MSc Project Steel Structures:

High Strength Steel Structures – FEM Validation Experiments

Yang Chen
Faculty CITG, Delft University of Technology
Delft, the Netherlands, 2014
Contents

Contents............................................................................................................................I
Notations ........................................................................................................................ III
List of figures...................................................................................................................IV
List of tables ................................................................................................................. VIII
1. Introduction............................................................................................................... 1
1.1 Background ..................................................................................................... 1
1.2 Goal of the research ........................................................................................ 1
1.3 Ductile failure criteria ....................................................................................... 1
1.4 Approach ........................................................................................................ 3
2. Test overview ............................................................................................................ 4
2.1 Joint geometry ................................................................................................ 4
2.2 Test parameters .............................................................................................. 5
2.3 Static test procedure........................................................................................ 6
2.4 Test rig and measurements .............................................................................. 6
3. Material properties ..................................................................................................... 8
3.1 Material specification ....................................................................................... 8
3.2 Material properties........................................................................................... 9
3.2.1 Stress strain relation ............................................................................. 9
3.2.2 Modified stress strain relation .............................................................. 13
3.2.3 Tensile test simulation (without necking) .............................................. 17
3.2.4 Tensile test simulation (with necking) ................................................... 21
3.2.5 Calibrated true stress strain input data ................................................. 31
4. FE validations of connections .................................................................................... 35
4.1 Connections type B ............................................................................................. 35
4.1.1 Specimen 3B1 (fail in plate) ................................................................. 36
4.1.2 Specimen 4B1 (fail in weld) ................................................................. 45
4.1.3 Specimen 1B1 ..................................................................................... 50
4.1.4 Specimen 2B4 ..................................................................................... 54
4.1.5 Specimen 5B1 ..................................................................................... 57
4.2 Connections type C ........................................................................................ 60
4.2.1 Specimen 1C1 ..................................................................................... 61
4.2.2 Specimen 2C1 ..................................................................................... 64
4.2.3 Specimen 3C1 ..................................................................................... 67
4.2.4 Specimen 4C1 ..................................................................................... 70
4.2.5 Specimen 5C1 ..................................................................................... 73
4.3 FEM result comparisons ................................................................................. 76
4.3.1 Connection 1 ...................................................................................... 76
4.3.2 Connection 2 ...................................................................................... 77
4.3.3 Connection 3 ...................................................................................... 78
4.3.4 Connection 4 ...................................................................................... 79
I
 4.3.5 Connection 5 ...................................................................................... 80
5. Conclusions and recommendations ............................................................................ 81
5.1 Conclusions ................................................................................................... 81
5.1.1 Validation of material input .................................................................. 81
5.1.2 Validation of specimen tests................................................................. 81
5.1.3 FEM result comparisons ....................................................................... 81
5.2 Recommendations ......................................................................................... 82
References ..................................................................................................................... 83
Appendices ..................................................................................................................... 84
A. Hardening rule ......................................................................................................... 84
B. Material input .......................................................................................................... 85

II
Notations

A0 – original cross sectional area

A1 – deformed cross sectional area

E – modulus of elasticity

F – reaction force

K – strength coefficient

l – specimen length

n – exponent in Ramberg Osgood relation

np – strain hardening exponent

Re – yield strength

Ru – ultimate strength

αm – ratio of failure strain to ultimate strain

αs – coefficient in parabolic relation of engineering stress strain based on the position of

ultimate stress and failure stress

βm – ratio of failure stress to ultimate stress

δ – longitudinal deformation

ε – strain

εeng – engineering strain

εeq,p – equivalent plastic strain

εf – failure strain

ε0 – characteristic failure strain

εtrue – true strain

εu – ultimate tensile strain

σ – stress

σeng – engineering stress

σeq – equivalent Von Mises stress

σh – hydrostatic stress

σtrue – true stress

ν – Poisson ratio
III
List of figures

Figure 1.1: Failure criterion of Lemaitre with a characteristic failure strain ...................... 2
Figure 2.1: Cross plate connection (Specimen type A) .................................................. 4
Figure 2.2: X-joint with low SCF (Specimen type B) ...................................................... 4
Figure 2.3: X-joint with high SCF (Specimen type C) ..................................................... 4
Figure 2.4: Static test ................................................................................................. 6
Figure 2.5: 10.000kN (left fig.) and 600kN (right fig.) test rig ........................................ 6
Figure 2.6: Instrumentation with strain gauges and displacement transducers ............... 7
Figure 3.1: Visualization of the characteristic material data for S690 .............................. 9
Figure 3.2: Geometry of simple coarse simulation....................................................... 10
Figure 3.3: Multilinear isotropic hardening material input in ANSYS.............................. 10
Figure 3.4: Material input .......................................................................................... 11
Figure 3.5: Boundary conditions ................................................................................ 11
Figure 3.6: Load-displacement comparison ................................................................ 12
Figure 3.7: Visualization of the characteristic material data for S690 12mm plate ......... 13
Figure 3.8: Resulting stress strain curves, S690 12mm plate ....................................... 14
Figure 3.9: Flow chart for creating the modified true stress strain curve ...................... 15
Figure 3.10: Comparison of true stress strain input ..................................................... 16
Figure 3.11: Simulation setup.................................................................................... 17
Figure 3.12: Comparison of engineering material input and FEM result, without necking18
Figure 3.13: Outline deformation, upper line .............................................................. 19
Figure 3.14: Comparison of the shapes at the required cross section ........................... 19
Figure 3.15: True stress strain comparison ................................................................. 20
Figure 3.16: 3D necking simulation specimen ............................................................. 21
Figure 3.17: Necking phenomena .............................................................................. 22
Figure 3.18: Comparison of engineering material input and FEM result, with necking ... 22
Figure 3.19: Comparison of 3D full FEM results and material input .............................. 23
Figure 3.20: Equivalent plastic strain ......................................................................... 23
Figure 3.21: Geometry I ........................................................................................... 25
Figure 3.22: Geometry II .......................................................................................... 25
Figure 3.23: Geometry III ......................................................................................... 26
Figure 3.24: Geometry IV ......................................................................................... 27
Figure 3.25: Load-displacement comparison............................................................... 27
Figure 3.26: FEM results of 120 elements .................................................................. 28
Figure 3.27: FEM results of 960 elements .................................................................. 29
Figure 3.28: FEM results of 4690 elements................................................................. 29
Figure 3.29: Load-displacement comparison............................................................... 30
Figure 3.30: Simulation geometry .............................................................................. 31
Figure 3.31: Boundary conditions .............................................................................. 31
Figure 3.32: Tensile test analysis, example, deformed shape and equivalent plastic strains
(3mm) .............................................................................................................. 32
Figure 3.33: Stress strain comparison for different element sizes ................................. 32

IV
Figure 3.34: Comparison of true stress strain input ..................................................... 33
Figure 3.35: FEM result after calibration, mesh size: 4 mm ......................................... 33
Figure 3.36: FEM result after calibration, mesh size: 3 mm ......................................... 33
Figure 3.37: FEM result after calibration, mesh size: 2 mm ......................................... 34
Figure 4.1: Specimen geometry B (with low SCF) ....................................................... 35
Figure 4.2: Side view of geometry B .......................................................................... 35
Figure 4.3: Specimen 3B1 ......................................................................................... 36
Figure 4.4: Measured geometry – top view (upper figure) and side vied (lower figure) . 36
Figure 4.5: FE-geometry specimen 3B1-3D full ........................................................... 37
Figure 4.6: FE-mesh specimen 3B1_refined near weld ................................................ 38
Figure 4.7: FE-mesh specimen 3B1_refined plate, 1 mm ............................................. 38
Figure 4.8: Boundary conditions ................................................................................ 38
Figure 4.9: FE-analyses results of Specimen 3B1 ........................................................ 39
Figure 4.10: Experimental results .............................................................................. 40
Figure 4.11: Determination of the upper and lower bounds of the experimental result .. 41
Figure 4.12: Comparison of experimental and numerical results .................................. 41
Figure 4.13: Results of 1 mm element size at a displacement of 30 mm (2*15) ............ 42
Figure 4.14: Results of 2 mm element size at a displacement of 30 mm (2*15) ............ 43
Figure 4.15: Lemaitre’s criterion check ....................................................................... 43
Figure 4.16: Equivalent plastic strain at a displacement of 1.5 mm (2*0.75, substep 5) 44
Figure 4.17: Equivalent plastic strain at failure (substep 56, 2*8.4=16.8 mm) .............. 44
Figure 4.18: Failure of specimen 4B1 ......................................................................... 45
Figure 4.19: Measured dimensions specimen 4B1 ....................................................... 45
Figure 4.20: Geometry and mesh specimen 4B1, 0.5 mm element size ........................ 46
Figure 4.21: Equivalent plastic strain at the end of FE-analysis .................................... 46
Figure 4.22: FE-analyses results of Specimen 4B1 ...................................................... 47
Figure 4.23: Equivalent plastic strain at a displacement of 0.96 mm (2*0.48) ............... 48
Figure 4.24: Lemaitre’s criterion check ....................................................................... 48
Figure 4.25: Comparison of experimental and numerical results .................................. 49
Figure 4.26: Specimen 1B1 ....................................................................................... 50
Figure 4.27: Measured dimensions of specimen 1B1 ................................................... 50
Figure 4.28: Geometry and mesh specimen 2B4, 2 mm element size ........................... 51
Figure 4.29: FE-analyses results of Specimen 1B1 ...................................................... 51
Figure 4.30: Determination of upper and lower bound of the experiment results .......... 52
Figure 4.31: Comparison of experimental and numerical results for specimen 1B1 ........ 52
Figure 4.32: Distribution of plastic strain in specimen (end of FEM analysis)................. 53
Figure 4.33: Specimen 2B4 ....................................................................................... 54
Figure 4.34: Measured dimensions specimen 2B4 ....................................................... 54
Figure 4.35: Geometry and mesh specimen 2B4, 5 mm element size at plate ............... 55
Figure 4.36: FE-analyses results of Specimen 2B4 ...................................................... 55
Figure 4.37: Comparison of experimental and numerical results for specimen 2B4 ........ 56
Figure 4.38: Over view of plastic strain at a displacement of 60 mm (2*30) ................. 56
Figure 4.39: Specimen 5B1 ....................................................................................... 57
Figure 4.40: Measured geometry specimen 5B1 ......................................................... 57
V
Figure 4.41: Geometry and mesh specimen 5B1, refined near weld, 3 mm................... 58
Figure 4.42: FE-analyses results of Specimen 5B1 ...................................................... 58
Figure 4.43: Comparison of experimental and numerical results for specimen 5B1 ........ 59
Figure 4.44: Equivalent plastic strain at a displacement of 3.6 mm .............................. 59
Figure 4.45: Specimen geometry C (with high SCF) .................................................... 60
Figure 4.46: Specimen 1C1 ....................................................................................... 61
Figure 4.47: Measured dimensions specimen 1C1 ....................................................... 61
Figure 4.48: Geometry and mesh specimen 1C1, refined near weld, 1 mm .................. 62
Figure 4.49: FE-analyses results of Specimen 1B1 ...................................................... 62
Figure 4.50: Equivalent plastic strain at the displacement of 14 mm ............................ 62
Figure 4.51: Determine of upper and lower bound of the experimental results ............. 63
Figure 4.52: Comparison of experimental and numerical results for specimen 1C1 ........ 63
Figure 4.53: Specimen 2C1 ....................................................................................... 64
Figure 4.54: Measured dimensions specimen 2C1 ....................................................... 64
Figure 4.55: Geometry and mesh specimen 2C1, refined near weld, 3 mm .................. 65
Figure 4.56: FE-analyses results of Specimen 2B1 ...................................................... 65
Figure 4.57: Equivalent plastic strain at the end of FE-analysis (3 mm element size) ..... 65
Figure 4.58: Comparison of experimental and numerical results for specimen 2C1 ........ 66
Figure 4.59: Specimen 3C1 ....................................................................................... 67
Figure 4.60: Measured dimensions specimen 3C1 ....................................................... 67
Figure 4.61: Geometry and mesh specimen 3C1, refined near weld, 0.7 mm ................ 68
Figure 4.62: FE-analyses results of Specimen 3B1 ...................................................... 68
Figure 4.63: Equivalent plastic strain at the end of FE-analysis (0.7 mm element size) .. 68
Figure 4.64: Comparison of experimental and numerical results for specimen 3C1 ........ 69
Figure 4.65: Stress concentration .............................................................................. 69
Figure 4.66: Specimen 4C1 ....................................................................................... 70
Figure 4.67: Measured dimensions specimen 4C1 ....................................................... 70
Figure 4.68: Geometry and mesh specimen 4C1, refined near weld, 0.7 mm ................ 71
Figure 4.69: FE-analyses results of Specimen 4B1 ...................................................... 71
Figure 4.70: Equivalent plastic strain at a displacement of 2 mm (0.7 mm element size)71
Figure 4.71: Comparison of experimental and numerical results for specimen 4C1 ........ 72
Figure 4.72: Specimen 5C1 ....................................................................................... 73
Figure 4.73: Measured dimensions specimen 5C1 ....................................................... 73
Figure 4.74: Geometry and mesh specimen 5C1, refined near weld, 3 mm ................... 74
Figure 4.75: FE-analyses results of Specimen 5B1 ...................................................... 74
Figure 4.76: Comparison of experimental and numerical results for specimen 5B1 ........ 75
Figure 4.77: Equivalent plastic strain at a displacement of 3 mm ................................. 75
Figure 4.78: Comparison connection 1 ....................................................................... 76
Figure 4.79: Comparison connection 2 ....................................................................... 77
Figure 4.80: Comparison connection 3 ....................................................................... 78
Figure 4.81: Comparison connection 4 ....................................................................... 79
Figure 4.82: Comparison connection 5 ....................................................................... 80
Figure A.1: Isotropic hardening ................................................................................. 84
Figure A.2: Kinematic hardening ................................................................................ 84
VI
Figure B.1: Simulation geometry for 40 mm thickness plate ........................................ 85
Figure B.2: Modified stress strain curves, S690 40 mm plate ....................................... 85
Figure B.3: Element size tests, S690 40 mm plate ...................................................... 85
Figure B.4: Comparison of true stress strain input after calibration, S690 40mm plate .. 86
Figure B.5: FEM result after calibration, S690 40 mm plate, mesh size: 10 mm ............ 86
Figure B.6: FEM result after calibration, S690 40 mm plate, mesh size: 10 mm ............ 86
Figure B.7: Simulation geometry for 10 mm thickness plate ........................................ 87
Figure B.8: Modified stress strain curves, S1100 10 mm plate ..................................... 87
Figure B.9: Element size tests, S1100 10 mm plate .................................................... 87
Figure B.10: Modified stress strain curves, S1100 40 mm plate ................................... 88
Figure B.11: Element size tests, S1100 40 mm plate ................................................... 88
Figure B.12: Modified stress strain curves, S690 12 mm undermatched weld ............... 89
Figure B.13: Element size tests, S690 12 mm undermatched weld ............................... 89
Figure B.14: Modified stress strain curves, S690 12 mm overmatched weld .................. 90
Figure B.15: Element size tests, S690 12 mm overmatched weld ................................. 90
Figure B.16: Modified stress strain curves, S690 40 mm overmatched weld .................. 91
Figure B.17: Element size tests, S690 40 mm overmatched weld ................................. 91
Figure B.18: Comparison of true stress strain input after calibration, S690 40 mm
overmatched weld ............................................................................................. 92
Figure B.19: FEM result after calibration, S690 40 mm overmatched weld, mesh size: 10
mm................................................................................................................... 92
Figure B.20: FEM result after calibration, S690 40 mm overmatched weld, mesh size: 5 mm
........................................................................................................................ 92
Figure B.21: Modified stress strain curves, S1100 10 mm undermatched weld .............. 93
Figure B.22: Element size tests, S1100 10 mm undermatched weld ............................. 93
Figure B.23: Comparison of true stress strain input after calibration, S1100 10 mm
undermatched weld ........................................................................................... 94
Figure B.24: FEM result after calibration, S1100 10 mm undermatched weld, mesh size: 4
mm................................................................................................................... 94
Figure B.25: FEM result after calibration, S1100 10 mm undermatched weld, mesh size: 2
mm................................................................................................................... 94
Figure B.26: Modified stress strain curves, S1100 40 mm undermatched weld .............. 95
Figure B.27: Element size tests, S1100 40 mm undermatched weld ............................. 95
Figure B.28: Comparison of true stress strain input after calibration, S1100 40 mm
undermatched weld ........................................................................................... 96
Figure B.29: FEM result after calibration, S1100 40 mm undermatched weld, mesh size: 10
mm................................................................................................................... 96
Figure B.30: FEM result after calibration, S1100 40 mm undermatched weld, mesh size: 5
mm................................................................................................................... 96

VII
List of tables

Table 2.1: Test plan .................................................................................................... 5

Table 3.1: Material specifications ................................................................................. 8
Table 3.2: Boundary condition simulation information ................................................. 24

VIII
1. Introduction

1.1 Background

Due to economic and technical advances, high strength steel (HSS) structures are more and
more used. However, the fabrication and design of HSS structures need special attention and
existing codes are lack of relevant design criterion (EN 1993-1-12 only gives additional rules to
steel grades S500 to S690). Therefore a research project called Integrity High Strength Steel
Structures, IHSSS, was carried out by TNO and Delft University of Technology for the
investigation of the applicability of HSS with yield strength up to 1100 MPa.

Normally, HSS has a high yield to tensile strength ratio which means that the deformation
capacity will be lower in connections and structural parts with (bolt) holes or other area
reducing effects (e.g. fatigue cracks). Furthermore, for HSS, especially the one has yield
strength as high as 1100 MPa, it cannot always be welded with overmatched weld material.
Therefore the deformation capacity of these joints may be limited due to failure of the weld
before yielding in the connection members occur. Because of these reasons, the safety of the
structure is declined and the application of a fully plastic design for HSS is restricted. Therefore
the deformation capacity of joints in HSS structures is essential and has been investigated in
this project.

In the IHSSS research project, experiments had been carried out on three types of welded
joints (Cross plate, X-joint with low stress concentration factor (SCF) and X-joint with high SCF).
However, the FEM validation analyses had only been carried out for a limited number of joints.
As a continuation on the basic FE-analyses, it is proposed to carry out validation analyses on
other joint geometries.

1.2 Goal of the research

Based on the former research performed by TNO, the behavior of a welded joint of HSS can
be nicely simulated by FEM. The aim of this master project is to continue the basic
FE-analyses, cross comparison the FEM results with a modern commercial FEM software and
carry out validation analyses on other joint types performed in the experiments. Then try to
finalize a method to accurately predict both strength and deformation capacity for different
types of (undermatched) welded joints for HSS by using FEM software.

1.3 Ductile failure criteria [7]

Ductile failure can be defined as a situation in which the plastic strains become too large.
Therefore, this strain level is considered as the failure criterion. Obviously, failure depends on
the material itself and the stress state (e.g. tension versus shear). High tensile stresses in all
directions (high triaxiality) will result in a low deformation capacity. Previous researches show
that the failure criterion of Lemaitre [1] provides decent results for high strength steel.

1
According to Lemaitre, failure occurs when the equivalent (“Von Mises”) plastic strain (ε eq.p)
exceeds the failure strain (εf).

The material is described by the Poisson’s ratio ν (=0.3 for steel) and the characteristic failure
strain ε0 (taken 1.0 in this report).

Whether the stress-state is either one-dimensional or three-dimensional has a substantial

influence. Therefore, triaxiality is introduced by including the hydrostatic pressure (σh) and the
equivalent stress (σeq).

Therefore, Lemaitre describes the failure criterion for the equivalent plastic strain as follows
[8]:

εf= (1.1)

In which the hydrostatic pressure σh equals:

σh= ( σ1+ σ2+ σ3) (1.2)

Whereas the equivalent, stress equals:

σeq= (1.3)

And σh/σeq is the so-called triaxiality, the allowable plastic strain decreases with increasing
triaxiality.

In previous analyses research, ε0 values ranging from 0.4 to 1.0. These values were based on
literature and can be regarded as a scatter band for ε0 values in several steels. And the results
showed that in general the overall deformation capacity does not vary much for the ε0 values
used [8].

Lemaitre’s failure criterion can be visualized in Figure 1.1 [8]. By introducing triaxiality,
Lemaitre’s criterion can give a better prediction than a one-dimensional failure criterion.

Figure 1.1: Failure criterion of Lemaitre with a characteristic failure strain

2
1.4 Approach

The approaches which are used in this report to solve the problem are listed as follows.

1) Literature study;

2) Study ANSYS;

3) Determine material behavior and input data for FE-analyses;

4) Model and simulate connections for Specimen series B;

5) Model and simulate connections for Specimen series C;

6) Compare the results and make conclusions.

3
2. Test overview

2.1 Joint geometry

The geometry tested in the experimental part must be practical and serve as a basis for the
modeling at the same time. Therefore 3 different joints with and without an SCF are tested in
the program.

The following geometry types are chosen (Figure 2.1-Figure 2.3) [3]:

 Type A: Cross plate connection with fillet welds loaded in shear

 Type B: X-joint with load carrying full penetration welds with low SCF

 Type C: X-joint with load carrying full penetration welds with high SCF

In type A the weld is loaded in shear. So there will be unequal stress distribution along the weld
and at the end of the plates there is an SCF. In type B there are load carrying full penetration
welds and there is no SCF along the weld. In type C there are also load carrying penetration
welds. Due to the rotated connection of the transverse plate on each side of the connecting
plate there is a high SCF along the weld.

Figure 2.1: Cross plate connection (Specimen type A)

Figure 2.2: X-joint with low SCF (Specimen type B)

Figure 2.3: X-joint with high SCF (Specimen type C)

4
2.2 Test parameters
The variables to be investigated in the test program are:

 Specimen type (A, B and C, connection A will not be investigated in this report);

 Test type (static, fatigue and ratcheting, the latter two will not be investigated in this
report);

 Material thickness (10/12 and 40 mm);

 Base material strength (690 and 1100 MPa);

 Relative weld strength (overmatched, only for S690 base material and undermatched, for
both S690 and S1100 base materials).

A review of the test program including 48 tests is shown in Table 2.1 [4]. Each test series
includes specimens of type A, B and C.

Table 2.1: Test plan

Number of
Test Series Re base / MPa Re weld / MPa Plate/mm Parameter
tests

I/VII * 12 690 900 12/4 Reference

II 9 690 900 40 Thickness

III/VIII * 12 690 490 12/4 Under matched

IV 9 1100 900 10 High strength

Thickness/ High
V 6 1100 900 40
strength

*: Results for A-type test specimens from Series I and A-type test specimens from Series III were included, even though
these test specimens presented wrong fabrication. Series VII and VIII are substitutes of those specimens, and are also
presented.

5
2.3 Static test procedure
In the static tests the specimens will be loaded until failure occurs. The tests will be
displacement controlled in order to get a stable curve to failure. The main result will be a
load-deformation curve (Figure 2.4) [3]. The strain gauges will be fixed on the specimen at
specific locations for detailed information. As given in Figure 2.4, it is expected that for an
overmatched weld the failure load and the deformation capacity will be higher.

Figure 2.4: Static test

2.4 Test rig and measurements

Two servo hydraulic test rigs are available to load test specimen to failure. The capacity of these
rigs are 600kN respectively 10000kN (Figure 2.5) [3]. Deformation of the weld and deformation
of the whole connection of all test specimens can be locally tested by displacement transducers.
Single strain gauges are used to measure the nominal stress in the connection plates. Strip
strain gauges are used to obtain information about the strain development at the hot spot
stress location of the welded connection. A typical example is shown in Figure 2.6 [3].

Figure 2.5: 10.000kN (left fig.) and 600kN (right fig.) test rig

6
Locally means that the measuring length includes the weld and the transverse plate (LVDT_1
and LVDT_2). The measuring length for the overall measurements is about 10 times the plate
thickness (LVDT_3 and LVDT_4). Single strain gauges are used to measure the nominal stress
in the connection plates. Strip strain gauges are used to obtain information about the strain
development at the hot spot stress location of the welded connection.

Figure 2.6: Instrumentation with strain gauges and displacement transducers

7
3. Material properties

3.1 Material specification

There are two parent materials (S690 and S1100) in the overall project. For the S690 material,
there are both undermatched and overmatched welds. For the S1100 only undermatched weld
is available. Real material properties like Re, Rm and εu are available and used in developing the
material input for FE-analyses. All the initial material properties are adopted from the TNO
report [8], and given in Table 3.1.

Table 3.1: Material specifications

S
Re Rm εu K np
Specimen Plate/weld T [mm] [Mpa]
[Mpa] [Mpa] [-] [Mpa] [-]
O/U

Plate 12 690 757 820 18% 1060 0.07

1
Weld Overmatched 864 1079 15% 1300 0.045

Plate 40 690 792 835 16.1% 1050 0.06

2
Weld Overmatched 864 1079 15% 1300 0.045

Plate 12 690 757 820 18% 1060 0.07

3
Weld Undermatched 500 600 15% 732 0.05

Plate 10 1100 1179 1432 11% 1500 0.003

4
Weld Undermatched 728 777 15% 950 0.05

Plate 40 1100 1106 1325 11% 1375 0.0001

5
Weld Undermatched 931 1061 14% 1250 0.037

Table 3.1 represents actual material behaviors under material tests and the load-displacement
relation has been scaled into engineering stress-strain.

In order to perform geometrical and physical non-linear FE-analyses, ANSYS requires the so
called true stress strain curve as an input. This curve is based on the material specifications
described in Table 3.1 (parameters K and np will be used later in creating the true stress strain
curve in the FE-analyses). In simulations of this report, same procedure and assumptions to
make the curve will be used based on [8], and the elastic modulus E is taken as 210000 N/mm 2.

8
3.2 Material properties

3.2.1 Stress strain relation

As the first step, a simple stress strain relation is adopted. In this section, the same procedure
and assumption based on [7] is used to create the material stress strain curve. S690 12 mm
plate material properties from Table 3.1 are numerically simulated in this section. The
engineering stress strain curve is generated from Ramberg-Osgood relation:

(3.1)

Where substitution of ultimate strength point (Rm at 1/3 εu) can result the parameter n.

Then based on engineering stress strain curves, true stress strain curves can be theoretically
determined. εeng and σeng are taken from Equation 3.1.

εtrue=ln(1+εeng) (3.2)

σtrue=σeng(1+εeng) (3.3)

The resulting stress strain curves are shown in Figure 3.1.

1600

1400

1200
Stress [N/mm2]

1000

800

600

400
Engineering curve
200
True curve
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Strain [-]

Figure 3.1: Visualization of the characteristic material data for S690

To validate the true stress strain curves obtained from previous work and test the nonlinear
analysis of the numerical software, tensile simulations are performed using the commercial
FE-software, ANSYS Workbench. These simulations are done in both 2D and 3D.

A very simple random size rectangular bar is tested. The analyses represent respectively plane
stress, plane strain and 3D-full situation. Due to the symmetry of geometry and loading, only
one-eighth of the material specimen is modeled (Figure 3.2).

9
 (a)

(b)
Figure 3.2: Geometry of simple coarse simulation
(a) 2D geometry, (b) 3D geometry

To include material nonlinearity, the true stress strain curve is used as the input. The isotropic
hardening rule is chosen for materials. More details about hardening rule can be found in
Appendix A.

For ANSYS, the input data requires both linear elastic material properties (black line, E & ν) and
metal plastic hardening data (red line, stress and plastic strain relation) (Figure 3.3, Figure 3.4):

Figure 3.3: Multilinear isotropic hardening material input in ANSYS

For plastic hardening part, the plastic strain can be calculated from Equation 3.4.

(3.4)

10
 Figure 3.4: Material input

The mesh is automatically generated and there is no preference for the size of elements, SOLID
186 (20 nodes brick) element is used in all the analyses. In the analyses, large deflection is
activated. The calculation procedure and time stepping method are program controlled. The
boundary conditions are only applied in the area of symmetry. Finally, for the loading,
displacement control is used where this load is increased gradually for each sub-step, and the
total longitudinal displacement is given as 30 mm (Figure 3.5). In these analyses necking
doesn’t happen.

Figure 3.5: Boundary conditions

The “Frictionless Support” in ANSYS is used to restrain normal displacement, so it could be used
on symmetrical area.

11
The resulting load-displacement curves are presented in Figure 3.6.

1200

1000

800
Load [kN]

600

400
eng_input
plane stress
200 plane strain
3D_full
0
0 5 10 15 20 25 30 35
Displacement [mm]
Figure 3.6: Load-displacement comparison

From Figure 3.6 it can be seen that:

1. In the elastic range the plane strain analysis is stiffer than the other analyses because
lateral contraction is restricted;
2. The plane stress situation and 3D full situation are in good agreement with the engineering
input data;
3. Because of the limitation of the material input and the geometry without imperfection,
necking doesn’t happen.
It is concluded that for this specific geometry and material input the plane stress and 3D full
analyses can provide good results.

12
3.2.2 Modified stress strain relation
Then a more sophisticated stress strain relation is used to simulate a more realistic material
behavior. This time a downward part is added to engineering stress strain relation after the
stress reaches Rm. In this section, the same procedure and assumption based on [8] is used to
create the material stress strain curve. The input data can be generated from the following
steps.

Firstly, 3 points will be constructed based on Table 3.1. The first point is Re positioned at strain
εe. The second point is Rm at the assumed strain of one third of ε u. The last one is the failure
stress (approximately 0.6 Rm) at εu. The results are presented in Figure 3.7.

900
800
700
Stress [N/mm2]

600
500
400
300
200
S690 original data
100
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure 3.7: Visualization of the characteristic material data for S690 12mm plate

Then based on Figure 3.7, the engineering stress strain curve can be made. The steps to create
this curve are described as follows.

1) Before the yield point, use the linear elastic relation.

(3.5)

2) The upward part of the curve can also be approximated by the Ramberg-Osgood
relation.

(3.6)

Where substitution of Rm and 1/3 εu can result the parameter n.

3) The downward part can be approximated by a parabolic relation.

(3.7)

Where

αm=1/3; βm=0.6

13
Finally, the true stress strain curve can be created based on engineering curve. The steps are as
follows.

1) Before the yield point, same with the engineering curve (Eq. 3.5).

2) For the upward part, the true curve can be theoretically determined from the engineering
curve. εeng and σeng are taken from Equation 3.6.

εtrue=ln(1+εeng) (3.8)

σtrue=σeng(1+εeng) (3.9)

3) For the downward part a formula called Power Law is used to generate the true stress
strain curve.

σtrue=K (3.10)

Theoretically, parameters K and np can be determined by an iterative approach with the

following steps:

1. Make an assumption for K and np in the Power Law;

2. Execute an FE-analysis resembling a tensile test;

3. Adapt the initial assumption for K and np to the outcome of the FE-analysis and start again
at step 1 for the next iteration step.

In this section, the K and np are directly adopted from Table 3.1. The resulting stress strain
curves are shown in Figure 3.8.

1100
1000
900
800
Stress [N/mm2]

700
600
500
400
300 S690_12mm plate_Original data

200 S690_12mm plate_Engineering

100 S690_12mm plate_True
0
0 0.04 0.08 0.12 0.16 0.2 0.24 0.28
Strain [-]
Figure 3.8: Resulting stress strain curves, S690 12mm plate

14
A flow chart to generate modified stress strain curves is given in Figure 3.9.

900
Plot input data: 800
700
1. Re positioned at εe 600
500
2. Rm positioned at 1/3 εu 400
3. 0.6 Rm positioned at εu 300
200
100
0
0 0.05 0.1 0.15 0.2

Engineering stress strain curve:

1. Linear relation

900
800
2. Fitted the upward part using 700
Ramberg Osgood relation: 600
500
400
300
200
3. Fitted the downward part 100
using Parabolic relation: 0
0 0.05 0.1 0.15 0.2

True stress strain curve: 1100

1000
1. Linear relation 900
800
700
600
2. Upward part using theoretic 500
400
relation: 300
200
εtrue=ln(1+εeng) 100
σtrue=σeng(1+εeng 0
0 0.05 0.1 0.15 0.2 0.25 0.3
3. Downward part using the
Power Law:
σtrue=K

Figure 3.9: Flow chart for creating the modified true stress strain curve

15
The comparison of true stress strain curve input used in Section 3.2.1 and the modified curve
based on S690 plate material is visualized in Figure 3.10.

1600

1400

1200
Stress [N/mm2]

1000

800

600
Old true curve
400
Modified true curve
200

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Strain [-]

Figure 3.10: Comparison of true stress strain input

16
3.2.3 Tensile test simulation (without necking)
To validate the modified true stress strain curves obtained from the former section, tensile tests
are performed as well.

The same geometries and boundary conditions as Section 3.2.1 are tested (Figure 3.11). The
analyses also represent respectively plane stress, plane strain and 3D situation. The mesh is
automatically generated.

(a)

(b)

(c)
Figure 3.11: Simulation setup
(a) 2D geometry, (b) 3D geometry, (c) boundary conditions

17
The reaction force (F) has been scaled to the nominal engineering stress σ (=F/A0, A0 is the
cross sectional area of the specimen before the test) in order to compare the results with
engineering material inputs. Figure 3.12 presents the FEM results as well as the 12 mm S690
plate material engineering input.

1200

1000
Stress [N/mm2]

800

600

400 Plane stress

Eng-input
200 Plane strain
3D full
0
0 5 10 15 20 25 30 35 40
Displacement [mm]

Figure 3.12: Comparison of engineering material input and FEM result, without necking

From Figure 3.12 we can see that

1) Same as before, in the elastic range the plane strain analysis is stiffer than the other
analyses because lateral contraction is restricted;

2) The plane stress situation and 3D full situation are still in good agreement, but since
necking doesn’t happen (because of there is no imperfection in geometry), they no longer
fitting the engineering input.

True stress strain relation has also been compared to have a more direct and accurate
comparison.

The deformed cross sectional area (A1) can be calculated from following steps (in order to have
a clear demonstration, the necking model from next section is shown here).

1) Exporting nodal deformation of the outlines of required cross section

(a)

18
 (b)
Figure 3.13: Outline deformation, upper line
(a)uy, (b)uz

2) Calculate the nodal coordinates after deformation

3) Put the coordinates into AutoCAD, plot a closed figure (Figure 3.14)

(a)

(b)
Figure 3.14: Comparison of the shapes at the required cross section
(a) FEM results, (b) Plot in AutoCAD

4) AutoCAD can calculate the area automatically.

19
The reaction force (F) has been scaled to the true stress σtrue (=F/A1) and the nominal true
strain is calculated from longitudinal displacement over original length (=δ/L).

1200

1000
stress [N/mm2]

800

600
3D full FEM results-true
400 stress strain
200 Input curve-true stress
strain curve
0
0 0.1 0.2 0.3 0.4 0.5
strain [-]
Figure 3.15: True stress strain comparison

From Figure 3.15 we can see that the 3D full simulation result is in good agreement with the
true stress strain input.

It is concluded that 3D full analyses provide good results when necking doesn’t happen.

20
3.2.4 Tensile test simulation (with necking)

3.2.4.1 General simulation

Then the specimen under necking situation is tested. In order to have necking in the
appropriate part of the tensile test specimen, an imperfection is introduced to the geometry.
This is done by reducing 2% of the height of the specimen part where the necking is expected
to occur, the reduction is gradually over the whole length. And the boundary conditions are still
like the ones used before (Figure 3.16).

(a)

(b)
Figure 3.16: 3D necking simulation specimen
(a) Geometry, (b) boundary conditions

The material input is still the true stress strain curve generated from Section 3.2.2. In this
general simulation the mesh is automatically generated (relatively coarse). The results are
present in Figure 3.17.

(a)

21
 (b)
Figure 3.17: Necking phenomena
(a) Directional deformation, (b) normal stress (longitudinal direction)

From Figure 3.17 we can see that the necking phenomena actually occur after calculating.

The minimum normal stress happens somewhere behind the necking area. However, the
minimum stress value is -674 MPa, which means that a small part of the specimen is under
compression, while the whole specimen is under axial tension. This phenomenon is confirmed
by another FEM software, Abaqus. One possible reason is that the given displacement in the
simulation is too large and the ‘static analysis’ is no longer applicable. In reality, the specimen
will fail before having such a large deformation.

The reaction force has also been scaled to the nominal engineering stress σ and compared with
engineering input (Figure 3.18).

1000

900

800

700
Stress [N/mm2]

600

500

400 Engineerin input

300 3D full

200 Plane strain

100
Plane stress
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure 3.18: Comparison of engineering material input and FEM result, with necking

22
From Figure 3.18 we can see that:

1) Same as before, in the elastic range the plane strain analysis is stiffer than the other
analyses;
2) The resulting engineering stress goes down after necking in all analytic situations, but they
all drop much quicker than material input;
3) 3D-full analysis gives the closest result to material input;
4) None of the analyses give the result in accordance with the material input.

Then true stress strain relation has also been compared in 3D-full analysis situation (Figure
3.19). The way to get cross sectional area and determine the true stress strain has been
described in Section 3.2.3.

2500
Input-True
2000 Input-Engineering
Stress [N/mm2]

3D full-True
1500
3D full-Engineering
1000

500

0
0 0.05 0.1 Strain [-] 0.15 0.2 0.25

Figure 3.19: Comparison of 3D full FEM results and material input

From Figure 3.19 we can see that the FEM resulting true stress strain doesn’t match the input
data anymore. This is because after necking, some of the elements will be highly distorted and
the nodal stress strain relation will be complicated within the element (localized plastic strain at
necking area can be more than 1, Figure 3.20). This means that the true stress strain can’t be
simply determined from the way described in Section 2.2.3. In ANSYS only the micro nodal
equivalent stress strain relation follows the actual input data, but not the macro structural
reaction. So for specimen under necking, the comparison has to be made based on engineering
results.

Figure 3.20: Equivalent plastic strain

23
3.2.4.2 Boundary condition (cross-sectional area) simulation
In order to find why the difference in Figure 3.18 happens, more simulations are made. Firstly,
the influence of boundary condition is tested. Every simulation setup is the same as Section
3.2.4.1 except boundary conditions.

In total four analyses are made with various restraints on the lower (y=0) surface in Y-direction
and on right (z=0) surface in the Z-direction. The x=0 end is fixed in the longitudinal direction
(X-direction). This means that the “real cross-section” of the geometry is according to the
values given in Table 3.2.

Table 3.2: Boundary condition simulation information

Restraint y=0 surface Restraint z=0 surface Resulting “real cross section”
Geometry
in Y-direction in Z-direction due to the symmetry conditions

I No No 20 mm x 30 mm

II Yes No 40 mm x 30 mm

III No Yes 20 mm x 60 mm

IV Yes Yes 40 mm x 60 mm

So, the cross section II and III are twice the cross section of the I geometry and the cross
section IV is four times the cross section of the A geometry. The boundary condition, directional
deformation, longitudinal normal stress are shown as follows (Figure 3.21-Figure 3.24).

1) Geometry I

(a)

(b)
24
 (c)
Figure 3.21: Geometry I
(a) Boundary condition, (b) directional deformation, (c) normal stress

2) Geometry II

(a)

(b)

(c)
Figure 3.22: Geometry II
(a) Boundary condition, (b) directional deformation, (c) normal stress

25
3) Geometry III

(a)

(b)

(c)
Figure 3.23: Geometry III
(a) Boundary condition, (b) directional deformation, (c) normal stress

4) Geometry IV

(a)

26
 (b)

(c)
Figure 3.24: Geometry IV
(a) Boundary condition, (b) directional deformation, (c) normal stress

The load-displacement comparison is shown in Figure 3.25.

6.00E+05

Geometry I
5.00E+05 Geometry II
Geometry III
4.00E+05 Geometry IV

3.00E+05

2.00E+05

1.00E+05

0.00E+00
0 5 10 15 20 25 30 35
Figure 3.25: Load-displacement comparison

During the necking process geometry with a larger cross-section will have a stiffer behavior due
to the fact that in the center of the cross section a more plane strain situation will occur. This is
in accordance with the simulation results.

27
3.2.4.3 Element size test
The influence of element size has also been tested. Every simulation setup is the same as
Section 3.2.4.1 except the element size.

There are three different element sizes tested in this section, resulting: 120 elements, 960
elements and 4690 elements. The directional deformation, normal stress as well as the mesh of
specimen are shown as follows (Figure 3.26-Figure 3.28).

1) 120 elements (element size: 10 mm)

(a)

(b)
Figure 3.26: FEM results of 120 elements
(a) Directional deformation, (b) normal stress

2) 960 elements (element size: 5 mm)

(a)

28
 (b)
Figure 3.27: FEM results of 960 elements
(a) Directional deformation, (b) normal stress

3) 4690 elements (element size: 3 mm)

(a)

(b)
Figure 3.28: FEM results of 4690 elements
(a) Directional deformation, (b) normal stress

29
The load-displacement comparison is given in Figure 3.29.

600000
120 elements

960 elements
500000
4690 elements

Engineering input
Reaction force [N]

400000 resulting curve

300000

200000

100000

0
0 5 10 15 20 25 30 35 40
displacement [mm]
Figure 3.29: Load-displacement comparison

It can be seen that for large directional deformation, the reaction forces have obvious
differences with different element sizes. It is well known that the smaller the elements the more
pronounced the influence of stress concentrations and other irregularities are present in the
results of analyses, so the influence of necking is more pronounced.

30
3.2.5 Calibrated true stress strain input data
From Section 3.2.4.2 and 3.2.4.3 it is obvious that both the cross sectional area and element
size can affect the FEM simulation results. So the difference in Figure 3.18 might because of
different simulative geometry and element size used in the simulation.

To prove this, another series of simulation are made.

3.2.5.1 Geometry & element size

The geometry used in actual specimen experiments is adopted this time (12 mm x 60 mm) and
this geometry is also similar to the one used in TNO report [8] when determine parameter K
and np in Table 3.1.

Due to the symmetry of the geometry, one eighth of the specimen is modeled (Figure 3.30),
boundary conditions are shown in Figure 3.31 and an example of the result is present in Figure
3.32.

Figure 3.30: Simulation geometry

Figure 3.31: Boundary conditions

31
 Figure 3.32: Tensile test analysis, example, deformed shape and equivalent plastic strains (3mm)

The influence of element size is also tested and the overall comparison is shown in Figure 3.33.
From this figure, it can be seen that the FEM resulting engineering stress strain is close to the
calculated engineering input and the relatively coarse mesh (which is more similar to the case
in TNO report [8]) gives a closer result.

900

800

700

600
Stress [N/mm2]

500
Engineering_input
400
1 mm element size
300
2 mm element size
200
3 mm element size
100 4 mm element size
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure 3.33: Stress strain comparison for different element sizes

3.2.5.2 Calibration
From Figure 3.33 it can be seen that the FEM results have small differences from engineering
input data when the strain is very large. So in order to have a more accurate simulation result,
a calibration of the true stress strain input for ANSYS is needed. Several reports such as [13]
have dealt with the problem of modeling the correct behavior after necking. However, the
values of true stress strain input data after necking are manually modified in this report. Several
calibrations are done based on different element sizes (Figure 3.34). After this data calibration
procedure, a more similar output curve could finally be established (Figure 3.35-Figure 3.37).

32
 1800

1600

1400

1200
Stress [N/mm]

1000
Modified true stress strain
800 input (before calibration)
True stress strain input
600
after calibration_2 mm
400 True stress strain input
after calibration_3 mm
200 True stress strain input
after calibration_4 mm
0
0 0.5 1 1.5 2
Strain[-]

Figure 3.34: Comparison of true stress strain input

900
800
700
Stress [N/mm]

600
500
400
300
200 Engineering input
100 FEM result_after calibration 4 mm
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure 3.35: FEM result after calibration, mesh size: 4 mm

900
800
700
Stress [N/mm]

600
500
400
300
200 Engineering input
100 FEM result_after calibration 3 mm
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure 3.36: FEM result after calibration, mesh size: 3 mm

33
 900
800
700

Stress [N/mm]
600
500
400
300
200
Engineering input
100 FEM result_after calibration 2mm
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure 3.37: FEM result after calibration, mesh size: 2 mm

Theoretically, a specific element size should use the corresponding calibrated material input.
However, for element sizes smaller than 2 mm, the calibrated true stress strain inputs only have
very small differences, so for any element sizes smaller than 2 mm, the 2 mm input is adopted
for subsequent numerical simulations.

Other material inputs can be found in Appendix B.

34
4. FE validations of connections

4.1 Connections type B

In this section, validations are carried out for five B-geometries. The B-geometry is an
X-connection with a low stress concentration factor (SCF) and the simplified geometry is
presented in Figure 4.1 [8]. This geometry has no geometrical discontinuity in the direction of
the weld and therefore it can be approximated by 2D geometries (Figure 4.2 [8]), however, it
has been proven that the 2D and 3D analytical results have relatively big differences when the
strain became very large (Figure 3.18). So in this chapter only 3D analyses are carried out.

Figure 4.1: Specimen geometry B (with low SCF)

Figure 4.2: Side view of geometry B

In the next sections the FEM results are compared to the experimental results. Specimen 3B1
and 4B1 are the first validation made and therefore extensively investigated with different
variables.

The comparison is mainly based on the load-displacement curve over the length of test gauges.
For specimen 1B1 and 3B1 the measuring equipment did not function until the end of the test
and an additional interpretation of the experimental results is necessary.

35
4.1.1 Specimen 3B1 (fail in plate)

4.1.1.1 Geometry
Specimen 3B1 is a 12 mm thick S690 undermatched weld specimen. The geometry is presented
in Figure 4.3 [8]. This specimen consists of two horizontal plates and one vertical plate. The
plates are connected by undermatched K-joint. In this case the K-joint is not fully penetrated
and has a small (vertical) incomplete penetration.

Figure 4.3: Specimen 3B1

The dimensions of the specimen are taken from report [8] (Figure 4.4). Figure 4.4 also includes
the displacement measurement (LVDT’s). The side view shows that upper right weld has the
smallest dimensions, so only this quarter of the geometry is modeled due to double symmetry.
A length of 74.5 mm, half of the measuring length of LVDT’s 3 and 4 is used.

Figure 4.4: Measured geometry – top view (upper figure) and side vied (lower figure)

36
4.1.1.2 Modeling
Note that the geometry studied in FEM presents a simplification of the actual situation: it
assumes a uniform and constant plate and weld over the whole specimen. The weld is
simplified to a triangular geometry. The influence of the heat affect zone (HAZ), weld beads,
surface irregularities and weld toe radii is not taken into consideration. However, the incomplete
penetration is taken into account by assuming a vertical gap of 2 mm over the plate thickness
(Figure 4.5).

Figure 4.5: FE-geometry specimen 3B1-3D full

In order to test the influence of element size, several different meshes are used. Some
examples of the mesh are presented in Figure 4.6 and Figure 4.7. SOLID 186 (20 nodes brick)
element is used for all analyses.

(a)

37
 (b)
Figure 4.6: FE-mesh specimen 3B1_refined near weld
(a) general mesh, (b) weld interface detail, 0.7 mm element size

Figure 4.7: FE-mesh specimen 3B1_refined plate, 1 mm

Again, the boundary conditions only applied in the area of symmetry, a directional displacement
is given as the external load (Figure 4.8). All the parts are connected by “bond connection”
except the 2 mm vertical defect (between the gray and green parts in Figure 4.7).

Figure 4.8: Boundary conditions

38
The resulting load-displacement curves are presented in Figure 4.9. Apart from the influence of
different mesh at different parts of the specimen, different material inputs are also compared
(besides of ‘original input’ all simulations are used calibrated input corresponds with element
size). And the influence of the number of substeps is also checked.

700

600

500

Original input_all 4 mm
Load [kN]

400
Coarse_all 4 mm
300 Refined_near weld 2 mm
Refined_near weld and plate 2 mm_25 substeps
200 Refined_plate 2 mm_25 substeps
Refined_plate 2 mm_100 substeps
100 Refined_near weld 0.7 mm
Refined_plate 1 mm
0
0 5 10 15 [mm]
Displacement 20 25 30

Figure 4.9: FE-analyses results of Specimen 3B1

Various observations can be made:

1) For all the analyses, the resulting curves are identical before necking and only have
relatively small differences when the directional deformation is very large;

2) For different element sizes with corresponding calibrated material input, the results only
have very small differences. However, with the original (before calibration) material input,
the deference is relatively big, so the calibration can give a more accurate simulation result;

3) The number of substeps also has a slight influence on the simulation result;

4) Even smaller element sizes are tested for materials both near weld and plate (respectively
0.7 mm and 1 mm). However, all the simulation results show that the failure is at the plate,
so the refinement at plate should give a more accurate result.

4.1.1.3 Interpretation of experimental results [8]

The method of interpretation of experimental results in this report is the same as [8]. The
length of the model is taken equal to the displacement gauge LVDT’s 3 and 4. The experimental
load-displacement curves of these LVDT’s are plotted in Figure 4.10. It also includes the
‘corrected bench displacement’, which is the displacement of the ends of test rig scaled with a
factor 0.15 to compensate the difference in measuring length.

From Figure 4.10, several observations can be made. Firstly, it can be seen that the elastic
branches are not identical, which indicates that a certain curvature over the width of the
specimen. Secondly, the figure also shows that the LVDT’s fell off the specimen before failure,
so a correction procedure is needed if we want to have information at failure. Finally, the
39
corrected bench result does not present accurate results at large displacement because after
the elastic phase, the plastic deformations occur locally near the necking point, most likely
within the range of the LVDT’s.

700

600

500
Load [kN]

400

300

LVDT_3
200
LVDT_4
100 Bench_corr(*0.15)

0
0 1 2 3 4 5 6
Displacement [mm]
Figure 4.10: Experimental results

However, the exact moment when necking starts is not clear, so the experimental results are
modified with the following lower bound and upper bound [8]:

Lower bound

1) Up to failure of LVDT3, the average results of LVDT’s 3 and 4;

2) Until the maximum load Fu is reached, the displacement increase of the corrected bench;

3) At displacements beyond Fu, the displacement increase of the not-corrected bench.

Upper bound

1) Up to failure of LVDT3, the average results of LVDT’s 3 and 4;

2) At displacements beyond failure of LVDT3, the displacement increase of the not-corrected

bench.

The results are presented in Figure 4.11. It shows that between the upper and lower bound
there is a very wide range, and the actual deformations beyond the failure of the LVDT’S are
substantial.

40
 700

600

500
Load [kN]
400

300
LVDT_3
200 LVDT_4
Bench_corr(*0.15)
100 Lower bound
Upper bound
0
0 5 10 15 20 25
Displacement [mm]

Figure 4.11: Determination of the upper and lower bounds of the experimental result

4.1.1.4 Comparison of experimental and FE-results

The FE-result compared to the experimental result is shown in Figure 4.12. In addition, the
failure point based on Lemaitre’s criterion is presented, using the value of ε0=1.

700

600

500
Load [kN]

400

300
Lower bound
200 Upper bound
FEM_3D full
100
Lemaitre e0=1
0
0 5 10 15 20 25
Displacement [mm]

Figure 4.12: Comparison of experimental and numerical results

Several remarks can be made with respect to this figure:

1) In the elastic range the analyses agree with the experiment, the yield strength is slightly
over predicted (3%). This difference might be attributed to the influence of HAZ;

2) The plastic capacity is also slightly overestimated by the FE-analysis (3%);

3) The FE-analysis failure result is positioned well within the range described by the upper and
lower bounds after necking. The failure load can be predicted accurately;

4) The failure mode (plate failure) is predicted very well. According to FE-analysis, failure

41
 starts at the center of the horizontal plate.

The influence of element sizes on Lemaitre’s criterion is also checked (Figure 4.13 - Figure 4.15).
It seems that with a smaller element size, the plastic strain near the necking point is more
concentrated and has a larger maximum value. Meanwhile, the triaxiality also has a difference.
So the Lemaitre’s criterion might give different results when using different element sizes.

(a)

(b)
Figure 4.13: Results of 1 mm element size at a displacement of 30 mm (2*15)
(a) Equivalent plastic strain (b) triaxiality

(a)

42
 (b)
Figure 4.14: Results of 2 mm element size at a displacement of 30 mm (2*15)
(a) Equivalent plastic strain (b) triaxiality

Figure 4.15 checks the Lemaitre’s criterion for two different element sizes. It can be seen that
at later substeps two curves are actually different. However, the failure seems to occur at a very
early stage, and the result shows that for both element sizes the failure happens in the 56th
substep out of 100 total steps. So for Specimen 3B1 (failure at plate) the element size doesn’t
affect the Lemaitre’s criterion result and the result is considered accurate.

Lemaitre criterion
2.5
Plate_1 mm
Equivalent plastic strain [-]

Plate_2 mm
2

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Triaxiality [-]

Figure 4.15: Lemaitre’s criterion check

43
Figure 4.16 and Figure 4.17 show equivalent plastic strain at different substeps for the 3D full
analysis.

Figure 4.16: Equivalent plastic strain at a displacement of 1.5 mm (2*0.75, substep 5)

Figure 4.17: Equivalent plastic strain at failure (substep 56, 2*8.4=16.8 mm)

From the analysis it could be concluded that the initial plasticity occurs in the weld (Figure 4.16).
This phenomenon can also be found with undermatched weld metal. However, the amount of
plasticity remained limited. It seems that the dimensions of the weld have the overall strength
exceeds that of the horizontal plate, even though it’s undermatched. Thus, in the analysis the
failure is due to necking of the plate and this is in accordance with the experimental result.

Based on the presented results, it is concluded that the FE-analyses provide comparable results
to the experiment if the failure happens in the plate. They present relatively accurate
load-displacement curves and the calculated deformation at failure is within the upper and
lower bounds.

44
4.1.2 Specimen 4B1 (fail in weld)

4.1.2.1 FEM-analyses
Specimen 4B1 is a 10 mm thick S1100 undermatched specimen. The plates are connected by
undermatched K-joint with incomplete penetration. The experimental result shows that the
failure is at the weld (Figure 4.18, [8]).

Figure 4.18: Failure of specimen 4B1

The measured dimensions are given in Figure 4.19. The side view shows that the lower left
weld has the smallest dimensions.

Figure 4.19: Measured dimensions specimen 4B1

45
The FEM geometry and the mesh are given in Figure 4.20 (refinement near weld, 0.5 mm).

Figure 4.20: Geometry and mesh specimen 4B1, 0.5 mm element size

Figure 4.21 shows the equivalent plastic strain result at the end of FE-analyses (at a
displacement of 6 mm) based on two different element sizes. From Figure 4.21 we can see that
numerical simulation predicts the specimen fails in plate if the element size is relatively big (2
mm). However, with smaller element size (0.5 mm) near weld, FE-result shows that the
specimen fails in weld, which is in accordance with the test result.

(a)

(b)
Figure 4.21: Equivalent plastic strain at the end of FE-analysis
(a) 2 mm element size, (b) 0.5 mm element size

46
More analyses were done based on several different element sizes (Figure 4.22). Note that the
analyses of specimen 3B1 show that if the specimen fails in plate, the element size won’t have
an obvious influence on the simulation results, so no refinement at plate is made.

The results show that with element size bigger than 1 mm, the specimen will fail in plate; with
element size smaller than 0.7 mm, the specimen will fail in weld.
800

700

600

500
Load [kN]

400

300 Coarse_2 mm
Refined_near weld 1 mm
200 Refined_near weld 0.7 mm
Refined_near weld 0.5 mm
100 Refined_even smaller element (not finish)
Lemaitre e0=1
0
0 2 4 6 8 10 12
Displacement [mm]
Figure 4.22: FE-analyses results of Specimen 4B1

From Figure 4.22 it seems that using different element sizes will give different
load-displacement FE-results. However, this is because of the given displacement in the
simulation is very large compared to actual deformation capacity of the specimen. Actually both
the experimental and the FE-analyses (using Lemaitre’s criterion, based on simulation result
with 0.5 mm element size) results show that the failure happens at a very small displacement,
when all the simulation results are still identical. So the element size is considered having very
little influence on the load-displacement curve.

Lemaitre’s criterion

No problems were encountered in the experiment with the LVDT’s. So the numerical results can
be compared directly with the average of the two LVDT’s.

From Section 4.1.1 it is known that the element size might affect the result of Lemaitre’s
criterion. From Figure 4.23 we can see that the equivalent plastic strain results have a very big
difference when using different element sizes at the same displacement. It is obvious that with
smaller element size, the equivalent plastic strain is more concentrated and has a much higher
value.

47
 (a)

(b)
Figure 4.23: Equivalent plastic strain at a displacement of 0.96 mm (2*0.48)
(a) 0.7 mm element size, (b) 0.5 mm element size

So the influence of element size on Lemaitre’s criterion is checked again using two similar
element sizes (0.5 mm and 0.7 mm) and the results are shown in Figure 4.24. From Figure 4.24
it can be observed that unlike specimen 3B1, for specimen 4B1 the difference happens in a very
early stage. The result shows that for 0.5 mm element size the failure happens in the 16th
substep out of 100 total steps (displacement: 0.96 mm), while for 0.7 mm element size the
failure happens at the 22th substep (displacement: 1.32 mm) (Figure 4.25). And for both
analyses, the failure occurs at the weld root at (near) the middle of specimen width.

1.8
1.6 lemaitre criterion
1.4 0.7 mm
Equialent plastic strain [-]

0.5 mm
1.2
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Triaxiality [-]
Figure 4.24: Lemaitre’s criterion check

48
Figure 4.25 shows the comparison of experimental and numerical results, the Lemaitre’s
criterion is also shown based on two checked element sizes.

700

600

500
Load [kN]

400

300
Experiment
200 FEM_3D full
100 Lemaitre_0.5 mm
Lemaitre_0.7 mm
0
0 0.5 1 1.5 2 2.5
Displacement [mm]
Figure 4.25: Comparison of experimental and numerical results

Several remarks can be made with respect to this figure:

1) In the elastic range the analyses agree with the experiment, the yield strength is over
predicted;

2) The plastic capacity is overestimated (less than 8 %) by the FE-analysis. However, the
simulative and the experimental results have similar hardening trend in plastic stage;

3) Failure deformation according to the Lemaitre’s criterion is highly influenced by the element
size used in the simulation;

4) The deformation capacity is underestimated by FE-analyses and smaller element size will
give a more conservative result;

5) The failure mode (weld failure) is predicted very well if the element size near weld is small
enough. According to FE-analyses, failure starts at (near) the middle of the weld root.

Based on the presented results, it can be concluded that the FE-analyses provide comparable
load-displacement results to the experiment if the failure happens in the weld. However, for
deformation capacity, the error between the numerical simulation (based on Lemaitre’s
criterion) and experimental results is big.

For specimens fail in weld, because of the complexity near weld, a refinement is necessary. It is
unfortunately that due to the limitation of the calculation device, 0.5 mm is the smallest
element size that can be used in the simulations. For more accurate simulation results, a
smaller element size is necessary.

49
4.1.3 Specimen 1B1

Specimen 1B1 is a 12 mm thick S690 overmatched specimen. The failed specimen is presented
in Figure 4.26, [8].

Figure 4.26: Specimen 1B1

The measured dimensions of the specimen are presented in Figure 4.27 ([8]). The dimensions
in the side view show that upper right weld has the smallest dimensions.

Figure 4.27: Measured dimensions of specimen 1B1

50
The geometry and the mesh are comparable to specimen 3B1 (Figure 4.28).

Figure 4.28: Geometry and mesh specimen 2B4, 2 mm element size

The element size effect is briefly checked and the results are given in Figure 4.29. It seems that
the element size won’t affect the load-displacement curve. And the following analysis is based
on the results of 2 mm element size near weld.

700

600

500
Load [kN]

400

300

200
3 mm

100 2 mm
1 mm
0
0 5 10 15 20 25 30 35
Displacement [mm]

Figure 4.29: FE-analyses results of Specimen 1B1

As the LVDT’s fell of the specimen, the same interpretation procedure as for 3B1 is carried out.
The experimental load-displacement curves, including the lower and upper bounds are given in
Figure 4.30, using the corrected data different from the report [4].

51
 600

500

400
Load [kN]

300
LVDT3
LVDT4
200
Average 3 & 4
0.166*S01
100
Lower bound
Upper bound
0
0 5 10 15 20 25 30
Displacement [mm]

Figure 4.30: Determination of upper and lower bound of the experiment results

The comparison of the FEM analyses with the experimental results is given in Figure 4.31.
700

600

500
Load [kN]

400

300
Lower bound
200 Upper bound
FEM 3D_full
100 Lemaitre e0=1

0
0 5 10 15 20 25 30
Displacement [mm]

Figure 4.31: Comparison of experimental and numerical results for specimen 1B1

The comparison between FEM and experimental result shows that:

1) In the elastic range the analysis agrees with the experiment;

2) The maximum load is over predicted by 8 % in the analysis;

3) The deformation capacity in the FEM analysis is between the upper and lower bounds.

The distribution of the plastic strain at the end of the FEM analysis is given in Figure 4.32. The
distribution of plastic strain at the end of the FEM analysis (at a displacement of 30 mm) shows
that the overmatched welded connection is not critical and failure can be expected in the plate.
This is confirmed by the experiment.

52
Figure 4.32: Distribution of plastic strain in specimen (end of FEM analysis)

53
4.1.4 Specimen 2B4

Specimen 2B4 is a 40 mm thick S690 overmatched specimen (Figure 4.33, [8]). The plates are
connected by means of an overmatched full penetration K-joint.

Figure 4.33: Specimen 2B4

The measured dimensions are given in Figure 4.34 [8]. The dimensions in the side view show
that the upper left weld has the smallest dimensions.

Figure 4.34: Measured dimensions specimen 2B4

54
For specimens with 40 mm plate thickness, it is modeled with a root opening of 5 mm (Figure
4.35).

Figure 4.35: Geometry and mesh specimen 2B4, 5 mm element size at plate

The element size effect is checked and the results are given in Figure 4.36, the refinement at
plate is not necessary based on the result of specimen 3B1. However, the 5 mm element size for
plate material is used for the following analysis.
8000

7000

6000

5000
Load [kN]

4000

3000
Coarse_all 10 mm
2000 Refined_plate_5 mm
Refined_near weld and plate 5 mm
1000 Refined_near weld 5 mm
Refined_plate 5 mm_100 substeps
0
0 10 20 30 40 50 60 70
Displacement [mm]

Figure 4.36: FE-analyses results of Specimen 2B4

55
No problems were encountered in the experiment with the LVDT’s. So, the numerical results
can be compared directly with the average of the two LVDT’s (Figure 4.37).

8000

7000

6000

5000
Load [kN]

4000

3000

2000
Experiment
1000 FEM_3D full
Lemaitre e0=1
0
0 10 20 30 40 50 60 70
Displacement [mm]
Figure 4.37: Comparison of experimental and numerical results for specimen 2B4

The comparison between FEM and experimental results showed that:

1) In the elastic range the analysis agrees with the experiment;

2) The maximum load is slightly over predicted (1 %) in the analysis;

3) The deformation capacity is underestimated by the Lemaitre’s criterion.

The distribution of the equivalent plastic strain at the end of the FE-analyses (at a displacement
of 60 mm) is given in Figure 4.38. It shows that the overmatched welded connection is not
critical and failure can be expected in the plate. This is confirmed by the experiment.

Figure 4.38: Over view of plastic strain at a displacement of 60 mm (2*30)

56
4.1.5 Specimen 5B1

Specimen 5B1 is a 40 mm thick S1100 undermatched specimen (Figure 4.39, [8]). The
dimensions in the side view (Figure 4.40, [8]) show that lower left weld has the smallest
dimensions.

Figure 4.39: Specimen 5B1

Figure 4.40: Measured geometry specimen 5B1

57
The geometry and the mesh are comparable to specimen 2B4 (Figure 4.41). The specimen did
not fail in the experiment because the load capacity was too high for the test rig.

Figure 4.41: Geometry and mesh specimen 5B1, refined near weld, 3 mm

From the analyses of specimen 4B1 we know that for specimen fails at weld (undermatched
weld connection), a refinement is necessary. However, the calculation device can’t finish the
simulation with the element size smaller than 3 mm because of “Insufficient memory was
available for the solver engine to obtain a solution”, the simulation will stop at the middle of
calculation and only partial of the results are available (Figure 4.42). So the following analysis
has to be taken with the smallest possible element size (3 mm).

12000

10000

8000
Load [kN]

6000

4000

4 mm
2000 3 mm
2 mm_not finished
0
0 5 10 15 20 25 30
Displacement [mm]
Figure 4.42: FE-analyses results of Specimen 5B1

The comparison of the numerical and experimental results is given in Figure 4.43. It seems that
in the elastic range the analysis result has a slightly higher stiffness than the experiment.
Because of the limitation of the element size, the deformation capacity seems quite inaccurate
(fails in plate at a very large displacement), based on the analyses of specimen 4B1, the
specimen could fail in weld if the element size is small enough.

58
12000

10000

8000

6000

4000
Experiment
FEM_3D full
2000
Lemaitre

0
0 5 10 15 20 25 30 35 40 45
Figure 4.43: Comparison of experimental and numerical results for specimen 5B1

The distribution of the equivalent plastic strain at a displacement of 3.6 mm (2*1.8) is given in
Figure 4.44.

Figure 4.44: Equivalent plastic strain at a displacement of 3.6 mm

59
4.2 Connections type C

In this section, validations are carried out for five C-geometries. The C-geometry is an
X-connection with a high stress concentration factor (SCF) and the simplified geometry is
presented in Figure 4.45. This geometry can not be approximated by 2D geometries, so 3D
geometries are used for analyses.

Figure 4.45: Specimen geometry C (with high SCF)

In the next sections the FEM results are compared to the experimental results. The simulation
procedures are similar to connection B.

The comparison is mainly based on the load-displacement curve over the length of test gauges.
For specimen 1C1 the measuring equipment did not function until the end of the test and an
additional interpretation of the experimental results is also made.

60
4.2.1 Specimen 1C1
Specimen 1C1 is a 12 mm thick S690 undermatched specimen (Figure 4.46, [8]). The
measured dimensions (Figure 4.47) show the smallest weld dimensions (9.05 mm).

Figure 4.46: Specimen 1C1

LVDT_4

LVDT_3

LVDT_7

HP_1
Figure 4.47: Measured dimensions specimen 1C1
61
The geometry and the mesh are given in Figure 4.48. Both the left and right parts are modeled
with the same (smallest) weld size. Previous section has already proven that if the failure
happens in plate, the element size at plate doesn’t really matter, so the refinement is made only
near the weld.

Figure 4.48: Geometry and mesh specimen 1C1, refined near weld, 1 mm

The influence of element size is checked and compared in Figure 4.49. The 1 mm element size
near weld is used for the following analysis.

700

600

500
Load [kN]

400

300

200
2 mm
100
1 mm
0
0 2 4 6 8 10 12 14 16
Displacement [mm]
Figure 4.49: FE-analyses results of Specimen 1B1

The FE-analyses show that for this specimen, the failure happens in the plate (Figure 4.57),
which is in accordance with the test result.

Figure 4.50: Equivalent plastic strain at the displacement of 14 mm

62
As the LVDT’s fell of the specimen, the same interpretation procedure as for 3B1 is carried out.
The experimental load-displacement curves, including the lower and upper bounds are given in
Figure 4.51. Because of gauge LVDT4 and HP1 provide questionable data, only LVDT3 and
LVDT7 results are used.
600
LVDT3
500 LVDT7
Average LVDT 3 & 7
400 0.23*S01
Lower bound
Load [kN]

300 Upper bound

200

100

0
0 5 10 15 20 25 30
Displacement [mm]

Figure 4.51: Determine of upper and lower bound of the experimental results

The comparison of the numerical and experimental results is given in Figure 4.52. It seems that
in the elastic range the analysis result has a slightly higher stiffness than the experimental
result. The maximum load is over predicted by 7 % in the FE-analysis. The deformation
capacity in the FE-analysis is between the upper and lower bounds.

700

600

500
Load [kN]

400

300
Lower bound
200
Upper bound
100 FEM_3D full
Lematire e0=1
0
0 5 10 15 20 25
Displacement [mm]

Figure 4.52: Comparison of experimental and numerical results for specimen 1C1

63
4.2.2 Specimen 2C1
Specimen 2C1 is a 40 mm thick S690 overmatched specimen (Figure 4.53, [8]). The measured
dimensions (Figure 4.54) show the smallest weld dimensions (25.8 mm).

Figure 4.53: Specimen 2C1

Elap-4

Elap-2

Elap-6

Elap-8
Figure 4.54: Measured dimensions specimen 2C1

64
The geometry and the mesh are comparable to specimen 2B4, both the left and right parts are
modeled with the same (smallest) weld size, the refinement is made near the weld (Figure
4.55).

Figure 4.55: Geometry and mesh specimen 2C1, refined near weld, 3 mm

The influence of element size is checked and compared in Figure 4.56. The 3 mm element size
near weld is used for the following analysis.
7000

6000

5000
Load [kN]

4000

3000

2000

1000 4 mm
3 mm
0
0 10 20 30 40 50 60 70
Displacement [mm]

Figure 4.56: FE-analyses results of Specimen 2B1

The FE-analyses show that for this specimen, the failure happens in the plate (Figure 4.57).

Figure 4.57: Equivalent plastic strain at the end of FE-analysis (3 mm element size)
65
The comparison of the numerical and experimental results is given in Figure 4.58. It seems that
in the elastic range the analysis agrees with the experiment. And the maximum load is over
predicted by the FE-analysis (less than 12%).

However, the ultimate deformation capacity has a dramatic difference with the experimental
result. Based on the analyses of specimen 3B1, we know that for specimens fail in plate, the
estimated deformation shouldn’t have such a big error. From Figure 4.53 we know that in the
test the specimen fails in the connection plate and adjacent weld. This phenomenon has never
been found from the other specimens. So the difference between experimental and analytical
results is possibly because of the test specimen has some defects and fails in an unexpected
position.

7000

6000

5000
Load [kN]

4000

3000

2000
Experiment_Average

1000 FEM_3D full

Lemaitre
0
0 10 20 30 40 50 60 70
Displacement [mm}

Figure 4.58: Comparison of experimental and numerical results for specimen 2C1

66
4.2.3 Specimen 3C1
Specimen 3C1 is a 12 mm thick S690 undermatched specimen (Figure 4.59). The measured
dimensions (Figure 4.54) show the smallest weld dimensions (9.45 mm).

Figure 4.59: Specimen 3C1

LVDT_4

LVDT_3
LVDT_7

HP_1
Figure 4.60: Measured dimensions specimen 3C1

67
The geometry and the mesh are shown in Figure 4.61, both the left and right parts are modeled
with the same (smallest) weld size, the refinement is made near weld.

Figure 4.61: Geometry and mesh specimen 3C1, refined near weld, 0.7 mm

The influence of element size is checked and compared in Figure 4.62. The 0.7 mm element
size near weld is used for the following analysis.
700

600

500
Load [kN]

400

300

200
1 mm
100 0.7 mm
0
0 2 4 6 8 10 12
Displacement [mm]
Figure 4.62: FE-analyses results of Specimen 3B1

The FE-analyses show that for this specimen, the failure happens in the weld (Figure 4.63),
which is in accordance with the experimental result.

Figure 4.63: Equivalent plastic strain at the end of FE-analysis (0.7 mm element size)
68
The comparison of the numerical and experimental results is given in Figure 4.64. It seems that
in the elastic range the analysis agrees with the experiment. And the maximum load is slightly
over predicted by the FE-analysis (2%).

The ultimate deformation capacity is also seems comparable to the experimental result.
However, based on the analyses of specimen 4B1, we know that for specimens fail in plate, the
estimated deformation using Lemaitre’s criterion is related to element size. So for more
accurate simulation results, a smaller element size is recommended.

700

600

500
Load [kN]

400

300

200

Experiment_Average
100
FEM_3D full
Lemaitre
0
0 2 4 6 8 10 12
Displacement [mm]
Figure 4.64: Comparison of experimental and numerical results for specimen 3C1
The stress concentration can be visualized in Figure 4.65, but the stress is more concentrated in
weld.

Figure 4.65: Stress concentration (normal stress at a displacement of 0.2 mm)

69
4.2.4 Specimen 4C1
Specimen 4C1 is a 10 mm thick S1100 undermatched specimen (Figure 4.66). The measured
dimensions (Figure 4.67) show the smallest weld dimensions (6 mm x 6 mm).

Figure 4.66: Specimen 4C1

LVDT_5

LVDT_6
LVDT_7

LVDT_8
Figure 4.67: Measured dimensions specimen 4C1

70
The geometry and the mesh are shown in Figure 4.68, both the left and right parts are modeled
with the same (smallest) weld size, the refinement is made near weld.

Figure 4.68: Geometry and mesh specimen 4C1, refined near weld, 0.7 mm

The influence of element size is checked and compared in Figure 4.69. The 0.7 mm element
size near weld is used for the following analysis.

700

600

500
Load [kN]

400

300

200

100 0.7 mm
1 mm
0
0 1 2 3 4 5
Displacement [mm]

Figure 4.69: FE-analyses results of Specimen 4B1

The FE-analyses show that for this specimen, the failure happens in the weld (Figure 4.70),
which is in accordance with the experimental result.

Figure 4.70: Equivalent plastic strain at a displacement of 2 mm (0.7 mm element size)

71
The comparison of the numerical and experimental results is given in Figure 4.71. It seems that
in the elastic range the analysis has a higher stiffness than the experiment.

The ultimate deformation capacity is under estimated by the FE-analysis. Based on the
Lemaitre’s criterion, the specimen fails at a very early stage and because of that, the maximum
load is also under estimated (around 9%). However, this FEM result is related to the element
size and considered not very accurate.

For more accurate simulation results, a smaller element size is recommended.

600

500

400
Load [kN]

300

200
Experiment_Average
FEM_3D full
100 Lemaitre

0
0 0.5 1 1.5 2 2.5
Displacement [mm]

Figure 4.71: Comparison of experimental and numerical results for specimen 4C1

72
4.2.5 Specimen 5C1
Specimen 5C1 is a 40 mm thick S1100 undermatched specimen (Figure 4.72). The measured
dimensions (Figure 4.73) show the smallest weld dimensions (25.8 mm).

Figure 4.72: Specimen 5C1

Elap-4

Elap-2

Elap-6

Elap-8
Figure 4.73: Measured dimensions specimen 5C1

73
The geometry and the mesh are shown in Figure 4.74, both the left and right parts are modeled
with the same (smallest) weld size, the refinement is made near weld.

Figure 4.74: Geometry and mesh specimen 5C1, refined near weld, 3 mm

The influence of element size is checked and compared in Figure 4.75. The 3 mm element size
near weld is used for the following analysis.

12000

10000

8000
Load [kN]

6000

4000

4 mm
2000
3 mm

0
0 5 10 15 20 25
Displacement [mm]

Figure 4.75: FE-analyses results of Specimen 5B1

The comparison of the numerical and experimental results is given in Figure 4.76. It seems that
in the elastic range the analysis result has a slightly higher stiffness than the experimental
result. Similar to specimen 5B1, due to the limitation of the element size, the deformation
capacity seems quite inaccurate (fails in plate at a quite large displacement), again, it is
considered that the specimen should fail in weld if the element size is small enough.

74
12000

10000

8000
Load [kN]

6000

4000
Experiment
2000 FEM_3D full
Lemaitre
0
0 5 10 15 20 25
Displacement [mm]

Figure 4.76: Comparison of experimental and numerical results for specimen 5B1

The distribution of the equivalent plastic strain at a displacement of 3 mm is given in Figure

4.77.

Figure 4.77: Equivalent plastic strain at a displacement of 3 mm

75
4.3 FEM result comparisons

Comparisons are made between B and C connections based on FE-analyses results to see the
influence of the stress concentration.

4.3.1 Connection 1
Connection 1 is an S690 12 mm overmatched weld. Both B and C connections fail in plate. From
Figure 4.78 several conclusions can be made:

1) In the elastic range two connections are identical;

2) The FE-analyses results show that the two connections have similar load-displacement
curves;

3) C (high SCF) connection has a slightly lower yield strength and ultimate strength than B
(low SCF) connection;

4) Failure according to Lemaitre’s criterion gives comparable deformation capacity for two
connections.

700

600

500
Load [kN]

400

300

200 1B1
1C1
100
Lemaitre
0
0 5 10 15 20 25
Displacement [mm]

Figure 4.78: Comparison connection 1

76
4.3.2 Connection 2
Connection 2 is an S690 40 mm overmatched weld. Both B and C connections fail in plate. From
Figure 4.79 several conclusions can be made:

1) In the elastic range two connections are identical;

2) The FE-analyses results show that the two connections have similar load-displacement
curves;

3) C (high SCF) connection has a slightly lower yield strength and almost the same ultimate
strength with B (low SCF) connection;

4) Failure according to Lemaitre’s criterion gives comparable deformation capacity for two
connections.

8000

7000

6000

5000
Load [kN]

4000

3000

2000 2B4
2C1
1000 Lemaitre

0
0 10 20 30 40 50 60 70
Displacement [mm]

Figure 4.79: Comparison connection 2

77
4.3.3 Connection 3
Connection 3 is an S690 12 mm undermatched weld. B connection fails in plate and C
connection fails in weld. From Figure 4.80 several conclusions can be made:

1) In the elastic range two connections are identical;

2) The FE-analyses results show that the two connections have similar load-displacement
curves before failure;

3) C (high SCF) connection has a slightly lower yield strength and ultimate strength than B
(low SCF) connection;

4) Failure according to Lemaitre’s criterion gives a very big difference. For specimen 3C1
which fails in weld, the deformation capacity is dramatically lower (around 70%, the
element size might cause some calculation error) than the specimen fails in plate (3B1).

700

600

500
Load [kN]

400

300

200
3B1
3C1
100
Lemaitre
0
0 5 10 15 20 25
Displacement [mm]

Figure 4.80: Comparison connection 3

78
4.3.4 Connection 4
Connection 4 is an S1100 10 mm undermatched weld. Both B and C connections fail in weld.
From Figure 4.81 several conclusions can be made:

1) In the elastic range two connections are not identical, C connection has a smaller stiffness
than B connection. This might because of the difference of the given weld dimensions;

2) The FE-analyses results show that the two connections have comparable
load-displacement curves;

3) C (high SCF) connection has a lower yield strength and ultimate strength than B (low SCF)
connection;

4) Failure according to Lemaitre’s criterion has a difference. However, both specimens fail in
weld, so both results are not accurate. To know an exact difference, a smaller element size
is necessary.

800

700

600

500
Load [kN]

400

300

200 4B1
4C1
100 Lemaitre
0
0 0.5 1 1.5 2 2.5
Displacement [mm]
Figure 4.81: Comparison connection 4

79
4.3.5 Connection 5
Connection 5 is an S1100 40 mm undermatched weld. Due to the limitation of the element size,
the failure deformation according to Lemaitre’s criterion seems quite inaccurate, so it’s not
given here. From Figure 4.82 several conclusions can be made:

1) In the elastic range two connections are identical;

2) The FE-analyses results show that the two connections have similar load-displacement
curves;

3) C (high SCF) connection has a slightly lower yield strength and ultimate strength than B
(low SCF) connection.

12000

10000

8000
Load [kN]

6000

4000

2000 5B1
5C1
0
0 5 10 15 20 25 30 35
Displacement [mm]

Figure 4.82: Comparison connection 5

80
5. Conclusions and recommendations

5.1 Conclusions

5.1.1 Validation of material input

 The material input can be numerically determined based on three real material
properties (Re, Rm and εu);

 The structural behavior is dependent on geometry and element size. The geometries
used in simulations should be exactly the same as the ones in tests;

 Compared to 2D analyses, 3D analyses can give a more accurate result;

 Calibration of the true stress strain input will give an accurate simulation result. The
true stress-strain input depends on the element size;

5.1.2 Validation of specimen tests

 The load-displacement FE-results are comparable to experimental results, the general

trend is predicted very well. If specimen fails in plate, the failure load can be predicted
accurately with the help of Lemaitre’s criterion;

 In general the FEM model over predicted the ultimate strength (except specimen 4C1);

 In general the overmatched specimen will fail in plate and undermatched specimen will
fail in weld. However, the undermatched specimen 3B1 fails in plate due to large weld
reinforcement;

 For specimens that fail in plate the Lemaitre’s criterion is less sensitive to element size,
the deformation capacity based on the criterion can be accurately calculated;

 If the specimen fails in weld, the result of Lemaitre’s criterion is highly related to the
element size. Due to the limited computer performance, the element size is not small
enough in this report. The deformation capacity can not be accurately predicted;

5.1.3 FEM result comparisons

 In general the specimens with high SCF have a slightly lower strength than specimens
with low SCF;

 If both specimen B and specimen C fail in plate, the SCF seems only have a very limited
influence on strength and deformation capacity;

 If the SCF affects the failure mode, then the deformation capacity will be dramatically
affected. Specimen fails in weld will have a much lower deformation capacity than
specimen fails in plate;

 For specimens that fail in weld, a smaller element size is needed to have an accurate
result on deformation capacity.

81
5.2 Recommendations

 A more specific material data is needed to generate a more accurate material input;

 More accurate experimental results without error will be helpful to reduce the
uncertainty of the validation results;

 The specimens’ imperfections and flaws which occur during fabrication should be taken
into account in the FE-analyses;

 HAZ, weld beads, surface irregularities and weld toe radii should also be taken into
consideration;

 A smaller element size for specimens fail in weld is necessary;

 A detailed experimental and theoretical study with small scale specimens to validate or
improve the failure criterion of Lemaitre is recommended;

 A parameter study on type B and C geometries with a more pronounced SCF for the C
specimens by reducing the plate thickness is recommended;

 Continue this investigation by validation connection type A.

82
References

[1] Lemaitre, J. (1985). A continuous damage mechanics model for ductile fracture. In Journal
of Engineering Materials and Technology, 107, 83.
[2] Dijkstra, O. D., & Kolstein, M. H. (2006). Numerical modeling of under matched welded high
strength steel connections. In Steel-a New and Traditional Material for Building: Proceedings of
the International Conference in Metal Structures, Poiana Brasov, Romania, September 20-22,
2006 (p. 119). Taylor & Francis.
[3] Kolstein, M. H., Bijlaard, F. S. K., & Dijkstra, O. D. (2006). Integrity of welded joints made of
steel grades S690 and S1100. In Steel-a New and Traditional Material for Building: Proceedings
of the International Conference in Metal Structures, Poiana Brasov, Romania, September 20-22,
2006 (p. 183). Taylor & Francis.
[4] Kolstein, M.H. (2005). Integrity of high strength steel structures, Part 1: Experiments’. In
Stevin report 6-05-6. Delft Stevin University of Technology, The Netherlands.
[5] Kolstein, M.H., Dijkstra, O.D. (2005). Integrity of high strength steel structures, Part 2:
Analysis of the experiments’. In Stevin report 6-05-7. Delft Stevin University of Technology, The
Netherlands.
[6] Integrity of high strength steel structures – Deformation capacity. In TNO report
2005-BCS-R0367.
[7] Integrity of high strength steel structures – pilot FEM analyses. In TNO report
2004-BC-R0054.
[8] Integrity of high strength steel structures – Validation FEM analyses. In TNO report
2005-BCS-R0446.
[9] Integrity of high strength steel structures – Analytical model for welded connection. In TNO
report 2005-BCS-R0366.
[10] Integrity of high strength steel structures – Parameter studies for welded X-Connections
using 2D FEM analyses. In TNO report 2005-BCS-R0368.
[11] Pakiding, L. (2007). Design criteria for high strength steel joints. Master report report.
Delft University of Technology, the Netherlands.

[12] Fabricius, A. (2009). Experimental and numerical analysis of the deformation capacity of
high strength steel. Master report report. Delft University of Technology, the Netherlands.

[13] Joun, M., Eom, J.G., & Lee, M.C. (2008). A new method for acquiring true stress-strain
curves over a large range of strains using a tensile test and finite element method. In
Mechanics of Materials, 2008, 40(7): 586-593.

83
Appendices

A. Hardening rule [11]

Hardening rule is a phenomenon where yield stress increases with further plastic straining.
There are several types of hardening rules which had been proposed in order to define the
modification of the yield surface during plastic deformation. For example, isotropic hardening,
kinematic hardening and mixed hardening which is the combination of both hardening. In the
following section only the first two will be explained.

A.1 Isotropic hardening

This hardening rule assumed that the initial yield surface expands uniformly without distortion
and translation as plastic flow occurs. The size of the yield surface is governed by the plastic
strain history. Figure A.1 shows the representation of this hardening rule.

Figure A.1: Isotropic hardening

A.2 Kinematic hardening

The kinematic hardening assumes that the yield surface translate in the stress space during
plastic deformation as a rigid body which means the size, shape and orientation of the surface
is the same as before plastic deformation occurs and it is shown in Figure A.2. This hardening
rule takes into account the Bauschinger effect which is not considered in isotropic hardening
rule.

Figure A.2: Kinematic hardening

84
B. Material input

B.1 S690 40 mm plate

Figure B.1: Simulation geometry for 40 mm thickness plate

1100
1000
900
800
Stress [N/mm2]

700
600
500
400
300 S690_40mm plate_Input
200 S690_40mm plate_Engineering
100 S690_40mm plate_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Strain [-]

Figure B.2: Modified stress strain curves, S690 40 mm plate

900
800
700
Stress [N/mm2]

600
500
400
300 Engineering_input
10 mm mesh size
200
5 mm element size
100 3 mm element size
0
0 0.05 0.1 0.15 0.2
Strain [-]
Figure B.3: Element size tests, S690 40 mm plate

85
 1200

1000

800
Stress [N/mm2]

600
Modified true stress strain
input (before calibration)
400 True stress strain input
after calibaration_10 mm
200 True stress strain after
calibration_5 mm
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Strain [-]
Figure B.4: Comparison of true stress strain input after calibration, S690 40mm plate

900
800
700
Stress [N/mm2]

600
500
400
300
Engineering_input
200
100 Fem result_after
calibration_10mm
0
0 0.05 0.1 0.15 0.2
Strain [-]

Figure B.5: FEM result after calibration, S690 40 mm plate, mesh size: 10 mm

900
800
700
Stress [N/mm2]

600
500
400
300
Engineering_input
200
100 Fem result_after
calibration_5mm
0
0 0.05 0.1 0.15 0.2
Strain [-]
Figure B.6: FEM result after calibration, S690 40 mm plate, mesh size: 10 mm

For any element size smaller than 5 mm, the 5 mm input is adapted for subsequent numerical
simulations.

86
B.2 S1100 10 mm plate

Figure B.7: Simulation geometry for 10 mm thickness plate

1600
1500
1400
1300
1200
1100
Stress [N/mm2]

1000
900
800
700
600
500
400 S1100_10mm plate_Input
300 S1100_10mm plate_Engineering
200
100 S1100_10mm plate_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22
Strain [-]
Figure B.8: Modified stress strain curves, S1100 10 mm plate
1600
1400
1200
Stress [N/mm2]

1000
800
600 Engineering input
400 1 mm element size
2 mm element size
200 3 mm element size
4 mm element size
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Strain [-]
Figure B.9: Element size tests, S1100 10 mm plate

ANSYS doesn’t allow a downward input for plastic hardening part of the material, so this material can’t be
calibrated, the modified true stress strain curve will be used for this material.

87
B.3 S1100 40 mm plate

The geometry is the same as S690 40 mm plate.

1500
1400
1300
1200
1100
1000
Stress [N/mm2]

900
800
700
600
500
400 S1100_40mm plate_Input
300
S1100_40mm plate_Engineering
200
100 S1100_40mm plate_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Strain [-]

Figure B.10: Modified stress strain curves, S1100 40 mm plate

1400

1200

1000
Stress [N/mm2]

800

600
engineering input
400
5 mm element size
200
10 mm element size
0
0 0.02 0.04 0.06 0.08 0.1 0.12
Strain [-]

Figure B.11: Element size tests, S1100 40 mm plate

ANSYS doesn’t allow a downward input for plastic hardening part of the material, so this material can’t be
calibrated, the modified true stress strain curve will be used for this material.

88
B.4 S690 12 mm undermatched weld

All the weld geometries are considered the same as corresponding plate sizes.

800

700

600
Stress [N/mm2]

500

400

300

200 S690_12mm undermatched weld_Input

S690_12mm undermatched weld_Engineering
100
S690_12mm undermatched Weld_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Strain [-]
Figure B.12: Modified stress strain curves, S690 12 mm undermatched weld

700

600

500
Stress [N/mm2]

400

300
engineering input
200
2 mm element size
100 3 mm element size
4 mm element size
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Strain [-]

Figure B.13: Element size tests, S690 12 mm undermatched weld

From the element size tests, it seems that a calibration is not necessary, so the modified true stress strain
curve will be used for this material.

89
B.5 S690 12 mm overmatched weld

1400
1300
1200
1100
1000
Stress [N/mm2]

900
800
700
600
500
400
S690_12&40mm overmatched weld_Input
300
200 S690_12&40mm overmatched weld_Engineering
100 S690_12&40mm overmatched weld_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Strain [-]

Figure B.14: Modified stress strain curves, S690 12 mm overmatched weld

1200

1000
Stress [N/mm2]

800

600

400
engineering input
200 2 mm element size
3 mm element size
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Strain [-]

Figure B.15: Element size tests, S690 12 mm overmatched weld

From the element size tests, it seems that a calibration is not necessary, so the modified true stress strain
curve will be used for this material.

90
B.6 S690 40 mm overmatched weld

1400
1300
1200
1100
1000
Stress [N/mm2]

900
800
700
600
500
400
S690_12&40mm overmatched weld_Input
300
200 S690_12&40mm overmatched weld_Engineering
100 S690_12&40mm overmatched weld_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Strain [-]
Figure B.16: Modified stress strain curves, S690 40 mm overmatched weld

1200

1000

800
Stress [N/mm2]

600

400
engineering input
200 5 mm element size
10 mm element size
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Strain [-]
Figure B.17: Element size tests, S690 40 mm overmatched weld

From the element size tests, it seems that for both 5 mm and 10 mm element sizes, only one calibration is
enough.

91
 1400

1200

1000
Stress [N/mm2]

800

600
Modified true stress strain
400 input (before calibration)
True stress strain input after
200 calibration_10 mm
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Strain [-]
Figure B.18: Comparison of true stress strain input after calibration, S690 40 mm overmatched weld
1200

1000

800
Stress [N/mm2]

600

400 engineering input

200 Fem result_after

calibration_10 mm
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Strain [-]
Figure B.19: FEM result after calibration, S690 40 mm overmatched weld, mesh size: 10 mm

1200

1000
Stress [N/mm2]

800

600

400
engineering input

200 Fem result_after

calibration_5 mm
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Strain [-]

Figure B.20: FEM result after calibration, S690 40 mm overmatched weld, mesh size: 5 mm

For any element sizes smaller than 10 mm, the 10 mm input is adapted for subsequent
numerical simulations.

92
B.7 S1100 10 mm undermatched weld

1000
900
800
700
Stress [N/mm2]

600
500
400
300
S1100_10mm undermatched weld_Input
200
S1100_10mm undermatched weld_Engineering
100
S1100_10mm undermatched weld_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Strain [-]

Figure B.21: Modified stress strain curves, S1100 10 mm undermatched weld

900
800
700
Stress [N/mm2]

600
500
400
engineering input
300
2 mm element size
200 3 mm element size
100 4 mm element size
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Strain [-]

Figure B.22: Element size tests, S1100 10 mm undermatched weld

From the element size tests, it seems that for 2 mm, 3mm and 4 mm element sizes, only one calibration is
enough.

93
 1200

1000

800
Stress [N/mm2]

600

400
Modified true stress strain
input(before calibration)
200 True stress strain input after
calibration_4 mm
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Strain [-]

Figure B.23: Comparison of true stress strain input after calibration, S1100 10 mm undermatched weld
900
800
700
Stress [N/mm2]

600
500
400
300
200 engineering input
100 4 mm_after calibration
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Strain [-]
Figure B.24: FEM result after calibration, S1100 10 mm undermatched weld, mesh size: 4 mm

900
800
700
Stress [N/mm2]

600
500
400
300
engineering input
200
100 2 mm_after calibration
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Strain [-]
Figure B.25: FEM result after calibration, S1100 10 mm undermatched weld, mesh size: 2 mm

For any element sizes smaller than 4 mm, the 4 mm input is adapted for subsequent numerical
simulations.
94
B.8 S1100 40 mm undermatched weld

1300
1200
1100
1000
900
Stress [N/mm2]

800
700
600
500
400
S1100_40mm undermatched weld_Input
300
200 S1100_40mm undermatched weld_Engineering
100 S1100_40mm undermatched weld_True
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Strain [-]
Figure B.26: Modified stress strain curves, S1100 40 mm undermatched weld
1200

1000

800
Stress [N/mm2]

600

400
engineering input
10 mm element size
200
5 mm element size

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Strain [-]
Figure B.27: Element size tests, S1100 40 mm undermatched weld

From the element size tests, it seems that for both 10 mm and 5 mm element sizes, only one calibration is
enough.

95
 1400

1200

1000
Stress [N/mm2]

800

600
Modified true stress strain
400
input(before calibration)
200 True stress strain input after
calibration_10 mm
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Strain [-]
Figure B.28: Comparison of true stress strain input after calibration, S1100 40 mm undermatched weld

1200

1000
Stress [N/mm2]

800

600

400
engineering input
200
10 mm_after calibration
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Strain [-]

Figure B.29: FEM result after calibration, S1100 40 mm undermatched weld, mesh size: 10 mm
1200

1000
Stress [N/mm2]

800

600

400
engineering input
200
5 mm_after calibration
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Strain [-]
Figure B.30: FEM result after calibration, S1100 40 mm undermatched weld, mesh size: 5 mm

For any element sizes smaller than 10 mm, the 10 mm input is adapted for subsequent
numerical simulations.

96