CHAPTER 3

Figure 3.6: Crack propagation in concrete at uniaxial tensile loading recreated from (Malm,
2016)
3.2.7 Fracture energy:

Crack opening displacement (w) is related to a factor called fracture energy (Gf) which is
represented by the area under curve of the descending part. Basically, the fracture energy (Gf)
is defined as the amount of energy necessary to create one unit area of a crack (Hallgren, 1996).

For a concrete with known material composition, the value for fracture energy (Gf) is normally
determined from uniaxial tension testing. Factors which has influence in the fracture energy
are; maximum aggregate size, concrete age and water cement (w/c) ratio (MC10, 2012).
However, if testing is not available, the value for Gf according to (MC10, 2012) for a normal
strength concrete is determined from equation:

3.1

Where:
fcm is the concrete mean compressive strength in (MPa)

In the earlier version of CEB/FIP Model Code (MC90, 1990), the fracture energy was related
to largest aggregate size and concrete grade as shown in Table 3.1 and equation (Malm, 2016).
Table 3.1: Fracture energy for different concrete grades and aggregate sizes as per (MC90,
1990), (Malm, 2016):

Gf (N/mm)

Dmax C12 C20 C30 C40 C50 C60 C70 C80

8 40 50 65 70 85 95 105 115

16 50 60 75 90 105 115 125 135

32 60 80 95 115 130 145 160 175

43
FINITE ELEMENT ANALYSIS

In addition, based on the descending part of the stress-crack width curve shown in Figure 3.6
and with the curve shape coefficient used in ATENA, the fracture energy (Gf ) is directly related
to the critical crack width (wc) through below equation:

3.2

Equation 3.2 presents the value for exponential crack opening curve. Linear and bilinear curves
could also be used to define crack opening (Malm, 2016):

For linear:

3.3

For bilinear:

3.4

An increase in fracture energy (Gf) or a decrease in tensile strength (ft) increases the value for
critical crack width (wc). The higher the value for (wc), the more energy is needed to propagate
macro crack. The crack width is not directly dependent on wc. However, a larger wc will give
more ductility as the crack will be able to carry stress for a larger deformation.

Modelling simplifications and assumptions in

ATENA 3D
After reviewing the drawings of the old building in Stockholm, it was found that there were a
variety of pile foundations with different sizes and dimensions used. Therefore, it was decided
to select a single pile cap model by referring to engineering codes and handbooks. Finally, the
dimensions of the models were adjusted according to engineering handbook by (Reynolds &
Steedman) (detailed calculations are in section 6.1).

For modelling, two software were used; ATENA engineering 3D (v.5) and ATENA studio
(v.5). ATENA engineering was used in the pre-processing stage where the material properties,
geometry, element types, loading, boundary conditions and mesh were introduced. Later, the
model was run using ATENA studio where the analysis was completed, and the required results
were extracted.

44
 CHAPTER 3

The following assumptions and simplifications were considered while creating the model:

a) Only vertical compressive force affects the pile cap i.e. lateral forces and moments
were not considered.
b) The stiffness of all the piles supporting the pile cap are equal and therefore the force
from column is equally divided between the four piles. Considering this assumption,
the pile cap is double symmetric i.e. only one fourth of the pile cap was modelled.
(Figure 3.8)
c) The pile cap was studied in isolation from structure and soil beneath, i.e. only short
column and short pile were considered for the model.
d) No transverse shear reinforcements were provided.

Figure 3.7: Full pile cap model created in ATENA 3D

Figure 3.8: Symmetric model created in ATENA 3D

45
FINITE ELEMENT ANALYSIS

Material properties definitions

To define material properties in ATENA 3D, there are three approaches:

a) direct; where one can select from the available material models in ATENA and also make
changes to parameters if required.

b) properties from file; where one can use material models created earlier.

c) properties from catalogue, where materials are defined based on standards, mainly EC2.

For the models created for the purpose of this thesis, all the materials were defined using direct
approach. In general, five materials were defined for the model: concrete, steel, reinforcement,
contact and springs.

3.4.1 Concrete material models

In the model created in ATENA, all the concrete parts are modelled with non-linear material
properties. The non-linear concrete material is named as CC3DNonLinCementitious which is
based on a fracture-plastic model. The significance of the fracture-plastic model is that it
captures the tensile and compressive behavior of the concrete at the same time and therefore
can simulate crushing, splitting, crack opening and closure very well.

Uniaxial failure criterion

To characterize the uniaxial behavior of concrete, the material CC3DNonLinCementitious in

ATENA 3D, uses the uniaxial stress-strain law.

Figure 3.9: Uniaxial stress-strain used by CC3DNonLinCementitious in ATENA. recreated

from (Cervenka et al., 2018)

46
 CHAPTER 3

Bi-axial failure criterion

The bi-axial state of stresses is different than that of uniaxial and is presented by a failure
envelop. The yielding of material occurs when the state of stresses reaches the boundary of the
envelope. To depict the bi-axial behavior, the material model in ATENA
(CC3DNonLinCementitious), uses the failure criteria suggested by Kupfer in (Kupfer et al.,
1969).

Figure 3.10: Biaxial failure criteria represented by an envelope recreated from (Cervenka et
al., 2018)

As it can be seen from the envelope, the compressive strength increases in case of bi-axial
compression and the in the mix state of stresses (compression and tension) the strength reduces.
The confinement effect in concrete is the reason for increase in case of bi-axial compression
and is said to increase the strength up to 16 % (Malm, 2016).

Triaxial failure criterion

In the tri-axial state of stresses, the concrete compressive strength increases considerably more
compared to bi-axial state (Malm, 2016). To employ the triaxial failure criterion, the model in
ATENA (CC3DNonLinCementitious) utilizes separate models for cracking and crushing in
concrete. The cracking is presented by Rankine fracturing model where the strains and stresses
in a structure are adapted to the direction of material. Crushing is presented by plasticity model
failure criterion is based on work by (Menetrey & William, 1995). The surfaces in
triaxial state of stresses resemble the shape of a cone and are related to eccentricity factor (e).
The value of (e) can vary between 0.5 (more circular edges) up to 1 (more triangular edges)
(Menetrey & William, 1995).

47
FINITE ELEMENT ANALYSIS

Figure 3.11: Triaxial state of stresses recreated and modified from (Menetrey & William,
1995)

3.4.2 Reinforcement model in ATENA

There are two approaches to model reinforcement in ATENA 3D; smeared approach and
discrete approach. In smeared approach, the reinforcement ratio is smeared over an element
whereas in discrete approach the reinforcement bars are introduced into the model. For the
models in the thesis, the discrete approach was used. The behavioral properties of the
reinforcement was selected to be bi-linear which corresponds to elastic-perfectly plastic
response. In addition, for further simplification, a perfect bond and anchorage of reinforcement
was assumed. This is normally sufficient for a ULS analysis, provide that the failure does not
depend on the anchorage of the bars. However, it is possible to model the bond between rebar
and concrete with a bond-slip.

Figure 3.12: Elastic perfectly plastic material response of reinforcement corresponding to bi-
linear material properties in ATENA 3D

48
 CHAPTER 3

Boundary conditions and loads

The pile cap was modelled in ATENA in such a way that represented a universal frame
consisting of hydraulic jack applying load on the pile cap. Steel plates were used both on
loading area (column top) and supports (bottom of pile). Regarding loading, two methods are
usually used in numerical analysis to define loading; a) load controlled; b) displacement
controlled. For the model analyzed, the latter (displacement controlled) method was used. This
was done on purpose to get a representation of the failure load and post failure response.
Therefore, instead of column force, a prescribed deformation was assigned on the top of column
which increased gradually. Corresponding to the prescribed displacement, ATENA calculated
a reaction force.

Other boundary conditions included symmetry and supports. As discussed previously, the pile
cap model was double symmetric. Therefore, only one quarter of the pile cap was modelled
with fixed boundary condition in two direction x and y respectively. To actualize this in
ATENA, surface supports were used on the face of pile cap, column, and loading plate in two
directions. The other benefit of using symmetry boundary is that you get a stable model where
rotation of the structure does not occur after loading.

The last boundary conditions included support at the bottom of concrete pile which resembles
hard soil layer. In the model, the bottom of the pile was assumed to be fully supported. The
behavior of the pile were, however, stiffer then they are on the site. The reason for a stiff
behavior is because short piles were assumed to reduce the analysis- time. This certainly had
effects on the behavior of the pile cap.

a) b)

Figure 3.13 a) symmetry and support boundary conditions in ATENA 3D

b) prescribed deformation as support in Z direction

49
FINITE ELEMENT ANALYSIS

The details about boundary conditions applied in the model are listed in table (3.2).

Table 3.2: Boundary conditions used in the model:

Structural Boundary Direction

element condition X Y Z
applied
Pile Cap Symmetry Fixed Fixed Free
Pile Support Free Free Fixed
Column Symmetry Fixed Fixed Free

Mesh and elements

When introducing elements in ATENA 3D, concrete and reinforcement bars are introduced
separately. When meshing is concerned, only the concrete part (solid element) is meshed. No
mesh is defined for the reinforcement bars in the pre-processing stage. As soon as the analysis
starts, the bar elements are counted as embedded elements within the mesh of solid elements.
Therefore, mesh was only introduced for the solid elements in the model. ATENA 3D has three
types of elements; brick, tetrahedron and pyramid (Cervenka et al., 2018).

Since the geometry for the pile foundation model is simple i.e. does not contain any
irregularities such as opening or refinement, therefore, only brick shaped elements were used.

Figure 3.14: Meshed model with brick elements: a) 10 cm mesh size; b) 5 cm mesh size

50
 CHAPTER 3

Another factor related to meshing is the mesh size. The mesh size affects the results directly,
i.e. if the mesh sizes are smaller (more elements are used), the results would be of more quality
(better results). However, the mesh size is also related to analysis time, i.e. the more elements,
the more the analysis time. To find an acceptable mesh size for the model a mesh convergence
analysis is performed where three mesh sizes are examined on a model. After obtaining and
studying the results, if the results for two mesh sizes are close enough the larger mesh size is
accepted. Care must be taken, if the mesh size is very large, the model might behave very stiff.
Therefore, the maximum element size in the model must be able to capture the fracture zone
process which can be found using the below equation (Malm, 2016):

3.5

Mesh convergence analysis is performed in section 4.3

Non-linear iterative solvers

In a non-linear analysis, the solutions cannot be obtained directly. An iterative procedure must
be used to solve the equations (Malm, 2016). ATENA presents two methods to solve non-linear
equations; a) Standard arc length method; b) Standard Newton-Raphson method. It is also
possible to change the solution parameters and define a new parameter based on requirements.

In the arc-length method the displacement and load both are iterated while the solution path is
kept constant. This method is more general when compared to Newton-Raphson, but it is not
useful for all cases i.e. in case of the body force it changes the weight of the structure.

In Newton-Raphson method, a tolerance limit for satisfaction of equilibrium solution is defined.

For this tolerance limit, a constant load increment is introduced for which the displacement is
iterated until equilibrium is reached. Therefore, it is beneficial in cases where the load value
must be met. For a displacement-controlled analysis, Newton-Raphson method is suggested by
(Cervenka et al., 2018) -
Raphson method was used. The limitation of this method is though that it is not possible to
capture any possible snap-back behaviour.

51
FINITE ELEMENT ANALYSIS

52
 CHAPTER 4

Models created
At first, four models of the pile were created. The models were identical with the only
difference between them being the position of reinforcement and the ratio of reinforcement.
The details of the models are presented in Table (4.1).

Table 4.1: Identification of models with varying reinforcement position and reinforcement
ratio
Model name: Model A Model B Model C Model D
Position of Top Bottom Top Bottom
reinforcement
Reinforcement area in 35.6 cm2 35.6 cm2 21.5 cm2 21.5 cm2
each direction of pile
cap

53
ANALYSIS AND RESULTS IN ATENA 3D

Figure 4.1: Reinforcement plan and layout in models A and C (placement of reinforcement at
the top of piles)

Figure 4.2: Reinforcement plan and layout in models B and D (placement of reinforcement at
the bottom of the pile cap)

54
 CHAPTER 4

Input data in model

4.2.1 Concrete

Since the aim in a non-linear analysis is to observe the failure of the model, therefore, the mean
values for strength of concrete materials were considered.

Table 4.2:Values for concrete material properties:

Material property Unit Pile cap Piles and column

- 0.2 0.2
Modulus of elasticity Ecm GPa 33 37
Tensile strength (Ft) MPa 2.7 3.7
Compressive strength (fck. cube) MPa 38 60
Compressive strength (Fck) MPa 30.4 50
Mean Strength (Fcm) MPa 38 58
Fracture energy (Gf) MN/m 67.8 92
Aggregate size: M 0.02 0.02

All the input values in ATENA presented in Table (4.2) are calculated based on Eurocode,
except the values for tensile strength (ft) and fracture energy (Gf). For tensile strength, the
Eurocode 2 uses the following equation:

4.1

Whereas ATENA relates the tensile strength to concrete cube strength (fcu):

4.2

For fracture energy (Gf) ATENA uses an equation recommended by (E. VOS, 1983) which
relates the fracture energy to concrete tensile strength:

4.3

Whereas the (MC90, 1990)and 2010 have different criterion for fracture energy as presented
earlier. (MC90, 1990) relates the fracture energy to maximum aggregate size as presented in
Table 3.1 and (MC10, 2012) relates it to concrete compressive strength as seen in equation 3.1.
The effect of different values of fracture energy are studied in the models.

55
ANALYSIS AND RESULTS IN ATENA 3D

4.2.2 Reinforcement:

Reinforcement is defined using bilinear law and the following properties are introduced in
ATENA:

Table 4.3: Values for reinforcement material properties

Material property Unit No of bars in each direction

14 bars in each direction
Elasticity Modulus (Es) (MPa) 210000
Yield strength (fy) (MPa) 500

The connection between the reinforcement and the surrounding concrete was accepted as
perfect i.e. no bond slips were defined.

4.2.3 Interface:

To reduce the over stiff behavior of the numerical model, an interface material was introduced
at the contacts between (column-pile cap) and (pile-pile cap). In ATENA 3D, this material is
called 3D interface and is defined by two types of parameters; a) physical parameters; b)
stiffness parameters. The physical parameters relate to the physical properties of interface such
as; friction, cohesion and tensile strength. The recommended value in a compression only
support for these parameters based on (Cervenka et al., 2018) are:

Tensile strength (ft) = Tensile strength of weaker material at the contact

Friction coefficient ( = 0.5

Cohesion (D) = ( (ft)

The stiffness parameters (Knn) and (Ktt) are only for numerical purposes. Each of them has two
sets of values: basic and minimal. The basic value represents closed state (rigid connection) and
the minimal value represents open contact. The recommended values for basic (initial)
stiffnesses according to (Cervenka et al., 2018) can be found using equations:

4.4

Where:

E- is the elastic modulus of weaker material

G- is the shear modulus of weaker material

t- is the thickness of the interface element which is assumed as 0.02 m

The values for residual (minimal) normal and shear stresses are estimated as (initial stiffness *
0.001) (Cervenka et al., 2018).

56
 CHAPTER 4

Table 4.4: 3D interface material properties

Material property Unit 3D Interface

Normal Stiffness (Knn) MN/m3 3000000
Tangential stiffness:(Ktt) MN/m3 3000000
Tensile strength (ft) MPa 2.7
Cohesion (D) MPa 1.35
Friction coefficient ( : - 0.5
Min. normal stiffness (Knn,min) MN/m3 3000
Min. tangential stiffness (Ktt,min) MN/m3 3000

4.2.4 Steel plate:

The steel plates used at the loading and support were defined as linear elastic material with the
below properties:

Table 4.5: Steel plate material properties

Material property Unit Value

Elasticity modulus (Es) MPa 210000
- 0.3

4.2.5 Input file:

After the pre-processing stage was competed in ATENA 3D, the model was saved as an input
file for ATENA Studio. After slight modifications in the input file based on recommendations
by (Cervenka et al., 2018), the analysis was run in ATENA studio.

Results:

4.3.1 Load deflection response

The load displacement diagram is a graphical representation of a collection of points which

correspond to certain load and displacement level. One can observe the response of a structure
with the increasing load level and extract the ultimate failure load and maximum deflection. As
stated previously, all the analysis performed were deformation controlled. A specific attribute
of a deformation-controlled analysis is that there is a drop in the load displacement diagram
after the concrete is cracked. As it can be seen in the load displacement diagram igure 4.3), the
load drop is clearly visible in all the curves. The displacement increment corresponding to each
load step was (0.1mm) in uncracked concrete and (0.02 mm) for the steps where the concrete
was cracked. This, to improve the convergence in the results obtained. Further on, the load in
(load displacement) diagram is factored by 4. This because only one-fourth of the pile cap
was modelled using symmetry boundary condition.

57