresponds, ifpossible, to the notation introduced by Ulfkjær
et al. (1991).
4.1.1 The UlfkJær, Drineker and Krenk Method.
The idea of modeiling the bending failure of concrete beams by the development of a fictitious
crack in an elastic layer with a thickness proportional to the beam height was introduced by
Ulfkjær, Brincker and Krenk (1990), and further refined by Ulfkjær, Krenk and Brincker
(1992). The foliowing presentation is equivalent to that, and some extra results are presented.
The model is based on the assumption that the complex stress field areund the crack is
modelled by simple spring-action in an elastic layer areund the crack, and outside the layer the
deformations are modelled by elastic beam theory.
Using a linear softening relation yields that the fracture energy is given by GF= Muuwc,
where, a11 , is the ultimate tensile stress and, wc:, is the critical crack opening dispiacement In
the elastic layer only bending stresses are assumed to be present and the stress is assumed to
depend Iinearly on the local elongation of the layer. By assuming a linear softening relation,
the constitutive relation of the layer becomes a bi-linear relation between the axial stress, a,
and the elongation, v, see Fig. 4.1. On theascending branch the elongation is linear elastic
v=vø and no crack opening is present. The linearresponseis given by ve =uh/E where, h, is
the thickness of the layer, and, E, is Young's modulus. On the descending branch the total
deformation v consists of two contributions v = ve + w, where, w, is the crack opening
59
J.P. Ulfkjer. Fncture Mecbanica of Concrete
u q
B<l B= l
v(y) Wc
(a) (b)
Fig. 4.1. Constitutive relation for the midsection of the layer a) When the layer is stable,
B< l .. b) When the layer is unstable B=l.
dispiacement The peak point corresponds to the deformation v = v11 , and total fracture
corresponds to v=vc· Therefore, the critical crack opening dispiacement corresponds to wc=vc•
If the elastic layer should be stable in dispiacement controlled loading the foliowing condition
must be fulfilled
(4.91)
In the present method it is convenient to define the brittleness number, in eq. (2.38), as
u;, (4.92)
B•--
2G~
corresponding to a characteristic size of the structure equal to h/2. The stability condition
(4.91) can then be written as
B< l (4.93)
TilUs, in this model the brittleness number, B, varles between zero corresponding to ideal
duetile behavior and one corresponding to ideal brittie behavior. The thickness h of the elastic
layer, is assumed to bealinear function of the beam depth h=kd. In general the factor, k, is
a function of the beam geometry, the beam size, the fictitious crack length and the real crack
length. In this model the factor, k, is assumed constantand is found using the DSS dernanding
that the peak load is predicted correctly.
As a first approximation only rigid body dispiacement is assumed of the beam parts outside
the elastic layer, see Fig. 4.2.
The calculations are, as in the numerical method, divided into three phases. Phase 1): Before
the tensile strength is reached in the tensile side of the beam, phase m: Development of a
60
4. Analytical Methods in Concrete Fracture
a)
h=kd
diD ~
l
b)
Fig. 4.2 a) The considered beam where the hatched areais the elastic layer. b) Deformed
beam where only rigid body dispiacements are considered.
l
u=uul a,
la
l
l
u<u u l
Phase I Phase II l Phase III
l
F 1
l
l
l
l
l
l
Fig. 4.3. The stress distributions of each phase a) Phase I where the stress distribution is
elastic. b) Phase n where the fictitious crack is developed. c) Phase m where the real crack
starts to grow. d) The load-dispiacement curve.
fictitious crack in the layer, and phase III): Crack propagation. The stress distribution in each
phase of the fracture processis illustrated in Fig. 4.3
Phase I.
In phase I a linear elastic constitutive relation is used for all parts of the Iayer ve = uh/E. By
simple geometric considerations it is seen that ve = f/J (d-2y) where V' is the rotation, d is the
61
J.P. Ulfkjaer. Fracture Mechanics of Concrete
I II III
Fig. 4.4 The moment rotation curve of the beam when only rigid body dispiacements are
considered.
beam depth and y is the vertical coordinate. The neutral axis is at the mid-point of the beam
corresponding to y = d/2. Instead of the bending moment M and the rotation tp it is convenient
to introduce the dimensionless bending moment
p.= M-6- (4.94)

u,f12t
and the corresponding dimensionless dispiacement
8= t{) dE • tp_E_ (4.95)

hu" ku"
giving the simple load-dispiacement relation
p.(8) = 8 (4.96)
In the limit situation of phase I the stress for y=O equals the tensile strength, and the
dimensionless bending moment equals one. Thus, in phase I the load-dispiacement curve is a
straight line between origo and (8,p.) = (1,1), see Fig. 4.4.
Phase n.
In phase n the size of the elastic tensile zone is determined by simple geometrical
considerations. When the fictitious crack develops, it is necessary to determine the crack
opening dispiacement By assuming that the stress in the fictitious crack is equal to the stress
in the elastic layer, the crack opening dispiacement becomes
(4.97)
where a1 is the length of the fictitious crack. Thus, this corresponds to a linear crack profile.
lf the linear softening relation is expressed as
(4.98)
62
4. Analytical Methods in Coocrete Fraelure
then the length of the fictitious crack can be determined by combining (4.98) with the
equilibrium condition (the resultant axial force equal to zero). The result reduces to
"f(Ø) =; (8) = 1-B-J(1-B)(! -B) (4.99)
The equivalent moment is determined by integrating the axial stresses
~ 8 [ {_~2 3 41
l' (6) - 6a1 (8) +4] - 3 ( • OO)
In order to stay in phase n the crack opening dispiacement at the hottom of the beam must be
smaller than the critical crack opening w(O) < wc, which by use of (4.97), (4.99) and (4.100)
can be reformulated as
1'(8) > l or l < 8 < 8c (4.101)
where
8 = l+{B (4.102)
c 2B
TitUs, during the development of the fictitious crack the moment increases from l to its
ultimate value and then decreases again. When the moment reaches the value l at the
descending branch corresponding to 8=flc the real crack starts to grow, see Fig. 4.4. The real
crack will therefore only propagate on the descending branch of the load dispiacement curve
as found by Harder (1991).
Pbase m.
In phase m the real crack starts to grow. The real crack length is termed a, see Fig. 4.3. The
size of the elastic tensile zone is determined by the condition that w(a+aJ = w11• The size of
the fictitious crack, ap is obtained by the condition w(a) = wc giving
l l-B (4.103)
af= 28B
The crack length a is determined through the equilibrium condition that the resultant axial
force is equal to zero
a 8c (4.104)
a=- • l - -
d 8
As in phase n the dimensionless bending moment is determined by integrating the axial
stresses. The result is
1'(6) - li r
When, tic' is suitably modified this result is general in the sense that it is valid for all softening
(4.105)
relations. The results for the moment-rotation curve including only rigid body dispiacements
of the beam parts are shown in Fig. 4.4.
63
J.P. Ulfkjær. Fracture Mecbanics of Concrete
Beam Depth, d [mm] 100.0

Beam Width, t [mm] 100.0
Beam Length, l [mm] 800.0
Notch depth, a; [mm] 0.0
Specifi~ Fracture Energy, GF [Nmm/mm2] 0.100
Tensile Strength, f!" [N/mm2] 3.0
Modulus of Elasticity, E [N/mm2] 20,000
Brittleness number, B 0.1125
Tabel 4.1 Geometry and materiais parameters for standard beam.
In Fig. 4.5
2.00 Slenderness-8.0
- - Numerical ' Scal.. 0.25
- - AnalyUcal
::s.
..;
1.75
' \
5 UiO \
8
o
:::::11 1.25
.,
"d
N
.,
:::1 1.00
~
o 0.75
z
0 .50
0.25
o.oo -f'T'TT"'rrTTT"l""T'T"rT'T'TT"'rrTTT"l""T'T"rT'T"TT'"I"'TTT'T'1rrTT'T'l
0.00 0.50 1.00 1.60 2.00
Rotation, rp (Thousands)
Fig. 4.5. Comparison between the analytical model and DSS using the standard beamat 4
different size scales.
Elastic defonnations in the beam parts outside the elastic layer are taken into account by
subtracting the elastic deformation, 8 =p.(B), from the elastic layer Ieaving only deformations
due to crack growth and adding the elastic deformations of the whole beam using a solution
for a Timoshenko beam, Timoshenleo S. {1955). The Timoshenko dispiacements are
64
Ml 2 (4.106)
~e = 12ElfJ(A)
where El is the bending stiffness of the beam, {J, is a factor describing the influence of shear
fJ = l +2.85A2 - O.B.f>...3 , and, :A, is the slendemess ratio "A=lld. Introducing the elastic
rotation similar to equation (3.63), (4.95)
(Je =
oe dE
2--- (4.107)
l huu
the relation (4.106) can be written in dimensionless form
(Je = 'YP. (4.108)
where
(4.109)
and the total deformation is then given by subtracting the elastic deformation in the layer and
then adding the deformations of the Timoshenleo beam
(4.110)
Hence, the complete moment rotation curve is fully determined by the brittleness number B,
and the slendemess ratio 'A.
The numerical results are obtained by the direct sub-structure method (DSS), Dahl and
Brincker (1989). Four-node elements and an element mesh with 21 nodes in the midsection
were used. Results for one beam geometry (slendemess ratio 'A= 8) are comparedat different
brittleness levels in order to see how wellthemodel predicts the load-dispiacement curve. lt
is assumed that the size of the elastic layer is proportional to the beam depth h = kb where
the factor k is assumed to be 0.5. A beam geometry similar to the RILEM beam and materlal
parameters corresponding to a normal strength concrete is chosen as standard beam, see
Tabel 4.1. With the chosen materlal parameters the maximum beam depth is according to
(4.93) 888 mm corresponding tothat all dimensions ofthe beamare scaled by a factor of 8.88
(scale factor = 8.88).
a comparison is shown between the analytical model and the numerical results for the standard
beamon 4 different size scales (0.25, 0.5, 1.0 and 3.0). 1t is observed that the shape of the
moment-rotation curves is almost identical and that themodel predicts the ultimate load quite
well. However, in the analytical model the snap-back effect is more pronounced which implies
that the analytical model is a littie too brittle.
In Fig. 4.6 results for the size of the fictitious crack are compare<i. 1t is seen that the size of
the fictitious crack calculated by the analytical model is slightly smaller than that calculated
by the numerical method before the real crack starts to grow (the ascending branch of the
curves) and larger at the descending branch. The small kinks on the numerical curve are due
to the discretization made in the numerical model. With a larger number of nodes in the
midsection these kinks would disappear. In Fig. 4.8 the real crack lengths for the two models
are compared. lt is seen that the real crack grows faster in the numerical model.
65
J.P. Ulflcjær. Fncture Mechanics of Concrete
1.00
- Analytical
·----- Numerical Scale=0.25
0.76
~ 0.50
'O
Q)
N
00 0.25
o.00 -+r"T"T"T"T'""r-T'"I'TrrT"T"T"T"T"T'""rT"T''T"T'"rT"T"T"T"T"T'""rT"T''T"T'"I"'T"T.,.,..,
0.00 0.50 1.00 1.50 2.00

Fig. 4.6. Size of the fictitious crack for the analytical model and DSS. The loops aredue to
snap-back.
- - Analytical
-- -
1.00
- Numerical
0.80
tS
:5bO 0.60
cQ)
-
. ):
(,)
0.40
Cl
r.
u 0.20
0.00
0.00 2 .00 4.00
Fig. 4. 7. Length of real crack for the analytical model and DSS.
In order to check the influence of the slendemess of the beam on the load dispiacement curves
two additional slendemess number are considered. The slendemess is regulated by changing
the beam length whereby a scale factor of 1.0 always corresponds to the beam height 100 mm.
The results are seen in Fig. 4.8 and Fig. 4.9. It is seen that the analytical predicted by the
analytical model still is almost identical to the numerical method. The stiffness of the beam
with the slendemess ratio equal to 4 is, bowever, not identical. This is not a surprise since the
Timoshenleo assumptions are not accurate enough for beams with low slendemess ratios.
The peak loads, #'max, predicted by the analytical and the numerical models are shown in
66
..
4. Analytical Methods in Concrete Frøeture
Slenderness-4.0
..;
s::
2.00
- - Numerical
- - AnalyUcal
-- Scale=0.25
G)
§ 1.50
:::1!
'd
G)
....
N
esE 1.00
o
z
0.50
0.00 --h""I"'TTTrr1""1"'TTT"rT"1"TTTT"r-i=iT'i",..,n-r"TTTT"'rT"I''TT"
0.00 0.50 1.00 1.50 2.00
Fig. 4.8 Load dispiacement curves for the slenderness ratio equal to 4.0.
2.00 Slendernessz16.0
- - Numerical
- - Analytlcal
1.75
:l
..;
s::G) 1.60
8
o
:::1! 1.25
'd
-E
l)
....N 1.00
liS
o 0.75
z
0.50
0.25
0.00 -f-r-T"T"T"T""T--r-r-r-T"T"T"T""T"""1r-T""r-T"~"T""T--.-r-r-T"T"T""T""1
0.00 1.00 2.00 3.00
Fig. 4.9 Load dispiacement curves for beam with slendemess = 16.0.
log-log scale in Fig. 4.10. Since there is no stress singularity included in the two models there
67
I.P. Ulfkjær. Fracture Mecbanics of Concreto
l
- - Analytical
- Numerical
.
8... ...... __
~ 1
0.001 0.01 0.1 1 lO 100

Size Scale
Fig. 4.10. Peak load at different size scales predicted by themodel and by the DSS.
is no size effect beyond the critical size of the models. In the numerical model this limit wi1l
depend upon the materlal parameters and the number o f nodes there are in the midsection (here
the critical size scale is approximately 20).
h~k(d-lj11) a, l• t l
Fig. 4.11. Modelof notehed beam.
The model is extended to notehed beams as indicated in Fig. 4.12 . The idea is to keep the
width/depth ratio of the elastic layer by setting the width of the layer equal to k times the
effective beam depth of the notehed beam section h = k(d-aJ, k = 0.5 where a; is the depth
of the noteh. The modifications thus introduced imply that the brittleness number B for the
layer is multiplied by a factor (1-a;fd). The total beam depth is still used in the formulas
(4.94), (3.63), (4.95) and (4.106) whereas the effective beam depth de = b-a; is used in all
other formulas. Results for different noteh depth' s are shown in Fig. 4.12 .
Comparlog numerical results with results for the analytical model it can be concluded that
deviations are relatively small. The errors introduced by the elastic layer and the assumption
68
2.00 - - Analytical
- - Numerical
Notcb lenetb 0.0
..;
.,c
ao
::lil1.00
.,
."
.....N
c;
8o
z
0.00 ~::.".,...,....-r"T"'li""'"'T"...,..,...,.....,...................,~~~:::;:;::;:;::;;:::;:.
0 .00 0.60 1.00 1.50 2.00
Rotation, ~ (Thousands)
Fig. 4.12. Moment dispiacement curves for notehed bearns with three different notch depths.
of wedge-like crack-opening are typically smaller than errors due to the simple linear softening
relation, Brincker and Dahl (1989).
When the size of the beam changes, the stress distribution in the partially fractured mid-section
changes and so does the shape of the load-dispiacement curve. In the foliowing a few closed
form solutions are given for these size effects predicted by the analytical model.
An important parameter describing the stress distribution in the partially fractured mid-section,
is the maximum size af,max of the fictitious crack. Since aajaø > O in phase II and aa( aø <
O in phase m, ~is !argest at the end of phase II. Thus, the maximum size of the fictitious
crack is found by combining eq. (4.99) and (4.101)
a.f.max • d(l-/i) (4.111)

Thus, for small duetile beams the relative size of the fictitious crack is large and for large
brittie beams the relative size of the fictitious crack approaches zero.
The way the load-dispiacement curve changes with size is more difficult to describe. One
important parameter of the load-dispiacement curve is the peak load, P.max· The peak load
might be obtained from the condition ap.!aB, =0. However, no simple expressions have been
derived for this case.
Another key-parameter for the load-dispiacement curve is the maximum slope S on the
descending branch. The slope is found by taking the derivative of eq. (4.110)
1 - ~ ap. •('Y-l) ap. (4.112)

ap. ae, aø,
from which
69
J.P. Ulflgær. Fraelure Mechanics of Concrete
....!!:.
ae,
ae]-
a ... ['Y-1+-
ap.
1
(4.113)
The steepest point on the descending branch of the load-dispiacement curve is at the transition
from phase n tophase m, i.e. for (J=(Jcr• Thus, the minimum value of ae/ap. is found from
eq. (4.105) and eq.(4.102) which together with eq. (4.113) yield the results
4B
S=----- (4.114)
1+{B -4('Y-1)B
The quantity S is a kind of brittleness number for the structure, as deseribed in Chapter 2. The
larger maximum slope on the descending branch, the more brittie the behavior of the beam wil1
be. The brittleness number varies between zero corresponding to ideal duetile behavior and
infinity corresponding to the case where the maximum slope becomes infinite. If the point of
infinity slope is exceeded, snap-back occurs, and the brittleness number S becomes
meaningless. Thus, the brittleness number S only describes the brittleness of struerures without
snap-back on the load-dispiacement curve.
2.00
Slenderness 8.0
~
..
~ 1.50
Q)
s
~ 1.00
"CC
Q)
-8
....
N
cø
.... 0.50
zo
0.00 0.50 1.00 1.50
Fig. 4.13. Moment dispiacement curve for the beam with the slendemess = 8.0, on the critical
size scale.
The maximum slope on the descending branch becomes infinite when the denaminator in eq.
(4.114) vanishes, i.e. when
1+/B-4('Y-1)B =O (4.115)
The solution to this equation defines a critical brittleness number for the elastic layer
70
4. Analytical Methoda in Coocrete Fracture
0.35
.
~ 0.3
~ 0.25
~ 0.2
ll 0.15
ø:l 0.1
"H
·a 0.05
Q
o 2 4 6 8 10 12 14 16 18 20
Slendemess ratio, l..
Fig. 4.14 The critical brittleness number as a fimetion of the slenderness.
2
B = [ l+J1+16('Y-1) ] (4.116)
er 8('Y-1)
If the brittleness number B of the layer is larger than the critical brittleness number Ber then
there is snap-back on the load-dispiacement curve. Otherwise there is no snap-back. The
critical brittleness number as dependent of the slenderness is seen in Fig. 4.14, and it is seen
as the slenderness increases the critical brittleness decreases. This was also seen in the
moment-rotation curves for the different slendemess numbers. For the standard beam the
critical brittleness number is found as Bcr=0.069 corresponding to a scale factor of0.615. The
case is illustrated Fig. 4.13 where the results for the analytical model are shown for B =Ber.
In the method presented by illfkjær et al. the equations arederived assuming that the beam
is loaded with a concentrated load at the midsection and that the strain distribution in the
elastic layer is constant. These two assumptions are conflicting, but this conflict is easily
removed by assuming that the beamis loaded in four point bending, whereby the equivalent
moment beoornes constant between the loads.
1t is also concluded that the model is not applicable for beams with a brittleness number equal
to or larger than one. This restrietion is not necessary, as it can be replaced by a condition
stating that phase II is absent for brittleness numbers equal to or larger than one. Beyond this
value the peak load can eventually be calculated by using LEFM, yielding the strong size
effect deseribed in chapter 2.
4.1.2 The Chuang and Mai Method.
71
I.P. Ulfkjær. Fracture Mecbanics of Concrete
This method was introduced by Chuang and Mai (1989) and is basedon the crack band model.
In the original paper four point bending specimens were considered and the descending branch
of the constitutive relation was a power relation of the type
(4.117)
where, u, is the stress, e, is the strain, uu, is the tensile strength, e~, is the strain at peak load,
ec, is the strain that corresponds to no stress transmission and, n, Is the softening coefficient.
Thus, all the parameters arematerlal constants, see Fig. 2.15. By setting the minor span (the
distance between the two Ioads) to zero and n=l, a three-point bending configuration with a
linear softening relation is obtained. The procedure for establishing the governing equations
is very similar to the previously presented method and the Chuang and Mai method will
therefore only be roughly presented.
Consider a beam with the length, L, depth, d, and thickness, t, withacrack band introduced
l t l
Fig. 4.15 The considered beam with a crack band around the midsection.
around the midsection with the thickness, hc, see Fig. 4.15. The calculations are again divided
into three phases, corresponding to those defined in the previous section. As a first
approximation the two bearn parts outside the crack band are assumed to perform rigid body
deformations. The strain distribution in each phase are shown in Fig. 4.16, the stress
distributions are as in the previous presented model, see Fig. 4.3.
Phase I
In phase I all the materlal is assumed to be linear elastic. By dernanding equilibrium it is seen
that the neutral axis is at x=d/2 whereby the normalized bending moment becomes
(4.118)
where, M, is the moment in the midsection and the normalized curvature, "• is defined as
72
4. Analytical Methods in Concrete FraelUre
Phase I Phase II Phase III
d
(a 11 -a1 )d
a 1d
l (a 11 -a1-a)d
......___~
a1d
ad
Fig. 4.16. Strain distribution in the three phases.
K = e•~ (4.119)
e,. x
where E• is the strain at the lowest part of the elastic tensile zone and a 11s the normalized size
E
of the elastic tensile zone and ~ = ..!. is a materlal parameter which retates the maximum
Ec
elastic strain to the separation strain. At the end of phase I the normalized moment and
normalized curvature becomes
p. = } K • 211 (4.120)
TilUs, as in the Ulfkjær, Brincker and Krenk model a linear relationship is obtained.
Phase n
In this phase the fictitious crack starts to develop. The normalized size of the elastic tensile
zone, a,, is obtained by the condition that the strain at the tip of the process zone equals, eP,
yielding: a" =71/K, and the size of the softening zone, ap is obtained by requiring equilibrium
(4.121)
It is immediately seen that the length of the process zone is only dependent on the curvature
and the materlal parameter, 11· This is in strong contradietion to the Ulfkjær, Brincker and
Krenk model, where, ap also is a lunetion of the size of the structure. Observe, however, the
remarkable similacity l:ietween equations (4.121) and (4.99).
The corresponding normalized moment can be expressed as
l'(•) =
21
( -::-a/ +2a!+6a1 a• +3aJ- (
1 ~~) [ 3a}+ 2 t] <4· 122)
Phase n ends when the strain in the lowest fibre equals, ec, yielding the following condition
to the curvature in phase n
(4.123)
73
J.P. Ultkjzr. Fraelure Mecbanics of Concrete
where "c = l + {;is the normalized curvature which ends phase n. This corresponds tothat
the normalized moment is one in both limits, completely equivalent to the illfkjær, Brincker
and Krenk method.
Phase m
The governing equations in the crack growth regime is obtained similarly to the previous
phases. The size of the elastic tensile zone is obtained by considering similarly triangles, the
size of the process zone is determined by the condition that the strain at the tip of the zone
must equal, E11,
(4.124)
the real crack length, a, is obtained by requiring equilibrium
a "c
= 1-- (4.125)
IC
1.6
1.4
Chuang et al.
:i 1.2 Ulfkjær et al.
ia o.s l
10.6
z 0.4
0.2
o
o o.s l l.S 2 2.5
Rotati~ ep ('lhousands)
Fig. 4.17 Moment rotation curves predicted by the Chuang and Mai model and The illfkjær,
Brincker and Krenk model.
and the equivalent normalized moment becomes
74
4. Analytical Metbods in Concrete Frøeture
2
K ] (4.126)
#l(K) =[ :
Again a remarkable similacity is seen between the two models, however, with the important
difference that the goveming parameters in the Chuang and Mai model are independent of the
beam size.
The displacement, ~' is obtained by adding the dispiacement in the crack band to the
dispiacement in the rest of the beam. Since the curvature in the crack band is constant, the
detleetion will be a circular arch with the radius 11K, the deformations in the parts outside the
crack band are calculated by assuming that the parts are cantilever beams loaded by, F, at the
end. The total detlection, ~' becomes
~ L . -1 h(/CEc l (4.127)
- =- tan[sm (-->1+-"1Ec'Y2l'
d 2d 2d 6
where
'Y2 =
f~-~ 13-2f ~r f~l2f ~-~l
+3 (4.128)
L
d
For the standard beam introduced in the previous section, the size of the crack band, h c, is set
to 50 mm (which is the same as the size of the elastic layer) whereby, e"=0.00015 and
Ec =0.00133.
= 2~ ,
The normalized moment-rotation curve, where the rotation is determined by ({)
predicted by the Chuang and Mai model and the Ulfkjær Brincker and Krenk model are shown
1
in Fig. 4.17. 1t is seen that the curves arealmost identical and the difference is attributed to
the difference in the method of calculating the elastic deformations. Normalized moment-
rotation curves, with the same materlal properties, at different size scales, predicted by the
Chuang and Mai method, are shown in, Fig. 4.18. lt is seen that the peak value is not affected
by the size scale and that the shape of both the ascending and the descending branch is
changed.
The only equations where the size of the beam is included are in the deflection equations. This
is due to the faet that during derivation of the governing equation an increase of the beam size
alters the normalized amount of elastic energy stored in the beam parts outside the crack. It
can therefore be conelurled that the Chuang and Mai model is not able to predict size effects,
but can only determine the difference in load carrying capacity for different materials. The
Chuang and Mai model will therefore not be considered further.
4.1.3 1be Llorca and Elices Method.
As was shown in chapter 2, virtual crack propagation can be considered by using LEFM. For
the three-point bending geometry this has been done by Catpinteri (1982), Carpinteri (1986),
Catpinteri (1989) and Biolzi et al. (1989). The Uorca and Elices method is an extension of
75
I.P. Ulflljer. FlliCture Mechanics of Concrete
1.6
1.4
. 1.2
:::1.
6a l
i 0.8
t
:z; 0.4
0.6
0.2
o
o o.s l 1.5 2 2.5
Rotation,~
Fig. 4.18 Normalized moment rotation curves, as deseribed by the Chuang and Mai model for
the standard beam at different size scales.
this method since it assumes a linear elastic beam withacrack with the length, a+a1 , and a
cohesive stress distribution acting on the crack edges deseribed by the FC-mcidel, see
Fig. 4.19.
The method is based on tindings of Foote et al. (1986), but the equations of the model
deseribed here arederived on basis of the work by Llorca and Elices (1990). The method is
in general applicable for all structures and all loading conditions as long as it is possible to
detennine the appropriate stress intensity factors. Here the method will betested on the three-
point bending configuration.
The basic hypothesis of the model is that the crack opening profile is a known function,
completcly deseribed by the length of the fictitious crack, af. Here it will be assumed that the
crack profile is a straight line deseribed by the function
for O ~ x s a + af (4.129)
where, a, is the real crack length, a~_, is the length of the fictitious crack and, wc• is the
critical crack opening dispiacement The beam is assumed to be linear elastic, loaded at the
load point with the force, F, and at the crack tip with cohesive stresses (internat loading)
deseribed by the constitutive relation,f(w), and the crack opening profile, w(x). Since the beam
is linear elastic the stress intensity factor, K, is determined by using the principle of
76
4. Analytical Melbods in Concrete Fracture
lo
r
x u(x)~
Uu Uu
, at
a
, ,
l
Fig. 4.19 The considered beam geometry, with a cohesive crack in front of an initial real crack
tip.
superposition
(4.130)
where, Kø, is the stress intensity factor due to the cohesive stresses and, KF, is the stress
intensity factor due to the externalload. One of the basic assumptions in the FC-model is that
there is no stress intensity at the crack tip, which implies that the stress intensity factor is equal
to zero
K • 0 ~ Kø = -KF (4.131)
For the three-point bending specimen with an initial notch of length, a, the stress intensity
factor for the extemalload, KF, can be determined by, ASTM (1974),
KF =
3lF /i YF(ald) (4.132)
2d2t
where the dimensionless shape function, YF(alb), for A=8 is given by
YF (a/d)=l.93-3.07(a/d)+l4.53(a/d)2 -25.11(a/d)3 +25.8(a/d)4 (4.133)
The stress intensity factor for a concentrated unit load at, x, on the crack edge for an infinitely
long beam is determined by, Tada et al. (1973)
(4.134)
where
77
y (a/d,xla) = 3.52(1 - x/a) _ 4.35 - 5.28xla

con (l - ald)312 Jl - a/d
(4.135)
312
+ { 1.30 - 0.30(x/a) + 0.83 - 1.76x/a}[l - (l - xla)ald]
J1 - (xla) 2
and by applying the principal of superposition the stress intensity factor for the stress
distribution, u(w{x}), becomes
ø+ Øf
a +af x {
= 2
w (x) } dx (4.136)
Ka(a,a)
J J. .(a +a)
ø
Ycon< d ,
a + af
) Cl
whereby thc externalload, F, can be determined by using eq. (4.131).
In order to obtain compatibility it is necessary to check if the assumed crack opening profile
corresponds to the one calculated by using dispiacement formulas according to LEFM. In the
foliowing the necessary dispiacement and crack opening dispiacement equations will be derived
using the energy principals deseribed by Llorca and Elices (1990a) and Bosco et al. (1990).
Instead of considering the entire cohesive stress distribution a concentrated force, F2 , at the
crack edge is considered together with the applied load, F1, with the corresponding
x
a
,
,
l
Fig. 4.20 Beam with two concentrated loads for determination of compliances.
displacements, u1 and, u2 , see Fig. 4.20. The total potential energy, A, stored in the system
is
ll = - .!.F1~1
2
- .!.F2c52
2
(4.137)
wherc thc dispiacements are given by the compliances
78
4. Analytical Metbods in Concrete Fracture
at • CuFt + C12F2 (4.138)

~ = C21F2 + C22F2
The energy release rate is given by
G=_ an
raa (4.139)
1 2 ac11 1 ac22 ac12
= 2F1 raa + 2F2 raa + F 1F 2 raa
where Maxwell-Bettis theorem has ben applied. The energy release rate can be written as
K2 (Kt + K~2
G • - "" ----=:---
E E (4.140)
2 2
K1 K2 2K K1 2
z -
E
+ -
E
+ ---==--
E
Since the terms in (4.140) are Iinearly independent the compliances are determined as
ø+ ør
c9{a + a) • ~ J Ki~ dw (4.141)

o
where, K;, and, JS, are the stress intensity factors for a unit load. The dispiacements can then
be determined by using eq. (4.138). When a distributed stress distribution is applied along the
crack faces the principle of superposition can be applied
a+ ae
aF = CFF(a + apF + J CF;;c(a + a_pX) u(x) tdx

ø
(4.142)
ø + ør
a(x) = CJCF(a + a_pX)F + J Cxt(a + ap:l,x) u(.i) t dl

a
where c.xi is the dispiacement in, x, from a unity load in, :l, is given by
ø + ac
CJ.a + ap:I,x) • 21
E I Kj(w,x)Kx(w,:f) dw (4.143)
max(x,:l)
The detleetion of the beam is then given as the sum of the deformation due to the crack and
the deformation of an uncracked beam.
By setting x=a1 in eq. (4.142) the crack tip opening displacement, CIOD, is calculated
!
ø+ ør
CIOD = a(a + a_p • Car(a + a)F + C'"u{w(x)}tdx (4.144)
where
79
I.P. Ulfkjær. Fraelure Mecbanics of Concrete
a+ a1
Cu(a,a) "" ; J KF (a + a)Kc011~,a + a) dx

o (4.145)
a + a1
Cax(a,apX) ,., ~ J Kc011(w,x) Kco11(w,a + a_p dw

x
By comparing the, CIOD, with the assumed value, wc, it is possible to check if the assumed
crack opening profile deseribed by, a1 , is correct. If the, CIOD, is higher than, wc' then, ap_
has to be increased, and vice versa. The actuallength of the fictitious crack is then calculated
foliowing an iteration procedure. The above formulas can only be calculated nurnerically, and
care should be taken with the singular stress intensity factors durlog the numerical integration.
Llorca and Elices suggest that by using this known value of, a1 , it is possible to calculate the
entire force-dispiacement relation. This seems to be a crude approximation. According to the
numerical methods and to the illfkjær, Brincker and Krenk method, where it was shown that,
a1 t initially is zero then growing to its maximum value and then decreasing, and i t was seen
that the maximum size oft ll;t' was obtained when CIOD=wc. TllUst in order to improve the
model Uorca and Elices have calculated, a1 t in each loading step in a subsequent paper,
Uorca and Elices (1990b), howevert without comparing the method with numerical results or
with the previous sketehed method and without giving the results of, a1 .
As was seen in the previous section, it is only at the descending branch the real crack starts
to grow, and it seems evident that the method should be extended, so the development of the
fictitious crack can be calculated. Actually the peak value is reached durlog the development
of the fictitious crack.
The method here suggesled is equivalent to the Llorca and Elices method, but is divided into
two phases. In the first phase the fictitious crack develops and in the second phase the real
crack grows. At the first loading steps the desired opening is not set to the critical opening,
but an opening which is a certain fraction of the critical opening, e. g. wdes =w;n, where, n,
is the fraction and wdes is the new desired opening. The above outlined iteration method is then
applied until the CIOD is equal to the desired value. The desired opening is then increased
and the iteration is performed again. This is done, n, times corresponding to that the desired
opening is equal to the critical opening. Then the second phase begins which corresponds to
the extended Uorca and Elices method. The method corresponds to that the loading in phase
I is crack tip opening dispiacement controlled and in phase n is crack length controlled. This
suggestion wil1 yield a considerable improvement of the Uorca and Elices method, and wil1
only yield a small increase in the total calculation time since it is known that in the initial
phase, ap therc is an increasing lunetion of the desired opening and a decreasing fimetion in
the second phase of the real crack length.
In the foliowing the different procedures of this method are comparedt see Fig. 4.21.
4.1.4 Model Evaluation.
Three methods basedon three different assumptions have been presented.
80
4. Analytical Methods in Concrete Fraelure
0.50
...,
~0.40
so
=s 0.30
'O
Q)
N
·~
«j 0.20
S Extended Llorca
$-t
o
Zo.to
o.00 -f-T""T""T"'T""T""T""T'""T""T'"T"T"T""T""T""T""T""T""T""T""T""T""r-r-lr-r-lr-r"'lr-r"'l
0 .00 1.00 2.00 3.00
Normalized Displacement.
Fig. 4.21 The different melbods as introduced by Llorca and the author.
The Ulfkjær et al. model is based on the development of a fictitious crack with a linear
softening relation in the midsection of an elastic layer. The layer is assumed to be Iinearly
dependent on the beam depth. The moment rotation curves calculated by themodel is atmost
identical to the ones calculated using a numerical model at different size scales and slendemess
ratios. The great advantages of the model are its simplicity and that the governing equations
are simple and explicit whereby the calculating time is minimal. Purther it is possible to derive
analytical results, e. g. the maximum size of the fictitious crack and the slope of the descending
branch. The disadvantage is that the model is only applicable for a linear or piece by piece
linear softening relation and that a size effect relation has not been found. However,
appmximate solutions can be obtained for other softening relations than the linear by assuming
that the dispiacement field in the midsection of the layer is linear.
The Chuang and Mai model is based on the crack band model, on the Bernoulli assumptions,
and on an exponential softening relation. The governing equations is derived on the basis of
the same stress distributions as in the Ulfkjær et al. model and by assuming a linear softening
relation the equation becomes atmost identical, with the important difference that the Chuang
and Mai model not is able of predicting size effects, which is a great disadvantage of the
model. The model is, however, able of comparing the load carrying capacity of identical
struerures with different softening relations.
The third model is based on LEFM and on the assumption that closing stresses acts on the
edges at the crack tip, deseribed by the fictitious crack model. By assuming that there is no
stress intensity factor at the crack tip, the externalload and the crack tip opening dispiacement
can be calculated, which shall equal the critical crack opening dispiacement By performing
an iterative procedure, which involves numerical integration of a double integral, the length
of the fictitious crack can be calculated. By using this length the load dispiacement curve can
81
J.P. Ulfkjer. Fracture Mechanics of Concrete
be calculated, this approximate procedure is uscd in order to make the method simple and fast.
The model is then extended by the author whereby the formation of the fictitious crack is
calculated.
Though the basic idea appears to be sound, results obtained by using this model is
disappointing. By using the original method the even descending branch is missing, and the
stiffness of the struerure is miscalculated. By using the extended version where the length of
the fictitious crack is calculated at each loading step, the descending branch is obtained, and
some similarity between a numerical method is observed. However, the stiffness and the peak
load is estimated unsatisfactory. In the procedure suggesled by the author the initial stiffness
is predicted better, but the shortcoming is that the peak load not is predicted with sufficient!y
accuracy. lt is therefore concluded that more research is needed in order to improve the
method.
4.2 Reinforced Concrete.
To the authors knowledge only two analytical models which describe crack propagation in
reinforced concrete exist. The two models are conceptual different since the first model, the
Catpinteri model, is based on LEFM whereas the later is based on the fictitious crack model
and an elastic layer in the midsection. The two models are presented, compared and evaluated.
4.1.1 The Carpinteri Method.
This method is based strictly on LEFM and was introduced by Carpinteri (1981),
Carpinteri(1984), Catpinteri (1985), Carpinteri (1988) and later refined by Bosco et al.(1991).
M,f&l/2( e----~---- )M,f&l/2

~------~~~~------~
H
'
w t
1-0
Fig. 4.22 Beam segment from reinforced beam made of a perfectly brittie material.
Consider the reinforced concrete beam segment in Fig. 4.22, with the depth, d, thickness, t,
crack length, a, and the teinforcement area, Ar, in the depth, r, loaded with the extemal
moment, M. The concrete is assumed to be linear elastic, perfectly brittie deseribed by the
modulus of elasticity, E, and the fracture toughness, K c, and the teinforcement is assumed to
be rigid, perfectly duetile deseribed by the yielding strength, uy.
82
Thc rcinforcement is now removed and an extemal steel force, F3 , is introduced which
corresponds to the stress state in the reinforcement. The stress intensity factor, K, for the
beam segment is then written as
(4.146)
where, KM, is the stress intensity factor due to the extemal moment and, KF, is the stress
l
intensity factor due to the steel force. Thc stress intensity factors are given in Tada et al.
(1973) as
(4.147)
where
YM<ald) = 6(1.99(a/d) - 2.47(a/d)312 + 12.97(a/d)512 -

23.17(a/d)712 + 24.8(a/d)912)
(4.148)
2
YF (a/d,rla) • - -Yco11(ald,rla)
l lr:d
where, Yco 11(ald,rla), is given in eq.(4.135). The additional rotation, q,, ofthe segment due
to the crack, produced by the two loads are given by the campliances
t/>(ald,hla) = CMM<.ald) M + CMF (ald,rla)F3
l
(4. 149)
and the crack opening displacement, w(x), at the center of the reinforcement is given by
w(a/d,rla) • CF~ald,r/a)M - CF.F,(ald,rla)F3 (4.150)
The campliances can be derived by using energy principles similar to the one shown in section
4.1.3, Bosco et al. (1990)
(4.151)
ald
CMF •
l
2
~t
d JYM<w)YF (CJJ,r/d) dw
l
rd
before the steel is yielding the crack opening dispiacement is assumed to be zero
83
J.P. Ulflgær. Fncturo Mecbanics of Concrete
w=O :::0
(4.152)
M = r 11 (a!d r/a)
Fjl '
where
11
r. _ CF.F.(a/d, r/a)
(4.153)
r (a/d, la) - CF,JJ.a!d,rla)d
from which the force in the reinforcement caused by the extemal moment is calculated. In the
limit of plastic flow the moment becomes
(4.154)
where F:
= uy-4.,. By considering eq.(4.146) the fracture moment, MF (the crack is extending
when K =KJ can be determined as
MF _ 1 YF,(ald,hla) F[ (4.155)
-....."...,.~--+ N-
K,p3121 YM YM<ald) P F:
where, F/, is the steel force at fracture and, NP, is a brittleness number defined by
N = uyJ112 Ar (4.156)
P Kc A
where, A=td, is the cross sectional area. Before the steel is yielding , F/, is determined
through eq.(4.153), whereby the fracture moment becomes
MF l
= y
for
(4.157)
KJl31z1
y - F,
M IT
r
and when the steel is yielding, F[ = F;, viz
for (4.158)
It is hereby possible to consider virtual crack propagation, where the controiling parameter is
the crack length, a/d.
In the limit state of virtua1 crack growth the ulti.mate moment becomes
Mil = F:(l - rid)

Mil (4.159)
-~ = N (l - rid)
Ktfl312t P
Bosco et al. concluded hereby that themoment-rotation curve wil1 consist of three phases, a
84
linear elastic part until the fracture moment is reached given by eq. (4.155) • As the crack
length increases the moment decreases until the minimum moment, Mmin• is reached. If the
minimum moment is less than the ultimate moment the fracture moment wi1l eventually
increase until it reaches the ultimate moment, which the author agrees upon. However, as will
be shown this conclusion can not be made with the presented shape functions.
In order to be able to make comparisens between experiments and the model the load
dispiacement curves should be considered. The load-dispiacement curves can be determined
by considering the sum of the dispiacement due to the cracking, 8c, caused by, </>, and the
elastic deformation of a Timoshenko beam without a crack, aet as Suggested by Hededal and
Kroon (1991)
(4.160)
In order to check the Carpinteri modelit has been implemented on computer. The results are,
however, disturbed by an erroneousness shape function, Ycon• which in Bosco et al. is given
as
y (a/d,xla) = 3.52(1 - x/a) _ 4.35 - 5.28xla
con (l _ a/d)312 (l _ ald)312
(4.161)
+ { 1.30 - 0.30(x/a)312 + 0.83 - 1.76x/a}[l - (l -x/a)]

(l - (x/a)2) - 112
25
Yp (e/h • o.~) v..
20
yp {~/h = 0.10)
15 Yp (e/h • 0.15)
10
o
0.1 0.2 0.3 0.4 0.5 0.6 0.7
t= o/h
RELATIVE CRACK DEPTH
Fig. 4.23 The shape tunetions as presentedin Bosco et al. (1990).
By using this shape tunetion the fracture moments calculated for virtual crack propagation are
negative, and they are therefore not consistent. The shape tunetions used by Bosco et al. are,
however, also presented graphically and are shown in Fig. 4.23. The shape tunetions
calculated using eq.(4.161) and eq.(4.148), (4.161) are shown in Fig. 4.22, Fig. 4.25, and it
85
J.P. Ulfkjær. Fradure Mechanics of Concrete
151.0 _M_ ~ 1.0 t.l

; 0.8 K.: bhll2 ::1 0.8 KIC bh l/2
Np = 0.0 (.!) Np = 0.10
lo.6 :z
~ 0.6
m
~o. ~ 0.4
l
~
% z
~ 0.2
O åf', /&p,o i3 --------- ~F j&ptD
::1
o 5 10 15 20 25 30 i5 o 5 10 15 20 25 30
Dlt.cENSIONLESS ROTATION Dlt.IENSIONLESS ROTATION
(a) (b)
ffi2 l.O M ~ 1.0 _t.l_

~ 0.8 KIC bh 312 ::~ 0.8 KIC bhl/2
Np = 0.26 Np = 0.53
~~ 0.6
(.!)
:z
~ 0.6
.....
m
~ 0.4 f\ ~ 0.4
z
~ 0.2 '---=-=-.-·----;r-f!Jipro ~ 0.2
z
~
iS o 5 10 15 20 25 30 i5 o 5 1o 15 20 25 30
Olt.IENSIONLESS ROTATION Olt.IENSIONLESS ROTATION
(c) (d)
~ 1.0 _t.l_
~ 0.8 KIC bhl/2 ------~~=-~~7
~
i 0.6
~o.
:z
~o.
~o~~5~1~o~1~5~2o~2~s~30
Dlt.tENSIONlESS ROTATION
(e)
Fig. 4.24 Moment rotation curves for different brittleness numbers as presentedin Bosco et
al. (1990).
is seen that YF is completely different from the one used by Bosco et al. (1990) while YM is
•
exactly the same.
For the standard beam the fracture toughness is calculated according to Kc = JEGF. The
other materlal parameters are shown in Tabel4.2. By using the presented method a normalized
moment-rotation curve for the reinforced standard beam is shown in Fig. 4.26, where the
rotation is normalized to the first cracking rotation, t/Jo-
The initial shape of the moment rotation curve is equivalent to Bosco et al., however, the rest
86
250
200
150
100
y
M
50
oo 0.1 0.2 0.3 0.4 0.5 0.6 0.7

crack length, afd
Fig. 4.25 Shape functions accordmg to Tada et al. (1973).
Fracture Toughness, Kc [Nmm-312 ] 44.7

Remforcement area, Ar [mm2] 29
Relative remforcement position, r/d 0.05
Brittleness number, N11 0.26
Yielding strength, uv [N/mm2] 400
Tabel 4.2 Parameters which describe the remforced standard beam.
of the curve is compietety different. The minimum moment is much larger using the presented
indeed different from the one shown in Fig. 4.23, anditis actually larger than the ultimate
moment. The second kink on the curve indicates where the steel starts to yield, and after that
rotation the moment was supposed to decrease, and as is seen that is not the case. When the
steel is yielding eq.(4.158), is applied. Since the first term is decreasing, and NP is constant
y
themoment-rotation curve is highly dependent by the ratio F., see Fig. 4.27. lf the shape
YM
of the moment rotation curves should be as found by Bosco et al., then this ratio should be
decreasing or at least constant.
87
I .P. Ultkjær. Fracture Mechanics of Concrete
1
M
0.9
Kc td3fl
0.8
0.7
1s 0.6
.
o.s
1 "'
z
0.4
0.3
0.2
-
0.1
o o s 10 15 20 25 30
Norma1ized rotation, t9/ t9
o
Fig. 4.26 Normalized moment-rotation curves by using the Carpinteri method.
The author feel that the basic ideas describing the Carpinteri model is sound, but it has not
been possible to determine, moment rotation curves as deseribed in Bosco et al. (1990). It
would indeed be nice to see the mathematical expressions used to describe the shape fimetions
used by Bosco et al. (1990).
4.2.2 The Hededal, Kroon, UlflQær and Brincker Method.
This method is an extension of the Ulfkjær, Brincker and Krenk Method with remforcement
introduced in the elastic layer. Themodel was developed in junction with theMasters thesis
of Hededal and Kroon (1991) and this presentalion follows their approach.
Consider the beamin Fig. 4.28, reinforced with the remforcement area, Ar, introduced in the
depth, r. The remforcement is assumed to be linear elastic, perfectly plastic deseribed by the
modulus of elasticity, E" and the yielding strength, uy. The concrete is deseribed as in section
4.1.1. The approach of obtaining the governing equations is identical to the method without
reinforcement. Titus, three phases are considered, however, each phase is divided into two:
a) the steel remains elastic or b) the steel yields. The strain condition in the steel in the elastic
layer is assumed to be equal to the strain condition in the layer. The stress distribution and the
shape of the load dispiacement curve are shown in Fig. 4.29.
The strain distribution, E(y), in the midsection is obtained by considering similar triangles
88
4. Analytical Melhads in Concrete Fracture
11
10
\
9 YM
8
7
6
s
4
3
1
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7
crack length, a/d
Fig. 4.27. The ratio between the two shape lunetions used in the Carpinteri method.
, h ,
l
Fig. 4.28 The considered beam used in the Hededahl et al. method.
(4.162)
where, 8, is defined in section, 4.1.1., a", is normalized position of the neutral line, E", is the
ultimate concrete strain and, y, is the coordinate.
Phase l.
In this phase the concrete is assumed to be linear elastic. By taking horizontal equilibrium the
position of the neutral axis, a", is determined as
89
J.P. ffitlgær. Fracture Mechanics of Concrete
l l
l~ l~
1u.A.:
u(O)
l
1
l u(O)
l
1
u.A.:
l
u=O
u=~
u.A.
l l
1 l
p l 1
l l
l l
l l
l
l
l
Fig. 4.29 Sketch of the load-dispiacement curve for reinforced concrete where the three phases
are indicated.
"' .!. [ l + 2fpa,] for

2 l + SP
(4.163)
a" = .!.[1-p(fy_!.]
2 8 (f"
for
E8 • A,
where, f = -E , IS the flexibility ratio between steel and concrete, p = - , is the
td
remforcement ratio and, a, = ~, is the normalized position of the reinforcement. The
equivalent moment becomes
f1s(8)
p(8) = 8(4 - 6a") - 6-pa, (4.164)
(fil
where, f18 is the steel stresses which is determined by using eq. (4.162). Phase I ends when
the strain in the hottom equals, e", yielding the foliowing condition to the normalized rotation
8~-1- (4.165)
2a"
Thus, the moment rotation curve is a linear curve in phase I if the steel is not yielding, and
bi-linear if the steel yields.
90
Phase n.
In phase n the fictitious crack develops. The size of the elastic tensile zone is found by the
condition that the strain at the fictitious crack tip is, e", yielding '1 =
2~. The position of the
neutral axis is determined by taking horizontal equilibrium
a1 = (l + iP)(l - B)
± {l + 1P)2(1 - B)2 - (l + 2tpar)(l - B) + (!(l - B)(l + l"P)
a1 • (l - B) ± .!.(1 - B) [l + u, p] for e">e1

8 a" (4.166)
and the equivalent moment becomes
3
2aj8) ] (4.167)
1'(8) = 8 [ - 6aj.8) + 4
l -B
which is completely equivalent to the un-reinforced model, except for the last term taking the
account of the steel and the modified equation for description of the size of the fictitious crack.
4.5
:l. 3.5
so
f 3
2 2.5 40
~ 2
29
·a 1.5
~
20
1
0.5
o o
0.5 1 1.5 2 2.5 3 3.5 4 4.5 s
Rotation, ep
Fig. 4.30 Moment-rotation curves for the Hededal et al. model with different remforcement
ratios.
91
J.P. Ulflcjer. Fracture Mecbanics of Concrete
3
Size Scale • 0.25
0.5
1.0
3.0
o o o.s l l.S 2 2.5 3

Rotation, ep
Fig. 4.31 Moment rotation curves for the reinforced standard beam, with constant
remforcement ratio at different size scales.
Phase m.
In phase m the real crack starts to grow. The size of the elastic tensile zone is as in phase n:
;B. The condition that the strain at the real crack tip is, e c, and by considering similar
triangles the size of the fictitious crack is detennined as
al z
9
[l ~B] i (4.168)
which is the same as in the un-reinforced situation. The length of the real crack is determined
be requiring horizontal equilibrium
a =l + l'P - ~B ± l'P(l'P + 2(1 - a,)) +B [~r for e3 ~ e1

(4.169)
l
a=l--±
288
2p2-
(f.
(fu
l +B [ - l
28 2B8
]2
92
and the equivalent moment becomes
1'(9) =9 [4 - 6 [ ~8 + "] + 2(a + ~~~2; 2Ba3] + 6 ::P"r (4.170)
As in the Ulfkjær et al. method, is this method basedon the assumption that the curvature is
constant in the elastic layer, and the model is therefore derived for four point bending.
Modifications must, however, be made in order to calculate the elastic displacement, according
to the four point bending geometry.
In Fig. 4.30 normalized moment-rotation curves as calculated using the method for different
reinforcement areas. it is seen that the shapes are as deseribed under the carpinteri method,
which is the expected behavior. In Fig. 4.31 normalized moment-rotation curves for the
standard beam at different size scales with constant teinforcement ratios are shown.
The governing equations are simple and explicit, which makes the calculation speed very high
compared to the Carpinteri method.
The peak moment is seen to be dependent on both the reinforcement area and on the size of
the structure.
It is seen that there is a small kink on the curves. This kink is due to that the constitutive
relation for the steel also have a kink and it was shown by Hededahl and Kroon (1991), that
the kink on the moment-rotation curve disappears when the kink on the constitutive condition
is removed, e.g. by making the constitutive relation parabolic-perfectly plastic.
There have not been made any comparison between the Carpinteri method and the Hededal
et al. method since it is assumed that the Carpinteri method is erroneousness as it is presented
here.
4.2.3 Model Evaluation.
The first method presented was the Carpinteri method. The method is strictly based on LEFM
and that the reinforcement is rigid-plastic. The model requires numerical integration of double
integrals. The results presented here are, however, disturbed by an erroneousness results
compared to the ones presented in tbe original paper, and further condusions seems
meaningless, until is solved if the errors are due to the shape functions or due to the model
concept.
The second method, the Hededal et al. method, is based on the development of a fictitious
crack with a linear softening relation in an elastic layer around the midsection of the beam and
that the reinforcement is linear elastic-perfectly plastic. The results fits very well with what
was expected. The model is powerful in the sense that the governing equations are expUcit and
the calculation speed is therefore very high.
93
J.P. Ulfkjer. Fncture Mechanics of Concrete
ASTM (1974). E399-74: Method of test for plane strain fracture toughness of metallic
materials.
Biolzi, L., Cangiano, S., Tognon, G. and Carpinteri, A. (1989), Snap-back Softening
Instability in High-Strength Concrete beams, Materials and Structures, 22, pp. 429-436.
Bosco, C. and Carpinteri, A. (1990), Fracture Mechanics Evaluation of Minimum
Remforcement in Concrete Structures, in Application of Fracture Mechanics to Concrete
Structures, Edited by A. Carpinteri, Elsevier Applied Science, Politecnico di Torino, Italy,
pp. 347-377.
Carpinteri, A. (1981), A Fracture Mechanics Model for Reinforced Concrete Collapse, I.ABSE
Colloquium, Report of the working commission, Delft University of Technology, Delft,
Vol. 34, pp.17-30.
Carpinteri, A. (1982), Notch Sensitivity in Fracture Testing of Aggregative Materials,

Engineering Fracture Mechanics, Vol. 16, No. 4 pp.467-481.
Carpinteri, A. (1984), Stability of Fracturing Process in RC Beams, Journal of Structural

Engineering, ASCE, Vol. 110, No. 3, pp. 544-558.
Carpinteri, A. (1985), Scale Effects in Fracture of Plain and Reinforced Concrete Structures,
in Fracture Mechanics of Concrete: Structural Application and Numerical Simulation,
Editors: G. C. Sih and A. Di Tommaso, pp. 95-140.
Carpinteri, A. (1986), Size Effects on The Brittleness of Concrete Structures, Il Cemento, 4,

pp.375-388.
Carpinteri, A. (1988), Snap-Back and hyper-strcngth in Lightly Reinforced Concrete Beams,

Magazine of Concrete Research, Vol. 40, No. 145, pp. 209-215.
Carpinteri, A. (1989), Fracture Mechanics and Post-Peak: Structural Behavior of Plain and
Reinforced Concrete, in Fracture Toughness and Fracture Energy,· Test Methods for
Concrete and Rock, Editors: H. Mihashi, H. Takahashi and F.H. Wittmann, A.A.
Balkema, Tohoku, Japan, pp. 13-28.
Chuang, T.-J. and Mai, Y.-W. (1989), Flexural Behavior of Strain Softening Solids,
International Journal Solids Structures, Vol. 25, No. 12, pp. 1427-1443.
Foote, R.M., Mai, Y-W. and Cotterell, B. (1986), Crack Resistance Curves in Strain
Softcning Materials, Journal ofthe Mechanics and Physics of Solids, 34, pp.593-607.
Harder, N.A. (1991), A Theorem on Brittleness, University of Aalborg, ISSN 0902-7513-

R9126, pp. 1-17.
Hededal, O. and Kroon, I.B. (1991), Lightly Reinforced High-Strength Concrete, M.Sc
'Jhesis Dl University of Aalborg, pp. 1-88.
94
Uorca, J. and Elices, M. (1990a), A Simplilied Model to Study Fracture Behavior in

Cohesive Materials, Cement and Concrete Research, Vol. 20, pp. 92-102.
Llorca, J. and Elices, M. (1990b), Fracture Toughness, Crack Growth Resistance and
Ductilily of Fibre-Reinforced Ceramics, Structural Ceramics -Processing, Microstruerure
and Properties, Editors. Bentzen et al. , Risø National Labaratory, pp. 403-411.
Tada, H., Paris, P. and Irwin, G. (1985), The Stress Analysis of Stress Handbook, Del
research Corporation.
Timoshenko, S. (1955), Strengthof Materials, Part l, Elementary 'Iheory and Problems,

Third edition, D. Van Nostrand Company Ine., New York, p. 174.
Ulfkjær, J.P., Brincker, R. and Krenk, S. (1990), Analytical Model for Moment-Rotation
Curves of Concrete Beams in Bending Fracture Behavior and Design of Materials and
Structures, Proceedings of the 8th European Coriference on Fracture-ECF8 (Editor D.
Firrao), Engineering Materials Advisory Services LTD, Vol. II pp.612-617.
Ulfkjær, J.P., Krenk, S. and Brincker, R. (1992), Analytical Model for Fictitious Crack
Propagation in Concrete Beams, submitted for publication in Journal of Engineering
Mechanics.
95
5 EXPERIMENTAL DETERMINATION OF
FRACTURE PROPERTIES OF IDGH-STRENGTH
CONCRETE.
With the different numerical and analytical methods presented, it is possible to analyze plain,
lightly reinforced and other struerures of nonnal and high-strength concrete where the tensile
strength is a goveming factor. It is, however, necessary to investigate if the size effects
introduced in chapter 2 are present. For three of the models presented in chapter 2 the
fracture parameters are determined, the results are presented and the size effects observed
are commented.
As was deseribed in chapter 2, it is very difficult to determine the constitutive parameters

in the fictitious crack model and for high-strength concrete, only the fracture energy has been
determined, and only by using the RILEM recommendation developed by Hillerborg and co-
workers, RILEM 50-FMC (1985). Instead indirect methods can be applied, e. g. the fracture
parameters are determined through three-point bending experiment. Here the experimental
results, in fonn of load dispiacement curves, are compared with numerical results obtained
by using the direct sub-structure method, deseribed in chapter 3, and the fracture parameters
are then assumed to be those who minimize a certain function, which in some way describes
the difference between the experiment and the numerical result. Three of such tunetions are
tested on 8 different beam geometries, and two different optimization techniques have been
used. This method of detennining the constitutive parameters is evaluated, and ronelusions
are drawn.
In order to measure the crack profile and to detect eauses for size effects, dye experiments
are performed on ordinary specimens and on specimens which are saw cut out.
5.1 Size Effect Experiments.
5.1.1 Materials.
A high strength concrete similar to the one used at the Great Belt project in Denmark has
been used. The !argest aggregate size in this concrete is, however, only 8 mm. Tabte 5.1
shows the mix of the concrete and Tabte 5.2 shows the mechanical properties. The modulus
of elasticity, the cylinder compressive strength and the cylinder splitting strength were
determined using conventional procedures.
96
S. Bxperimaatal Determinalioa. of F neture Proportie~ of High-strenath Concrete
11 \
Servo Hydraulic
Machine
Fig. 5.5 Testing equipment for three-point bending testing.
Cement Fly Sillea Water Plast. Super Sand Gravel

ash fume (*) Plast.
312 44 29 122 1.56 8.58 614 1238
(*) Melamine-Naphthalene based plasticizer; Unitsare [kg/m3]
Table 5 .l Mix proportions.
The experiments were carried out on 8 different bearn sizes as shown in Tabte 5.3. These
geometries were chosen for two reasons. Firstiy, in order to be able to study statistical and
structural size effects separate!y, and secondly because of the limitations of the testing
equipment. The bearns were made out of three batches (40 l each) and beams of 8 different
sizes were cast on the same day. A total amount of 40 bearns were cast over a period of 4
weeks. The bearns were narned alter size, cast day and batch number (e. g. A-3a is beam size
A cast on day 3 from the first batch). The day alter easting the bearns were striped from
their molds and cured 7 days in water (37°C). Then a notch ofrectangular cross section was
saw cut in the beams and the bearns were Curther cured under water (20°C) for 7 days.
97
J.P. Ultkjer. Fncture mechanics of coocrete
Control Panel
/
••
Fig. S.6 Ph~ of reference bar.
Cylinder compres- Cylinder splitting Young's modulus

sive strength strength
Mean S.Dev Mean S.Dev. Mean S.Dev.
81,7 S.O% 5.5S 8.1% 42660 2.S%
Unitsare [MPa]
Table S.2 Meehanical properties oftested concrete.
Hereafter, the beams were storedin the laboratory at relative humidity of 100% at 20°C until
the day of testing.
5.1.2 Testing equipment and procedure.
The beams were submitted to three-point bending in a servo controlled materiais testing
system, see Fig. S .S. In order to measure the true beam detleetion a reference bar was placed
on each side of the bearn, and the beam detleetion was measured as the distance from the
98
5. Experim.ental Determination of Frøeture Properties of High-Strength Concrete
o o 100 200 300 400 500 600 700 800 900 1000
Dispiacement [J.Lm]
Fig. S. 7 Difference between the true dispiacement and the piston dispiacement for beam type
H-Sc.
Type Length Depth Width Notch depth

[mm] [mm] [mm] [mm]
A 840.0 so.o so.o 5.0
B 840.0 so.o 100.0 s.o
c 840.0 100.0 so.o 10.0
D 840.0 100.0 100.0 10.0
E 840.0 100.0 200.0 10.0
F 840.0 200.0 so.o 20.0
G 840.0 200.0 100.0 20.0
H 840.0 200.0 200.0 30.0
Table S.3 Sizes of the beams tested, the dimensions was all within +/- 2.0 mm.
load point to the reference bar using two LVDT's withabase of 20.0 mm and a sensitivity
of O.S V/mm, see Fig. S.6. The piston dispiacement was measured using the buiid-in LVDT
99
1.P. Ulfljær. Fraelure mecbanics of concrete
with a base of 5.0 mm and a sensitivity of 2.0 V/mm. The difference between the 'true
dispiacement and the piston displacement, which usually is assumed to be the beam
deflection, is seen in Fig. S. 7, for beam type H~Sc. The crack opening dispiacement was
measured using a cli~gauge withabase of2.0 mm and with a sensitivity of5.0 V/mm, see
Fig. 5.6. The load was measured using a 50.0 kN load cell with a sensitivity of 0.2 V/kN.
A schematic of the test set~up is shown in Fig. 5.8.
All signals together with the time, t, were recorded using an analog to digital converter and
an AT Personal Computer. The test was controlled by a feedback signal that included
contributions from both the piston dispiacement and the COD. The feedback signal, ~, was
Beam support~
Spherical BearingsY ~age Beam support
Support fitting~ · ,. Ø Q- Spherical Bearing
Clip Gage
~ fltting LVDT .
Fitting \ ~ 200 mm x 200 mm
. ~ ~VDT Specimen
Load Cell
Fig. 5.8 Schematic of the test set~up.
created by analog addition of the corresponding signals:

~=a~coD+fJ~p (5.171)
where, ~c00 , is the crack opening dispiacement and, ~ , is the piston dispiacement The
weight factors, a, and, {3, were chosen to a=lO.O, 5.0 and 2.5 and {3=1.0, where a is
decreasing with the beam depth.
The idea of using this combined signal as feedback is that each term in (5.171) plays a
dominant role in the different stages of the test. Before the fracture process starts the piston
dispiacement plays the dominant role. When the fracture process starts the COD is increasing
rapidly. In this way a stable test is obtained, see Fig. 5.9 , where the load-COD curve is
plotted together with the load~piston dispiacement curve and the feed~back signal for beam
100
S. Experimental Determination of Fracture Properties of High-Strength Concrete
4
x10
2.5
COD Piston
2
Feedback
~ 1.5
e
~ l
o.s
o o 200 400 600 800 100012001400160018002000

Dispiacement [llm]
Fig. 5.9 Displacement, crack opening dispiacement and feed-back signal for bearn type G-4a.
G-4a. It is obvious from this example that if the test was piston dispiacement controlled an
instability would occur immediately after the peak load.
The reference signal, a linear ramp, was generated with the same AT PC using a digital to
analog converter. The controlarrangement is shown in Fig. 5.10.
The measured load dispiacement curves are shown in Appendix A2, categorized according
to the geometry. The peak load and the dispiacement at peak load are shown in Appendix
A3.
5.2 Fracture Parameter Results.
Fracture parameters from three different models are calculated:
A: The bending'tensile strength according to Bernoulli (modulus of rupture).

B: The critical stress intensity factor Kc according to LEFM.
C: The modulus of elasticity, E, the fracture energy, Gp, the uniaxial tensile strength,
u", and the coordinates of the kink of a bi-linear er-w relation, according to the Fe-
model.
101
J.P. Ulfkjer. Fracturo medumics of coocrete
D/A converter t-
D A/D converter J
IBY AT
'
AT bus
CUp gage,clca uca• ccac5c t-

~ ~
l l
l : Ub•Cb6b
Load cell,4Sp_, t
Acluator~ ~LVDT,cSb
H
Up•CpcSp
ou mpply ~rLVDT,cSJ
~ : Ut•CtcSt
:UOVolt
Analog addition
Ut•aUJ+bUCG
Servo controller
...
Control
I•C(Ur-Ut)
l
~
lUr(rampfunctlon)
tiDVolt
Fig. 5.10 Control Arrangement.
5.2.1 1be Modulus of Rupture.
The modulus of rupture is calculated as the bending tensile strength of a Bernoulli-Euler

beam
3Fmaxl
CTm = ----=-
2(d - ai)2t
(5.172)
where, Fmax• is the peak load when the load dispiacement curve is adjusted for the beam
weight. The modulus of rupture for all the beams are shown in Appendix A3, together with
the mean values, the standard deviation and coefficient of variation. In Fig. 5.11 the mean
value of the modulus of rupture is drawn as a tunetion of the beam depth, for the different
beam thickness. It is seen that except for the smallest beam (d = 50 mm, t = 50 mm) the
same trend is seen as in the Reagel and Willis (1931) experiments, thus, the modulus of
rupture is decreasing with the beam depth. The decrease in modulus of rupture by doubling
the beam depth is 4% for the thielmesses 50 mm and 100 mm and 10% for the beam
thickness 200 mm which is of the same order as in the experiments by Reagel and Willis
(1931).
102
S. Experimental Determination of Fraelure Properties of High-Strength Concrete
8
7.8
oe
. 7.6
l't:
7.4
7.2
7
t•200mm
~ 6.8
~
t• SO mm.
6.6
6.4
6.2 60 80 100 120
140 160 180 200
40
Beamdep~ d
Fig. 5.11 Mean values of modulus of rupture as function of the beam depth for the three
beam thickness.
The reason for the unexpected behavior of test series A could be that the specimens have
been too small to make a homogeneous structure of the specimens, however, visual
inspection of the specimens did not indicate any difference in the structure of the smaller
specimens than that of the larger. lt is therefore assumed that the decrease in strength of the
smallest beam size is due to some edge effect, which e.g. is introduced by the mould.
5.2.2 1be Fracture Toughness.
The fracture toughness, Kc is calculated according to ASTM:
K =3Fmaxl{if
c 2 bd2 d
[a] (5.173)
where
2 3 4
f [:] =1.93-3.07 [:] +14.53 [:] -25.11 [:] +25.8 [:] (5.174)
Fmax is the peak load and d,t and a is the beam depth, thickness and notch depth, respective-
ly.
103
The critical stress intensity factor for each beamis shown in Appendix A3, and the mean
values are plotled in Fig. 5.12. It is seen that the fracture toughness is strongly increasing
with the beam depth, and is almost independent of the beam depth. The increase by doubling
the beam depths is 53% for t=50 mm, 46% for t=lOO mm and 65% for t=200 mm. Thus,
a distinct size effect is observed, when LEFM is used.
2.6
2.4
2.2
,.......
2
l ~ 1.6
1.8 t •lOOmm
~u
1.4
1.2
t • 50 mm
l
0.8 40 60 80 100 120 140 160 180 200
Beam depth, d [mm]
Fig. 5.12 Mean value of the critical stress intensity factor.
5.2.3 Fracture Energies According to RILEM.
According totheRILEM recommendation, RILEM TC-50 (1985), the ftacture energy should
be determined on a specimen with the dimensions l = 800 mm, d = 100 mm, t= 100 and
initial notch depth a; = SO mm, the maximum aggregate size should not be larger than 16
mm. In this test series the initial notch depth ratio was chosen to be only 1110 of the
normalized beam depth, because as stated in chapter 2, LEFM is only applicable for small
notch depth. By using this reduced notch depth the amount of energy dissipated outside the
ftacture will increase. Therefore it can be expected that a certain size effect will be measured
by using the RILEM method, since the energy dissipated in the bulk is not taken into
consideration by using this method.
Different areas are calculated in connection with the RILEM method. Consider the load-
dispiacement curve in Fig. 5.13. The experiments are usually terminated before the load is
decreased to zero, due to the faet that the piston dispiacement speed is very slow at the
104
S. Experimeotal Determination of Fracture Properties of Hiah-Strength Concrete
descending branch. The experiment wi1l therefore end at the load, F1 , and the corresponding
displacement, ~1 , and the remaining area under the load-dispiacement curve must be
estimated.
According to the Ulfkjær et al. model the descending branch in Phase m is deseribed by
(5.175)
the remaining contribution to the fracture energy, A 1 , can then be determined by

oot?-__
A
l
... l
t(d - O;)
I...!.d8
~2
(5.176)
l
By applying the condition
the remaining fracture energy becomes

F, • [::r (5.177)
Ft~t
Åt,., ~~~ (5.178)
t(d - a;)
The fracture energies (the sum of the area under the measured curve and the remaining area
divided by the ligament) and the remaining fracture energy, A 1 , for each experiment are
shown in Appendix A3. The mean value of the fracture energies for each geometry is shown
in Fig. 5.14. It is seen that the fracture energy is increasing with the beam depth, and is
almost independent of the beam thickness. The increase by doubling the beam depth is for
t= 50 mm 19%, for t= 100 mm 8% and for t= 200 mm 16%. The fracture energy for test
series A, is quite low, which is due to the experiment A-Sa, where the fracture energy is
very low. Visual inspection of the cracked surfaces did not indicate that this specimen in any
sense was weaker than the others. The reason for the size dependency of the fracture energy
is attributed to the faet that energy dissipation in the bulk is not taken into consideration.
5.2.4 Fradure Parameters Accordin& to the Fietitlous Crack Model and Data fitting.
The fracture energy, the tensile strength, the modulus of elasticity, the coordinates of the
kink onabi-linear er-w relation were estimated by salving the minimizing of an objective
function which in some manner describes the difference between the measured load
dispiacement relation and one calculated using a numerical method. Three different objective
functions and two different optimization procedures were tested in order to obtain the best
estimate of the constitutive parameters. The functions will be presented in the order they
were tested by the author, and the shortcomings of each function will be given.
105
J.P. Ulflgær. Fraelure mechanics of concrete
Objectiye fimetion 1:
The first objective tunetion to be tested were
(5.179)
z= [uu *c E f
F/
where is the experimentalload deflection curve at the applied dispiacement ~i' which were
z
calculated at equidistant dispiacements and is the state vector. No bounds were imposed
on the elements in the state vector. P/< z)
is a load deflection curve determined using the
direct sub-structure method (DSS-method),Brincker and Dahl (1988), deseribed in chapter
3. Using the finp element method for determination of influence coefficient, four-node
Fig. 5.13 Areas calcuiated in order to determine the fracture energy, using the RILEM
method.
element were chosen and the number of nodes in the midsection were 21. The u-w relation
u
was assumed to be bi-linear with the normalized kink-coordinate at (0.16,0.31). u, .Vc and E
are the estimated uniaxial tensile strength, critical crack opening dispiacement and the
modulus of elasticity modulus respectively. The term (F/;2 is a weight, that emphasizes the
importancee of the peak.
The optimization problem was solved using the Nelder and Mead algorithm, see e. g. Gill et
al. (1981) or Vanderplaats (1984) on a 386 20 MHzPersonal Computer implemented in a
MATLAB shell, MATLAB (1989) and it took about l hour for each optimization. However,
in order to avoid local minimaseveral runs were necessary. Using this objective lunetion
106
5. Experimental Determination of Fracturc Propertics of High-Strength Concrete
145
140
1
g
135
t • lOOmm
130
r.:f 125
r 120
115
t • 200 mm
j 110
105 t • SO mm
100
95
40 60 80 100 120 140 160 180 200
Beam dep~ d [mm]
Fig. 5.14 Fracture energies determined according totheRILEM recommendation.
initially rather good results were obtained and it was concluded that the method was rather
good. However, as the results became more, and more scattered and as it was observed that
the fltting results were very poor, especially at the descending branch. It was therefore
concluded that another objective lunetion was needed, so that the descending branch could
be fitted better.
Objective tunetion TI:

The second objective lunetion is similar to the first, however, the normalized eoordinates of
the kink point (w1/wc ,&1/fr,J of the strain softening relation is taken into account
(5.180)
z.,. [ fr., (;F E wllwc frtlfrl ]T

and the state vector is subjected to the following constraints
107
J.P. Ulfkjær. Fractwe mecbanics of concrete
o< < 10.0 MPa
o< < 80,000 MPa (5.181)

O < li~c < 0.5 mm
O < lil1/ltlc, &1/&11 < 0,5
F/
Another important difference is that the loads 1, were calculated at equidistance points at
the load-dispiacement curve and not just on the dispiacement curve. By doing so the
descending branch was weighted equally as much as the ascending branch, and even snap
back is taken into account by this method. The number of points, N, was set to 350. In order
to increase the calculation speed the objective tunetion was implemented using the C
programming language, and instead of the Nelder and Mead scheme the NLPQL scheme
developed by Schittkowski (1980) implemented in the programming language FORTRAN was
used.
The NLPQL algorithm is a very effective method where each iteration consists of two steps.
The first is the determination of a search direction by salving a quadratic optimization
problem formed by a quadtatic approximation of the Lagrange tunetion of the non-linear
optimization problem and a linearization of the constraints at the current design point. The
second step is a line search with an augmented Lagrangian merit tunction, Enevoldsen et al.
(1990). NLPQL requires estimates of the gradients of the objective tunetion and the
constraints, and is in this case done numerically.
In this way a more complicated objective tunetion with two additional parameters is used
with almost the same calculation time as the previous tunction. However, as in the case of
the first objective function it is necessary to try more than one start guess, in order to find
the true minimum.
In most cases this object function yields very satisfactory results, however, it is in some
cases not possible to obtain an acceptable agreement at the tail at the descending branch,
which usually results in a too large fracture energy. An example of such an optimization
result is seen in Fig. 5.15.
Objectiye filDetion m:
In order to solve the above outlined problem a third objective tunetion was designed
r-
2 2
N
(P/I(z)-F/12 -'-
F.II t/
E
i•l Fmax
+ '
8max
(5.182)
z. . [ ;," GF E lYtflYc ;,1/&" ]T

subjected to the following constraints
108
S. Experimental Determination of Fracture Properties of High-Strength Concrete
1600
1400
Experiment
1200
Numerical
~ 1000
J 800
600
400
200
o o 100 200 300 400 soo 600 700 800 900 1000
Dispiacement [p.m]
Fig. 5.15 Optimization result where the second objective tunetion fails.
1600
1400
Experiment
1200
Numerical
1000
~
§ 800
~ 600
400
200
o o 100 200 300 400 500 600 700 800 900 1000
Dispiacement [p.m]
Fig. 5.16 Optimization result from using the third objective function.
o< CTu
A
< 10.0 MPa

A
o< E < 80,000 MPa (5.18fag

o< lYc < 0.5 mm
o< lYtllYc, u1/(1u < 0,5
I.P. Ulfkjær. Fracture mecbanics of concrete
which are equivalent to objective function n, but weighted with the square of the normalized
dispiacement
t/
-'-
]2 . By using this objective function the optimization result in Fig. 5.15
[ 8max
changes to Fig. 5.16.
-0.00 0.33
l l l
l l l
l l l
l l l
l l l
0.67 l l l 0.67
l l
l l 11
l l
l l ~
l l l
0.33 l l l 0.33
y l
l
l
l
l
l
l
l
l
l
~ -0.00
l
l \ l
l
l l \
l \ l'
l
l
-0.00
\ \ \ l
ll '\ '\ \
\ \ \
-0.33 \\ ' \
\\
\
\
\ -0.33
\ \ ' \
\
'\ ''
' \ \
. ..
\
' .. ..
-0.67
\
.. ''........................... ' ' ...... . . -0.67
..... ............ ... ...
Fig. 5.17 Contour plot of the object function for beam A-4a.
A contour plot of the objective function for beam A-4a is plotted in Fig. S .17 where the
optimal values of E = 43086 MPa, ltr1/ltrc = 0.14 and 1/C114 = 0.22 are used. The u
corresponding surface plot is shown in Fig. 5.18. The plot consist ofa square grid of 51*51
points and it took over 26 hours to perform the calculations on a 386 33 MHz Personal
Computer. The solid contour lines are plotled with equidistance o f 3, 000 and the dashed
contour lines are plotled with equidistance of 30,000. The two variables x and y are defined
sothat -1.0 at the x-axis corresponds to (fil= 10 MPa and 1.0 corresponds to f114 = 10.0 MPa,
at the y-axis -1.0 corresponds to wc= 0.05 mm and 1.0 corresponds to wc= 1.0 mm. The
minima of the function is at (x,y) = ( 0.08, -0.84) and is equal to 107.96 the maximais at
(x,y) = (-1.0, -1.0) and is equal to 318260. The minimum corresponds to that wc = 0.09
mm and (fil= 5.86 MPa, the values shown in Appendix A3 found by NLPQL are wc= 0.09
mm and (fil = 5.42 MPa.
It is seen that the objective function is 'banana' shaped with a very deep valley. At the
bottom of the valley the objective function is very flat for constant brittleness numbers, and
a lot of local minima are present. It is therefore very difficult to find the true minimum of
the objective function and often several start guesses are necessary in order to get a good fit.
110
S. Experimental Determination of Fneture Properties of High-strenath Concrete
.,.,
_.,
,,,
t·.~,
..
.,.,
..
Fig. 5.18 Surface plot of the object tunetion for beam A-4a.
The optimization results for all the beams are shown in Appendix A4, and the estimated
parameters are shown in Appendix A3.
The mean value of the estimated fracture energies are shown in Fig. 5.19, and the trend
observed in the RILEM method is also observed here, whieh is due to the faet that the titting
is made so the eurve wi1l be as equal as possible and that the energy dissipation in the bulk
is not talren into consideration in this method. The increase in the fraeture energy by
doubling the beam depth is 15% for t= 50 mm and t= 200mmand 10% for t= 100.
The mean values of the uniaxial tensile strength, cr.,, are shown in Fig. 5.20. No distinet
trend is observed from the estimated uniaxial tensile strength. The tensile strength is smallest
for the small specimens, are then inereasing and then decreasing a little. It is observed that
the uniaxial tensile strength is smaller than the modulus of rupture, whieh is in agreement
with the results presentedin ehapter 4.
The mean values of the normalized kink point of the softening relation are shown in
Fig. 5.21. It is also seen that the normalized kink coordinate is decreasing with the beam size
and that the values are lower that that of NSC {(0, 16,0.31)), indicating that HSC is more
brittie than NSC.
111
I.P. Ulflgær. Fraelure ~nics of concrete
I 14S
~ 140
c~ 13S
.. 130
i 125
120
J
]
11S
110
J 105
100
40 60 80 100 120 140 160 180 200
Beam depth, d [mm]
Fig. 5.19 Estimated fracture energies.
..... 6.5
t•200mm
l t>~'
6
t
5.S
s
1 4.5
t - 100 mm
l 4
40 60 80 100 120 140 160
Beam depth, d [mm]
180 200
Fig. S .20 Estimated uniaxial tensile strength.
When this work was finished the author became aware of a work by Roelfstra and Wittmann
(1986) and Bruhwiler et al. (1990) who also have determined the fracture parameters through
112
5. Experimeotal Determination of Fraelure Properties of High-Strength Concrete
0.3 0.3
0.28 -Y; l wc 0.28 al l &u
0.26 0.26
0.24 0.24
0.22 t•SOmm 0.22
0.2 0.2
0.18 0.18
0 .16 0.16
0.14 0.14
0.12 0.12
0.1 0.1 so
so 100 l SO 200 100 l SO 200
Fig. 5.21 Estimated results for the normalized lånk coordinate of the softening relation.
a method similar to theoneunder objective tunetionIon NSC. They did, however, not use
a weight on the objective function, and their estimation of the peak load is not as that good
as with objective tunetion I.
5.2.5 Conelosions on Size Effect Experiments.
For all the three models it were seen that there were a significant size effect. The size effect
on the modulus of rupture was of the same arder as that of experiments performed by Reagel
and Willis (1931). The increase on the fraeture toughness with the increasing beam depth was
very large anditis concluded that LEFM, is not applicable for HSC concrete of labaratory
size. The size effect on the fraeture energies wase also significant, and the RILEM method
for detennining the fraeture encrgy seems inappropriate. By using the data fltting technique
the uniaxial tensile strengths and the lcink coordinate on the descending branch were
determined and it was observed that the tensile strength was both decreasing and increasing
with the beam depth. The normalized lånk coordinates were decreasing with the beam depth.
For all parameters it was concluded that they were independent of the beam thickness. It
appears, thus, that edge effects are present, since the Weibull in same way is outbalanced.
113
•
J.P. Ulftgær. Fracture mernani.cs of concrete
5.3 Dye Experiments.
In order to determine the crack profile and to investigate why the thickness of the beam does
not influence on any of the determined fracture parameters, a series of dye experiments
similar to the ones performed by Swartz and Refai (1989) has been performed, with emphasis
on studying edge effects.
In the method developed by Swartz and Rafai (1989), two Aluminum sheets were glued to
the beam using silicone. The specimen was then loaded beyond the peak load, where after
the dye, Blue Vanish (a fluid soap), was poured into the reservoir. The beamisthen load·
cycled, with the maximum load being app. 113 of the peak load. The reservoirs were then
removed and the beam was split whereby the cracked profile corresponding to the unloading
load is revealed. The crack was then measured immediately after testing, since the dye
disappears alter about an hour. The crack profile measured by Swartz was rather U-shaped
indicating the possibility of the existence of a weak zone at the edges of the specimen. The
reason for this weak zone could be that the structure of the concrete is not the same at the
edges of the beam. Another reason could be that the stress intensity factor for a crack on the
edge of the struerure is up to a factor of 2.0 of that of an internal crack. That means that a
crack at the surface will propagate for a load which is only the half of that for an internat
crack. A third explanation could be that the used dying technic is responsible for the shape
of the measured crack profiles.
Almost the same testing procedure test set-up as used by Swartz is utilized in this
investigation, see Fig. 5.22. Though, the dye used was instead Rotamin, a red colored dye
used in women lip-sticks, and is in the concentrated form used here almost impossible to
remove. A specimen which was dyed approximately a year ago, placed in the authors office,
has not changed at all since the day of dyeing!. The drying procedure in this investigation
was done by using a hot air canon whereby the drying time was very short. Also the
reservoirs were not glued on to the specimen, but were hinged to the top of the beam, and
then a littie silicone was used to make the connection waterproof. It is thereby assumed that
the stiffness of the beam has not changed, due to the reservoirs.
In order to check if the crack profiles change as the structure of the surfaces changes, four
specimens which were saw cut out were produced, and four reference specimens of beam
type D were cast. The procedure just outlined was then followed. The crack profile measured
and drawn. The profiles are shown in appendix AS, and as seen it is not possible to see any
difference in the measureel crack profiles.
5.4 Conelosions of Chapter S.
114
S. Experimeotal Determination of Føcture Properties of Hip-Streugth Concrete
Fig. 5.22 Test set up used in the dyeing experiments, with the aluminum reservoir.
An experimental investigation on 40 high-strength concrete beams in three point bending have

been perfonned. The beams consisled of eight different geometries where the beam depth and
the thickness were varied. The experiments were made in a cernbination of piston
displacement and crack opening displacement control, whereby the same controiling
parameter was used throughout the entire experiment. The fracture parameters according to
three different models were determined. That is the Bernoulli beam theory, linear elastic
fracture mechanics, and the fictitious crack model.
Explicit fonnulas are available for the two first theories. The results for the Bernoulli beam
theory showed that the bending tensile strength or the modulus of rupture is decreasing with
the beam depth, but not with the beam thickness. The decrease was approximately 10% by
doubling the beam depth. The fracture toughness determined according to LEFM displayed
a significant size effect, as the increase was approximately 50~ by doubling the beam depth.
Also here there was no significant size effect by increasing the beam thickness. These results
are parallel to other similar investigations.
Two different methods were used in connection with the fictitious crack model. The RILEM
method and a novel method were the fracture parameters determined through curve fitting.
The fracture energy calculated by using the RILEM method displayed a size effect when the
beam depth was increased by 15% No size effect was observed by increasing the beam
115
J.P. Ulfkjær. Fncturo mechanica of concrete
thickness. The data titting was done by minimizing a tunetion which describes the difference
between the results from a numerical method and the experimental results. Three such
lunctions, objective functions, were testedand the third one was the peak load and the tail
were weighted gave the best results. The method is disturbed by the faet that the objective
lunetion has a very deep flat valley with many local minima, which malres it very difficult
to find the true minimum of the lunetion even when powerful optimization techniques are
applied. This problem malres it also very time consuming to determine the fracture
parameters.
The materlal parameters determined using the data titting technique showed the size effects
as the other related parameters. The normalized kink coordinates on the softening relation,
which only can be used by using this technique were decreasing with the beam size and were
in general smaller than those of normal strength concrete indicating that normal strengtil
concrete is more brittie than high strength concrete.
For all the parameters it was observed that the thickness of the beam was not a goveming
factor and instead extra experiments were performed in order to check for edge effects due
to the mould. Dying experiments were performed on standard specimens and on specimens
saw cut out. No difference in the cracking profiles was observed.
ASTM (1974), E399-74: Method of test for plane strain fracture toughness of metallic
materials.
Bruhwiler, E., Wang, J. and Wittmann, F.H. (1990) Fracture of AAC as Influenced by
Specimen Dimension and Moisture, Journal ofMaterial in Civil Engineering, ASCE, Vol.
2, No. 3, pp. 136-146.
Enevoldsen, I., Sørensen, J.D. and Sigurdsson, G. (1990), Reliability·Based Shape

Optimization Using Stochastic Finite Element Methods, 3rd IFIP WG 7.5 Working
Conj'erence, Berkeley, USA, pp. 75-89.
Gill, P.E., Murray, W. and Wright, M.H. (1981), Practical Optimization, Academic Press,
Ine.
Hillerborg, A.(1985), Influence of Beam Size on Concrete Fracture Energy Determined

According to a Draft RILEM Recommendation; Repart to RILEM TCSO-FMC. Rapport
TVBN-3021, pp. 1-15.
Karihaloo, B.L. and Nallathambi, P. (1990), Final Repart of Sub-committee A (Chairman

B.L. Karihaloo) Notehed Beam Test Mode I Fracture Toughness RILEM TC89
116
S. Experimeotal Determination of Fracture Properties of High-Strenath Concrete
FMT,(Cbairman S.P. Shah), 1-86.
MATLAB (1989), MATLAB™ for MS-DOS Personal Computers, The MA11I Works Ine.,
21 Eliot Street, , South Natick MA 01760, USA. pp 1-365.
Reagel, F.V. and Willis, T.F. (1931), The Effect of the Dimensions ofTest Specimens on
the Flexural Strengthof Concrete, Public Roads, Vol. 12, No. 2, pp.37-46.
RILEM 50-FMC (1985), Determination of the Fracture Energy of Mortar and Concrete by
Means of Three-Point Bend testson notehed beams, Materials and Structures, Vol. 18,
No. 106, pp. 285-290.
Roelfstra, P.E. and Wittmann F.H. (1986), Numerical Method to Link Strain Softening with
Pailure of Concrete, in Fracture Toughness and Fracture Energy of Concrete, Edited by
F.H. Wittmann, Elsevier, pp. 163-175.
Swartz, S.E. and Refai, T. (1989), Cracked Surface Revealed by Dye and its Utility in
Determining Fracture Parameters, in Fracture Toughness and Fracture Energy - Test
Methods for Concrete and Rock, Eds. .Mihashi et al., pp.509-520.
Schittkowski, K. (1985), A FORTRAN subroutine Solving Constrained Non-Linear

Programming Problems, Annals of Operation Research, Vol. 5, pp. 485-500.
Vanderplaats, G.N. (1984), Numerical Optimization Techniques for Engineering Design,

McGraw-Hill Book Campany.
117
1.P. mflcjer. Fracture mccbaoicø of coocreto
6 CONCLUSIONS.
This chaptcr has three sections. The first contains a summary of the different chaptcrs of the
thesis. The second section gives an overall conclusion of the thesis. This conclusion is based
on the evaluation made at the end of chapters 2-5. Finally future research needs and
applications are presented and discussed.
6.1 Summary of Thesis.
Chapter 1
Chaptcr l contains tlle introduetion to this thesis. The scope of the thesis is partly to
investigate different numerical and analytical models based on fracture mechanical ideas,
which are able to predict size effects, and partiy to perform an experimental investigation on
high-strengtil concrete.
Chapter 2
A description of the factors which influence the strengtil and cracking of concrete and high
strength concret.e is made. Then basic linear fracture mechanics is outlined followed by a
description and evaluation of the models used to describe concrete fracture in tension. The
chapter ends with a description of the different types of size effects. Three examples which
discuss the two terms 'size effect' and 'brittleness' and the importance of a stiff test rig.
Finally some brittleness numbers are defined.
Chgpter 3
In chapter 3 the most well-known numerical metllods which use the fictitious crack to
describe fracture in concrete are presented. Two of tlle metllods are cernbined into a power
method which is stable for all brittleness numbers and which is able of calculating the entire
load-dispiacement curve even for very duetile beams. This method is used extensively in the
rest of tlle thesis.
ChiPter 4
Since analytical metllods are very time consuming different analytical models have been
developed. Three methods for plain concrete are presented, where one of the methods is
developed by the autllor. The metllod is based on three different fracture models. Also two
models applicable for lightly reinforced concrete are presented.
118
S. Experimeatal Determination of Fracture Properties of High.Streogth Concrete
An experimental investigation is performed in a elosed-loop testing setup where high-strength

concrete in three-point bending is tcsted. In order to determine the materlal parameters in the
fictitious crack model an estimation procedure based on the solving of an optimization
problem is developed. The materlal parameters and the load-dispiacement curves obtained
by using this procedure are presented. Different size effects are observed, anditis concluded
that cdge effects play a dominant role. Dyeing experiments were performed on ordinary
specimens and specimens which are saw-cut out.
6.2 General Conclusions.
One of the major problems solved in this thesis is the stabillty problem deseribed in chapter
3. The direct sub-structure method is thereby complete, and is a strong tool, when analyzing
structures of quasi brittie materials. The only limit of the method is now the number of nodes
used in the midsection. When a fictitious crack develops, the normalized size of this zone is
dependent on the brittieness of the structure. The more brittie the structure the smaller the
fracture process zone, and consequently many nodes are necessary for describing this zone.
The analytical methods deseribed in chapter 4 show very different results. The analytical
method developed by the author, R. Brincker and S. Krenk, seems to be themost promising.
This method is capable of predicting the size effect on the modulus of rupture. Purther more
it is based on basic principles and the goveming equations are explicit and simple. These
properties of the model make it a very powerful tool, which is applicable for the designing
engineer. The method is also extended to reinforced concrete, where the results look very
promising.
The large experimental investigation on high-strength concrete seems to have been too small
However, somegeneral results were obtained. It is observed that the decrease in the modulus
of rupture with the increase of beam depth is of the same order as that of normal strengtil
concrete. The fracture toughness is increasing with the beam depth, making LEFM
inadequate for high-strength, even though high-strength is considered to be a brittie material.
The data titting performed by finding the minimum of three different tunetion which
describes the difference between a numerical and an experimental load dispiacement curve
turned out to be very time consuming. If the constitutive parameters are wanted for high
strength concret this method is then the only applicable. The method is, however, not
recommendcd.
6.3 Future Perspective.
The foliowing items and problems are suggestions for future research in the field of fracture
119
mechanics of concrete
• Developing analytical methods where the softening relation is bi-linear.

• Perform experiments with high-strength where the beam depths and span are
varying with more than a factor 5
• Development of direct experimental techniques whereby the softening relation
can be determined for ultra brittie materials.
• Development of simple models, where the bulk dissipation is included.
• Development of models, which describes the bonding between steel and
concrete.
• Development of stochastic methods, whereby the reliability and lifetime of
cracked structures can be calculated.
120
APPENDIX Al
NOTATION.
The foliowing symbols are used in this thesis:
a - Crack length.
ae - Effective crack length.
aef - Effective crack length.
~ - Fictitious crack length.
ai - Initial crack length.
llo - Distance between nodes.
B - Brittleness number.
Bo - Materlal geometry dependent parameter in the size ejfect law.
c - Compliance, or greens function.
- Injluence coefficients.
~bD - Crack tip opening displacement.
d - Characteristic size of struerure (e. g. beam depth).
:; -
=
Length of vinual cut in structure.
Modulus of elasticity.
Edisp - Dissipated energy.
F - Force.
Fmax - Peak load.
/f) - Materlal function.
g f) = Geometry function.
G - The energy reJease rate.
Gc - The crack drlving force.
GF - The fracture energy.
l - Moment of lnertia.
K - The stress intensity factor.
Kc - The critical stress intensity factor.
l - Length of structure.
tJ - Elongation of structure.
LEFM - Linear elastic fracture mechanics.
R - Resistance towards crack growth.
r - Position of reinforcement.
sJ - Nodal force.
s" - Uldmate nodal force.
t = Thickness of structure.
u, v - Dispiacement of vinual suiface.
w - Crack opening displacement.
Wc - Crltical crack opening displacement.
121
J.P. Ultkjer. Føcture medtanics of coocre&e
x,y - CaTtesian coordi1lllles

a - Nonnalized crack length.
af - Nonnalized jictitious crack length.
ar - Nonnalized position of reitiforcement.
a" - Nonnalized position of neutral axis.
13 - Brittleness number.
'Y - Crackpath.
~ - Displacement.
~1M% - Dispiacement at peak load.
E = Strain.
Ee - Linear elastic strain.
E,. - Nonlinear part of strain.
Epeølc - Strain at peak load.
Eru - Residual strain
f/> f) - Energy density.
n - Total potential energy of the considered system.
De = The elastic energy conrent in structure.
DF - The potential of external forces.
Dg = The lånetic energy.
DC - The fracture potential.
- Position of neutral axis.
"
(l - Stress, or arbitrary external load.
(lij - Stress tensor.
um - Modulus of rupture.
(III - Tensile strength.
u, - Yield strength.
Ø,r - Polar coordi1Ulles.
j
122
APPENDIXA2
LOAD DISPLACEMENT CURVES.
On the foliowing pages the load dispiacement curves are plotted. The beams are plotted
according to beam size.
123
a_1c, a_2a, a_4a and a_Sa
700 Max a_1c = 543.1

Max a 2a = 663.7
600 Max a_4a = 683.8
Max a_5a = 616.9
500
-z
u
1:
400
& 300
200
100 -.
100 200 300 400 500 600 700 800
Dispiacement [m E..()]
Figure A2.1: Load dispiacement curvcs for bcam series A.
b_2a, b_4aand b_Sa

1800
Max b 2a = 1197.9
1600 Max b 4a = 1514.9
..--i
-' • Max b_5a = 1701.0
1400
.'
// \'
1200 ·' [
~
u
1000 li
1: •.. \
& 800 \,~ ..
..,\
.............
600 .........
·-.
~ .............. ..
............
200 300 400 500 600 700 800 900
Dispiacement [m E..()]
Figure A2.2: Load dispiacement curves for beam series B.
124
Appendix A2. Load Dispiacement Curves
c_1c. c_2a, c_3a, c_4a and c_Sa
3000
Max c_1c = 2036.6
Max c_2a = 2843.7
2500 Max c_3a = 2831.5
Max c_4a = 3031.6
z-2000 - Max c_5a = 3001.8
.......
lU
l:! 1500
&
1000
100 200 300 400 500 600 700 800 900
Dispiacement [m E-6]
Pigure A2.3: Load dispiacement curves for beam series C.
d_1b, d_2a, d_3a, d_4a and d_Sa
6000
Max d_1b = 4502.9
Max d_2a = 5601.9
5000 Max d_3a = 4592.6
Max d_4a = 6116.8
z 4000 - - Max d_Sa = 6308.3
lU
l:!
& 3000
2000
1000
100 200 300 400 500

Dispiacement [m E-6)
Pigure A2.4: Load dispiacement curves for beam series D.
125
J.P. Ulfkjær. Fracture mecbanjca of c:oncrcte
e_ 1c, e_2b, e_3b, e_4b and e_Sb
12000
Max e_lc = 9258.6
Max e_2b = 9542.9
10000 Max e_3b = 10747.6
Max e_4b = 12125.4
z 8000 - Max e_Sb = 12034.6
4000
·-.
--- ... _.. __ ._
·---- ....... _
100 200 300 400 500 600 700 800
Pigure A2.5: Load dispiacement curves for beam series E.

f_lc, f_2b, f_3b, f_4b and f_Sb
toooo ~---~------~------~----~----~----~--~----~----~~
Max f le = 8864.9
8000 Max f 2b = 7965.7
Max f 3b = 8551.7
Max f_4b = 9155.5
Max f_5b = 9173.7
z-6000
........
50 100 150 200 250 300 350 400 450

Pigure A2.6: Load dispiacement curves for beam series F.
126
Appendix A2. Load Dispiacement Curves
g_3a, g_4aand g_5a
. Max g_3a = 18468.0

""o
o" Max g_4a = 22324.5
2 /l
o
o 'o
Max g_5a = 20515.0
o o
• o
o '
o '
o '
o ' ''
oo
' ''o
1.5 : l
! \\.
1
..
.
o
\,
........
o..s ~ ..........,
····························
·-·-------- ...------
100 200 300 400 500 600 700 800 900
Figure A2. 7: Load dispiacement curves for beam series G.
X1(}4 h_2c_2, h_3c, h_4c and h_5c
Max h_2c_2 = 30802.9

3.5 [\ Max h_3c = 35945.6
:,' ~ \
i: ', 1 Max h_4c = 35383.7
~
3 l
• : 1
\
Max h_5c = 33964.7

o \
o '
'' '\
2.5 o '
' '
~ :
z...... ' '
\1
u 2 '.
o 1
~ ', ~
æ •'',t,
\\.
1.5 \ ~.\
.
\. \~
'
.....'\..
l \
....:.......
.....:.. , ..
' •,
0.5 "' ... ::.... ~ ......................
~- ......._.=:..":,::_-:,:::..":.::.~:.....;:.::.":..:.........................________ ..,__ _
100 200 300 400 500 600 700

Figure A2. 8: Load dispiacement curves for beam series H.
127
APPENDIX A3
FRACTURE PARAMETERS.
On the foliowing pages the fracture parameters determined in chapter 5 are presented.
In table A3.1 and A3.2 the modulus of rupture, The fracture toughness, the peak load and
the peak dispiacements are shown.
In Table A3.3 and A3.4 the Fracture energy, the remaining fracture energy and the fracture
energy obtained through optimization are presented.
In tabte A3.5 and A3.6 the parameters determined through the optimization technique
deseribed in chapter 5 are shown.
128
Appendix A3. Frøeture Parameters
Specimcn Mod. of ru:f."'rc Kc Pcak load Pcalc dispiacement

[N/mm] [N/mm3n1 [N] [p m]
taDJe J\~ . l ~som trea rrag_lfore para meters IOr oeam ype J\!.6·
Al 1.14 284
A2 7.18 0.87 663 294
A3 ~ ~ ~ ~
A4 7.43 0.94 684 318
A5 6.68 0.78 617 271
Mean± S.dev 6.94 ± 0.43 0.93 ± 0.15 627 ± 62.5 295 ± 33.2
Cocff.of Var. 0.062 0.16 0.10 0.11
Bl ~ ~ ~ ~
82 6.51 0.917 1198 262

B3 ~ ~ ~ ~
B4 8.21 1.13 1.515 314

B.S 8.94 1.18 1710 311
Mean± S.dev 7.89 ± 1.25 1.08 ± 0.14 1474 ± 258 296 ± 29.2
Cocff.of Var. 0.16 0.13 0.18 0.09
Cl 6.43 1.39 2160 135

C2 7.97 1..57 2843 161
C3 8.25 1.69 2831 154
C4 8.54 1.69 3032 162
es 8.45 1.67 3002 115

Mean± S.Dev 7.93 ± 0.87 1.60 ± 0.13 2773 ± 355 157 ± 14.6
Coeff.of Var. 0.11 0.08 0.13 0.09
Dl 6.58 1.52 4502 124

D2 7.97 1.65 5601 152
Dl 6.54 1.36 4593 126
D4 8 ..53 1.62 6117 167
DS 8.74 1.66 6308 162
Mean± S.dev 7.67 ± 1.05 1.56 ± 0.13 .5424 ± 842 146 ± 20.1
Coeff.ofVar. 0.14 0.08 0.16 0.14
129
J.P. Ulfkjær. Fraelure mecbanics of concrete
Spcclmcn Mod. o f ru~rc Kc Peakload Peak diapiacement

[N/mm [N/mm3n1 [N1 (pm]
El 6.9 1.52 9259 133

E2 6.87 1.37 9543 131
E3 1.56 1.41 10748 165
E4 8.66 1.68 12125 178
ES 8.44 1.53 12034 165
Mean± S,dev 7.69 ± 0.84 1.50 ± 0.12 10741 ± 1343 154 ± 21.1
Cocff.of Var. 0.11 0.081 0.13 0.14
Fl 6.37 1.92 8865 90.8
F2 5.85 1.79 7966 ~
F3 6.13 1.78 8552 69.4
F4 6.62 2.03 9155 75.6

F5 6.73 2.07 9174 73.7
Mean± S.dev 6.34 ± 0.36 1.92 ± 0.13 8742 ± 503 77.4 ± 9.32
Cocff.ofVar. 0.051 0.069 0.051 0.120
G1 ~ ~ ~ ~
G2 ~ ~ ~ ~
G3 6.65 1.96 18468 15.5
G4 7.82 2.24 22324 80.3
GS 7.17 2.02 20515 63.6
Mean± S.Dcv 7.21 ± 0.59 2.07 ± 0.15 20436 ± 1929 73.1 ± 8.60
Cocff.of Var. 0.081 0.071 0.094 0.12
Hl ~ ~ ~ ~
H2 6.33 2.31 30803 69.4
H3 6.98 2.33 35946 82.3
H4 7.36 2.75 35383 81.5

HS 6.9 2.47 33964 78.4
Mean± S.dev 6.89 ± 0.42 2.47 ± 0.20 34024 ± 2304 77.9 ± 5.91
Cocff.of Var. 0.062 0.12 0.068 0.076
Table A3.2 Estimated fracture parameters for beam type E-H.
130
Appendix A3. Fracture Parameters
Spccimcn opIIIl ,
... G'p
Al 115 31.2 105
A2 140.0 41.1 126.2
A3 ~ ~ ~
A4 93.7 27.7 85.0
AS 40.5 0.0 87.0
Meani/S.dev 98.8 ± 42.3 ~ 100.8 ± 19.2

Coeff.of Var. 0 .42 ~ 0.19
Bl ~ ~ ~
B2 122.2 34.0 117.5
B3 ~ ~ ~
B4 123.5 36.6 120.0
BS 114.8 25.4 117.0
Mean//S.dev 120.2 ± 4.69 ~ 118 ± 1.6

Coeff.of Var. 0.04 ~ 0.01
Cl 140.7 24.1 132.1

C2 109.6 34.0 111.0
C3 123.0 31.2 103.6
C4 114.0 32.9 102.0
es 184.0 17.8 188.0
MeaniS.Dev 134.3 ± 30.3 ~ 127.3 ± 36.0

Coeff.of Var.
Dl
0.23
145.2
"
56.0
0.28
136
D2 140.5 30.7 118
D3 93.0 13.5 81.0
D4 130 29.9 124.2
DS 135.4 17.3 126
Meani/S.dev 128.8 ± 20.1 ± 21.2

Coeff.of Var. 0.16
" 117.0
" 0.11
Tabte A3.3 Fracture Energies for beam types A-D.
131
I.P. Ulflcjær. Fracture medumics of concrete
Spccimcn al• Al G'p

El 119.4 34.7 108
El 107 15.6 99
E3 100.9 0.79 lOS
E4 163.4 30.1 159.4
ES 131 27.9 128
± 24.7 ± 24.6
Meani/S.dcv
Coeff.of Var.
124.3
0.20
"
Il
119.9
0.21
Pl 151.1 33.1 159

P2 113.8 19.5 107.1
P3 161 34.4 154

P4 120.8 9.17 121.3
PS 130.5 53.3 115
Meani/S.dcv 136.8 ± 21.6 Il 131.3 ± 23.6

Coeff.ofVar. 0.16 Il 0.18
G1
02
" " "Il
G3 138.7
" "
13.5 143
G4 153.6 9.62 154
G5 125.1 44.8 128.3
139.1 ± 14.3 ± 12.9

Mean/S.Dcv
" 141.8
Cocff.ofVar.
Hl
0.10
" 0.09
H2 127.1
" "
15.1 123.2
"
H3 152.6 19.2 142
H4 168.6 26.6 159.8
HS 126.5 21.8 124.6
143.7 ± 20.6 ± 17.2

Mcan/IS.dev
0.14
" 137.4
Cocff.of Var.
" 0.13
Tabte A3.4 Fracture Energies for beam types E-H.
132
Appendix Al. Fractute Parameters
Spccimcn ø. w' z' y•

• ~
Q
Al 3.60 0.12 0.19 0.28 1.14
A2 3.97 0.12 0.22 0.31 1.12
A3
A4
"
5.42
"
0.086
"
0.14
"
0.22
"
1.01
AS ~
Mcan±S.dev
"
4.33 ± 0.96
"
0.11 ± 0.02 0.18 ± 0.04
"
0.27 ± 0.05
"
1.09 ± 0.07
Cocff.ofVu. 0.22 0.18 0.22 0.17 0.06
Bl ~
B2
"
3.31 0 .15
"
0.20
"
0.26
"
1.20
83
B4
"
5.14
"
0.12
"
0.15
"
0.24
"
1.23
BS 5.19 0.15 0.17 0.13 1.35
Mcan±S.dev 4.55 ± 1.07 0.14 ± 0.02 0 .17 ± 0.03 0.21 ± 0.07 1.26 ± 0.08
Cocff.of Var. 0.24 0.12 0.15 0.33 0.06
Cl 4.15 0.14 0.18 0.29 1.26

C2 6.14 0.19 0.12 0.07 1.18
C3 6.91 0.10 0.14 0.16 1.32
C4 7 .87 0 .08 0.16 0.16 1.26
es 5 .53 0.18 0.18 0.19 1.26
Mcan±S.Dcv 6.12 ± 1.41 0.14 ± 0.05 0.16 ± 0.03 0.17 ± 0.08 1.26 ± 0.05
Cocff.ofVar. 0.23 0.35 0.17 0.45 0.04
Dl 4 .12 0.16 0.18 0.23 1.31
Dl 6 .47 0 .11 0.15 0.20 1.38
D3 6.25 0.07 0.13 0.27 1.27
D4 6.69 0.15 0.14 0.12 1.23
DS 5.71 0.16 0.14 0.14 1.25
Mcan±S.dev 5.86 ± 1.03 0.13 ± 0.04 0.15 ± 0.02 0.19 ± 0.06 1.29 ± 0.06
Cocff.ofVu. 0.18 0 .30 0.13 0.32 o.os
Table A3.5 Estimated fracture pararneters for beam type A-D.
133
J.P. Ultkjer. Fracture meclJanica of concrete
Spccimcn fl11' wc ' x' y' a
El 5.11 0.09 0.19 0.30 1.15
E1 6.15 0.11 0.11 0.10 1.13

E3 6.63 0.09 0.16 0.19 1.13
B4 6.61 0.18 0.11 0.15 1.19
ES 6.80 0.18 0.11 0.10 1.11

'
Mean±S.dcv 6.26 ± 0.68 0.13 ± 0.05 0.14 ± 0.03 0.19 ± 0.07 1.10 ± 0.05
Coeff.ofVu. 0.11 0.35 0.14 0.39 0.04
Pl 5.95 0.26 0.09 0.12 1.06
P2 5.13 0.16 0.09 0.17 1.72

P3 5.24 0.17 0.14 0.21 1.37
P4 6.18 0.15 0.12 0.15 1.39
PS 5.90 0.08 0.19 0.27 1.47
Mean±S.dcv 5.1 ± 0.44 0.16 ± 0.06 0.13 ± 0.04 0.18 ± 0.06 1.40 :1: 0.24
Cocff.ofVu. 0.08 0.39 0.33 0.32 0.17
Gl ~ ~ ~ ~ ~
G2 ~ ~ ~ ~ ~
G3 5.48 0.21 0.11 0.14 1.47
G4 7.57 0.16 0.11 0.14 1.51

GS 6.17 0.20 0.13 0.09 1.28
Mcan±S.Dcv 6.41 ± 1.06 0.19 ± 0.026 0.12 ± 0.012 0.12 ± 0.03 1.42 ±0.12
Cocff.ofVar. 0.17 0.14 0.10 0.13 0.09
Hl ~ ~ ~ ~ ~
H2 5.38 0.18 0.12 0.14 1.49

H3 5.11 0.14 0.15 0.08 1.34
H4 5.15 0.28 0.11 0.09 1.43
HS 6.31 0.17 0.12 0.11 1.32
Mcan±S.dcv 5.64 ± 0.52 0.22 ± 0.05 0.13 ± 0.02 0.11 :1: 0.03 1.40 ± 0.08
Cocff.of Var. 0.09 0.14 0.14 0.15 0.06
Table A3.6 Estimated fracture parameters for beam type E-H.
134
APPENDIX A4
OPTIMIZATION RESULTS.
On the following pages the fitted load dispiacement curves are plotteet The experimental
curve is the dashed curve and the solid curve is the fitted curve.
135
I.P. mtJgær. Fracture mecbanics of concrete
a_1c
600
500
400
J 300
200
100
100 200 300 400 soo 600 700 800

Dispiacement [m ~
Figure A2.1: Fitted results for beam A-le.

a_2a
700
600
!500
~ 400
l 300
200
100
100 200 300 400 !500 600 700 800

Dispiacement [m ~
Pigure A2.2: Pitted results for beam A-2a.
136
Appendix A4. Optimization Results.
a_4a
700~----~------~----~r-----~------~----~
soo
200
100
100 200 300 400 soo 600

Displacement [m B-6]
Figure A2.3: Fitted results for bearn A-4a.
a_sa
700
600 . ...
soo
~ 400
J 300
200
100 .. .··.
....
·· ... ......
so 100 ISO 200 2SO 300 350 400 4SO
Disp1accment [m B-6]
Figure A2.4: Fitted results for bearn A-Sa.
137
1.P. Ulfkjær. Fracture mecltanics of concrete
b_2a
100 200 300 400 soo 600 700 800 900

Displacemcat [m B-6]
Figure A2.5: Fitted results for beam B-2a.
1~r-----~------~----~b-4~·~----~----~~----~
1400
1200
æ: 1000
J 800
400
200
100 200 300 400 soo 600

Dispiacement [m B-6]
Figure A2.6: Fitted results for beam B-4a.
138
1800.---~----~----~--b~SF•----~----~--~----~
1600
1400
1200
g1000
J 800
600
400
200
100 200 300 400 soo 600 700 800

Disp!acemeot [m E-6]
Figure A2. 7: Fitted results for beam B-5a.

c_1c
~OOr-----~----~----~~--~----~------~--~
2000
~ 1500
J 1000
soo
.. .
100 200 300 400 soo 600 700

Dispiacement [m E-6)
Pigure A2.8: Fitted results for beam C-le.
139
J.P. Ulflgær. Fracture mecbanics of concrete
c_2a
3000
..
2500
2000
~
J lSOO
1000
soo
100 l SO 200 250 300 350 400 4SO 500

Figure A2.9: Fitted results for beam C-2.

c_3a
3000
2500
2000
~
J 1500
\
\
\
\
\
1000
''~ ......
......,,
....,,. ,
...._
soo
00 so 100 150 200 250 300 3SO 400 450

140
3SOOr-----~------~-----c~-~4a------~------~------,
3000
2SOO
~ 2000
J 1SOO
1000
soo
0 o·L------s~o~-----1~oo------1~s~o------2~oo~-----~~o----~3oo
eJa
3SOOr----r----~--~----~~~----~--~----~---.
3000
~00
~ 2000
J lSOO
1000
soo
0
o ~ ~ ~ ~ soo ~ ~ ~ 900
Dispiacement [m :S.6]
141
J.P. Ulfkjær. FraelUre mechanics of coocrete
~r---~-----r----~---dL~lb____~----~--~----~
500
50 100 150 200 250 300 3SO 400

Figure A2.13: Fitted results for beam D-l.
4000
2000
1000
oo 50 100 150 200 250 300 350 400 450 soo

Dispiacement (ID E-6)
Figure A2.14: Fitted results for beam D-2.
142
4!500
4000
3500
2000
1500
1000
500
. ....
OO SO 100 150 200 2SO 300 350 400 4SO SOO
Figure A2.1S: Fitted results for beam D-3.
7000~----~----~----~----~-----d~-~4·----~--------~----~----~----~
6000
2000
50 100 1$0 200 2$0 300 3$0 400 450 500

Displacemeat [m ~
143
J.P. UIOcjær. Fracture DJl"dCbanics of concrete
d_5a
7000
6000
5000
~4000
J3000
2000
... .... ..
00 100 200 300 400 soo 600 700

c_1c
10000
g()()()
8000
7000
~6000
J 5000
4000
3000
2000
1000
100 150 200 250 300 350

Figure A2.18: Fitted results for beam E-l.
144
e_2b
l
9000
8000
7000
~6000
j 5000
4000
3000
...... ............
00 100 200 300 400 soo 600
Dispiacement [m ~6]
Figure A2.19: Fitted results for beam E-2.

e_3b
12000
10000
8000
~
16000
4000
.........
.............._
... ----........._....___
00 100 200 300 400 soo 600 700 800
Pigure A2.20: Pitted results for beam E-3.
145
1.P. Ulfkjær. Fracture mecbanics of concrete
e_4b
1~~----------~----~~~--~----~----~-----.
300 400 soo 600 700

Displacem.eut [m B--6)
12000
10000
4000
2000
oo so 100 150 200 250 300 350 400 450 500

Displacemeut [m B-6]
146
8000
7000
3000
2000
1000
so 100 1SO 200 250 300 3SO 400

Figure A2.23: Fitted results for beam F-l.
7000
6000
3000
...........
00~--~----~----~----~--~----~~--~--~
so 100 1SO 200 250 300 3SO 400
Figure A2.24: Fitted results for beam F-2b.
147
J.P. Ulflcjær. Fracture mechanica of concrete
f 3b
9000
8000
7000
6000
!: sooo
j4(XX)
3000
0 o~--~s~o----~~~oo~--~.s~o----~~~--~~~o~--~3~oo~--~3so
Displac:em.em [m Ml
100001r---~----~--~--~~f~-~4b~--------~----~--~
9000
8000
7000
3000
.... .
0 o·t_--~so~--~~oo~~~~s~o--~~~--~~~o--~3~oo~~3;so~~400~==~4so
Dispiacement [m M]
148
x104 g_3a
2
1.8
1.6
1.4
~ 1.2
J l
0.8
0.6
0.4
0.2
............
oo 100 200 300 400 500 600 700 800 900
Dispiacement [m ~
Figure A2.27: Fitted results for beam G-3a.

4
l.S xlO g_4a
~ 1.5
! 1
100 300 400 500 600

Displacernemt [m B-6]
Figure A2.28: Fitted results for beam G4-a.
149
J.P. Ulflcjær. Fracture mechanics of coocrete
~5a
2.5 ...........__--.----.---.----=r----.---.------.------,
0 .5
20 40 60 100 120 140 160

Displacem.eot [m E-6]
Figure A2.29: Fitted results for beam G-Sa .
l.5 x10
• 2c2
2.5
..... ..... ... .

······ ...... .......
100 200 300 400 600
Figure A2.30: Fitted results for beam H-2c.
150
xlO4 h_3c
4
3.5
2.5
!:
J 2
1.5
0.5
. ...... .... ....... ... . .... .... ... ...
% 100 200 300 400 500 600 700
Displaccmcnt [m E-6]
xlO4 4c
4
3.5
1.5
~ ~ ~ ~ ~ ~ ~ 400 ~
151
4
3 .5 x10
U c
2.5
50 100 150 200 250 300

152
APPENDIX A5
CRACKINO PROFILES.
On the foliowing pages the measured cracking profiles deseribed in chapter 5 are shown. The
specimens Sl-S4 are the standard specimens, and specimens Kl-K4 are the saw cut
specimens. The profiles are meassured for both fracture surfaces, and are atmost identical.
The dashed line indicates the position of the notch, and the solid line shows the dye edge,
which is assumed to be near the real crack tip.
153
J.P. Ulfkjær. Fracture mecbanica of coocret.e
~·----~1~00~------~
51
·o·. ·d . • . · · .Q.·
• <l • • • •
• 4 • • • • • • •
....
.~ . . . . . . .·.·<l··ri
• • · • <1. •.
. . .
• c::J
• . . '4.
• •
52
• • ••
. . .
4
.
• •
.
·o .. 4. • • <:J • . • •. • •
~· ........ . 'Q
.
•
•
•
.• . . <l. . • •. • () •
• • • •• <l. • .
. . : . ~ :. 4 .. . . ~
. . .. ~·. <:1 • <> .. ..
• •• • • : •
'.;·<l .·: ~·· . .. ::, '.

. . . . 4 . • •. : .•<l. . .. .".:
• <\
~ . . . . . . . 53
• 4 • • 4
• '4 • •
..; : .. . ~ . . .... ~ ....
A
.<J . . . . . . . . <l..... <l.
' • • • '
s4
Figure AS .l: Dyeing profiles for the standard specimen type D.
154
Appendix AS. Craddoa Profiles
100
•
. . -·-• ·-·
~. Q • • • • . <l'
.....:........._·_._.
. ..
·. ·.. 4 .. ·. . . 4 •. • ... ~ •• ~ •. 4. • •• :
• • • • • • • Q
... . . ... ·.· ·.~. ·.
-~ • : • • 11 • : : • : : •• . .o . . . . . . .. .
...
•
. ~· ......\) . .
• •• .::1
: •••. A • •• •• 6·
'<J ·... ·. ~ .. ·. ·.~_·:.·
.4. . . . ~.: . ri.
... • .d ~ · · · ~ ·.. · ~.. K1
... . . .
. ·.., .. · .. ·. . . <l >.
. .. . . • . <1 . • . • ... .
'V
• • <1 •
Å ° ~ 4
~·· : :<l
A ' ' , '
.<:\ . • ""
0
Il.& - • •
. . ... .. -6. : . ·. Q . . ••
K2
.4 .. '/\ . . . . . • : . .. • • • • • ' • A • •
• • <.J • • A , . ·. - ~· . . ~·- .. ~ .. .
•
4. . . •«L
.
. •
• . . . • 'J . •
..... . .. ·.. ·~·
·••• • • •• • ••• o.' • •• 4: () • q • • • • •
o·. .. 4 .
• •
,4 ' .: ~. ·.:'.<:J .' · .. ... ·· 4· . ·. .<l

•• l
... .
p. • • • • • K3
. . . . 4 . . 4 . • . <l •
'<l ... . . . : . . . .
. . . . ~ . . . . <J . . . <l. .
. . . . . . . . . . K4
F'~gure AS.2; Dyeing profiles for tbe saw cut specimen, of size D.
155
APPENDIX A6
RESUME IN DANISH.
Titlen på afhandlingen er
Brudmekanik for beton
Som titlen beskriver omhandler projektet brudmekanik for beton. Projektet er en del af et
større forskningsprogram under Statens Teknisk Naturvidenskabelige Forskningsråd med
titlen:
"Højkvalitetsbetoner i 90'eme"
Brudmekanik er en forholdsvis ny disciplin og de udviklede teorier har fortrinsvis været rettet

mod stål hvor en lineær teori ofte giver gode resultater. For beton har lineær brudmekanik
generelt vist sig at give dårlige resultater og i stedet er der udviklet en række nye modeller.
Den mest kendte af disse er den fiktive revnes model udviklet af Arne Hillerborg. Denne
model har gjort fagområdet bredt tilgængeligt og indenfor de seneste år har stadig flere
fundet interesse for fagområdet, der er inde i en rivende udvikling. Brudmekanik er en mere
nøjagtig metode at beskrive konstruktioner på end de mere almindelige anvendte teknikker
(elasticitet teori og plasticitetsteori), og kan derfor beskrive visse fænomener som de
førnævnte teorier ikke kan forklare. I Danmark er der dog stadig kun il, der har beskæftiget
sig indgående med emnet (H.H. Bache, H. Stang og N.A. Harder). Det var på denne
baggrund projektet blev startet.
I denne afhandling er der især fokuseret på fænomenet 'størrelseseffekter', hvor betegnelsen

størrelseseffekter er opfattet bredt (størrelseseffekter på vilkårlige materiale parametre).
I kapitel 2 af afhandlingen gives en grundig beskrivelse af brudprocessen i beton og

højstyrkebeton gående fra mikro- til makroniveau. Derefter følger en beskrivelse af de mest
anvendte brudmodeller for beton. Kapitlet afsluttes med forskellige eksempler, der ved hjælp
af lineær elastisk brudmekanik illustrerer begrebet størrelseseffekter.
Efter beskrivelsen af disse modeller er der valgt at fokusere på den fiktive revnes model. Der
er desuden foretaget den afgrænsning, at der udelukkende betragtes bjælker udsat for tre
punkts bøjning.
I kapitel 3 beskrives det hvorledes den fiktive revnes model kan anvendes i tilknytning til
numeriske metoder. De to mest kendte metoder udviklet afP.E. Petersson og A. Carpinteri
præsenteres. En ny metode baseret på randelementmetoden præsenteres. Der er i forbindelse
med projektet lagt et stort arbejde i at udvikle og implementere denne metode.
156
Appendix I. Notation
Da numeriske modeller ikke altid er særlig hensigtsmæssige er der udviklet en analytisk

metode hvor den fiktive revnes model er anvendt således at den fuldstændige arbejdskurve
kan beregnes. Modellen beskrives i detaljer og sammenlignes med to modeller, der er
udviklet af Y.W. Mai, baseret på revnebåndsmodellen, samt en metode af J. Llorca, der er
baseret på den fiktive revnes model. Modellen er endvidere udvidet til at gælde for armeret
beton.
I forbindelse med projektet er der udført forsøg med omkring 50 bjælker af uarmeret og
svagt armeret beton i en nyudviklet servostyret revneåbningkontrolleret forsøgsopstilling. Der
er genemført forsøg til belysning af størrelseseffekter og forsøg med henblik Ø at bestemme
revne profiler i udsavede prøveemner. Der er endvidere forsøgt at bestemme materialepara-
metre i den fiktive revnes model ved at tilpasse arbejdskurver bestemt ved den numeriske
metode med arbejdskurver målt ved forsøg. Dette er gjort ved anvendelse af et generelt ikke-
lineært optimerings program. Forsøg og forsøgsresultater er beskrevet i kapitelS.
Rapporten afsluttes med en konklusion der opsummerer de opnåede resultater, samt giver
forslag til den videre forskning indenfor området.
157

Fracture Mechanics of Concrete PDF

Uploaded by

Copyright:

Available Formats

Fracture Mechanics of Concrete PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fracture Mechanics of Concrete PDF

Uploaded by

Copyright:

Available Formats

What is the scope of the thesis?

What is the scope of the thesis?

What fracture models are described?

What fracture models are described?

Aalborg Universitet

Fracture Mechanics of Concrete

Ulfkjær, Jens Peder

Link to publication from Aalborg University

Citation for published version (APA):

Downloaded from vbn.aau.dk on: januar 23, 2019

JENS PEDER ULFKJÆR

Jens Peder t.ntkjær

Aalborg, July 1992 Jens Peder Ulfkjær

2 FRACTURE AND FRACTURE MODELS OF

3 NUMERICAL METHODS IN CONCRErE

4 ANALYTICAL METHODS IN CONCRETE

1.1 Background and Motives.

Therefore, it would be welcome if powerful numerical techniques, simple accurate analytical

1.2 Scope of Thesis.

The main scqpe of the thesis is:

Developmenl ofnwnerical and analyticalfracture mechanical techniques, examination ofthe

The development of approximate analytical methods where the calculation time

To develop experimental techniques which can perform stable three-point

Perform experiments with high strength concrete in order to investigate size

Develop methods for indirect determination of the constitutive parameters in

The foliowing limitations are general in this thesis.

The compressive strength is assumed to be so large that compressive faiture

The fictitious crack model will be used.

lt is in general assumed that the softening relation is piece by piece linear.

The structure outside the crack is assumed to be modelled satisfactory by

In chapter 3~5 only the three-point bending geometry is considere(i.

1.3 Readers Guide.

The chapters 3-5 are the mainbodyin the thesis.

In chapter S an experimental investigation of 8 different beam geomelries of a high strength

An overall summary and condusion is given in chapter 6.

2 FRACTURE AND FRACTURE MODELS OF

2.1 Fracture of Concrete and ffigh Strength Concrete.

Concrete is a heterogene()US anisatropie non-linear inelastic composite materlal tllat consists

Micro cracks : Cracks that can only be observed by an electron microscope.

2.1.1 Concrete on the Micro-level.

1 According to the Danish code: DS-SBC 227

elinker Con- Hydration prod- Morphology

Table 2: Constituents of cement.

Fig. 2.2:. Rate of heat evolution curve, Herholdt et al. (1985).

Trans1tlon Zone Bulk Cement Poste

1.1.11be Fraelure Process in Compression.

Level Materlal Strong Particles Weak Matrix

Crackin& on the meso-level

A!ld1lional growth cl bond cracks.

Pre·exist ng bond " acks

The amount of cracking is significantly less in HSC than in NSC.

2.1.3 The Fradure Process in Tension.

Unitormly distribu ted Localized zone

Fig. 2.9: Concrete rod subjected to dispiacement controlled loading.

Edisp = I J Ji(e,JAde,.dx+ J J h.(e,JAde,.dx

2.2 Linear Elastic Fracture Mechanics.

2.2.1 Basic LEFM.

Fig. 2.12:. Arbitrary specimen with an initial crack, energy approach.

However, if crack growth is considered then eq. (2.3) changes to