Improving the fatigue life

of high strength steel

welded structures
by post weld treatments
and specific filler material
(FATWELDHSS)

Research and
Innovation EUR26929 EN
EUROPEAN COMMISSION
Directorate-General for Research and Innovation
Directorate D Key Enabling Technologies
Unit D.4 Coal and Steel

E-mail: rtd-steel-coal@ec.europa.eu
RTD-PUBLICATIONS@ec.europa.eu

Contact: RFCS Publications

European Commission
B-1049 Brussels
 European Commission

Research Fund for Coal and Steel

Improving the fatigue life of high strength steel
welded structures by post weld treatments
and specific filler material
(FATWELDHSS)

Ir. Sofie Vanrostenberghe

OCAS NV, J. Kennedylaan 3, 9060 Zelzate, Belgium

Ir. Mattias Clarin

SSAB, SE-78184, Borlnge, Sweden

Ing. Yang Shin

Volvo Construction Equipment (VCE), SE-631 85, ESKILSTUNA, Sweden

Ing. Benny Droesbeke

Belgian Welding Institute (BIL), Technologiepark 935, 9052 Zwijnaarde, Belgium

Mr. Vincent van der Mee

Lincoln Electric Europe, Nieuwe Dukenburgseweg 20, 6534 AD, Nijmegen, The Netherlands

Mr. Matthew Dor

The Welding Institute (TWI), Granta Park Great Abington, CB21 6AL, Cambridge, United Kingdom

Prof. Gary Marquis

Aalto University, Otakaari 1, 02015 TKK, Espoo, Finland

Dipl Ing Jari Parantainen

Stresstech Oy, Tikkutehtaantie 1, 40800, Vaajakoski, Finland

Dr. -Ing Thomas Kannengiesser

BAM, Unter den Eichen 87, 12205, Berlin, Germany

Dr. Zuheir Barsoum

KTH, Teknikringen 8, 100 44, Stockholm, Sweden

Grant Agreement RFSR-CT-2010-00032

1/07/10 - 31/12/13

Final report

Directorate-General for Research and Innovation

2015 EUR 26929 EN

LEGAL NOTICE
Neither the European Commission nor any person acting on behalf of the Commission is
responsible for the use which might be made of the following information.
The views expressed in this publication are the sole responsibility of the authors and do not
necessarily reflect the views of the European Commission.

Europe Direct is a service to help you find answers

to your questions about the European Union

Freephone number (*):

00 800 6 7 8 9 10 11
(*) Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.

More information on the European Union is available on the Internet (http://europa.eu).

Cataloguing data can be found at the end of this publication.

Luxembourg: Publications Office of the European Union, 2015

ISBN 978-92-79-44077-9
doi:10.2777/705266

European Union, 2015

Reproduction is authorised provided the source is acknowledged.

Printed in Luxembourg

Printed on white chlorine-free paper

Table of Contents

1 Final summary ................................................................................................................................. 5

2 Nomenclature.................................................................................................................................. 9

3 Scientific and technical description of the results ........................................................................ 11

3.1 Objectives of the project ....................................................................................................... 11

3.2 Results per task ..................................................................................................................... 12

3.2.1 WP1A: Sampling / Welding ........................................................................................... 12

3.2.2 WP1B: Case studies ....................................................................................................... 17

3.2.3 WP2: Weld Treatments and Stress Measurements ...................................................... 27

3.2.4 WP3: Fatigue Testing ..................................................................................................... 51

3.2.5 WP4: Modelling ............................................................................................................. 65

3.2.6 WP5: Guidelines ............................................................................................................ 77

3.3 Conclusions............................................................................................................................ 87

3.4 Exploitation and impact of the research results ................................................................... 89

4 List of References .......................................................................................................................... 90

5 Appendices .................................................................................................................................... 94

3
1 Final summary

The general objective of this project is the development of welded joints with improved fatigue
properties in High Strength Steels (HSS) in the thickness range of 5-20 mm. Therefore, technical
know-how on and understanding of the effect of various fatigue improvement methods will be
developed. The specific objectives of the project are:
Quantification and explanation of the effect of high frequency mechanical impact (HFMI)
techniques (e.g. ultrasonic impact treatment (UIT)) on the fatigue properties.
Determining the suitability of diode laser dressing as a novel technique for re-melting the
weld toe region in order to remove sharp weld toe discontinuities that provide fatigue crack
initiation sites and possibly to introduce compressive stresses at the weld toe, leading to an
improvement in fatigue life.
Development and optimisation of low transformation temperature (LTT) filler wires to
generate compressive residual stresses at the welded joint to prolong the fatigue life.
Establishing best practice for measuring and profiling the residual stresses at the weld zone
surface using conventional X-ray diffraction (XRD) as well as the novel Barkhausen noise
(BN) technique. This is required to allow a thorough understanding of the effect the
improvement techniques investigated on the residual stresses and fatigue properties.
Developing weld simulation capabilities to predict the residual stresses at the welded joint
and in its vicinity based on the weld geometry, material properties and welding
parameters, which will be coupled to fatigue life analysis using local methods. The resulting
product will be a predictive tool capable of estimating the welding residual stresses and
their effect on fatigue.
Development of guidelines for improving the fatigue properties of welded joints, which will
facilitate the introduction of HSS grades in industrial applications leading to weight, cost
and energy reduction.

A description of the main results obtained is given in the following paragraphs.

Welding of fatigue test specimens (WP1a) has been carried out in various steel grade/thickness
combinations, after welding procedure development, ensuring optimal weld quality for testing.
Steel grades with yield strengths in the range of 700 to 960 MPa were used. The test specimen
geometry was flat sheets with longitudinal welded attachments. In total, over 520 welded
specimens were produced and tested. Welding Procedure Specifications (WPS) are available for
each steel grade/thickness combination. Furthermore, temperature distributions were measured on
several specimens for input into the modelling tools.

Two demonstrators (dipper arm and bogie beam) were selected and re-designed for weight
reduction using high strength steels (WP1b). These demonstrators were selected considering
various aspects that reflect the project objectives of introduction of high strength steel grades and
their limitations due to fatigue properties at the weld. For the dipper arm, the weight was reduced
by 18%; for the bogie beam a weight reduction of 24% was achieved. Identification and analysis of
critical welded joints was done during the re-design phase and it was decided to treat critical weld
toes using the HFMI technique. The fatigue life of both components was estimated based on the
draft guidance document for HFMI treated welded joints from IIW Commission XIII.
Several prototypes of the dipper arm and bogie beam were manufactured in high strength steel,
with critical weld toes treated by means of Pneumatic Impact Treatment (PIT). Optimal parameters
for the treatment were used based on recommendations from the supplier and knowledge available
in the consortium.
Rig testing of the dipper arm was simplified and a first prototype was tested which ran for 100% of
its design lifetime without failure of the component. Rig testing of the re-designed bogie beam was
carried out with the current bogie beam as a reference: the re-designed bogie beam showed a
fatigue life about three times higher than the reference case. Furthermore, the experimental rig
test results agree well with the fatigue life analysis results. Finally, the economic potential of the
optimised components was calculated which showed that it is possible to obtain a cost reduction on
both the direct product cost and the product life cycle cost. The knowledge and experience
gathered within the project will be implemented by VCE in future projects. More specifically, the
analysis and re-design procedure, as well as the use of HSS and HFMI treatment for welded joints
will be considered for new products. For instance, the optimisation result obtained for the bogie
beam is considered as a valuable solution for next generation products.

The application of Ultrasonic Impact Treatment (UIT) was carried out by Applied Ultrasonics
employing different treatment conditions (Task 2.1). A total of 170 specimens were HFMI treated,
using four sets of parameters. Additionally, quality checks (width and depth measurement) of the
HFMI treated groove were done. This data was used as input in guidelines for HFMI treatments
developed in WP5.

5
Re-melting of weld toes to improve the surface profile and thus reduce stress concentrations is a
known fatigue improvement techniques. In the study reported here, a novel approach was taken,
using a diode laser as heat source (Task 2.2). Different diode laser treatment conditions were
tested, using the diode laser power, spot size and shape, and the travel speed as main parameters.
Characterisation work of diode laser re-melted specimens was done, including weld toe radius
measurements and hardness measurements on cross sections. It was shown that laser re-melting
hardens the surface layer in the re-melted zone. Furthermore, residual stress measurements
showed that tensile stresses decreased for up to 10 mm from the weld toe.
The work performed showed that it is possible to perform diode laser re-melting of weld toes
successfully, using different focused spot shapes and dimensions, and parameters sets. However, it
became apparent that in order to create a smooth profile, a certain heat input/power density is
required, favouring smaller spot sizes.

Concerning Low Temperature Transformation (LTT) filler wire development (task 2.3), several LTT
alloys were evaluated in terms of cracking behaviour, weld microstructure, transformation
temperatures and mechanical properties. Several iteration steps using levitation melting (OCAS)
and filler wire production (Lincoln) were needed, for varying the chemical composition, mainly in
terms of the elements C, Ni, Cr and Mn. Two alloys were selected for further fatigue testing: LTT-S
(Cr-Ni-Fe based) and LTT-C (Cr-Mn-Fe based). Both these alloys best fulfilled the toughness
requirements and have a good hot and cold cracking resistance. Furthermore, they show
acceptable Ms temperatures, as well as appropriate strength levels. Lincoln has manufactured both
wires in a diameter of 1.2 mm and they were shipped to BIL for welding of fatigue specimens.

Regarding the ability of the LTT filler wires to control the residual stresses in specific fatigue
samples with T joints, it can be stated that significantly lowered residual stresses adjacent to the
weld toe may be formed compared to a conventional high strength filler wire. Significant stress
relief due to the martensite formation during cooling was proven by in-situ reaction force
measurements. This is particularly the case when a relatively high interpass temperature (180C)
is applied. Furthermore, the positive effect of LTT filler wires on fatigue behaviour was confirmed in
the fatigue test programme developed in the frame of this project: the benefit in fatigue strength
for Constant Amplitude Loading (CAL) ranges from a factor of 1.08 (R=0.5) to 2.06 (R= 0.1).

The work performed on development of LTT filler wires has made it possible to make correlations
between the filler metal composition and related phase transformation temperatures on the
magnitude of stresses developed and thus fatigue improvements at the welded joint. LTT welding
consumables offer the potential to achieve improved fatigue strength without time and cost
extensive post-weld treatments. The results can thus be utilised in a variety of steel processing
industries where components subjected to fatigue loading are made from HSS.

Task 2.4 focuses on measurement of residual stresses developed due to HFMI treatment, laser re-
dressing and LTT filler wires using the X-ray diffraction (XRD) technique.
For HFMI treated specimens, surface residual stress measurements revealed that high compressive
stresses are present in the treated zone near the weld toe. For laser re-dressed specimens it was
shown that there is a decrease in tensile residual stresses compared to the as-welded condition.
Surface measurements on LTT welded specimens showed only a slight decrease in tensile residual
stresses compared to the as-welded condition.
Within this project, an electrochemical polishing device has been developed for residual stress
depth distribution. Residual stress depth distribution measurements using the polishing device
coupled with XRD measurements were done on HFMI treated specimens. It was shown that
residual stresses are compressive to a depth of about 1 mm.
Extensive work was done on measurement of residual stresses and relaxation of HFMI treated
welded joints. It was concluded that compressive residual stresses induced by HFMI were quite
stable for a number of cycles with some rate of relaxation in constant amplitude fatigue tests.
Overloads in variable amplitude loading were observed to be the cause of the prominent relaxation
in variable amplitude testing.
In situ monitoring of stresses using the Barkhausen Noise (BN) technique was completed on HFMI
treated samples. Although there was quite a large scatter in the BN results, it is believed that the
method shows promise for future monitoring of critical welded joints.

Characterisation data of the selected steel grades was generated in Task 2.5. OCAS performed
tensile testing on two steel grades (S700MC and S960MC) at various temperatures (ranging from
room temperature up to 1000C). This data was used to validate the material models generated
using JMAT Pro software.

6
Fatigue testing was conducted under both constant amplitude cyclic loading (CAL) and variable
amplitude spectrum loading (VAL), with the objective to establish the levels of improvement in the
fatigue lives of the welded attachments.
Before launching the test programme, several round robin exercises were carried out to ensure
consistency of results between the testing partners involved. The VAL spectrum selected included
both tensile and compressive applied stresses and therefore, in the majority of cases, anti-buckling
frames were used. The degree of scatter in VAL tests using the equivalent stress range was similar
to the scatter in the CAL tests.

The main conclusions drawn from this programme of work are as follows:

Constant Amplitude Loading:

- For HFMI treated samples, there was no consistent effect of either of the two variables
investigated, plate thickness and HFMI condition.
- For laser re-melting, it was found that a more focused spot size (using less laser power but
a higher power density) is preferable to a higher power, larger focus size and thus lower
power density.
- For the LTT specimens, no significant effect was observed from the two wires or the plate
thickness investigated in this series of testing.
- A comparison of the techniques investigated suggests that at R=0.1, HFMI and LTT give
the greatest benefit in fatigue strength, by a factor of 2.36 and 2.06 respectively, with
laser re-melting giving an increase by a factor of 1.5. At R=0.5, HFMI again offers the
greatest benefit with an increase of 1.69 times versus a factor of 1.08 by LTT.

Variable Amplitude Loading:

- Comparing the performance of the as-welded specimens with the Class F2 curve in
BS7608, HFMI offers significant improvement in life compared with LTT for the 5mm
thickness S700MC and S960MC material tested. For 10mm thickness S700MC, LTT offered
the greatest improvement. No comparative tests were performed for laser re-melting;
however, based on the results it could be assumed that the performance of laser re-melting
under the applied VAL spectrum would be similar to that of the LTT technique.

The assessment guidelines for HFMI produced by the IIW are supported by the data produced.

In the modelling work package included in this project, KTH has developed an efficient simulation
approach for the prediction of welding residual stresses. This approach, which includes the
implementation of the gradual weld bead deposition method, was validated on a large and complex
welded structure (i.e. bogie beam, one of the demonstrators selected by VCE). Data from
experimental measurements, including temperature histories and residual stress measurements,
was used as reference data for the FE models. The FE model of the bogie beam incorporates both
thermal (moving heat source method) and mechanical (gradual bead deposition) analysis in order
to predict residual stresses. Predicted temperature histories were in very good agreement with
experimental measurements. Qualitatively good agreement was found between the experiments
and the numerically predicted welding residual stresses.
The work done on residual stress relaxation revealed that compressive residual stresses were
stable for most of the fatigue life of the specimen in CAL testing. However, for the VAL tested
specimens, the residual stresses were observed to relax slowly after consuming 10-20% of their
fatigue life until failure.
Additionally, the effect of LTT filler wires on the formation of residual stresses was analysed using
an FEA approach. It was possible to predict a reduction in tensile residual stresses at the weld toe
by incorporating expansion from the austenite to martensite phase transformation. Furthermore, it
was observed that the Ms and Mf temperatures had a significant effect on the final state of the
residual stresses. Further investigations are ongoing to refine the modelling tools to predict the
behaviour of the welded joint under different loading conditions.

The objective of guideline development (WP5) has been to combine the experimental and analytical
results produced within this project with other available expertise, data and published information
to develop practical guidance for enhancing the fatigue strength of high strength steel welded
structures. The guidelines were developed in cooperation with Commission XIII of the International
Institute of Welding (IIW). IIW Fatigue Recommendations are widely used internationally and
provide important guidance on design of many types of welded structures.

7
The document developed provides practical information on 1) how to implement the most
promising improvement technologies for high strength steels, 2) assuring the quality of the
improved weld, 3) design fatigue curves for improved welded joints using different improvement
methods and steel grades, 4) limiting factors for improved welded joints, e.g., joint geometry,
material thickness, type of loading, environment, etc. 5) qualification procedures and tests for new
methods that are developed/matured following the publication of the document and 6) general
information on good design practice for fatigue resistant structures.

A guidance document on Improving the fatigue life of high strength steel welded structures by
post weld treatments has been written within this project and has been distributed to experts in
the IIW Commission XIII. The guidance document is based on experimental evidence published
within the IIW and presented in the open international literature. The guidance document is
considered to apply to steel plate thicknesses from 5 to 50 mm and for yield strength levels from
235 to 960 MPa.

The design proposal addresses 1) the effect of material strength, 2) special requirements for low
stress concentration weld details, 3) high R-ratio loading conditions and 4) variable amplitude
loading. A yield strength correction method was proposed and verified. A design recommendation
including one fatigue class increase in strength (about 12.5%) for every 200 MPa increase in static
yield strength was proposed and shown to be conservative with respect to all available data.
A proposal for procedures and quality assurance for HFMI treated welded joints in steel has also
been presented. It was developed based on discussions, presentations and experimental evidence
published within Commission XIII of the IIW.

Dissemination of the obtained project results was made through several scientific papers.
Furthermore, the achievements made were presented during the Fatigue Design 2013 conference,
held in Senlis (France), as well as during several IIW meetings.

8
2 Nomenclature
CAL Constant Amplitude Loading
fy Yield strength
fy,o Reference yield strength
FAT IIW fatigue class, i.e., the nominal stress range in MPa corresponding to 97.7% survival
probability at 2106 cycles to failure (a discrete variable with 10-15% increase in stress
between steps)
HFMI High Frequency Mechanical Impact
HSS High Strength Steel
LTT Low Transformation Temperature
ko Strength magnification factor for high frequency mechanical impact treatment for steel fy
= fy,o
kR Strength magnification adjustment considering R-ratio
ky Strength magnification adjustment considering yield strength
Kt Stress concentration factor
m Slope of the SN line
Ms Martensite start temperature
Mf Martensite finish temperature
Nf Cycles to failure
PIT Pneumatic Impact Treatment

S Nominal stress range

Sc Yield strength corrected nominal stress range

S* Fictitious nominal stress range for specimen i at 2106 cycles to failure

UIT Ultrasonic Impact Treatment

VAL Variable Amplitude Loading
XN Improvement factor in life for HFMI treated welded joints at S equal to the FAT class of
the as-welded joint: Nf = XN2106

Strength correction coefficient for yield after high frequency mechanical impact

Strength correction coefficient for high frequency mechanical impact

N Standard deviation in log (Nf)

subscripts

A In the as-welded condition

K
Characteristic value corresponding to 97.7% survival probability at 2106 cycles to
failure (a continuous variable)

H Following high frequency mechanical peening

i value for specimen i

m Mean value corresponding to 50% survival probability at 2106 cycles to failure

9
3 Scientific and technical description of the results

3.1 Objectives of the project

The general objective of this project is the development of welded joints with improved fatigue
properties in High Strength Steels (HSS) in the thickness range of 5-20 mm. Therefore, technical
know-how on and understanding of the effect of various fatigue improvement methods will be
developed. The specific objectives of the project are:
Quantification and explanation of the effect of high frequency mechanical impact (HFMI)
techniques (e.g. ultrasonic impact treatment (UIT)) on the fatigue properties.
Determining the suitability of diode laser dressing as a novel technique for re-melting the
weld toe region in order to remove sharp weld toe discontinuities that provide fatigue crack
initiation sites and possibly to introduce compressive stresses at the weld toe, leading to an
improvement in fatigue life.
Development and optimisation of low transformation temperature (LTT) filler wires to
generate compressive residual stresses at the welded joint to prolong the fatigue life.
Establishing best practice for measuring and profiling the residual stresses at the weld zone
surface using conventional X-ray diffraction (XRD) as well as the novel Barkhausen noise
(BN) technique. This is required to allow a thorough understanding of the effect the
improvement techniques investigated on the residual stresses and fatigue properties.
Developing weld simulation capabilities to predict the residual stresses at the welded joint
and in its vicinity based on the weld geometry, material properties and welding
parameters, which will be coupled to fatigue life analysis using local methods. The resulting
product will be a predictive tool capable of estimating the welding residual stresses and
their effect on fatigue.
Development of guidelines for improving the fatigue properties of welded joints, which will
facilitate the introduction of HSS grades in industrial applications leading to weight, cost
and energy reduction.

11
3.2 Results per task

3.2.1 WP1A: Sampling / Welding

3.2.1.1 Introduction
This section covers welding and optimisation of the specimens in different steel grade and
thickness combinations with conventional and LTT filler material. An overview of the selected steel
grades and thickness combinations is given in Table 1.

Table 1: Different grades and thicknesses used in the project

Material
grade Thickness supplier / brand name
S700 MC 5 mm AM S700 MC
S700 MC 10 mm AM S700 MC
S690 QL 10 mm SSAB Weldox 700
S690 QL 20 mm AM Supralsim 690
S960 MC 5 mm SSAB Domex 960
S960 QL 10 mm AM SuperElso 960
S960 QL 15 mm SSAB Weldox 960

3.2.1.2 Equipment
The welding was performed with following equipment:

- Kuka KRC 15 robot with KR C2 control unit (Figure 1)

- Power source Fronius TransPuls Synergic 5000 (TPS 5000): liquid cooled, pulse welding
500 A (40%) and 360 A (100%) and wire feeder Fronius VR7000 (Figure 1)
- Welding torch Fronius Robacta liquid cooled; push pull; up to 700 A with a gas nozzle of 19
mm
- Welding table Siegmund with fixing possibilities each 50 mm (Figure 1)
- Weld and stitch calibre (Figure 2)

Figure 1: Overview of equipment used for welding activities

Stitch Side clamps

clamp

Central bar
40 x 40 mm

Figure 2: Weld and stitch calibre used for welding 15 and 20 mm thick samples

12
3.2.1.3 Test specimen
In view of the fact that the fatigue performance of many welded structures is invariably limited by
the inherently poor fatigue strengths of T joints, it was proposed to confine attention to the
improvement of T joints. In particular, the test specimens consist of dog bone plates with
longitudinal welded attachments, as shown in Figure 3, to be loaded parallel to the attachments.
The dimensions of the longitudinal welded attachments are 150 x 40 mm with a thickness equal to
the thickness of the base plate.

Figure 3: Welded fatigue test specimen

The criteria used in the selection of the shape of welded test specimen are the following:
a) It is a small-scale specimen that behaves like a real welded structure in that it embodies high
tensile residual stresses from welding acting in the same direction as the fatigue loading.
b) Fatigue test results obtained from such specimens generally exhibit little scatter, an aspect that
is further enhanced by the use of mechanised robot welding in the production of the specimens to
achieve good repeatability.
c) It has been used in many previous investigations of weld improvement methods which facilitates
comparisons between the present results and relevant published data.

The plate specimens were tested axially in suitable testing machines equipped to grip the ends of
the specimens. In the case of specimens with improved welded joints, two potential problems with
such a loading arrangement are the risks of fatigue cracking in the weld itself or where the plate is
gripped, rather than from the test detail of interest, the weld toe in this case. The first risk is slight
and, in view of the very high tolerance to welding defects of this type of welded joint, exclusively
confined to fatigue cracking from the weld root. To avoid it, full-penetration welded joints were
used near the attachment ends (for approximately 30mm). To avoid the second potential problem,
the specimens were waisted in width, so that the stress level in the test section is higher than that
in the gripped part. In addition, to avoid the risk of fatigue cracking from the plate edge in the
waisted part, all edges were ground smooth.

The improvement methods studied in the frame of this project are only relevant to fatigue failures
initiating from the weld toe. The selected laboratory design is therefore very well suited because it
generated fatigue cracks starting from the weld toe in all tested cases. In many industrial
applications welded structures fail due to fatigue cracks initiating from the weld toe. Referring to
the case studies and more specifically the dipper arm (see also 3.2.2.2), a redesign was done to
lower the stress level at critical points and to transfer the critical points from the weld root side to
the weld toe. Based on analysis results of the dipper arm, welded joint W3 was critical in terms of
fatigue induced weld toe cracking. Therefore, welded joint W3 was chosen to verify the HFMI
treatment effect in the simplified rig test.

13
3.2.1.4 Welding trials
General

For each of the different grades / thickness combinations and filler materials, welding trials have
been performed in order to fix the optimum welding parameters in terms of geometrical
arrangement, welding sequence, welding current and voltage, welding wire and welding speed. The
guidelines for maximum heat input, preheat, maximum interpass temperature provided by the
material supplier, were taken into account. During the trials it became clear that besides the
welding parameters, also other elements had a significant influence on the weld quality and needed
optimisation:

- geometrical shape of the weld preparation on the longitudinal attachments

- gap between the base plate and the bottom of the longitudinal attachments
- orientation of the welding gun
- stick-out

Before welding, the steel sheets were cleaned and the oxide layer removed. The longitudinal
attachments were chamfered with a weld preparation depending on the thickness of the sample.
Figure 4 gives an overview of the final welding preparations used for welding the different grades /
thickness combinations.

Figure 4: Overview of the different welding preparations depending on the thickness

Metallographic sections
Once the welding appearance was acceptable and the welding parameters within the limits
provided by the material supplier, 16 metallographic sections were made for each grade / thickness
combination in the critical areas to verify the weld quality and repeatability [1] - [2].

Preheat
According to the recommendations of the material suppliers, the 15 mm S960 QL and 20 mm S690
QL needs preheating of 100C in order to avoid cold cracking and high hardness. The base plate
and longitudinal attachments were preheated with an induction coil wrapped around the parts. The
temperature was controlled with a thermo couple and the parts were heated to 150C to
compensate the drop in temperature that occurs during the time the coil was removed (see Figure
5).

Figure 5: Preheating to 150C with induction preheat

14
Hardness measurements
For each grade / thickness combination, hardness measurements (HV10) were done at several
locations. The measured hardness for each grade was compared with the highest allowed hardness
mentioned in the standard EN ISO 15614-1 [3]. For all grade / thickness combinations, the
hardness in the base material remained below the maximum values mentioned in the EN ISO
15614-1 standard.

Registration and measurement of the welding parameters

During and after optimisation, the welding variables for each run were registered in tables in order
to be able to tune the parameters.

Figure 4 shows the final welding sequences for the different grade / thickness combinations. After
registration of all welding parameters, a Welding Procedure Specification (WPS) according the
standard EN ISO 15609-1 [4] was made for each grade / thickness combination.

3.2.1.5 Specimen welding

All specimens were welded using the optimised welding parameters mentioned in the WPS. Also
interpass temperature monitoring and interpass cleaning was performed in order to achieve the
best welding quality and ensure maximum repeatability. In total 388 samples were welded with
conventional filler material and 138 samples were welded with LTT filler material.
Furthermore, the temperature distribution was measured on several samples for use in welding
simulations (see also section 3.2.5). The measurements were performed at four locations (see
Figure 6) with thermocouples attached to a temperature recorder.

Figure 6: Drawing showing locations for temperature measurement

15
3.2.2 WP1B: Case studies

3.2.2.1 Objectives
Good design practice can lead to state-of-the-art welded structures with 10-20% weight savings as
compared to reference structures in conventional steel grades. In construction equipment, the
exact value depends on the manufacturing quality, stress ratio, stiffness requirements and load
time history. Two demonstrators were selected considering various aspects that reflect the project
objectives of introduction of high strength steel grades and their limitations due to reduced fatigue
properties at the welded joints. The first component was the dipper arm of the EW180D wheeled
excavator (see Figure 7); the second component was the bogie beam of the A30F articulated
hauler (see Figure 8).

Figure 7: (a) EW180D wheeled excavator (b) Dipper arm from EW180D

Figure 8: (a) Position of bogie beam (b) Bogie beam from AF30F

The following tasks were accomplished:

Design, fatigue analysis, fabrication and fatigue testing of two generic components in HSS;
Definition of the quality requirements for the generic and the reference component;
Definition of the type and extent of post weld improvement for the new generic structures;
and
Analysis of the impact on Life Cycle Cost for weight reduction of construction equipment.

3.2.2.2 Dipper arm

Optimization of dipper arm

Dipper arm from EW180D wheeled excavator was optimized and the weight was reduced from 469
kg to 389 kg, i.e. by 17 %. For the current dipper arm commodity steel (S355) was used, in
thicknesses as shown in Figure 9 a). The re-designed dipper arm uses S700MC steel in the highly
stressed areas as shown in Figure 9 b).

17
Figure 9: (a) Current dipper arm (b) New design of dipper arm

Plate thickness were changed for several parts such as web, flange plates to utilize high strength
steel. However as the current design has already full penetration for welded joints some weld roots
becomes critical because of thinner plate thickness.
Some redesign was done to lower stress at critical points or at least to transfer the critical points
from the weld root side to the weld toe by changing the shape critical weld toes were treated with
HFMI method.

Figure 10: (a) Box shape of current dipper arm (b) Box shape of new design of dipper arm

An inverted box type is introduced in order to lower stress in the critical welded joints as shown in
Figure 10. Cylinder lugs are included in the rear side plate.
After several iterations the new design was done and its fatigue life was estimated identifying
several critical welded joints as indicated in Figure 9 b). The approach followed for analysis of
fatigue life specified several yield strength classes and was based on the draft guidance document
for HFMI treated welded joints from IIW Commission XIII (see also WP5: Guidelines).
Welded joint W3 is chosen to verify HFMI treatment effect in the simplified rig test. Therefore,
special attention was paid to W3 during rig testing.

Manufacturing of dipper arm

Manufacturing of the dipper arm prototype was done by VCE subcontractor PPS in Slovakia.
700 MPa yield strength steel materials were provided by ArcelorMittal and SSAB and the
commodity steel grade (355 MPa material) was sourced locally.
Detailed examination of the critical welded joints was done to examine the quality of the prototype.
Following inspection, the results were proven to be satisfactory. Figure 11 is the Welding Procedure
Specification (WPS) of welded joint W3 and its test weld result.

Figure 11: (a) WPS for W3 (b) test weld result of W3

18
Having secured all welding parameters for the critical welded joints, the prototype was
manufactured.
In order to secure the quality of the weld end of W3 a temporary backing plate was used. The
prototype of the dipper arm was inspected with Magnetic and Ultrasonic inspection method and
was approved.
The critical weld toes were HFMI treated by means of PIT (pneumatic impact treatment) as shown
in Figure 12. The parameters used for the treatment were recommended by the equipment
manufacturer and in accordance with the knowledge gathered in the project. A frequency of 90 Hz
and a needle end radius of 2 mm was used.

Figure 12: (a) HFMI treated critical (b) HFMI treated W3

welded joints

Rig test
The ideal rig will consist of complete structure, multiple actuators and real load time history.
However, simplification was necessary and to get the best result from rig in spite of the
simplification, it was agreed to test the work piece under variable amplitude loading. The simplified
rig test with single actuator for the dipper arm (see Figure 15a) will allow the determination of the
fatigue life at one spot which matches the real load case. Fatigue life target for the dipper arm is
8000 hours, which corresponds to 530 rig test hours.
Two dipper arm prototypes were delivered to VCE in Eskilstuna, Sweden to perform the rig test.

How to define loads for rig testing

The signal for the simplified rig test was created and also used to provide an estimate on fatigue
life. Measurement data of the dipper arm in real life conditions, made at different customer
locations, were available for further analysis. The measured data were split into predefined ROTs
(real application tasks) for relevant signals.
The signals were evaluated and duty was calculated. Monte Carlo simulation was performed to
improve data quality. Figure 13 shows the loading scheme and boundary conditions of the dipper
arm in the rig test; the force signal of the actuator is shown on the right (4300 repetitions). A
global stress distribution for the dipper arm in lifting load case is shown in Figure 14.

Figure 13: FE model of dipper arm (left) force signal of actuator in rig test (right)

19
 Figure 14: Global stress distribution of dipper arm - lifting load case (max. principal stress)

Figure 15: (a) Rig test of dipper arm (b) Close up view of inside crack

Rig test results

The rig test for dipper arm (#1) ran for 100% of its design goal hours with variable amplitude load
signal, having taken over 5 weeks in the rig, and yet no crack was observed. In order to reach
crack initiation, according to the analysis, the rig needs to run multiple times longer. As it would
require too long time, it was judged not practical and it was decided to continue with Constant
Amplitude Load (CAL).

The second dipper arm (#2) was only tested with constant amplitudes, with the same ranges as
the first specimen. Detailed inspection of both dipper arms was done after the rig test and on both
dipper arms fatigue cracks occurred at ends of accumulations of weld material inside the box
structure as shown in Figure 15 b. Thanks to the HFMI post treatment effect the critical point was
moved (re-located) from the weld toe of W3 to the root of alongside weld.

Conclusions
Fatigue analysis and analysis of static strength has been performed for the dipper arm of EW180
excavator in the FATWELDHSS project in order to show the potential of HSS in practical
applications (construction equipment business). Overall a weight reduction of app. 17 % could be
reached and the design was proven through calculations and rig tests.

20
 Estimated fatigue life met the target on current design and exceeded the target 10 times
on the component level for the new design.
New design rig test result:
#1- 100% of duty with VA + rest of duty with CA(10 times duty)
#2- Total duty with CA (85 % duty of #1)
The rig tested fatigue life of the new dipper arm exceeded the target.

3.2.2.3 Bogie beam

Optimization of bogie beam

The other selected object was the bogie beam from A30F articulated hauler. For the current bogie
beam S355 material was used, in thicknesses as shown in Figure 16 a). The re-designed bogie
beam uses S700MC and S690QL steel in the highly stressed areas as shown in Figure 16.

Figure 16: (a) Current bogie beam (b) New design

A weight reduction of 24% was achieved (the weight of the current design 145 kg was reduced to
110 kg in the new design). As shown in Figure 17 the new design includes the implementation of
thinner high strength steel plates (S700 MC and S690 QL), the use of weld assembly on both ends
instead of casted parts, implementation of smooth radius and introduction of chamfer to improve
weld root fatigue life. Detailed analysis was done, including the comparison between current and
new design fatigue life [2].

Figure 17: Critical welded joints on bogie beam

Based on analysis results, several critical welded joints were identified as shown in Figure 17 for
the current and new design respectively.

21
Manufacturing of bogie beam
Two prototype bogie beams were manufactured using high strength steels S700MC and S690QL as
shown in Figure 18.

Figure 18: (a) Prototype of bogie beam (b) Chamfer around sleeve

Detailed inspection of weld toes was done, as well as Ultrasonic Testing for weld root and results
were approved. HFMI treatment was applied on the critical weld toes by means of PIT with
frequency 90Hz and needle end radius 2 mm as shown in Figure 19.

Figure 19: HFMI treatment of bogie beam

Rig test

The loading scheme and boundary conditions are shown in Figure 20. The rig load ranges are
shown in Figure 21.

Figure 20: Loading and boundary conditions for bogie beam rig testing

22
Figure 21: Range Pair distribution for the command load ranges and the achieved rig load ranges

Figure 22: Stress distribution of bogie beam for equivalent load of 312kN (Von Mises stress)

Several stress distribution analyses were made. A typical result is shown in Figure 22.
Three A30 bogie beams were tested, one of the current design using commodity steel and two
made by high strength steel with HFMI treated welded joints as shown in Figure 23. The fatigue life
target for the specimens was 1000 simulated field hours on the heavy duty proving ground of
Volvo CE Sweden.

After rig testing, the three components were inspected:

- Current design: A crack in the rubber bearing sleeve weld was observed after 330 hr rig
test (see Figure 24). This was caused by imperfect geometrical weld conditions. However,
the weld met the penetration depth requirement.
- New design: The fatigue life of the tested prototypes was 845 and 1163 hr respectively.
The cracks of the two HSS bogie beams were caused by geometrical notch effects from the
backing plate tack welded joints as shown in Figure 25.

Figure 23: (a) Rig test for reference (b) Rig test for new design

23
 Figure 24: Location of cracks in current bogie beam

Figure 25: (a) Location of inside cracks of new bogie beam (b) Crack at tack welding

Conclusions

Re-design of the bogie beam was done, using HSS steel in thinner gauges and HFMI treatment of
critical weld toes. The bogie beam prototypes were then rig tested in order to verify the analysis
results and compared with results obtained on the current design.
In summary, the following conclusions can be drawn:
Calculated fatigue life of the new bogie beam was approx. 3 times fatigue life of the
current design (800 vs. 270 h)
Results of rig test validate the analysis results (for current design 270 vs. tested 330 h)
The rig tested fatigue life of the new bogie beam was approx. 3 times fatigue life of
current design (845 h/1163 h for new design vs. 330 h for current design)

3.2.2.4 Technical and economic potential

Product cost new dipper arm design: it was estimated that based on current pricing for raw
materials product cost for one dipper arm would increase approximately 100 (or 6 %
increase) when compared to the current serial production. This is mainly due to more
expensive high strength steel and HFMI post treatment.
Impact on exploitation cost dipper arm: For the same machine configuration and conditions
(counterweight and power), the vehicle will consume same fuel but increase handling
capacity by 5 %. Thus, fuel efficiency improves by 5 % for the duty cycle mass
excavation. This means for an assumed machine life time of 10000 operating hours, 2000
liter less fuel consumption. With assumed average diesel cost of 1.30 / ltr, customers
could save 2600 of fuel cost.
Product cost bogie beam: it was estimated that based on current pricing for raw materials
product cost for the redesigned bogie beam would be cost neutral compared to the current
serial production.
Impact on exploitation cost bogie beam: Weight reduction of the bogie beam will increase
the load capacity in some portion, assume 50 % of weight reduction for a bogie. A weight
reduction of the body would correspond to 100 %, a bogie beam is positioned lower and is
thus not so influencing. Earlier investigations state that replacing structural weight with
more payload is worth 29 / kg in life cycle cost. So total cost saving is approximately
35 kg*2*0,5*29 / kg. The total cost saving would be approximately 1000 per machine
over 5 years service time (with typical 2000 hours/year) of operation.

24
3.2.2.5 Conclusions and future outlook

Table 2 below summarizes all full scale fatigue tests performed on the two case studies.

Table 2: Summary of full scale fatigue tests performed on dipper arm and bogie beam

Demonstrator Case Target life Rig test life Calculated life Remark
[h] [h] [h]
10 times on the Over 10 times on
Re-design #1 8000 component the critical weld Crack in parent material
Dipper arm
8,5 times on the Over 10 times on
Re-design #2 8000 component the critical weld Crack in parent material
Crack at critical position
current design 1000 330 270 (center sleeve)
Crack at different
Bogie beam
Re-design #1 1000 845 793 position
Crack at different
Re-design #2 1000 1163 793 position

Based on the results obtained, it can be concluded that:

It is possible to achieve a weight reduction of approx. 20% on the selected demonstrators.

Experimental results indicate that there is probably room for further optimisation:

o Use of lower strength grade for dipper arm to reach the design target (e.g. S460 MC)
o Reduction of number of welded joints to be treated using HFMI for bogie beam

The re-designed light-weight components show considerable increase in fatigue life

compared to the current designs.
Experimental rig test results agree well with the analysis.
HFMI treatment is a very promising method if structural weight reduction are investigated /
considered for or if fatigue life should be improved.
Possible cost reduction was calculated both on direct product cost and product life cycle
cost.

Future outlook on exploitation of results for Volvo CE:

Introduce the HFMI treatment procedures in a work shop and real production in order to
get experience of the method in serial production.
Repeat the work done on a bigger component, where weight reduction plays a major role
and estimate the impact on production cost, cost in service and productivity (e.g. fuel
consumption) and environmental influence.
Continue to study the influence from HFMI in combination with HSS to weld quality and
fatigue on real components.
Redesign in such way that the critical points are transferred from the weld root side to the
weld toe, in order to be able to treat the weld with HFMI method.

Potential applications:
The optimization result of the EW180D dipper arm will be implemented by similar design
approach in the EWR150E dipper arm and introduction of the concept into first and second
boom is under investigation.
The optimization result of the A30F bogie beam is one of the influential cost reduction
solutions for next generation product.

25
3.2.3 WP2: Weld Treatments and Stress Measurements

3.2.3.1 High Frequency Mechanical Impact (HFMI) treatment

Welded joints used in steel structures such as cranes and bridges are often the cause of failure
accidents as a result of fatigue cracks at the weld toe initiated by weld geometries and defects.
Welded joints represent regions of stress concentration and possess high tensile residual stresses.
Ultrasonic impact treatment (UIT), which is a peening method, is a novel post weld improvement
method which modifies the stress distribution near the weld to produce beneficial residual
compressive stresses and modify the local geometry of the weld toe to eliminate initial defects and
decreases local stress concentration. The area being treated being treated experience excessive
plastic deformation leading to material work hardening and introduction of favourable compressive
stresses.

In this project, ultrasonic treatment is investigated because of multiple reasons. Firstly, the
technique showed promising results in other steel grades and further understanding and
implementation in high strength steels will lead to the possibility of introduction of high strength
steel in various industrial applications. Secondly, the time required for implementation is rather
short and the technique is easy to use (the required force is only 30N on the UIT tool, whereas
ordinary hammer peening requires 200N against the material being treated). This will be reflected
in the shape (removal of stress concentration) and dimension of the area treated as well as the
stresses at the weld toe (reduction of tensile stresses of introduction of compressive ones), and
may therefore impact on the fatigue improvement achieved.

Equipment
In the current RFCS-project, UIT treatment was revisited with a special focus on high strength steel
grades at different thicknesses. The fundamental difference with regard to the work reported till
date was the choice of steel grades and the examination of effect of various treatment parameters
on the resulting fatigue life improvement. For our study, we focused on various steel grades
S690QL, S700MC, S960MC and S960QL at various thicknesses and various treatment parameters.

To perform the trials, Esonix UIT equipment developed by Applied Ultrasonics was used for
treatment (shown in Figure 26). The equipment consists of a generator, a cooling unit and a hand
tool. The pins mounted on the tip of the hand tool are vibrated by ultrasonic waves generated in
the transducer. The radius of curvature of the pin tip varied from 3 to 8 mm. The resonant
frequency of the pin is 27 kHz and the amplitude of the pin was varied between 24 m to 28 or
29m.

Figure 26: equipment used for UIT treatment experiments:(left) Ultrasonic Esonix impact
treatment equipment (UIT), (middle) Illustration of the application of the UIT pin on the weld toe,
(right) Schematic illustration of the UIT equipment various components

Different pin radiuses and amplitude functions as well as sequences of treatment were selected to
investigate the effect of them on the subsequent fatigue life. It should be noted that during UIT
there is no needed pressure on the tool. Thus, the impact energy and hence the treatment quality
at a given point is independent of the pressure applied on the tool by the operator and is
determined by UIT conditions only. The treatment is performed in a single pass by the operator
pushing the tool and holder (3.5 kg weight) towards the weld toe and moving along the weld toe.
The tool holder isolates the operator from the vibration and there is noise emitted during the
treatment. The UIT tool impresses a radius into the surface at the weld toe which is about the
same as the tool component used. The UIT conditions employed in this work are summarized in
Table 3.

27
 Table 3: UIT conditions employed in this project

The frequency of the ultrasonic transducer (responsible for deformation

27kHz
resistance and material relaxation during impact)
Ultrasonic oscillation amplitude of the transducer under unloaded 24m
conditions (varies depending on condition used) Max.(28-29m)
Impact frequency under loaded conditions (responsible for the depth of
100-120 Hz
the plastic deformation process)
Impact amplitude under loaded conditions (at 100-120 Hz) Up to 1.5mm
Ultrasonic oscillation amplitude under loaded conditions (during impact) 30m
Indenter: Hardness HRC 6264
Radius (varies based on experimental condition) 3 and 4.8mm
Tool is pressed against the treated surface with a force defined by
weight of the tool only. This does not affect the treatment quality and
depends on the UIT condition employed. Power of the tool 1200 W
Range of took oscillation angle during treatment relative to the initial
35-55
position of 45
Number of pins 3 or 4
Treatment rate Not less than
0.2 m/min

Different pin radiuses, power levels and sequences of UIT treatments were employed to determine
the effect of different treatment parameters and sequences of treatment on the resulting fatigue
life. In addition, the treatment time for each sample was recorded and used as a guideline to
ensure consistency among various samples.

General
Ultrasonic impact treatment procedure aimed at developing different parameter sets, for example
focusing on pin radius, amplitude used during treatment or combining different sequences of
treatment which in turn may affect productivity. These procedures were then tested for their
suitability for fatigue improvement.
The samples were treated using Esonix equipment to the UIT procedure developed by Applied
Ultrasonics. According to this procedure, a 27 kHz hand held tool was used and the size of the
indenter and treatment conditions could be selected based on the welded joint type and fatigue
test conditions. It should be noted that all indenters were made of the same material and their
working surface had the same hardness and roughness. Table 4 indicates the main treatment
conditions employed in this project.

Table 4: Experimental treatment conditions employed in this project

Conditions Pin radius Number of Amplitude Remarks

number (mm) pins level (m)
C1 3 4 Max. (28 or 29) Base line condition for HSS
C2 4.8 3 Max. (28 or 29) Used to determine the effect of
pin radius (in comparison to
C1) on fatigue
C3 3 4 24 Used to determine the effect of
amplitude level (in comparison
to C1) on fatigue life
C4 3 and 4.8 3 and 4 Max. (28 or 29) Treated in 2 sequences as
mm follows:
3 pins x 3mm pin radius, then
3 pins x 4.8mm pin radius

28
3.2.3.2 Laser re-melting
Re-melting of weld toes to improve the surface profile and thus reduce stress concentrations is a
known fatigue improvement technique. In the study reported here, a novel approach was taken,
using a diode laser as heat source. Different diode laser treatment conditions were tested, using
the diode laser power, spot size and shape, and the travel speed as main parameters. T joint
specimens in S690QL and S960QL in 10 mm thickness were treated and fatigue tested with
constant amplitude loading. In response to those initial results, further samples were welded and
treated and variable amplitude fatigue testing was undertaken.

In fusion welds, a sharp transition from the weld metal to the heat-affected zone/parent metal is
often found, particularly in those with considerable overthickness/excess weld metal. This transition
acts as stress raiser and can thus impair the welds performance, in particular the fatigue
performance. The transition at the toe can be characterised by the toe angle and radius (both of
these are illustrated in Figure 27), with a small toe angle and radius being detrimental.

One of the well-established ways of improving a welds fatigue performance is therefore via
smoothing of the transition between the base metal/HAZ and the weld metal (at the weld toe) via
mechanical (e.g. grinding or peening) or thermal (via re-melting (dressing) of the weld toe area)
means. Traditionally, toe re-melting or dressing is performed using TIG or plasma welding with the
suitability of both of these processes resulting from their ability to create just a small amount of
molten metal (and this fairly superficially) without the need for filler metal addition to run the
process. However, both are generally performed manually and are very sensitive TIG in particular
to the torch-to-workpiece distance which controls the voltage and thereby significantly affects the
process. In addition, the achievable travel speeds tend to be limited to a few hundreds of
millimetres per minute, which is generally slow in comparison to welding travel speeds used for
example for MAG welding. Re-melting can then become a bottleneck in production, lowering
productivity and raising manufacturing costs.

Figure 27: Macro-section of a bead-on-plate MAG weld with drawn on it the toe angle (left) and
toe radius (right).

Diode laser re-melting set-up

Introductory trials for proof of concept and to select suitable optics were performed at German
manufacturer of diode lasers Laserline GmbH in Mlheim-Krlich (Germany). For these trials, runs
were made on flat plate and on bead-on-plate MAG welded joints. For the main body of work, a
Laserline 6 kW LDF 6000-100 fibre-delivered diode laser source was rented and installed for a two-
month period at OCAS NV at its facility in Zelzate near Gent (Belgium).

The laser used operates in the wavelength range of 900-1070 nm with a beam parameter
product of 110 mm*mrad, allowing it to be used in combination with a 1 mm diameter fibre optic
cable for beam delivery. Different collimating and focusing optic combinations were used, giving
focused spot sizes of approx. 5x6 mm and 10x13 mm for rectangular and 5 mm, 3 mm and
2.5 mm diameter for circular spots. The focusing optics at OCAS were mounted on a Panasonic
TAWERS arc-welding robot for following the 2D shape of the weld toe. The axis of the laser beam
was at 20 to the normal to the sample surface (so in welding terms, approx. the PB position).
Because of the complexity of movement required from the robot arm (exacerbated by the long
distance from the tool centre point to the last robot axis), the travel speed for toe dressing was
limited to 2 m/min. During the experiments at Laserline (both campaigns), a gantry system was
used that allowed only vertical beam orientation (and therefore the laser beam was kept
perpendicular to the main plate surface).

29
Re-melting procedure

Re-melting procedure development aimed at developing different parameter sets, for example
focusing on the depth of re-melting, or on the travel speed and thus productivity. Evaluation of the
performance was in first instance done via external visual appearance followed by cross-sectioning
for promising parameter sets. The main evaluation criterion was smoothness of the treated weld
toe with a secondary criterion being minimization of the amount of molten material and heat input,
and maximization of the travel speed. In addition, hardness testing, residual stress measurements
and 3D surface measurement before and after fatigue testing were periodically performed.

The main parameters varied for the initial trials on flat plate and bead-on-plate melt runs
performed at Laserline were the laser power, travel speed and focused spot shape and dimensions.
The ranges investigated were 2 to 6.5 kW of laser power (measured at the laser source) and 0.25
to 1.5 m/min travel speed. With proper parameters, weld toe smoothing could be achieved
consistently, an example of which can be seen in the cross-section in Figure 28. This weld was
treated using a 5x6 rectangular spot with a laser power of 2 kW but a travel speed of only
0.25 m/min.

Figure 28: Example of a cross-section of a bead-on-plate MAG weld of which the weld toe on the
left has been re-melted using a diode laser.

Procedure development on fatigue specimens

During the trials at OCAS, it quickly became apparent that parameter settings developed on bead-
on-plate welded joints on flat plate were not directly transferable to the T joints on the fatigue
samples, most likely due to the more inclined weld top bead surfaces. The actual procedure
development was therefore performed on welded samples that had been rejected for fatigue
testing because of weld irregularities and/or imperfections. Five re-melting parameter sets were
developed, which can be found in Table 5. As can be seen in the table, in most cases different
travel speeds were specified for the straight parts of the trajectory and the movement around the
corners. This was in response to the finding that even though one travel speed was programmed
the robot did not manage to keep the speed around the corners. It was then attempted via
setting of different speeds for the straight parts and corners, and timing of the robot to get an even
speed and movement all along the weld.

Table 5: Re-melting parameters used for constant amplitude fatigue testing samples
in S690QL & S960QL.

Trial Focal length Focused spot shape, Laser power Travel speed [m/min]*
ID [mm] size [mm] [kW] Straight Corner
C1 300 Rectangular, 13x10 3 0.75 0.75
C2 300 Circular, 3 2 0.50 2,00
C3 250 Circular, 5 2 0.50 1.50
C4 150 Rectangular, 5x6 2 0.50 1.50
C5 250 Circular, 5 2 1.50 1.50
* Travel speed was set so as to maintain near-constant speed all around the weld

Characterisation of treated welded joints

Treated weld cross-sections for some of the different spot sizes can be seen in Figure 29. The
cross-sections were taken in the plane through the middle of the welded attachments which in the
pictures b) to d) can be seen on the left, above and below the central plate. It can be seen from
the photographs that, for the parameters used, the largest spot size (13x10 mm rectangular) gave
a very fine layer of melting, resulting in the tendency to ball up and thereby not giving the
smoothest weld profile. Better results were achieved with the smaller circular spot sizes.

30
a) Macro-section location b)S960QL-C1 c) S960QL-C2 d) S960QL-C3
13x10 mm 3 mm 5 mm

Figure 29: Macro-sections of laser re-melted welded fatigue samples using different spot sizes (re-
molten zones are indicated by the arrows on the cross-sections).

Hardness testing (HV0.2) was performed on several macro-sections, with traverses taken just
below the surface (at approx. 0.1 and 0.3 mm from the surface) [5]. The hardness varies for the
different conditions, but is lowest for the 10x13 mm spot size (with the maximum ~475 HV0.2)
and higher for the smaller spot sizes (3 mm diameter circular) where the hardness reaches
~500 HV0.2.

Furthermore, residual stress measurements were performed using an X-ray measurement device
(Xstress robot) on samples in the as-welded and laser re-melted (C2) condition as shown in Figure
30. It is clear from the stress measurements that tensile residual stresses near the weld toe are
present on as-welded specimens. Laser re-melting (condition C2) tended to decrease the level of
tensile residual stress for a distance of up to 10 mm from the weld toe whereas further away from
the toe, it left stress levels comparable to those for the as-welded case.

a) Measurement direction b) Average residual stress measured

Figure 30: Stress measurement direction (a) and results (b) on as-welded and laser re-melted
specimens.

Discussion
In the experiments described here, smooth dressed weld toes could be achieved with very little
heat input and at appreciable travel speeds using a diode laser. However, the trials also showed
some caveats with regard to the employment of diode laser re-melting in practice.

Firstly, there is the issue of sample temperature. It was found in the experiments that the
sample needs to be sufficiently hot to allow the shallow molten zone around the weld pool to wet
the base material properly, thereby creating the desired smooth transition from base to weld metal
(avoiding stress concentrations). This heat can either be introduced via the chosen diode laser re-
melting parameters, but this tends to also reduce the travel speed. If re-melting is performed on a
hot sample as would be the case with the diode laser head travelling directly after a welding
torch diode laser re-melting tends to give good results even at lower heat inputs and/or higher
travel speeds. The sample could also be made hot by having a low heat input diode laser pass prior

31
to the actual re-melting pass, but this will impact severely on the productivity and was therefore
not further investigated. Although in a sense it is good that laser re-melting performs well when
performed on a hot sample, because this means an inherent suitability for welding and dressing in
a single pass, it also means that if for whatever reason re-melting would have to take place after
the sample has (fully) cooled, different parameters would have to be used. This means parameters
depend on the sample temperature.

Secondly, because of the required heat input, it was found that even with the diode laser, re-
melting was most comfortably performed at travel speeds that are not that far above those often
used for TIG or plasma dressing (in the range of a few tenths of metres per minute) when used on
a weld at room temperature. In spite of this fact, some re-melting parameters were also found that
used higher travel speeds (in this project up to 1.5 m/min), but this also requires higher laser
power levels, with an associated higher cost. In terms of the optics to choose, particularly when
aiming for higher travel speeds, best results were found when using a small, circular spot,
indicating that it is not the size of the treated zone that is important, but mostly the sheer fact that
the angle and radius are increased at the weld toe.

The type of weld treated also had a significant influence which was noticed when trying
parameter settings that worked well on bead-on-plate welded joints in the PA position on T joints in
PB to PC position. The parameters proved not directly transferable, possibly indicating the influence
of weld surface orientation or heat sink.

Conclusion

The work performed has allowed the following conclusions to be drawn:-

Diode laser re-melting of weld toes can be performed successfully, using different focused spot
shapes and dimensions and parameter sets. However, it became apparent that in order to create
a smooth profile, a certain heat input/ heat concentration is required, favouring smaller spot
sizes.
Weld toes treated with suitable parameters in macro-section showed a smooth profile and
hardness testing showed hardening in the re-melted zone to values of 450-500 HV0.2.
Residual stress measurements on a sample treated with the best performing procedure showed
decreased tensile stresses up to 10 mm near the weld toe, and comparable stress levels further
away from the weld toe.
The greatest fatigue benefit was found for condition C2 using a small, circular spot and the
lowest benefit for condition C1, using a large, rectangular spot.

32
3.2.3.3 Development and application of Low Transformation Temperature (LTT) filler
materials

Objectives

The work described in this section was focused on the development of LTT filler materials to induce
compressive residual stresses at the weld toe for improved fatigue properties. Following approach
was used:

An initial screening of possible LTT alloy compositions including variations of the chemical elements
C, Ni, Cr and Mn was conducted on levitation melted samples, as well as metal cored filler wires in
parallel. The latter served also to provide information about strength and toughness to be expected
for the weld metals. The results have shown that an evaluation of the alloys had to incorporate
implicitly investigations on welded samples as differences were found between levitation melts and
all weld metal regarding chemical composition and therefore transformation behaviour. Cracking
tests allowed to identify alloys not appropriate for further investigations.

A second iteration step using again levitation melts and filler wires was necessary in order to
recommend alloy compositions for welding of the fatigue specimens. The evaluation was based on
Martensite start temperature Ms, strength, toughness and cracking behaviour. Hot cracking as well
as cold cracking tests clarified that the weldability is a crucial factor.

Finally, two alloy compositions were recommended for fatigue testing. They were manufactured as
metal cored wires with a diameter of 1.2 mm. The wires are both iron based showing on the one
hand a modified chromium-nickel ratio leading to Ms below 250C. On the other hand a chemical
composition was designed showing chromium and manganese as main alloying elements. Ms of the
latter is situated below 150C.

During the project it became apparent that the development of special filler wires needs a few
iteration steps including theoretical as well as practical considerations. Despite the fact that the
filler wires fulfil the project requirements, further modifications are still possible.

The main objective was to develop and manufacture cored wires that would facilitate to study the
amount of tensile or compressive residual stresses produced by these wires and also their effect on
the fatigue strength of the weld joint under constant amplitude loading and variable amplitude
loading. This objective was achieved within the project and the outcome is described in the
following paragraphs.

Design criteria for LTT weld filler materials

The first steps in designing LTT weld filler materials were based on own investigations and results
found in the literature. The alloy design is an iterative process as the prediction of the material
properties on basis of the chemical composition is complex. The Ms-temperature as the essential
feature was evaluated first. Based on the experience from former studies an Ms-temperature range
between ~150C and ~300C was aimed for. Ms above 300 C would cause formation of
detrimental shrinkage stresses after the martensite transformation has finished. On the other hand
Ms below 150 C may lead to unwanted amounts of retained austenite. Beside the Ms-temperature
the strength of the LTT alloys is an important property. To match the strength of the base material
a minimum strength of 700 MPa should be achieved. Another important issue concerns the
toughness of the weld metal. A Charpy impact value of 27 J at 0C was aimed for. This material
property is hardly predictable and can only be evaluated after appropriate welding experiments.
Concerning the solidification mode, ferrite should be aimed for. The reason is that this phase is less
susceptible for solidification cracking than austenite. Finally, the weldability is another important
criterion for the applicability of a LTT weld filler material. To ensure these requirements a few
iteration steps were necessary.

Selection of LTT alloying concepts

Six basic chemical compositions were proposed as possible LTT filler materials (see Table 6). The
alloys stem from concepts to be found in the literature or were based on own investigations. Mainly
the chemical elements Ni, Cr, C and Mn were varied as they are known to largely influence the Ms-
temperature. The influence of the chemical composition on Ms, the content of retained austenite
and the solidification mode was preliminary estimated using analytical predictions.

33
 Table 6: Proposed alloying concepts

LTT wire Alloying elements varied

C Cr Ni Mn
008 0.01 12.0 4.0 0.50
009 0.04 8.0 6.0 0.50
010 0.08 8.0 6.0 0.50
011 0.08 10.0 - 9.00
012 0.08 8.0 - 0.50
013 0.08 1.0 12.0 0.50

First series of LTT filler wires

The six compositions proposed were evaluated concerning the Ms-temperature after manufacturing
by levitation melting (OCAS). In parallel metal cored wires in diameter 1.6 mm were manufactured
and tested by Lincoln. Test were conducted for apparent and tapped density, flow rate, grain size
distribution, weldability, diffusible hydrogen content, chemical composition and mechanical
properties. Additionally, weldments as well as metal cored wires were supplied to BAM for further
investigations concerning cracking susceptibility, microstructure and Ms-temperature. Results of
characterisation and discussion can be found in [6].

Evaluation of the results and recommendations:

LTT008 showed Ms on the upper limit of the recommended range (150C-300C). Further
increase of Ni should decrease Ms to an appropriate level.
LTT009 showed a martensite formation at temperatures above the desired level. Indeed a
further increase of the Ni content would decrease Ms, but the experience has shown that Ni
contents higher than 6% may encourage solidification cracking.
The same applies for LTT010. Further decrease in Ms may be achieved by alloying some
carbon.
As already indicated by the microstructure the Ms-temperature of LTT011 was quite low.
Decreasing the Mn content would be appropriate to increase Ms at least to temperatures
above 150C.
From viewpoint of the transformation behaviour Ms of LTT012 is too high for further
consideration. Additional decrease of Ms by alloying of Cr or C is unpractical due to ferrite
stabilization of Cr and carbide formation due to C. Also the hardness would be drastically
increased.
The Ms-temperature of LTT013 is in the desired range below 300C. Nevertheless, the
transformation behaviour is not clarified due to deviations between the individual
measurements. The results of the first weldability/cracking tests have shown that LTT013 is
not appropriate and will not be followed further.

The results showed that the intended high strength above 700 MPa was reached by all alloys
tested. The impact toughness varies in a range between 22 J and 52 J. Alloys LTT008, LTT010 and
LTT011 were selected for further modification. The objective was to fine tune the composition
further. For LTT008 the Ni as well as the C content was varied stepwise between 4.5% and 6.0%.
Higher C content was aimed for as well. Also the C content of LTT010 was varied between 0.10%
and 0.12%. On the one hand this may decrease Ms further. On the other hand possible tempering
effects should be investigated in terms of impact toughness. To increase the Ms-temperature of
LTT011 amounts of Mn from 4% to 6% were aimed for. The carbon content was varied also
between 0.04% and 0.06%.

Second series of LTT filler wires

Based on the evaluation of the results for the first series of LTT filler wires with BAM, Lincoln and
OCAS a second series of ten metal cored wires in diameter 1.6 mm have been manufactured and
tested on chemical compositions by Lincoln. The alloys were manufactured additionally as casts by
OCAS and tested in terms of Ms-temperature by dilatometry.

Based on the results of Ms-temperature measurements, the following wires were selected for
further examination (Table 7).

34
 Table 7: Wires selected for further examinations

Main alloying elements

LTT No.
C Cr Ni Mn

014 0.036 12.2 4.8 0.62

015 0.036 11.6 5.4 0.60

017 0.035 11.8 6.6 0.60

018 0.073 6.3 7.9 0.56

019 0.085 7.1 8.2 0.60

020 0.096 6.8 7.7 0.60

021 0.111 6.5 7.9 0.59

022 0.072 11.8 0.02 7.0

024 0.070 11.9 0.02 5.4

025 0.032 11.4 0.02 5.0

Alloys no. 014, 015 and 017 are chromium nickel based with lower carbon contents. The nickel
content was varied between 4.8% and 6.6% in order to optimize the Ms-temperature. Varying the
carbon content within the wires was waived for feasibility reasons. Alloys no. 018 to 021 are also
chromium nickel based but with higher carbon and nickel contents and lowered chromium contents.
Carbon was varied stepwise between 0.073% and 0.111% aiming at better mechanical properties.
In alloys no. 022, 024, 025 manganese has been varied between 5% and 7%. In case of 5%
manganese, carbon was additionally varied between 0.032% and 0.07%. The aim was to increase
the Ms-temperature compared to the first LTT series.
The wires were manufactured by Lincoln and shipped to BAM. Furthermore, butt welded plates with
dimensions 400 mm x 240 mm x 20 mm (L x W x T) using each wire were produced and forwarded
by Lincoln to BAM in order to test the mechanical properties. Tensile tests as well as Charpy impact
tests were conducted on samples made from all weld metal. Table 8 shows the results. Values of
alloy LTT011 from the first LTT-test series are included. These values were best for all alloys
suitable so far. As strength values were remarkable high in some cases the yield point could not be
quantified due to experimental limitations. Note, that LLT alloys no. 018 to 021 were not tested
due to extensive cracking.

Table 8: Mechanical properties of second LTT series

LTT no. Rp0.2 Rm A Z CVN Av

(MPa) (MPa) (%) (%) (J @ 0C)
014 882 1106 11 38 24
015 900 1123 12 42 20
017 810 1153 11 32 20
022 328 (Defect) 1074 5 8 24
024 611 1207 4 8 14
025 746 1159 11 31 16
011 - 1400 11 19 24

not determined due to experimental limitations

Cracking behaviour of second series LTT alloys

A detailed test program was conducted at BAM to finally evaluate the cracking behaviour. Hot
cracking, especially solidification cracking, is a major issue of high alloyed steels. The reasons are
closely related to the solidification mode. For that reason the Modified Varestraint Transvarestraint
(MVT) test was used to evaluate the hot cracking susceptibility of each LTT alloy. Figure 31 shows
a schematic of the test. Radius R is 125 mm, which equals 4% bending strain on the sample
surface. TIG re-melting is used during testing. For that reason each single LTT wire was welded
prior to this in a U-shaped groove of special samples. Subsequently, excess weld metal was
machined to achieve test samples with similar geometries. The samples were tested in the
Varestraint mode. That means bending around the weld transverse axis was applied during TIG-
welding (see Figure 31).

35
 Figure 31: Schematic of MVT-test (left) and sample geometry (centre + right)

Each tested sample was evaluated on the basis of its cumulative crack length on the upper surface
by counting each crack to be observed using a magnification of 25 in an optical microscope. In
Figure 32 the cumulative crack lengths measured for each LTT sample are shown. It becomes
clear, that the high carbon chromium nickel wires (018-021) are highly susceptible to solidification
cracking. Minor additional cracks were observed for the HAZ, which are believed to be liquation
cracks. On the other hand, all low nickel and all manganese based alloys passed the test without
significant cracking.

Figure 32: Cumulative solidification crack length found during MVT-test (left) and solidification
cracks located in the centre of the sample (right)

For evaluating the cold cracking behaviour the modified Controlled Thermal Severity (CTS)-test
was applied. This test allows for testing the cracking susceptibility of T joints. Figure 33 shows a
schematic view. The design of the test induces high notch-stresses at the root of two deposited T
joints. Macro sections of test welded joints are used to evaluate cracking as exemplarily shown in
Figure 34.

36
 Figure 33: Schematic of CTS-test (left) and deposition of the first test weld (right)

Figure 34: Example of crack free (left) and cracked macro section (centre + right)

The TEKKEN-test was used for final evaluation of selected wires which have passed the CTS-test
conditions. The TEKKEN-test is a very rigid test for butt welded joints. Due to its design, high notch
stresses are induced at the root side of a single pass weld. A schematic view of the test is shown in
Figure 35. The tests revealed that cracking was present in every LTT weld, as shown exemplarily in
Figure 35.

Figure 35: Schematic of TEKKEN-test (top) and surface crack in test weld (bottom)

Based on results of the CTS-test, the wires 018 to 021 showed complete cracking. Only minor
cracking was observed for the low nickel wires (014-017). Root cracking appeared for the
manganese based wires (022-025). All wires selected showed complete cracking under rigid
restraint of the TEKKEN-test subsequently applied. Despite negative result in the TEKKEN-test the
wires are expected to show crack free welded joints during welding of the fatigue samples within
this project. It has to be considered that the fatigue samples only consist of T joints which are
deposited in multi-layer technique. Due to tempering and relatively high interpass temperature
applied indeed cracking could be avoided. Cracking behaviour in butt welded joints will be focused
on in future work.

37
Furthermore, detailed characterisation of the selected LTT alloys was done (microstructural
analysis, hardness measurements on weld metal cross sections, and Ms temperature
determination). Results and discussion can be found in [6].

Based on the results the alloys no 014 and 011 (first LTT series) were selected finally for fatigue
testing. The reason is that both alloys best fulfil the toughness requirement. The alloys have good
hot and cold cracking resistance. Furthermore, they show acceptable Ms-temperatures as well as
appropriate strength.

Final LTT series

Lincoln has manufactured both wires in a diameter of 1.2 mm. Their weld metals showed an
acceptable level of diffusible hydrogen (ISO 3690) around 3 ml/100 g. The consortium agreed to
call the wires:

series S (Cr-Ni-Fe-based), former LTT014 and

series C (Cr-Mn-Fe-based), former LTT011.

The wires were coiled and packaged in vacuum sealed Al/PE foil. From both wires, 14 coils (196 kg)
were shipped to BIL. Two coils of each were forwarded to BAM for further evaluation. Additionally,
per request of BIL, Lincoln has welded extra plates for further investigation. Table 9 shows the final
chemical compositions.

Table 9: Chemical composition of wires S and C

Wire C Ni Cr Mn Si Ti Al S
series 0.042 4.7 12.1 0.64 0.31 0.008 0.015 0.015
S
series 0.079 0.03 11.2 6.8 0.25 0.007 0.014 0.015
C

BAM conducted investigations concerning the ability of the selected final wires for residual stress
reduction in the weld area. For this purpose specimens as shown in Figure 36 were welded in a
special large scale testing facility. The samples were similar to fatigue tested ones in the project.
As base material high strength steel S960 in quenched and tempered condition was used. Welding
was conducted manually using the parameters shown in Table 10. For reasons of comparison a
conventional high strength filler material G89 according to EN ISO 16834 was applied.

Figure 36: Sample geometry and welding sequence used for test welded joints (thickness 15 mm)

Table 10: Welding parameters for test welded joints (in brackets values for the root)

LTT S LTT C
Welding current 25 V {21 V} 25 V {21 V}
Welding voltage 170-176 A {134-138 A} 160-165 A {124-126 A}
Wire speed 5.7 m/min {4.2 m/min} 5.7 m/min {4.2 m/min}
In this investigation, a new test facility was used. The three-dimensional servo-hydraulically
controlled system has a load capacity of up to 2 MN. It can be used to apply a defined restraint on
large scale samples. Welding heat input and cooling are inducing reaction forces in the system due
to the restraint. In the present case the sample was fixed on both ends and therefore shrinkage
was disabled. During welding the component load was monitored by six independently operated
hydraulic cylinders. This allows for evaluation of the occurring forces and bending moments which
are influenced by the applied welding process parameters. The reaction forces were measured
along the longitudinal axis of the sample. The measured bending moment is located around the
samples transverse axis, as indicated in Figure 37 (right). The interpass temperature was varied
between 100C and 180C and monitored by thermocouples.

38
Figure 37: 2 MN test facility (left) and sample with thermocouples as well as schematic of the
measured loads (right)

Detailed discussion of the measured reaction forces and bending moments can be found in [6].
The resulting reaction stresses can be calculated from the reaction forces related to the sample
cross section and additionally from the bending moments related to the elastic section modulus of
the plate following equation below:

Fy Mx
y = + (1)
A Wx

y - reaction stress on top surface in samples longitudinal direction

Fy - reaction force in samples longitudinal direction
A - joints cross section
Mx - bending moment around the samples transverse axis
Wx - elastic section modulus around the samples transverse axis

Figure 38 summarizes the reaction stresses found for each combination of filler wire and interpass
temperature. It becomes clear that both LTT wires resulted in a significantly lowered specimen load
if an interpass temperature of 180C was applied. This is mainly due to the lower bending moment
induced during welding.

Figure 38: Reaction stresses determined from reaction forces and bending moments for LTT wires
S and C, respectively conventional wire G89; interpass temperatures 100C and 180C

Additionally, local residual stresses were measured in the heat affected zone of each sample by X-
ray diffraction. For reasons of accessibility only longitudinal residual stresses were determined
starting in a distance of approximately 1 mm from the weld fusion line (weld toe).

39
Figure 39 shows the residual stress gradients found in the joints for different interpass
temperatures. In case of an interpass temperature of 100C the maximum residual stresses of
about 250 MPa were located at the weld toe independent of the filler wire. With increasing distance
to the weld the residual stresses decrease. The decrease is higher in case of the LTT welded joints.
With distances higher than 5 mm from the weld toe the residuals stress gradients became
heterogeneous but tending to decrease further showing even compressive values.

Figure 39: Residual stresses in the HAZ of filler wires S, C and G89; interpass temperature 100C
(left) and 180C (right)

With an interpass temperature of 180C, the residual stresses in the LTT welded joints are around
zero at the weld toe, while the conventional G89 joint shows tensile residual stresses around 200
MPa. With increasing distance to the weld the residual stresses in the G89 joint decrease rapidly to
compressive values. The joint welded with filler LTT S remains at a zero stress level close to the
fusion line and shows slight compressive residual stresses at higher distances from the weld. On
the other hand the joint welded with LTT C has a tensile maximum at about 4 mm from the weld
toe. With increasing distance the residual stresses decrease to about zero. As already observed for
the reaction stresses, the lowest residual stress levels (adjacent to the weld toe) were found for
the LTT filler wires when an interpass temperature of 180C was maintained.

Conclusions

From the results presented the following conclusions can be drawn:

Weld filler wires showing LTT characteristics could be successfully designed based on approaches
using chromium and nickel as well as chromium and manganese as main alloying elements.
Adapting these alloying concepts allowed designing filler wires which meet the in-project
requirements concerning Ms-temperature, strength and barely in toughness. While the Ms-
temperature can be estimated roughly by existing analytical formula prediction of the strength and
the toughness is difficult.

It turned out, that the weldability is an important issue not to be predictable by theoretical
considerations. Especially the solidification cracking susceptibility has to be evaluated based on
welding trials. As a result the LTT filler wires developed within the project are less susceptible to
solidification cracking. The LTT filler wire containing chromium and manganese as main alloying
elements shows tempering effects during multilayer welding. Therefore single layer welded joints
not tempered may show higher hardness and should therefore be avoided.

Regarding the ability of the LTT filler wires to control the residual stresses in specific fatigue
samples with T joints, it can be stated that significantly lowered residual stresses adjacent to the
weld toe may be formed compared to a conventional high strength filler wire. Significant stress
relief due to the martensite formation during cooling was proven by in-situ reaction force
measurements. This is particular the case when a comparatively high interpass temperature
(180C) is applied what leads to a significantly lowered bending moment.
Furthermore, the positive effect of LTT filler wires on fatigue behaviour was confirmed in the
fatigue test programme developed in the frame of this project: the benefit in fatigue strength for
Constant Amplitude Loading (CAL) ranges from a factor of 1.08 (R=0.5) to 2.06 (R= 0.1). All
fatigue results are discussed in detail in section 3.2.4.

It is recommended to use rather high interpass temperatures during welding of the LTT fillers in
order to exploit the stress relief effect due to martensite formation.

40
Specific objectives for Lincoln for this project were:

First, understanding and development of LTT electrodes and facilitate to define correlation between
composition, phase transformation temperature, magnitude of stresses developed and fatigue
improvements at the welded joints.

Secondly, development LTT electrodes that will lead to an improvement in fatigue life. This will
entail the understanding of the effect of varying the alloying elements on the phase transformation
temperatures and subsequent stresses developed at the weld and correlate this to stress relaxation
measurements during fatigue testing and subsequent improvements.
The correlation between composition and phase transformation has been established. Metal cored
wires have been developed and manufactured, followed by testing through other partners in the
consortium. It is concluded that with application of LTT consumables, improvement of fatigue life
can be established. This is reported in section 3.2.4.

Exploitation and impact of the research results

LTT welding consumables offer the potential to achieve improved fatigue strength without time and
cost extensive post weld treatments. The results can thus be utilized in a variety of steel
processing industries where fatigue stressed components are made of high strength steel.
Additionally, LTT fillers open up the possibility to produce favourable residual stresses even in
areas of a welded joint which are difficult to access. Moreover, volume-like compressive residual
stress fields in the weld metal and HAZ may be produced contrary to conventional treatment
methods which are mainly limited to surface areas. The freedom of design is increased as the
accessibility for weld treatment methods may be neglected during manufacturing. The adjustment
of the residual stresses (level and distribution) is highly reproducible compared to strongly user-
dependent (manual) weld treatment methods. This allows the economical and efficient use of high-
strength steels.

With the results obtained in this RFCS, many economic sectors of the industry may benefit from an
improved fatigue life in welded high strength steel structures, mainly load carrying constructions
from several types of construction machinery, through application of LTT consumables. The main
benefit will be in critical regions in these constructions, with impact on life cycle cost. A typical
market segment is with heavy equipment manufacturers like Volvo, Caterpillar or John Deere, etc.

It is planned to disseminate the achievements made through next IIW Annual Assembly in July
2014 in form of a presentation and a related publication, which will be submitted to the Journal
Welding in the World after the IIW Annual Assembly. The Document is expected to be published in
early 2015.

41
3.2.3.4 Residual stress distribution measurements and post weld improvement methods
Fatigue cracks in fabricated steel structures occur predominantly at welded joints, where stress
concentrations due to the joint geometry and tensile residual stresses are relatively high. Fatigue
life improvement techniques, which rely on improving the stress field and/or the surface geometry
in and around the welded joints, are generally known to be beneficial. One of most critical point of
the weld is the toe due to combined effect of stress concentration effect of the toe geometry and
high tensile residual stresses.

The objective of the work described in this section was to measure residual stresses after welding
and weld improvement technologies in the toe and in the Heat Affected Zone (HAZ). Another
important factor is how permanent the residual stresses are during fatigue loading. Stress changes
during fatigue loading have been measured by X-ray diffraction. Furthermore, a new method called
Barkhausen noise, was used to monitor the residual stress changes during the fatigue.

In summary, following stress measurement actions were performed:

- Residual stress measurements of HFMI and laser re-melting post-treatment methods. Both
surface and depth distributions were made.

- Residual stress measurement on welded specimens using LTT filler wire.

- Stress measurement before and after fatigue testing to determine the stress relaxation.

- In-situ monitoring during testing using Barkhausen noise.

Measurement methods

X-ray diffraction (XRD) is the most used residual stress measurement method: a) it is widely
available b) relatively fast and c) does not need any calibration. X-ray diffraction is limited only to
measure the stresses in the surface. The method gathers information in a depth of about 7m. To
measure depth distribution the material has to be polished electrochemically. A procedure to do
this electro-polishing was developed within this work package. Another important value which is
can be collected using in XRD measurement is the Full Width at Half Maximum (FWHM) of the
diffraction peak shown in Figure 1. FWHM describes the density of lattice defects such as
dislocations. The higher the density of defects, the higher is the value of FWHM.

Figure 40: Full Width at Half Maximum definition

Measurement results

Surface measurement distributions were started as close as possible from the toe towards base
material. The first 10 mm was measured with a step size of 1mm and after that a step size of 2
mm was used up to 20 mm. The sample geometry and measurement location, as well as
measurement arrangement and direction are shown in Figure 41 and Figure 42.

43
Figure 41: The sample geometry and the measurement location. From most of samples all four
locations A, B, C and D were measured

Collimator
a) b)
Figure 42: Measurement arrangement a) starting point of measurement X-ray collimator and b)
measurement directions.

As welded surface measurement

Residual stress measurements were done for all samples on four different locations. Those
locations were the ends of the welded longitudinal attachments marked as A, B, C and D. X-ray
measurement parameters were: radiation CrK, ferrite lattice plane (211), spot size 2mm.
Following elastic constants were used to calculate stress values: Youngs modulus 211000 MPa and
a Poissons ratio of 0.3.

Direction phi = 90 Direction phi = 0

250 100

200 50
0 A
150 -50 0 5 10 15 20
B
Stress MPa

A
Stress MPa

100 -100
B -150 C
50
C -200 D
0 D -250
0 5 10 15 20 -300
-50
-350
-100 -400
Distance mm Distance mm

Figure 43: As welded stress values of S690QL 10 mm specimen (locations as described in Figure
41)

3.4
3.2
3
FWHM deg

2.8 A
2.6 B
2.4 C
2.2 D
2
0 5 10 15 20
Distance mm

Figure 44: As welded FWHM values of S690QL 10 mm specimen

44
In total seven as-welded samples were measured. All had similar tensile stress distributions.
Maximum tensile stress varied from 100 to 600 MPa. Tempering effect of heat from the welding
was affecting to about 8 mm distance from the toe based on FWHM value changes, see Figure 44.
From Figure 43 it is clear that highest tensile stresses were following phi = 90 direction, stresses
following phi = 0 were lower tensile or up to compressive stresses.

HFMI treated surface measurement

By HFMI the steel surface toe area is severely deformed i.e. the material in the treated area
undergoes important work hardening. This leads to a high hardness of the treated zone, indicated
by high FWHM values. There are three factors in HFMI treatment which affect the fatigue strength:
geometry of the tow, compressive stresses and work hardening. All these three factors typically
improve fatigue strength.

Surface residual stress measurements were done on HFMI treated specimens and results are
shown for S700MC in Figure 45. For comparison data on a lower strength grade is included as-well
as residual stress measurements on specimens in as-welded state (i.e. untreated condition). From
these results it is clear that HFMI treatment creates compressive residual stress up to 4 mm from
the weld toe. Measurements were done on several specimens leading to the same conclusions as
stated above: compresses stresses are created systematically in all HFMI treated specimens.

800 4.5
S700MC-10-12
600 4 S420MC-05-46
400
FWHM deg

3.5 S700MC-10-41
Stress MPa

200
3 S420MC-05-19
0 S700MC-10-12
2.5
-200 S420MC-05-46
-400 S700MC-10-41 2

-600 S420MC-5-19 1.5

0 2 4 6 8 10 12
-800
0 4 8 12 Distance mm
Distance mm
Figure 45: HFMI treated specimens (420 and 700MPa strength): stress values (left), FWHM values
(right). Average values of four different locations, results in as-welded condition as reference.

HFMI treated depth distribution measurements

One important issue is to know how large the impact volume of the HFMI treatment is. This
requires measurement data of the stress depth distribution. Since the XRD method only has a
depth penetration of 7 m, the material had to be removed step wise to be able to measure the
depth distribution of the residual stresses. Material removal is changing the stress state of the new
surface, but this change is considered small if the removed amount of material is small compared
to the total volume of the sample. Within this work package, a special procedure using
electropolishing, was developed. Figure 46 shows a typical polished area. The stress distribution
was measured along a 20 mm measurement path after each polish step, in the same way as in the
surface measurements.

Figure 46: Electropolished area 5*20 mm, depth 1.5 to 2.0 mm

45
 Figure 47: Depth distributions of HFMI treated S700 MC specimens

From results shown in Figure 47, it can be concluded that residual stresses are compressive to a
depth of about 1 mm. Compensating tensile stresses are not located on the surface but deeper in
the material. This is beneficial for fatigue strength. Distributions are similar in all measured cases.
Maximum depth of the effect due to HFMI treatment in terms of hardness is about 0.5 mm based
on FWHM values (see Figure 48).

3.5
FWHM deg

3 25

2.5 29
33
2
37
1.5
0 0.5 1 1.5 2
Depth mm

Figure 48: Depth distribution of FWHM values in the zero position of HFMI treated S700 MC
specimens

Residual stress depth measurements after fatigue test

How stable are the compressive stress state during cycling / fatigue loading? This is one of main
questions when HFMI treatment is used to improve fatigue strength. There were two different
approaches to measure relaxation of stresses: a) measure stresses before and after the fatigue
testing and 2) interrupt the test and measure the stresses.

Interrupted stress measurements by X-ray diffraction were done by KTH. These results are
reported in section 3.2.5.4.

Before and after fatigue testing several samples were measured to find out how much residual
stresses of HFMI treatments have changed or relaxed during fatigue tests. Measurements were
done in the unbroken end of the samples. Results for S700MC samples are shown in Figure 49.

46
 700 300
600 Before 200
500 After 100

Stress MPa

Stress MPa
400
0
300 0 2 4 6 8 10
-100
200
100 -200 Before
0 -300 After
-100 0 5 10 15 20 25
-400
Location mm Location mm

(a) As welded S700MC sample (b) HFMI treaded S700MC sample

Figure 49: Relaxation of surface residual stresses (a) as welded S700MC sample (CA, R= 0.1,
Fmax=250kN, Smax=312.5MPa) and (b) HFMI treated S700MC sample (VA, R = -1, Fmax
=300kN, Smax=375MPa).

Tensile stresses of as-welded and compressive stresses of HFMI treated samples have become
significantly lower. The amount of stress relaxation varies between the samples. This might be
attributed to the fatigue parameters used or to a non-uniform loading of the specimen. Same clear
trend of relaxing stresses can be observed in the depth measurements which have been done after
fatigue tests. Both compressive and tensile stresses are relaxed almost totally when comparing
stress distributions in Figure 47 and Figure 50.

Figure 50: Residual stress depth distribution profile for sample S700MC (VA, R=-1, Fmax=300kN,
Smax=375MPa) and for sample S700 MC (VA, R=-1, Fmax=300kN, Smax=375MPa). Samples are
HFMI treated and fatigue tested

Laser re-melting

Laser re-meting belongs to the group of methods which improves the weld toe geometry. Other
commonly used methods are TIG dressing or burr grinding. Laser treatment is an interesting
method since it may be easier to control the treatment parameters. Laser treatment also changes
the residual stress state which again is known to affect the fatigue strength, as can be seen in
Figure 51.

4.5
200
Lase re-melt, 600QT 4
150
Laser re-melt, 900MC
FWHM deg

100 3.5
Stress MPa

As welded, 900MC
50 3
0 Laser re-melt, 600QT
2.5
-50 Lase re-melt, 900MC
2
-100 As welded, 900MC
1.5
-150
0 5 10 15 20 25
0 5 10 15 20 25
Distance mm Distance mm

Figure 51: Laser re-melted sample (690QL, 10 mm and 960MC, 5 mm) and for comparison as-
welded sample (960MC, 5 mm): Stress distribution (left) and FWHM values (right)

Laser re-melting hardens the surface layer which can be seen from FWHM distribution in Figure 51,
increasing values next to the toe. Furthermore, tensile residual stresses decrease after laser
dressing as can be seen in Figure 51.

47
Low Transformation Temperature (LTT) filler material

Three LTT samples were measured for evaluate the benefits of using LTT filler materials for welded
joints in HSS. Results of residual stress measurements are shown in Figure 52.

300 3.2
7MC-05-C-01
200 3
9MC-05-C-01

FWHM deg
100 2.8
9MC-05-S-07
Stress MPa

0 2.6
0 5 10 15 20 25 7MC-05-C-01
-100 2.4
9MC-05-S-01
-200 2.2 9MC-05-S-07
-300 2
a 0 5 b
10 15 20 25
-400
Distance mm Distance mm

Figure 52: Surface stress and FWHM distribution of three LTT samples (S700 MC LTT C; S 960
MC LTT C and LTT S) in 5mm thickness

Base material stresses and FWHM distributions are similar to as-welded samples (see Figure 43 and
Figure 44). Tensile stresses next to the weld toe are slightly lower compared to as-welded
specimens.

In-situ monitoring of stresses

How the residual stresses change during fatigue testing is commonly tested mainly by interrupting
the fatigue test and then measure the stress state. The predominant stress measuring technique
used is X-ray diffraction. However, this is a time consuming procedure.
The Barkhausen Noise (BN) method is known to be strongly stress sensitive, but the weak point is
microstructural sensitiveness. However, the BN noise is fast method and values can be recorded in
fraction of seconds. This would make it possible to follow up stress changes continuously during
fatigue testing / cyclic loading. Within this project BN in-situ measurements were made during
fatigue testing in purpose to examine stress relaxation or crack initiation.

The motive of these tests was to investigate if there were some noticeable changes in BN when
monitoring the signal during fatigue testing. During the tests one special designed BN fixture was
attached to one welded stiffener and the signal was captured directly on the HFMI treatment
groove (see Figure 53). In those samples where the sensor was in the same location as the sample
broke, the trend of the measured data had similarities with each other. Usually the signal stayed
very stable from beginning of the tests until somewhere in the middle of the test where it started
to increase slowly. In Figure 54 and Figure 55, the data from end part of those fatigue tests where
changes in monitored signal occurred is shown. In all presented cases, the sensor was located at
the broken end of the specimen.

Figure 53: Picture of the measurement setup of in-situ Barkhausen noise monitoring during
fatigue test.

48
Figure 54: The increasing behaviour of the measured signal in the end part of fatigue test until
sample broke down.

Figure 55: Data from end part of fatigue test. The overall trend has some general similarities with
other test data.

Conclusion

Extensive residual stress measurement work was conducted within this project. The results of the
as-welded measurements are in accordance with literature and with common known facts that
welding creates harmful tensile stresses and tends to soften the material in the heat affected zone.
It was also discovered that the effect of the welding has an impact in about 10 mm distance from
the weld toe regardless sample thickness or pre-processing of sample surface.

It seems that laser re-melting and use of LTT wires reduced tensile residual stresses to some
extent. Thus, the benefits of laser re-melting might be limited to its capability to improve weld toe
geometry.

The HFMI method showed significant improvements in residual stress state compared to an as-
welded condition. With this method residual stresses were clearly shifted to compressive side in the
critical area near weld toe. It also improved the durability of the samples due to the work
hardening effect due to HFMI treatment. Depth profiles revealed more significant details about
phenomena due to HFMI method. It seems that HFMI method can generate relatively high
compressive stresses up to 1 mm depth or possibly even deeper. Besides the efficiency of creating
compressive stresses it also creates a very steep stress gradient deep inside the base material.
Stress measurements before and after fatigue testing revealed that important stress relaxation
takes places for the specimens tested in VA, which is confirmed also by experiments described in
section 3.2.5.4.

In the in-situ measurements changes in the recorded signal were observed. However, too few tests
were conducted to make any exact conclusion from these tests. Nevertheless, it was clear that the
Barkhausen Noise changes during the fatigue testing. The main reason for these changes is
believed to be the relaxation of stresses as previously described within this report. Although there
was quite large scatter in the BN results, it is believed that the method shows promising
possibilities for future monitoring of critical welded joints.

49
3.2.4 WP3: Fatigue Testing
The overall objective of the work described in this section is to establish the levels of improvement
in the lives of welded attachments that can be achieved by application of the selected improvement
techniques, with particular reference to the potential extra benefit to be derived from the use of
high-strength steels.

The fatigue test specimens consisted of steel plates with longitudinal welded attachments, as
shown in Figure 3.

The specimens were fatigue tested axially in various fatigue testing machines in five laboratories.
In some cases the specimens were gripped directly in the testing machine while in others it was
necessary to use pin loading. Consequently, the specimens incorporated 50 mm diameter holes at
each end to facilitate pin loading.

The weld detail in the test specimen is classified as FAT 63 in the International Institute of Welding
(IIW) fatigue design recommendations [7]. The corresponding design S-N curve was compared
with the present test results for reference purposes. The corresponding design curve in BS 7608
[8] is Class F2, which has the same slope as IIW FAT 63 but is slightly lower.

Cyclic loading

Fatigue testing was conducted under both constant amplitude cyclic loading (CAL) and variable
amplitude spectrum loading (VAL).

The CAL tests were performed under tensile loading at stress ratios, maximum/minimum applied
stress R = 0.1 or 0.5. Applied stress ranges were chosen to generate lives in the range of 104 to
107 cycles. Tests exceeding 107 cycles without any evidence of fatigue cracking were classed as un-
failed or Run outs.

In the case of the VAL tests, the applied loading cycled from tension to compression about a mean
stress of zero, so that R=-1. The maximum applied stress range for the spectrum was determined
as a percentage of the material yield strength; this was typically in the range of 35% to 65%. The
spectrum used for the VAL programme was designated SP3_Short (edited) and contained 200,000
turning points. A sample of the stress-time history is shown in Figure 56. Prior to the application of
this spectrum, SP3_Long was considered (500,000 turning points) but abandoned in favour of the
aforementioned spectrum. Further details are given when the results of the tests are discussed.
Percentage of applied stress

Turning points

Figure 56: Variable amplitude loading sequence (SP3_Short (edited)) used in the current
programme.

51
Equivalent constant amplitude stress range

In order to compare fatigue test results obtained under both constant and variable amplitude
loading it is useful to define the constant amplitude stress that is equivalent in terms of fatigue
damage to the applied variable amplitude stress spectrum. For an applied stress spectrum
consisting of ni cycles at stress range si, and assuming that Miners rule is correct, this is given by:

( )
1
s m .n i m
Seq = i [1]
n i

where m is the slope of the constant amplitude S-N curve for the detail concerned, expressed in
the form:

S m .N = A [2]
where S is the constant amplitude stress range and A is a constant.

It should be noted that the above definition of the equivalent stress is based on the assumption
that the constant amplitude S-N curve with slope m extends to the lowest stress ranges in the
spectrum. As will be seen, some of the lower stress levels applied in the present tests were
probably below the fatigue limit for the test specimens. In practice (eg BS 7608), such stresses are
usually assumed to be less damaging than implied by the S-N curve, such that their effect is
accounted for if the S-N curve is extrapolated below the fatigue limit at the shallower slope of
m+2. Making such an assumption increases Seq slightly, depending on the magnitude of the fatigue
limit.

Good agreement between constant and variable amplitude test results with the latter expressed in
terms of the equivalent stress range, would indicate that Miners rule was correct for the weld
details and stress spectrum considered, whereas disagreement would indicate that Miners rule was
not accurate.

A detailed analysis of fatigue test results, including the fatigue test matrix can be found in [9].
Below a summary of the fatigue test results is given.

Fatigue Test Results Constant Amplitude Loading

As-welded specimens
The endurance data for specimens tested in the as-welded condition are plotted in the form of an
S-N diagram with double logarithmic axes in Figure 57. The results are considered in terms of
nominal applied stress ranges. In all cases the fatigue failure mode was as expected, consisting of
fatigue crack growth through the main plate thickness from one or more of the weld toes around
the ends of the attachments.

Initially, S-N curves of the form of Eq. 2 were fitted to the results from each individual test series
by the method of least squares linear regression, treating log N as the dependent variable. Only
results from failed specimens were included. Free regression analysis suggested a slope m close to
3, the slope of the IIW FAT 63 curve, and so the data were re-analysed assuming this value. The
results, including the standard deviation of log N, are presented in Table 23 (see Appendix). Since
there is no consistent influence of steel grade, plate thickness or R on the results, regression
analysis was also performed on the combined data and the results are included in Table 23. The
resulting mean S-N curve is also included in Figure 57, together with the IIW FAT 63 design curve
for comparison.

52
 Figure 57: Fatigue test results obtained from as-welded specimens under CAL.

Laser re-melting treated specimens

The test results obtained from the laser re-melted specimens are presented in Figure 58 in
comparison with the mean curve obtained from the as-welded specimens and the IIW FAT 63
design curve.

Regression analysis was performed to fit mean S-N curves to each test series, as summarised in
Table 24. The slopes varied from 2.59 to 4.30 but for ease of comparison with the S-N curve for
as-welded joints the regression analyses were repeated assuming m = 3. The same analysis was
performed for the combined data and the resulting mean curve with a slope of m = 3, which was
close to the free regression value of 2.85.

Figure 58: Fatigue test results obtained at R=0.1 from laser re-melted specimens under CAL.

Comparing the results from laser re-melted specimens with those from as-welded ones, there is a
marked improvement in fatigue life. The treatment that provided the highest improvement for both
steel types (S690 and S960) in the same thickness was Condition 2 (C2), the improvement in
fatigue strength at 2 x 106 cycles being between 44 and 47%. However, there was considerable
overlap between the results and comparing the mean curve fitted to the combined toe failure data
with that for the as-welded specimens indicates an improvement of 40% at 2 x 106 cycles.

53
The 95% confidence limits on the estimated slope indicated that it would be justifiable to assume
m =3, as with the as-results for as-welded specimens. The corresponding fatigue strength at 2x106
cycles is included in Table 24 (see Appendix). Although there was good agreement between the
results for Conditions 2 to 5, it will be noted that Condition 1 consistently gave the lowest results.

This suggested that based on a small data-base for each individual treatment condition, a more
focused spot size with less laser power but more power density is preferable to higher power,
larger focus size and less power density.

High Frequency Mechanical Impact treated specimens

The same procedure as that followed in the evaluation of the results for laser re-melted specimens
was used in an initial evaluation of the results from HFMI treated specimens. Thus, all the results
are compared with the S-N curve for as-welded specimens and the IIW FAT 63 design curve in
Figure 59, while Table 25 (see Appendix) summarises the results of the regression analyses.

Figure 59: Fatigue test results for the HFMI treated specimens under CAL.

Referring to Figure 59 it will be evident that the results are very widely scattered with little clear
indication of which factors had the most influence on fatigue life. In an attempt to clarify the
situation separate plots were produced for each steel, which can found in [9].

Apart from the influence of R no consistent effect of either of the other two variables, plate
thickness or HFMI condition, was evident. Therefore, curves were fitted to the combined results
obtained at R=0.1 or 0.5 and in most cases it will be seen that it would be justifiable to assume a
slope of m = 5. One exception is the curve fitted to the R=0.1 data from the S690 steel but since
there were so few results for each condition little confidence can be placed in the fitted curve. The
other exception was S960 steel results for R=0.5. However, excluding the exceptionally low result
obtained at 90MPa from a 5mm condition C1 specimen, leads to a mean curve with a slope
between 3.5 and 5.3, justifying the choice of 5. Therefore, S-N curves were fitted to each set of
data assuming m = 5 and the resulting fatigue strength at 2 x 106 cycles compared, as shown in
Table 25.

To facilitate estimates of the improvement in fatigue strength from the three weld toe treatments,
the results were analysed further without distinguishing between the steel types, as indicated in
Table 25. Recalling that for the as-welded condition the fatigue strength at 2x106 cycles (m = 3)
was 78MPa, there is an increase in strength of around 1.7 times for R=0.5 and 2.3 for R=0.1.

Comparing the results of the S700MC 10mm thickness PIT specimens with the same grade and
thickness but UIT treated (Condition C1), shows no real effect on fatigue strength at 2x106 cycles.

54
LTT filler wire specimens
The results of tests performed on specimens treated using LTT filler wire are presented in Figure
60, together with the curve for as-welded specimens and the IIW FAT 63 design curve.

Figure 60: Fatigue test results for the LTT filler wire specimens under CAL.

As before, linear regression analysis was performed for each series, on condition that there were at
least three valid results, and the results are given in Table 26 (see Appendix).

Comparing the results for identical tests there does not appear to be any significant difference
between the LTT C and LTT S wires. However, as with HFMI, the other residual stress-based
improvement technique, the main distinction seems to be related to the applied stress ratio, the
results for R=0.1 being higher than those for R=0.5.
In each case there is no evidence of a consistent influence of LTT wire type or plate thickness, the
clearest distinction being R. It will be evident that it would be justifiable to assume a slope of m =
5 for the data obtained with R=0.1 and 3 for the R=0.5 data. To facilitate comparisons between the
results, the regression analyses were repeated with these assumptions and the results are included
in Table 26.

With regard to the influence of the steel on the benefit of the LTT treatment it will be evident from
Table 26 that there was no influence, all three steels giving essentially the same fatigue strengths
for the two R values. With regard to the benefit obtained from the LTT treatment, and recalling that
the fatigue strength of the as-welded specimens (m = 3 value) at 2x106 cycles was 78MPa, it will
also be evident that there was no significant improvement for R=0.5. In contrast, at R=0.1 the
fatigue strength at 2x106 cycles was doubled.

Comparison of the benefits from the three improvement techniques for CAL
The improvements in fatigue strength obtained from the three weld toe treatments are
summarised in Table 11.

Although the fatigue strength at 2 x 106 has been used as a reference value it is only useful for
identifying the improvement in fatigue strength if the S-N curves for the as-welded and treated
specimens have the same slope, that is m = 3. As was noted, this was not generally the case and
the results from treated specimens supported the shallower slope of m = 5. Therefore, Table 11
gives the fatigue strength improvement factors at three endurances, 104, 2 x 106 and 107 cycles.
These were obtained by dividing the mean fatigue strength obtained from the S-N curve for the
improved specimen results, fitted assuming the slope indicated in the Table, by the mean fatigue
strength for the as-welded specimens at the same endurance. In this respect the fatigue strengths
for the as-welded specimens were 453, 78 and 45 MPa at 104, 2 x 106 and 107 cycles respectively.
Referring to Table 11, in the case of weld toe laser re-melting the most useful comparison is
between the curve fitted assuming a slope of m = 3 to all the results from specimens failing from
the treated toe. Thus, for R=0.1 the improvement obtained was by a factor of 1.5, that is 50%.
Since the fitted curve is parallel to the curve for as-welded specimens the improvement is
independent of endurance.

55
However, it is generally found that the benefit of weld toe improvement techniques decreases with
increase in applied stress and therefore the improvement found here should only be assumed to
apply over the range of applied stress ranges used in the tests. Another restriction concerns the
applied stress ratio. The present tests were all performed under R=0.1 but experience indicates
that the benefit could be less at higher positive stress ratios.

It was anticipated that benefit from the other two improvement techniques, where the
improvement arises mainly as a result of the introduction of compressive residual stress, would
depend on applied stress ratio and tests were performed at R=0.1 and 0.5. Consequently, in both
cases a distinction is drawn between the improvements in fatigue strength for these two R values.
In the case of HFMI, the appropriate S-N curves are those obtained from the combined data at
R=0.1 or 0.5, fitted assuming a slope of m = 5. As anticipated, the improvement is greater for
R=0.1 than 0.5 and it decreases with increase in applied stress range (corresponding to a decrease
in endurance). For R=0.5 there is no improvement at N=104 cycles.

In the case of the LTT treatment, although it proved to be justifiable to assume a slope of m = 5
for the treated specimen results obtained with R=0.1, the value was close to 3 for R=0.5.
Furthermore, in contrast to the use of HFMI, the benefit for R=0.5 was negligible, amounting to
just 6%. However, there was significant improvement at R=0.1, increasing with decrease in
applied stress range, but it was marginally less than that obtained from HFMI.

Table 11: Comparison of regression analysis for CAL tests

Free slope and

Fatigue strength improvement at N cycles
95% confidence Assumed
Specimen condition limits slope N = 104 N = 2x106 N = 107

Laser re-melted, R=0.1 3.79 0.76 3 1.5 1.5 1.5

HFMI, R=0.1 3.72 1.12 5 1.17 2.36 2.96

HFMI, R=0.5 4.23 0.82 5 0.84 1.69 2.13

LTT, R=0.1 4.42 0.96 5 1.03 2.06 2.60

LTT, R=0.5 3.22 0.31 3 1.08 1.08 1.08

Assessment guidelines for HFMI data under CAL

As part of the project framework, a new guidance document has been produced by the IIW giving
the proposed fatigue assessment guidelines for fatigue strength improvement by HFMI [10]. This is
discussed in 3.2.6.

The assessment is based on the assumption that the HFMI data produced conforms to a slope of 5
and are therefore forced to that slope. Based on the material yield strength, the maximum fatigue
class improvement level that can be claimed is given, this is subsequently penalised according to
the applied stress ratio.

In the present case the HFMI data are first split into two groups of yield strength with the
S690/S700 in one group and the S960 in the other. The two groups are then split further according
to stress ratio. Table 12 summarises the groups and the appropriate increase in fatigue class from
FAT 63 allowing for penalties on stress ratio.

Table 12: Maximum FAT Class based on IIW Assessment Guidelines

Maximum Penalties due Permissible Resulting

Steel Stress increase in to applied increase in IIW FAT
grade ratio, R FAT Class stress ratio FAT Class Class
0.1 0 6 125
S690/S700 6
0.5 3 3 90
0.1 0 8 160
S960 8
0.5 3 5 112

The data represented in Figure 59 were re-analysed using linear regression analysis to determine
the slopes (m) for each set of data, as shown in Table 13.

56
 Table 13: Slope (m) of HFMI data

Steel grade Stress ratio, R Slope (m)

S690/S700 0.1 2.68
S960 0.1 4.57
S690/700 0.5 4.79
S960 0.5 3.96

Where the free slope of the data was closer to 3 than 5, both forced mean curves are included in
the Figures. Where the free slope was closer to 5, only the mean curve forced to 5 was included.

For S690/700 material at R=0.1 (Figure 61), one or two data points fall below the FAT 125 design
curve, however, the majority of the data sit above supporting the guidelines. As discussed earlier,
it is justifiable to assume a slope of 5 for the S700 HFMI data. In the case of the S690 data, there
were too few results for each condition to fully form a conclusion. Therefore, with the data set
combined as in Figure 61, the assumption of a slope of 5 is justifiable; however, more data at the
higher endurance end would be of benefit.

Figure 61: S690/S700 HFMI data at R=0.1.

The same type of analysis was done for each set of data represented in Table 13. It can be
concluded that by comparing the HFMI data according to the guidelines produced by the IIW, the
recommendations would seem justified. However, in some cases further data at higher endurances
are of benefit in order to assess the level of conservatism typically at, or, beyond 107 cycles.

Fatigue Test Results Variable Amplitude Loading

Round robin exercise was done to validate the selected spectrum (see [9]).

As-welded specimens
As mentioned earlier, a convenient way to consider variable amplitude fatigue test results is in
terms of the equivalent constant amplitude stress range. As the maximum stress range in the
programme of tests varied it was necessary to calculate the equivalent stress range for each
individual test. This value is included in Table 27 and used to plot the test results in Figure 62. Also
shown are the constant amplitude test results used to define the S-N curve, the IIW FAT 63 design
curve and the Class F2 mean curve.

The corresponding value(s) of the Miners rule summation (n/N) at failure, calculated using N
values from the mean constant amplitude S-N curve (m=3) and the Class F2 mean curve are also
included in Table 27.

All specimens exhibited identical modes of failure to those previously observed under CAL, with
multiple weld toe cracking. A typical example of a failed specimen is given in Figure 63a-b. Faceted
crack growth was observed around the weld return.

57
Figure 62: Fatigue test results obtained from as-welded specimens under variable amplitude
loading.

Faceted growth

Weld toe failure

a)

b)
Figure 63: Typical example of a failed as-welded specimen tested under variable amplitude
loading:

a) View of specimen plate surface showing weld toe failure;

b) Fracture faces showing faceted growth around the weld return.

58
HFMI treated specimens
The results for the HFMI VAL tests are presented in Table 28 (see Appendix) and plotted in Figure
64. The mean constant amplitude S-N curve (m=3 and m=5) for HFMI treated specimens, the IIW
FAT 63 design curve and the Class F2 mean curve are also shown for comparison. As with other
VAL tests performed, the equivalent stress for each test was calculated and is included in Table 28.

Figure 64: Fatigue test results obtained from HFMI treated specimens under variable amplitude
loading.

For the tests performed it was found that the average n/N for the 10mm thickness S960QL
specimens in condition C1 gave the greatest increase in life compared with the HFMI mean CA S-N
curve, by an average factor of 12.12. The lowest increase was associated with the S700MC 10mm
thickness specimens in condition C1 with an average factor of just 1.41.

Laser re-melted specimens

Due to the generally better performance of condition C2 seen in the results for CAL tests, condition
C2 was chosen for the VAL test specimens. The results for the laser re-melted VAL tests are given
in Table 29 (see Appendix) and presented in Figure 65 based on the equivalent stress range
derived for each material grade. Also shown for comparison is the mean laser re-melted constant
amplitude S-N curve (m=3 and 5), the IIW FAT 63 design curve and the Class F2 mean curve.

Figure 65: Fatigue test results obtained from laser re-melted specimens under variable amplitude
loading.

59
For the two material grades tested it was found that the average n/N for the S960QL specimens
gave the greatest increase in life compared with the laser re-melted mean CA S-N curve by an
average factor of 3.65; the S690QL material giving an increase of 2.37.

LTT filler wire specimens

The results of the LTT tests are given in terms of equivalent stress range in Figure 66 and include
the mean LTT constant amplitude S-N curve (m=3 and 5), the IIW FAT 63 design curve and the
Class F2 mean curve. They are also presented in Table 30 (see Appendix) and include the
associated Miners summation.

Figure 66: Fatigue test results obtained from LTT treated specimens under variable amplitude
loading.

Comparison of the benefits from the three improvement techniques for VAL
The results of tests performed under VAL give average summations (n/N) of life compared with
their respective CAL tests. In order to compare the improvement techniques with one another the
results are compared against the relevant BS7608 design curve, for Class F2. These results are
summarised in Table 14.

Table 14: A comparison of average n/N at failure between improvement techniques with the
Class F2 mean curve given in BS7608

Average n/N at failure

Improvement Material Thickness, based on Class F2 mean
technique grade mm curve
S700MC 5 7.18
S700MC 10 3.20
HFMI S960MC 5 9.38
S690QL 11.35
10
S960QL 27.48
S700MC 5 -
S700MC 10 -
Laser re-
S960MC 5 -
melted
S690QL 5.05
10
S960QL 7.78
S700MC 5 2.36
S700MC 10 4.02
LTT S960MC 5 4.23
S690QL 5.10
10
S960QL 10.37
*Note this was not the lowest factor determined for this improvement technique.

60
It can be seen that for all thicknesses and material grades tested (with the exception of S700MC
10mm thickness), HFMI gave the greatest benefit under VAL when compared with the Class F2
mean curve. Where comparisons can be made, this resulted in a 204% increase in life for S700MC
5mm thickness, 122% for S960MC 5mm, 123% for S690QL 10mm and 165% for S960QL 10mm
when compared with LTT. Against the laser re-melted technique, the increases were larger still at
125% for S690QL 10mm and 253% for S960QL 10mm thickness.

In the case of 10mm S700MC, HFMI did not offer the greatest benefit. Instead this was found in
the LTT specimens, with an increase in life of 26% over HFMI.

Assessment guidelines for HFMI data under CAL

A new guidance document has been produced by the IIW giving the proposed fatigue assessment
guidelines for fatigue strength improvement by HFMI.

Table 12 shows the resulting FAT Class following grouping of the material yield strengths and
penalties due to the applied stress ratio. In the case of the HFMI VAL data, the FAT Class relating
to a stress ratio of R=0.1 has been considered, representing the highest FAT Class that can be
obtained for the material grade and joint configuration.

Figure 67 and Figure 68 show the VA HFMI data plotted in their respective groups. The IIW FAT 63
design curve and the allowable IIW design curve (m=5) are again plotted for comparison. The data
was analysed using linear regression analysis to determine the slopes (m) for each set of data as
follows:

S690/700 m = 5.22

S960 m = 2.87

Figure 67: S690/700 VA HFMI data.

Figure 68: S960 VA HFMI data.

In both cases the limited data support the guidelines produced by the IIW. In the case of the S960
specimens, even though the free slope of the data is closer to m=3, the plot would suggest that
the assumption of a slope of m=5 for the FAT 160 curve is justified.

61
Discussion
Fatigue tests under constant amplitude loading
In some of the CA tests performed (particularly HFMI) a significant number of variables have been
considered. The number of tests performed has not sufficiently covered the variables explored but
has given some indication as to the level of performance of the various techniques and optimum
parameters to be employed.

The fatigue tests performed under CAL all produced mean curves above the IIW FAT 63 design
curve, which was to be expected. Comparing improvement techniques found that HFMI gave the
greatest benefit recording the highest fatigue strengths at 2x106 cycles.

In general, techniques such as burr machining, needle peening and hammer peening increase the
fatigue strength of a welded joint by 30% [8], [11], [12], or a factor of approximately 2.2 on life.
The level of improvement by the techniques applied support that found in guidance literature; with
HFMI giving the greatest level of improvement (a factor of 2.36 on strength for R=0.1 data and
1.69 for R=0.5 data).

Work in the field on high strength steels [13], reported fatigue strength improvements by HFMI of
>150% for CAL and >100% for VAL. Whilst the former is considerably higher than that observed in
the present study for CAL, it is probably significant that the tests performed were under alternating
tension-compression (i.e R=-1) loading, whereas the present tests were all performed under fully
tensile loading. It is generally found that the benefit of weld toe improvement techniques that
depend on the introduction of compressive residual stress, like HFMI, is greater for negative R
values than positive, especially under CAL.
In terms of the re-melting technique applied, another more common approach is TIG dressing. In
Figure 58 published data on S700 material has also been included to provide some degree of
comparison between the two re-melting techniques. The findings of which, based on a small
population of data, showed almost no difference between lives generated, with the TIG dressing
data comfortably sitting about the laser re-melting mean curve.

In general, the most significant influence in the results of improved joints under CAL has been that
of mean stress, with tests performed at R=0.1 indicating higher fatigue strengths than those at
R=0.5. With regards to stress, it was considered that an increase in fatigue strength would be as a
result of an increase in material yield strength. Under CAL this has not been the case, with a
number of tests performed using S690QL outperforming the higher strength S960 material.

As discussed earlier, the HFMI data is in agreement with the IIW guidance document [10]. Where
on occasion a slope of 3 is closer to the free slope than 5, it has been discussed that as the
available data for m = 3 was too few (S690 R=0.1) combining with the S700 R=0.1 data justifiably
resulted in a slope of 5, however, more data at the higher endurance end would be of benefit to
assess the level of conservatism. This was also the case for the S690/700 R=0.5 data.

A review of high R ratio HFMI data for low strength steels (S355) for a variety of joint
configurations has recently been performed (Mikkola et al, 2013). Here high strength steel such as
S700 and S960 has been included, but normalised to compare with the S355 material grade. The
review again shows that the data agree well with the IIW guidance, although as recommended
above, more data in the high endurance end (>2x106 cycles) would be of benefit.

Fatigue tests under variable amplitude loading

The majority of tests performed were typically above the respective mean of the CA data when
expressed in terms of the equivalent constant amplitude stress range, calculated using Miners rule.
This suggests that Miners rule was not accurate for the applied loading spectrum and hence
resulted in over conservatism, thus, safer lives (Figure 62, Figure 64, Figure 65 and Figure 66). In
the as-welded and HFMI treated conditions, S700MC 10mm thickness (condition C1 in the case of
HFMI), were in good agreement, indicating that Miners rule was accurate for the applied loading
spectrum in these instances.

As discussed earlier, when comparing the three improvement techniques with the performance of
the as-welded VAL results against the Class F2 curve in BS7608, HFMI offers significant
improvements in increases in life compared with LTT for the 5mm thickness S700MC and S960MC
material tested.
With respect to HFMI, whilst the performance was good, it is considered that in both cases (CAL
and VAL), the performances would have been better if the improvement process had been
performed in a more controlled manner. Indeed in some cases individual strike marks were
observed on the surface in an irregular manner.

62
Again as in the case for the CA tests discussed above, the increase in fatigue strength was not
necessarily a result of the increase in yield strength. When referring to the Tables for each
improvement technique in Appendix, lower strength steels on occasion outperform the higher
strength.

When considering the HFMI guidance document, the limited data under VAL supports the maximum
FAT Class increase for the material grade and joint configuration. The applied stress ratio for the
spectrum was R=-1, however, the penalties relating to a stress ratio of R=0.1 was used therefore
reducing the level of conservatism.

Conclusions

A total of 487 specimens (excluding round robin exercises) have been fatigue tested in the
programme of work. The main conclusions drawn from this programme of work are as follows.

Constant amplitude loading:

Under HFMI there was little clear evidence from the results of each test series as to
which factor had the greatest influence on fatigue life. For S690 there were too few results for each
variable considered. For S700 and S960 the greatest influence was stress ratio with lower results
at R=0.5 than R=0.1. Other than stress ratio there was no other consistent effect of either of the
two variables investigated, plate thickness and HFMI condition. For the R=0.5 data the increase in
fatigue strength at 2x106 cycles over the as-welded condition was 1.7 times. For R=0.1 this
increased further to 2.3 times.

For laser re-melting it was found that a more focused spot size with less laser power
but more power density is preferable to higher power, larger focus size and less power density (i.e.
condition C2).

For LTT specimens no significant difference was observed for the two wires or plate
thickness investigated in this series of testing. The clearest distinction was that of stress ratio.
Whilst there was no significant improvement in the data at R=0.5, an increase in fatigue strength
twice that of the as-welded condition at 2x106 cycles was observed for the R=0.1 data.

A comparison of the techniques investigated suggests that at R=0.1, HFMI and LTT
give the greatest benefit in fatigue strength when compared with the as-welded condition, by a
factor of 2.36 and 2.06 respectively, with laser re-melting giving rise to an increase by a factor of
1.5. At R=0.5, HFMI again offers the greatest benefit with an increase of 1.69 times versus a factor
of 1.08 by LTT.

The assessment guidelines for HFMI produced by the IIW are supported by the data
produced. However, it has been mentioned that further data at the higher endurance end (>2x106
cycles) would be of benefit to assess the level of conservatism in the guidelines.

Variable amplitude loading:

Comparing the performance of the as-welded results against the Class F2 curve in BS7608,
HFMI offers significant improvements in increases in life compared with LTT for the 5mm thickness
S700MC and S960MC material tested. For 10mm thickness S700MC LTT offered the greatest
improvement. No comparative tests were performed by laser re-melting; however, based on the
results it could be assumed that the performance of laser re-melting under the applied VA
spectrum would be similar to that of the LTT technique.

With respect to HFMI, whilst the performance was good, it is considered that in both cases
(CAL and VAL), the performances would have been better if the improvement process had been
performed in a more controlled manner. Better control on its application may have resulted in
further improvements in life.

The assessment guidelines for HFMI produced by the IIW are supported by the data
produced.

63
3.2.5 WP4: Modelling
The work described in this section focussed on the development of engineering tools for residual
stress assessment in the fatigue design of welded structures in high strength steel applications.
This was done using finite element techniques in order to predict the residual stresses and their
effect on fatigue. Further, the research focussed on analysing local weld geometry for different
welding procedure specifications (WPS) to assure and control the weld quality. The residual stress
predictions were validated by using residual stress measurement techniques, e.g. x-ray diffraction
technique. Furthermore, the residual stress relaxation was studied during fatigue loading
experimentally and numerically, particularly for variable amplitude fatigue loading. Fracture
mechanics methods were systematically applied to evaluate the expected effect of various
geometric parameters, residual stress states and loading types on the fatigue strength of selected
weld joints.

3.2.5.1 Finite element welding simulation procedure

Welding is the most commonly used joining process in the manufacturing industry. Along with the
advantage of giving flexibility in the design of structure, welding also creates weld defects and
residual stresses in the structure. Welding residual stresses can have either detrimental or
favourable effects on the fatigue life of the structure depending upon their magnitude, distribution
and loading conditions. Often welding tensile residual stresses are detrimental and in the
magnitude of the materials yield stress and, these stresses are the main cause for the fatigue
failure when the service stresses are superimposed on the already present tensile residual stresses.
Therefore it is important to have knowledge of the formation, distribution and magnitude of the
welding residual stresses.

Finite element simulations for weld residual stresses involve in general many phenomena e.g. non-
linear temperature dependent material behaviour, micro-structural phase transformation, 3D
nature of the weld pool and the welding process which makes the welding simulations quite
cumbersome. Different simplifications can be made by excluding different factors, but still the
welding simulations are CPU time demanding and complex.

Therefore, in this work package the main aim was to develop guidelines for welding simulations,
which can effectively reduce the CPU computational time and also simplify the implementation of
these procedures in every day weld design work. In this regard an alternative approach named as
Rapid Dumping is developed. This approach is based on sequential thermo-mechanical analysis
which makes use of the moving heat source in thermal analysis and the block dumping approach in
mechanical analysis for prediction of welding residual stresses in 3D. Some of the results are
highlighted in this report. The detailed information regarding the approach can be found in [14].

Simulation Scheme

All of the welding simulations are carried out using finite element software ANSYS. Pre-processing
and post-processing has been done in ANSYS. For welding simulations by using rapid dumping
approach, an uncoupled thermo-mechanical analysis has been done i.e. firstly thermal analysis has
been carried out followed by elastic-plastic stress analysis (mechanical analysis).
For mechanical analysis, the transient temperature distributions is first stored for all time steps and
then this temperature distributions are used as load in the mechanical analysis for determination of
weld residual stresses. Same type of meshing is used in thermal and mechanical analysis. The
scheme for welding simulations is shown in Figure 69.

65
 Thermal Analysis Mechanical Analysis
Temperature field, HAZ Residual Stresses and
Heat Source And Fusion Zone Deformations

Thermal Mechanical
Material Material
Properties Properties

Figure 69: Simulation scheme in welding analysis

Material Properties

Thermal and mechanical simulations require temperature dependent material properties of the
base material and filler material. In this work, a simulation software JMat Pro has been used for the
evaluation of temperature dependent material properties for thermal and mechanical simulations.
Following material properties are validated with experiments and also compared with literature.

Thermal Analysis
Density
Thermal Conductivity
Heat Capacity
Mechanical Analysis:
Yield Stress
Modulus of Elasticity
Coefficient of Thermal Expansion and Poissons Ratio

The simulation of material properties were in good agreement with the experiments. The detailed
results are presented in [14].

Implementation of Rapid Dumping approach (small welded structures)

The proposed approach is implemented on selected weld joints i.e. longitudinal stiffener joints. In
order to further illustrate the effectiveness of the rapid dumping approach it is also implemented on
butt welded, double T butt welded and multi pass tube-flange joints. The welding residual stresses
obtained by using rapid dumping approach are compared with the experimentally measured
residual stresses and the results have shown quite good agreement with the experimental
measurements. Some of the results obtained by using rapid dumping approach are shown in below
Figure 70.

Path A Path B

Path C Path D

Figure 70: Comparison of experimental and FE (using rapid dumping) predicted residual stresses
at four different paths in the longitudinal stiffener joint.

66
The predicted residual stresses show qualitatively good agreement with the measured residual
stresses and, also, showing the similar trends, as shown in Figure 70. The residual stress prediction
along the path A and B is not following the experimental results exactly. The difference in the
numerical and experimental results can be attributed to different sources, for example, residual
stresses during the formation and cutting of the plates, thermal analysis is only validated by
matching the temperature history at three points, the mechanical properties are not accurate
especially at higher temperatures, selection of mechanical boundary conditions in numerical
simulations. Also, micro-structural evolution model is not included in the analysis. All these factors
can result in deviation of the experimental results from the numerical predictions. The peak value
of the residual stresses is very well captured at all four paths. Low peak value of residual stress is
obtained for path A and C when compared to path B and D. A possible explanation is that the
residual stresses are affected by start/stop location and also the welding path. In Figure 70, it is
believed that a higher peak value of residual stress is predicted because the start/stop location was
close to paths B and D.

Conclusions Rapid Dumping approach (small welded structures):

JMat Pro software can be used for good approximation of temperature dependent material
properties.
Proposed simplified moving heat source showed good agreement between measured and
predicted temperature histories.
An efficient sequential thermo-mechanical welding simulation approach called rapid dumping
is developed in which it is suggested to use moving heat source method in thermal analysis
and, in mechanical analysis use final cooling load step for entire weld bead instead of using
final cooling load step for individual activated block. By using rapid dumping the CPU
computational time is reduced by 90-95% as compared to gradual weld bead deposition.
The experimentally measured residual stresses showed qualitatively good agreement with
the residual stresses predicted by rapid dumping approach for double T butt welded joint.

Implementation of Rapid Dumping approach (large welded structures)

Despite the advancements in the computational capabilities of finite element (FE) simulations for
prediction of welding residual stresses on large welded structures, these are still very uncommon.
The highly nonlinear and transient nature of the welding process makes the FE simulation
computationally intensive; and a large and complex welded structure would make it even more
complex and time consuming. Moreover, an accurate representation of welding residual stress in a
structure demands three dimensional FE simulation as well as incorporation of the entire structure
surrounding the local weld zone. Thus, most of the welding simulations found in literature focused
on simple and small structures like butt welded plates, T joints etc. This motivated to implement
the rapid dumping approach on a large welded structure and for this purpose, the bogie beam
structure is analysed. Furthermore, the bogie beam structure is analysed with other available
approaches in literature in order to devise an efficient FE simulation framework for prediction of
residual stresses in large welded structures. For validation of the numerical results the
experimentally measured residual stresses using X-ray diffraction method are compared with the
numerically predicted residual stresses.

The radial residual stresses were measured along four paths B1-B4 on bogie beam as shown in
Figure 71. In order to avoid the collision of X-ray detectors with the specimen the first
measurement point was taken at a distance of 1mm from the weld toe. The distance of sub-
sequent measurement points from the weld toe are shown in Figure 71.

67
Figure 71: Different measurement positions using r-ray diffraction along path B1-B4

The detailed information regarding implementation of rapid dumping and other alternative
approaches i.e. gradual weld bead deposition, block dumping and substructuring on large welded
structures can be found in [15]. Some of the results are shown in Figure 72.

Figure 72: Comparison of experimental and FE measured residual stresses using different
approaches along four different paths B1-B4.

Qualitatively good agreement is found between experimentally measured and numerically predicted
welding residual stresses by using gradual weld bead deposition and rapid dumping approach along
all four paths (B1-B4) as shown in Figure 72. The difference in predicted and experimentally
measured residual stresses can be accredited to different factors, for example, simplified
assumptions in FE simulation, unavailable data of material properties at elevated temperatures,
residual stresses during cutting and forming of webs and flanges etc. Also, the temperature
distributions are only validated around the center sleeve of the bogie beam.

68
Moreover, the bogie beam structure has been cut at the edges in order to place it in the fixture for
measurement of residual stresses using X-ray diffraction. Cutting of bogie beam structure at the
edges could have relaxed the residual stresses. All of these factors can be considered as a possible
reason for variation in numerical and experimental results.

In Figure 72 it can be seen that the tensile residual stresses are obtained along paths B1, B2 and
B4 during x-ray diffraction measurements. Quite high magnitude of tensile residual stresses
(approximately equal to materials yield strength) is obtained near the weld toe along path B1.
Since path B1 is at weld start/stop location therefore it is believed that high temperature values at
that location resulted in high tensile residual stresses. Furthermore, compressive residual stresses
are obtained along path B3 during experimental measurements. The effects of weld sequence and
boundary conditions can be considered as possible reason for the formation of compressive residual
stresses.

Except for residual stresses along path B3 the prediction by both approaches are approximately the
same and showing similar trends. Overall, the prediction of residual stresses by gradual weld bead
deposition approach is more accurate than the rapid dumping approach since it depicts the realistic
behaviour of the welding process. The only drawback is that it is computationally very expensive.
By using rapid dumping approach the computational time is reduced by 60% during the mechanical
analysis when compared with gradual weld bead deposition approach.

The bogie beam is further analysed with substructuring technique. It is applied on bogie beam
structure with rapid dumping approach in sequential thermo-mechanical simulation. The prediction
of residual stresses by rapid dumping and substructuring is almost identical as shown in Figure 72.
The only difficulty in carrying out analysis with substructuring technique is the selection of linear
and nonlinear region. In this study a separate simulation was required to identify the interface
location between two regions. The substructuring technique using rapid dumping has further
reduced the computational time by 20% when compared to rapid dumping approach.

The residual stresses predicted by block dumping approach are completely erroneous as shown in
Figure 72. Compressive residual stresses are predicted along all four paths B1-B4 with identical
magnitude and trend. Since, the welding of bogie beam is continuous i.e. subsequent weld
sequence is started as soon as the first weld sequence is finished. Therefore, in block dumping
approach the succeeding weld bead is activated as soon as the proceeding weld bead is deposited.
This will result in global increase of temperature in the bogie beam structure consequently predicts
false stresses. Appropriate predictions using block dumping approach can be made if cooling time is
provided between the heating of two consecutive weld beads, but this will increase the
computational time considerably.

In the end the FE simulation is carried out by analysing weld sequences around the center sleeve
only. The predicted residual stresses along all four paths are conservative but showing to some
extent similar trends when compared with experiments as shown in Figure 73. Although, this result
highlights the importance of the inclusion of all the weld sequences during FE simulation of a large
structure but it is encouraged to analyse firstly the weld sequences in area of interest as a starting
point in the initial design stage if a large structure involves many welded joints.

69
Figure 73: Comparison of the experimental and FE measured residual stresses by analyzing the
welded joints around the center sleeve only.

Conclusions Rapid Dumping approach (large welded structures)

The radial residual stresses predicted by rapid dumping approach around the center sleeve
of bogie beam structure showed qualitatively good agreement with the experimental
measurements. The computational time during the mechanical analysis is reduced by 60%
when compared with gradual weld bead deposition approach. However, more accurate
results are obtained with gradual weld bead deposition approach.
Substructuring technique using rapid dumping approach is successfully implemented on
bogie beam structure. The technique has produced almost identical results and further
reduced the computational time by 20% when compared with rapid dumping approach.
Erroneous results are obtained by using block dumping approach because of the continuous
activation of weld beads. Appropriate results can be achieved if cooling time is provided
between two weld beads deposition in the FE simulation. But this will result in large
computational time.

3.2.5.2 Effect of LTT filler wires on formation of residual stresses

When high strength steels are welded, phase transformation (martensitic transformation) occurs
during the cooling phase. It is known that this phase transformation influences the evolution of
distortion and residual stresses in the welded structures. Consequently it is important to
incorporate this phase transformation in the material model for welding simulation which occurs
during the cooling cycle.
By using LTT filler wires the martensitic transformation is forced to occur near the ambient room
temperature and due to expansion during transformation a reduction in welding residual stresses
was obtained. In order to implement martensitic phase transformation in the numerical
simulations, a method specified in [16] is used. A brief introduction of that method is outlined
below. The theory and figures in this section are taken from the reference paper for a better
understanding of the method. This method is implemented on T joints because of unavailability of
material data for developed LTT filler wires.
In Figure 74 the starting martensitic transformation temperature (Ms) is 480oC and ending
martensitic transformation temperature (Mf) is 310oC for HT780 steel.

70
Figure 74: Temperature dependencies of mechanical properties

In Figure 74(a) the range of phase transformation during cooling is expressed by the dotted lines.
In this temperature range the Youngs modulus and yield strength cannot be specified. When
tensile stress was applied during the phase transformation range a remarkable decrease in the
strength and extraordinary ductility is observed. Considering this fact of transformation super-
plasticity, Youngs modulus in the range of phase transformation is idealized as shown by dotted
line in Figure 74(a). The yield strength and thermal expansion have been idealized as shown by
dotted lines in Figure 74(a).

Stress extraction Welding

points Direction
130mm

300mm
Figure 75: Dimensions and welding directions for T joint

In order to apply the aforementioned method, T joint of S700MC steel grade is considered. The
dimensions and welding directions are shown in Figure 75. In this study only the effect of
transformation expansion is taken into account. The mechanical properties of S700MC are shown in
Figure 76 and Figure 77.

Figure 76: Comparison of experimental and JMat Pro evaluated Youngs modulus and yield
stress.

71
 Figure 77: JMatPro evaluated Thermal expansion

The effect of martensitic transformation by using LTT filler wires is analysed in three different
alternatives.
Case 1: No phase transformation was taken into account.
Case 2: Martensitic transformation occurred between Ms 435C and Mf 330C
Case 3: Martensitic transformation occurred between Ms 240C and Mf 115C

The idealization of thermal expansion during the martensitic transformation range in case 2 and
case 3 is shown in Figure 78. The value of thermal expansion during the transformation was taken
from literature [16].

Figure 78: Idealization of thermal expansion during martensitic phase transformation at different
Ms and Mf temperatures

Sequential thermo-mechanical simulation was carried out using ANSYS. An assumption of

autogenous weldments i.e. base plate and filler material had the same properties. In all three
cases the longitudinal stresses were extracted from the center top side of the specimen. And also,
a node exactly at the weld toe is selected and time history of the longitudinal stresses and
temperature at that particular node is plotted as shown in Figure 79.

Figure 79: Comparison of longitudinal stress and time history variation of stress and temperature
for all three cases(Case 1: No phase transformation was taken into account; Case 2: Martensitic
transformation occurred between Ms 435C and Mf 330C; Case 3: Martensitic transformation
occurred between Ms 240C and Mf 115C)

72
In Figure 79 it can be seen that very high tensile longitudinal residual stresses are obtained at the
weld toe when no phase transformation was taken into account. Also, during the time history plot
of longitudinal residual stress by selecting a node at the weld toe it can be seen that the final state
of residual stress is tensile with a magnitude of approximately 650 MPa.
Martensitic phase transformation is assumed to take place between 450C and 330C. This is
incorporated in the numerical simulations and a reduction in the tensile longitudinal residual stress
is obtained as shown in figure 15. A further reduction in the tensile residual stress is obtained when
the Ms and Mf temperature ranges are changed i.e. 240C and 115C as shown in Figure 79.

Conclusions:

Reduction in the tensile residual stresses at the weld toe was obtained by incorporating
transformation expansion.

It is observed that Ms and Mf temperature had a significant effect on the final state of
residual stresses.

3.2.5.3 Finite element simulation of weld improvement techniques

Very limited work has been found in literature in the field of numerical simulations for
implementation of ultrasonic peening treatment on high strength steels. Most of the work found in
the literature is mostly related to shot peening or hammer peening treatment methods. Recently,
some research work has been carried out at Montan University in Leoben [17], Austria regarding
the finite element simulation for ultrasonic peening treatment methods. A brief outline of this work
mentioned in [17] is summarised below.

Subsequent to the complete thermo-mechanical simulation the weld induced residual stress
condition is converted to the finite element software Abaqus at which a quasi-static HFMI post
treatment simulation is performed. For simplification purposes the hardened pin is modelled as a
non-deformable rigid with a penetration depth at the weld toe of 0.1 mm equal to measurements
at fatigue test specimens. Between each impact a distance of 0.4 mm is selected which leads to an
overlap implying a smooth curvature of the post treated weld toe. Thermal effects due to heat
dissipation from plastic strain by friction between the contact surfaces are not considered. In
opposition to the structural weld simulation, a combined isotropic/kinematic hardening model is
applied which corresponds to the type of hardening for the investigated mild steel S355 [17].

Subsequent to the HFMI treatment simulation the elements of the lateral run-on/off parts are
removed involving stress redistribution at the edges following to the real manufacturing process.
The resulting simulated residual stresses are in good accordance to the measurements. A detailed
description and further results of the investigated numerical HFMI simulation are given in [18].

The FE simulations for implementation of ultrasonic peening treatment are very time consuming as
it is a highly nonlinear analysis, which involves contact and dynamic analysis. Therefore within our
research group it was suggested that the numerical simulations shall be carried out on a simple
plate which is assumed to be residual stress free and the residual stresses from a thermo-
mechanical welding simulation are to be mapped on that plate and the reduction in the stresses are
to be observed. The work on weld improvement simulation is still in progress, but not finalized yet
and hence not reported in details here.

3.2.5.4 Residual stress relaxation

Various factors e.g. weld defects and residual stresses affect the fatigue strength of welded joints.
The presence of residual stresses can have a beneficial or detrimental affect depending on their
nature i.e. tensile or compressive. Limited knowledge of the behaviour of residual stresses during
fatigue is available. In this part of the project the Royal Institute of Technology (KTH) has
investigated the behaviour of compressive residual stresses induced in high strength steels (HSS)
welded joints by high frequency mechanical impact treatment (HFMI) during fatigue testing. Stress
concentration factors at the weld toes for the specimens have been calculated using finite element
analysis (FEA). The specimens were tested with constant amplitude (CA) and variable amplitude
(VA) loading. Residual stresses were measured at different number of cycles during the fatigue
testing using X-ray diffraction technique (XStress robot system by Stresstech). It was observed
that the induced residual stresses were quite stable with some relaxation in constant and variable
amplitude loading.

73
Test Specimens

Dimensions of the specimens and longitudinal attachment investigated are shown in Figure 3.
These specimens were treated in different HFMI conditions (see Table 4).

Table 15: Specimen details

Fatigue testing
Specimen number Steel grade Thickness [mm] Peening method
method
1 S690QL 10 HFMI C3 CA
2 S700MC 10 HFMI C2 CA
3 S690QL 10 HFMI C1 VA
4 S700MC 10 HFMI C2 VA
5 S960QL 10 HFMI C2 VA

Experimental set up

Fatigue tests

Fatigue testing has been done using a SCHENK HYDROPULS PSB 250 machine. Both constant and
variable amplitude fatigue tests have been carried out. Shorter version of VA spectrum developed
in the project was used to test three specimens with VA loading testing. Summary of the fatigue
test set-up and results is given in the published articles [19].

Residual stress measurement parameters

Each specimen was divided into four areas, two faces and two sides on each face. Residual stresses
were measured at three points on each side in three different directions before, during and after
fatigue testing. After measurements in three different directions principal stresses are calculated
and analysed for residual stress relaxation behaviour. This is shown in Figure 80 and Figure 81.

Figure 80: Specimen distribution of measuring residual stresses

Figure 81: (a) Residual stress measurement direction (b) First measurement point and measuring
distance

74
Results: residual stress relaxation

Residual stress measurements taken before, after and during fatigue testing for specimen 1 tested
with CA loading and specimen 3 tested with VA loading is given in Figure 82 and Figure 83. Here
the scatter for every point on average is 75MPa.

Figure 82: Specimen no; 1 Residual stress measurement, Gusset Failure side1, Constant
amplitude

Figure 83: Specimen no; 3 Residual stress measurement, Weld toe Failure side1, Variable
amplitude

In order to calculate local stresses acting at the weld toe, stress concentration factors were
calculated using FEA and the corresponding weld toe radii using silicon imprints.
Table 16 shows the mean stress concentration factors of the four sides for each specimen.

Table 16: Mean stress concentration factors

Mean stress
Specimen Standard
concentration
number deviation
factor, Kt
1 2.12 0.114
2 2.02 0.064
3 1.93 0.032
4 2.20 0.238
5 2.05 0.168

Conclusions Residual Stress Relaxation

In most of the specimens investigated in this part of the project compressive residual stresses were
observed in the HFMI treated area. However, in Specimen 4, faces A and B at side 1, less
compressive residual stresses were observed [19]. The probable reason is the process time and
way of HFMI treatment. It can also be because of under or over peened surface. Compressive
residual stresses were observed to be stable for most of the fatigue life of the specimen in constant
amplitude testing. However, for the specimens tested with variable amplitude loading the residual
stresses were observed to relax with a slow rate after consuming 10-20% of their fatigue life until
failure. This relaxation may affect the overall fatigue life of the welded joints.

75
3.2.6 WP5: Guidelines

The objective is to combine the experimental and analytical results produced within this project
with other available expertise, data and published information to develop practical guidance for
enhancing the fatigue strength of high strength steel welded structures. The guideline will be
developed in cooperation with Commission XIII of the International Institute of Welding (IIW). IIW
Fatigue Recommendations are widely used internationally and provide important guidance on
design of many types of welded structures. The document will provide practical information on 1)
how to implement the most promising improvement technologies for high strength steels, 2)
assuring the quality of the improved weld, 3) design fatigue curves for improved welded joints
based on improvement method and steel grade, 4) limiting factors for improved welded joints,
e.g., joint geometry, material thickness, type of loading, environment, etc. 5) qualification
procedures and tests for new methods that become mature following the publication of the
document and 6) general information on good design practice for fatigue resistant structures.

Accomplished tasks
Activities are divided into following tasks
1. Drafting of guidance document
2. Procedures and quality assurance
3. Guidance on fatigue design
4. Verification of guidance document based on case studies

Numerous studies have observed that the fatigue strength of improved welded joints increases
with material yield strength (fy) [20], [21], [22]. Thus, the work has been performed based on the
comprehensive evaluation of published data for high frequency mechanical impact (HFMI) treated
welded joints. In total, 228 experimental results for three weld geometries subject to stress ratio of
R=0.1 axial loading have been reviewed.

A design recommendation including one fatigue class increase in strength (about 12.5%) for every
200 MPa increase in static yield strength is proposed and are shown to be conservative with
respect to all available data. Stress analysis methods based on the nominal stress (NS), the
structural hot spot stress (SHSS) and the effective notch stress (ENS) are all given. Special
considerations for low stress concentration details are also discussed. Special cautions are given for
high R-ratio or variable amplitude fatigue and potential alternate failure locations. Lessons learned
concerning appropriate HFMI procedures and quality assurance measures are presented. Due to
differences in the HFMI tools and the wide variety of potential applications, certain details of a
proper treatment procedures and quantitative quality control measures are presented generally.

3.2.6.1 Guidance on fatigue design

Published Data
In a review of published experimental data on the fatigue strength of welded joints improved by
High Frequency Mechanical Impact (HFMI) peening methods, Yldrm and Marquis [23] identified
19 publications containing fatigue data for welded steel joints improved by one of the seven
available HFMI methods. Some of these studies contained multiple materials, improvement
techniques, stress conditions or specimen types. Thus, a total of 46 data sets for four specimen
types were reviewed. Only data for axially loaded test specimens loaded at R=0.1 loading were
analysed. It has been observed that a variety of alternate failure modes are observed for HFMI
treated welded joints depending on the type of fatigue loading [24]. Wherever possible, failure
modes other than at the weld toe and run-outs have been excluded.

The yield stress of steel grades varies from 260 to 960 MPa, and specimen thickness varies from 5
to 30 mm. The 228 data points from 24 data sets analysed in the current study are summarized in
Table 31 to

Table 33 (see Appendix). These tables show the specimen type, thickness, stress concentrations,
fy, HFMI method, and number of test specimens for each data set. In cases where fy was not
specifically reported, values were taken from published datasheets [25], [26], [27].

The IIW fatigue class (FAT), i.e., the nominal stress range in MPa corresponding to 97.7% survival
probability at 2106 cycles to failure, of each specimen type is taken from the IIW
Recommendation [28] and Sm ,which is a characteristic mean value corresponding to 50%
survival probability at 2106 cycles to failure, are typical values from IIW recommendations [29].

77
Data Assessment
Several different hypotheses were investigated as a mean of establishing the empirical relationship
between yield strength and fatigue strength for HFMI treated welded joints. These included a
strength magnification factor, ky ,which increases linearly with yield strength, Eq. (1), was used
with a product form, Eq. (2) and an exponential form Eq. (3).

When assessing the data, it was assumed that 1) the slope of SN curves, m, for HFMI improved
welded joints was m=5, 2) that fatigue strength values from
Table 34 (see Appendix) are valid and 3) the best fit for the data resulted in minimum N being the
standard deviation of number of cycles in logarithm. In the current study only data for R = 0.1 was
evaluated and no statement can yet be made concerning stress ratio.

ky = ( fy - fy,o)/ fy,o (1)

SH = SA ko (1+ ky) (2)

SH = SA (ko )1/(1- ky) (3)

While all of the proposed equations resulted in a reduction in N with respect to the data with no
yield strength compensation, the best fit for the data was found using Eq. (3) with ky defined by Eq.
(1) with strength correction coefficient, , for yield after HFMI impact. In Eq. (3), ko is the stress
magnification factor for HFMI treatment for steel with a yield strength that is equivalent to the
reference yield strength (fy,0) , SA and SH are the stress ranges in the as-welded and HFMI-
treated condition, respectively.
For a single fatigue test result for a HFMI treated specimen tested at Si and with fatigue life Nf,i,
the fictitious nominal stress range for specimen i at Nf = 2106 , Si*, was computed using Eq. (4)
with m=5. In these equations Nf is the number of cycles to failure.
(4)
Si* = ((Si)m Nf,i / 2106)1/m

For evaluating test data, the mean fatigue strength from

Table 34, Sm, for a particular weld joint type was used. Based on Eq. (3), the yield strength
corrected nominal stress range for specimen i at 2106 cycles to failure, Si , is given by Eq. (5).
c

With Eq. (5) ky was defined as in Eq. (1) and fy,0 = 355 MPa.
(5)
Sic = (Sm,A)ky (Si*)1-ky

Although the yield strength correction method was initially developed for fatigue strength
assessments of HFMI-improved welded joints based on the NS approach, the method was also
suggested for local assessment approaches. Yldrm et al [20] studied an evaluation of published
HFMI-treated fillet welded joints using the local assessment methods like the SHSS [28] and the
ENS [29] methods as defined by the IIW.

Design recommendations
After evaluating all test data together based on specimen geometry, design proposals for HFMI
welded joints were made and then presented by Marquis et al [30]. The recommendation in the NS
method includes a four (4) fatigue class increase in strength for joints fabricated from fy 355 MPa
steel with respect to the nominal fatigue class in the as-welded condition. One additional fatigue
class increase in fatigue strength for every 200 MPa increase in static yield strength is proposed.
The specific fatigue class increase is defined for N=2x106 cycles and assumes an S-N slope m = 5
for HFMI treated welded joints and m = 3 for welded joints in the as-welded state. Figure 84 shows
a proposed increase in FAT class count as a function of yield strength. The solid line presents the
proposed increase and the broken line represents the increase for hammer peened and needle
peened welded joints in the current IIW guidelines [28].
Table 17 presents the existing IIW recommended FAT classes for the three joint geometries
evaluated in this study in the as-welded condition and followed by hammer or needle peening. The
table also shows the proposed FAT classes for HFMI treated joints as a function of fy. As an
example, the characteristic nominal stress S-N curves for HFMI-improved welded joints in HSS
(750 MPa < fy 950 MPa) is shown in Figure 85. For other steel grades the proposed S-N lines are
presented in [30].

78
Table 17: Existing IIW FAT classes for as-welded and hammer or needle peened welded joints
and the proposed FAT classes for HFMI treated joints as a function of fy.

fy (MPa) longitudinal welds cruciform welds butt welds

as-welded, m1 = 3
all fy 71 80 90
improved by hammer or needle peening, m1 = 3
fy 355 90 100 112
355 < fy 100 112 125
improved by HFMI, m1 = 5
235 < fy 355 112 125* 140*
355 < fy 550 125 140 160
550 < fy 750 140 160 180
750 < fy 950 160 180* -
950 < fy 180 - -
* no data available
Proposal for HFMI treated welds, m=5
8
maximum possible improvement,

7
6
# of FAT classes

5
4
IIW guideline for needle or hammer peening, m=3
3
2
1
0
0 1 2 3 4 5
235-355 355-550 550-750 750-950 over 950
fy (MPa)
Figure 84: Proposed maximum increase in the number of FAT classes as a function of fy [23].

Figure 85: Characteristic nominal stress S-N curves for HFMI-improved welded joints for high-
strength steels, 750 MPa < fy 950 (MPa) for R 0.15. The value in ( ) represents the FAT class
of the joint in the as-welded state according to Hobbacher [28].

In 2006 the IIW published fatigue design recommendations based on the use of the structural
hotspot stress which includes proposals for design S-N curves expressed in terms of the hotspot
stress range [31]. The HFMI fatigue strength enhancement factors in this current document can
also be applied to those curves. The connection between structural hotspot stress and
improvement factors for HFMI-treated joints has been studied by Yldrm et al. [32]. For HFMI-
improved welded joints the corresponding S-N characteristic curves are given in Table 18 for
different steel grades.

79
Table 18: Existing IIW FAT classes for the structural hotspot approach for as-welded and improved
joints and the proposed FAT classes for HFMI treated joints as a function of fy.

Load-carrying fillet welds Non-load-carrying fillet welds

fy (MPa) FAT FAT
as-welded, m=3
all fy 90 100
improved by hammer or needle peening, m = 3
fy 355 112 125
355 < fy 125 140
improved by HFMI, m = 5
fy 355 140 160
355 < fy 550 160 180
550 < fy 750 180 200
750 < fy 950 200 225
fy > 950 225 250

Characteristic structural hotspot stress S-N curves for non-load carrying and load-carrying HFMI-
improved welded joints are shown in Figure 86. For comparison S-N curves for as-welded [28] and
hammer-peened or needle-peened welded joints [33] are also shown. As can be seen, for the
example case of improved joints in fy = 355 MPa steel, the proposed HFMI lines are more
conservative than the hammer or needle peening lines for N < 300 000 cycles due to the different
slope.

(a) (b)
Figure 86: Characteristic structural hot-spot stress S-N curves for non-load carrying (a) and load-
carrying (b) HFMI-improved welded joints for R 0.15. For comparison S-N curves for as-welded
[24] and hammer-peened or needle-peened welded joints [26] are also shown.

analysis using an artificial notch radius = 1 mm using procedures as described by Fricke [29]. In
With respect to the effective notch method and HFMI, the design proposal is to perform the stress

theory, HFMI-treated welded joints with larger radii should perform better in fatigue. However,
there is a complex interaction between treatment parameters, toe radius, microstructure of the
treated zone and residual stresses which is not fully understood. According to Yldrm et al. [32]
the extra benefit of modelling a weld toe notch with radius equal to the actual radius of the HFMI-
treated zone is therefore difficult to define. For HFMI-improved welded joints the corresponding S-
N characteristic curves are given in Table 19 for different steel grades.

Table 19: The existing IIW characteristic curve based on the effective notch stress approach for
as-welded joints and proposed characteristic curves based on the effective notch stress approach
for HFMI-improved joints as a function of fy.

fy (MPa) Effective notch stress characteristic curve

modelled using f = 1 mm
as-welded, m =3
all fy 225
improved by HFMI, m = 5
235 < fy 355 320
355 < fy 550 360
550 < fy 750 400
750 < fy 950 450
950 < fy 500

80
Characteristic effective notch stress S-N curves is shown in Figure 87. For comparison the FAT 225
S-N curves for as-welded joints is also shown. As can be seen, for the example case of improved
joints in 355 < fy 550 MPa steel, the proposed HFMI lines are more conservative than the
hammer or needle peening lines for N < 60 000 cycles due to the different slope. The curves in
Figure 87 are based on the assumption that notch factor is computed in terms of the maximum
principal stress range in the notch. If von Mises stress is used, S-N curves must be reduced by one
fatigue class.

Figure 87: Characteristic effective notch stress S-N curves for HFMI-improved welded joints for R
0.15. For comparison the as-welded FAT 225 S-N curve is also shown.

Loading effects
In the IIW recommendations for fatigue design of welded components in the as-welded state, the
nominal normal stress range is assumed to remain below 1.5 * fy while the nominal shear stress
range must remain below 1.5 * fy / 3 [28]. In the structural hotspot stress approach, hotspot
stress range is assumed to remain below 2 * fy [31]. In practice this requirement is quite
comparable to that of the nominal stress. For welded structures improved by needle peening or
hammer peening, the guidelines state that the techniques are not suitable for R > 0.5 or when max
> 0.8 fy [33]. These restrictions are intended to influence the fatigue design assessment of
structures when the beneficial compressive residual stress state due to post-weld improvement
may not be stable. Weich et al. [34] have found that fy has a strong influence on the stability of
residual stresses for HFMI-treated welded joints. These relationships demonstrate the critical
importance of fy in the design of welded components subjected to high mean stresses, low-cycle
fatigue loading or large stress cycles as part of a variable amplitude load history.

Stress ratio:
A large research initiative has recently been completed in Germany on life extension of both new
and existing structures [35] [36]. Various aspects of HFMI treatment were studied and several
design recommendations were developed. In terms of R-ratio effects, the following relation was
proposed,

kR = 1.075 0.75R for 0.1 R 0.5

(6)
kR = 1.0 for R < 0.1

where kR is a reduction factor for stress ratio which is multiplied by the characteristic fatigue
strength determined using R = 0.1 fatigue testing. Based on Eq. (6), the R-ratio correction
relationships for needle and hammer peening in the current IIW guidelines were established.
Whereas Eq. (6) represents a continuous function, the IIW formulation was expressed in terms of
discrete fatigue classes as seen in Table 20.

Table 20: Maximum number of fatigue classes in fatigue strength improvement that can be
claimed for needed or hammer-peened welded joints [34].

R ratio Maximum fatigue class improvement

R 0.15 3
0.15 < R 0.28 2
0.28 < R 0.4 1
R > 0.4 No fatigue improvement can be claimed
without verification testing

81
In contrast to needle- or hammer-peened welded joints where the maximum FAT improvement is
three (3), HFMI-treated welded joints can have up to 8 FAT classes of improvement depending on
the material strength, welded joint geometry, etc. It is therefore easier to express the stress ratio
influence as a penalty with respect to the maximum increase in the number of FAT classes as a
function of fy as presented in Figure 84. These are given in Table 21.

Table 21: Minimum reduction in the number of FAT classes in fatigue strength improvement for
HFMI-treated welded joints as presented in Fig. 1 based on ratio R.

R ratio Minimum FAT class reduction

R 0.15 No reduction due to stress ratio.
0.15 < R 0.28 One FAT class reduction
0.28 < R 0.4 Two FAT classes reduction
0.4 < R 0.52 Three FAT classes reduction
0.52 < R No data available. The degree of improvement must be
confirmed by testing

While it has not been studied systematically, the previously mentioned limitation that max 0.8fy
should apply also to HFMI-treated welded joints. This information is summarized in Figure 88 which
shows the limitation on maximum constant amplitude stress range, , as a function of R for
several values of fy. In the case R < -0.125 loading, the horizontal lines correspond to = 0.9fy.
If exceeds the stress range values in this figure, benefit from HFMI treatment cannot be claimed
without supporting tests. This table illustrates the benefit of HFMI especially for the higher-yield-
strength materials. For low-fy materials with high R, HFMI can improve the so-called fatigue limit
but it is not expected to improve fatigue strength in the finite life regime. A computation example
involving high-R-ratio fatigue loading is given in the guideline [30].

Figure 88: Limitation on maximum constant amplitude stress range, , that can be applied to a
weld in order to claim benefit from HFMI treatment (in MPa).

Variable amplitude loading:

In many design situations it is convenient to transform the variable amplitude history into an
equivalent constant amplitude value based on rainflow counting and Eq. (7) [37].

1


(7)

In this equation k is the stress range associated with the knee computed at N=1107, Ni is the
number of cycles of stress i where i > k, Nj is the number of cycles of stress j where i
< k, m is the slope of S-N line above the knee point, m = 2m -1 is the slope of the S-N line
below the knee point and D is the damage sum, e.g., D = 0.5. As previously discussed m = 5 and
m = 9 are proposed for HFMI treated welded joints. Yldrm and Marquis [38] showed that Eq. (3)
can be used to correlated variable amplitude and constant amplitude fatigue data for welded joints
improved using HFMI.

82
It has frequently been observed that during variable amplitude fatigue loading, the largest stress
cycles in the load spectrum are computed to have only a small contribution to the computed
damage sum. However, the significance of the large cycles cannot be underestimated because they
strongly influence the damage produced by the smaller cycles [39] [40]. This increased damage
produced by the small cycles has led to IIW to adopt a damage sum D = 0.5 for variable amplitude
loading.
However, it has been noted that for load histories with dramatically changing mean stresses
between cycles, the damage sum may be as low as D = 0.2 [28] [41] [42]. In the case of HFMI
treated welded joints, a significant portion of the fatigue strength improvement is attributed to the
presence of compressive residual stresses. Therefore, any change in the residual stress state which
might occur during variable amplitude loading could have significant influence on the resulting
fatigue strength. For example, it has been observed that, for HFMI-treated high-strength steel
welded joints, the failure mode changes depending on whether the fatigue loading has a constant
or variable amplitude [43]. Variable amplitude loading using a load history which contained some
large stress range cycles led to failure in the HFMI groove while constant amplitude loading
resulted in other failure locations.
The information contained in Figure 88 can also be used to assess whether the fatigue design
guidelines outlined in this proposal can be used for an HFMI-treated welded joint subjected to
variable amplitude loading. The stress amplitude and R ratio in the loading spectrum can be
compared with Figure 88. If the largest cycle exceeds the limit value for a given fy, then the benefit
of HFMI cannot be claimed without fatigue testing. A computation example involving variable
amplitude loading is given in the guidelines [30].

Special Considerations for Low Stress Concentration Joints:

In the context of the nominal stress based approach to fatigue design of HFMI-treated welded
joints, it has previously been noted that the highest allowable S-N curve corresponds to FAT 180 m
= 5. When the structural hot-spot stress method is used a computation problem may arise for
welded details with relatively small structural stress concentrations, i.e., the hotspot stress system
must be limited so as not to result in a S-N curve greater than FAT 180 in the nominal stress
system.
By defining the structural hotspot stress concentration as the ratio of structural stress to nominal
stress

K

(8)

the minimum KS for HSS should be limited according to the values given in Table 22.

Table 22: Minimum structural hotspot stress concentrations for HFMI-improved welded
connections are also shown.

fy (MPa) Load-carrying fillet welds Non-load carrying fillet welds

fy 550 no restriction no restriction
550 < fy 750 no restriction KS,min = 1.15
750 < fy 950 KS,min = 1.15 KS,min = 1.25
fy > 950 KS,min = 1.25 KS,min = 1.40

When the effective notch stress method is used a computation problem may arise for welded
details with relatively small structural stress concentrations, i.e., the effective notch stress system
must be limited so as not to achieve an S-N curve greater than FAT 180 in the nominal stress
system. Fricke [29] defines the effective notch stress concentration as the ratio of effective notch
stress to structural stress.

K

(9)

Fricke proposes that a minimum Kw of at least 1.6 should be used for low stress concentration
details. This means that the structural hotspot stress concentration, KS, and the effective notch

element calculations with = 1 mm is less than 1.6, then a minimum value of Kw, min = 1.6 should
stress concentration, Kw, at a weld toe need to be checked. If the computed Kw based on finite

be used for fatigue assessment. No additional requirements are needed for HFMI-treated welded
joints because Kw is defined with respect to S. It is sufficient to guarantee that the KS,min for HSS
should be limited according to the values given in Table 22.

83
3.2.6.2 Quality Assurance Guidelines
Improvement techniques described defined by the IIW are intended to be used both for increasing
the fatigue strength of new structures and repair or upgrading of existing structures. The IIW has
consistently emphasized that, especially with respect to new structures, weld improvement
techniques should never be implemented to compensate for poor design or bad fabrication
practices. Instead, improvement measures should be implemented as a means of providing
additional strength after other measures have been taken. Because HFMI is normally specified as a
fatigue strength improvement technology for new structures or during repair and retrofitting
operations, it is always essential to consult fatigue experts to ensure that all critical regions in a
structure identified and properly treated. Most fatigue loaded structures will normally have only a
limited number of locations that are critical from a fatigue point of view. Proper identification of
these regions is also important to avoid extra costs and treatment of regions which are not fatigue
critical. Additionally, the possibility of a failure starting at some other location must always be
considered. For instance, if the failure origin is merely shifted from the weld toe to the root there
may be no significant improvement in fatigue life.

Prior to HFMI treatment the weld cap and adjacent parent material shall be fully de-slagged and
wire brushed or ground to remove all traces of oxide, scale, spatter and other foreign material.
HFMI treatment of a convex weld profile or of a weld with a large weld angle can cause the
plastically deformed metal to fold over the original weld toe and leave a crack-like lap feature that
resembles a cold lap [44]. The weld bead profile should meet the acceptance limits for weld profile
quality level B in ISO 5817 [45]. This requirement does not imply that the weld must fulfil all
quality level B criteria in ISO 5817.
However, weld profile-related quality criteria in ISO 5817 need to be evaluated. These include
Undercuts (imperfection 1.7), Excessive overfill (imperfection 1.19), Excessive concavity
(imperfection 1.10) and Overlaps (imperfection 1.13). If the weld profile does not comply with
these acceptance limits, light grinding before treatment may be desired. It should be noted,
however, that HFMI treatment is most effective when the weld toe region itself is treated. Thus,
grinding operations which make it difficult for the HFMI operator to distinguish the exact location of
the weld toe should be avoided. Decisions on the need for weld grinding and the proper grinding
procedure should be agreed on with an experienced HFMI operator.

The need for proper weld profile before HFMI is illustrated in Figure 89 a) which illustrates the
formation of a crack-like defect due to improper contact between the indenter and weld toe.
Surface inspection of such a defect reveals a dark crack-like line in the middle of the otherwise
smooth and shiny HFMI groove as seen in Figure 89 b). Figure 89 c) shows section micrographs of
these defects. The resulting fatigue performance of a welded joint with such defects may actually
be less than that of the original as-welded joint. The same type of flaw has been observed in
welded joints with adequate profiles but with improper indenter se-lection or too severe treatment,
i.e. too many passes over the same region. For specific applications it may be needed to consult
with the HFMI tool manufacturer in order to select the proper treatment procedures and optimal
indenter configuration to avoid crack-like defects.

The depth of the HFMI groove is an excellent indicator of the extent of HFMI treatment. Depending
on the yield strength of the steel and the size of the indenters, typically the optimum HFMI groove
will be 0.2 - 0.6 mm in depth and the width will be 3 - 6 mm [46] [47] [48], see Figure 90.
However, it should be noted that no single groove dimension is optimal in all situations. A welded
structure with relatively deep undercuts at the weld toe may require light grinding of the weld toe
before HFMI and will have a deeper groove following HFMI.
Also, HFMI grooves in high strength steel structures will typically be shallower and narrower than
grooves in low strength steel. Groove depth can be checked relatively easily by using simple depth
gauges such as is shown in Figure 91. Callipers can be used to measure the width of the groove.
The centre of the HFMI groove should correspond to the fusion line of the weld. The portion of the
HFMI groove in the weld metal must be between 25% and 75% of the total HFMI groove width
[47].

84
a)

b)

shiny defect-free defect

HFMI groove

c)

weld metal base metal

defect

weld
HAZ metal

defect base metal

Figure 89: a) Potential introduction of crack-like defect due to HFMI treatment of a weld with a
steep angle or with too large of an indenter and b) resulting groove for a properly treated (left) and
improperly treated weld toe (right) and c) micrographs of the induced crack-like defects due to
improper HFMI treatment [49].

85
Figure 90: The HFMI indentation depth following treatment should be 0.2-0.6 mm while the
resulting width is typically 2-5 mm.

gap

Figure 91: Depth inspection using simple gauges [50]. The gap between the base plate and the
gauge indicates that 0.2 mm has not been achieved.

A HFMI Procedure Specification (HFMI-PS) similar to a Welding Procedure Specification (WPS)

should be prepared for the HFMI treatment. The HFMI-PS includes information concerning the
component being treated; base and filler material; HFMI equipment type and power settings;
number, size and shape of the indenters to be used; special inspection requirements including
HFMI groove dimension, etc. Lopez Martinez and Haagensen have developed a HFMI-PS template
which is prepared for each weld in a structure [51] [31].

86
3.3 Conclusions
The programme of work undertaken in this project has led to the following conclusions:

Welding of fatigue test specimens (WP1a) has been carried out in various steel grade/thickness
combinations, including parameters selection, ensuring optimal weld quality, testing and welding.
In total, over 520 welded specimens were produced.

Two demonstrators (dipper arm and bogie beam) were selected and re-designed to reduce weight
using high strength steels (WP1b). For the dipper arm, the weight was reduced by 18%; for the
bogie beam a weight reduction of 24% was achieved. The fatigue life of both components was
estimated based on the guidance document for HFMI treated welded joints from IIW Commission
XIII.
Several prototypes of dipper arm and bogie beam were manufactured in high strength steel, with
critical weld toes treated by means of Pneumatic Impact Treatment (PIT).
Rig testing of several prototypes was done and a significant increase in fatigue life was obtained:
the re-designed bogie beam shows a fatigue life of about 3 times higher compared to the reference
case. Finally, the economic potential of the optimised components was calculated which showed
that is possible to obtain a cost reduction on both direct product cost and product life cycle cost.

The application of Ultrasonic Impact Treatment (UIT) was carried out by Applied Ultrasonics
employing different treatment conditions (Task 2.1). A total of 170 specimens was HFMI treated,
using 4 sets of parameters.

Different diode laser treatment conditions were tested, using diode laser power, spot size and
shape, and the travel speed as main parameters. The work performed has allowed to show that it
is possible to perform diode laser re-melting of weld toes successfully, using different focused spot
shapes and dimensions and parameters sets. However, it became apparent that in order to create
a smooth profile, a certain heat input / heat concentration is required, favouring smaller spot sizes.

Concerning Low Temperature Transformation (LTT) filler wire development (task 2.3), two alloys
were selected for further fatigue testing: LTT-S (Cr-Ni-Fe based) and LTT-C (Cr-Mn-Fe based).
Both alloys best fulfil toughness requirements and have a good hot and cold cracking resistance, as
well as acceptable Ms temperatures and appropriate strength levels.
Regarding the ability of the LTT filler wires to control the residual stresses in specific fatigue
samples with T joints, it can be stated that significantly lowered residual stresses adjacent to the
weld toe may be formed compared to a conventional high strength filler wire. Significant stress
relief due to the martensite formation during cooling was proven by in-situ reaction force
measurements. This is particularly the case when a relatively high interpass temperature (180C)
is applied. Furthermore, the positive effect of LTT filler wires on fatigue behaviour was confirmed in
the fatigue test programme developed in the frame of this project: the benefit in fatigue strength
for Constant Amplitude Loading (CAL) ranges from a factor of 1.08 (R=0.5) to 2.06 (R= 0.1).

Measurement of residual stresses was done (task 2.4) on specimens in HFMI treated and laser re-
dressed condition and LTT welded using X-ray diffraction (XRD) technique.
For HFMI treated specimens, surface residual stress measurements revealed that high compressive
stresses are present in the treated zone near the weld toe. Concerning relaxation of residual
stresses in HFMI treated welded joints, it was concluded that compressive residual stresses induced
by HFMI were quite stable for a number of cycles with some rate of relaxation in constant
amplitude fatigue tests. For laser re-dressed specimens it was shown that there is a decrease in
tensile residual stresses as compared to the as-welded condition. Surface measurements on LTT
welded specimens showed only a slight decrease in tensile residual stresses compared to the as-
welded condition.
Within this project, an electrochemical polishing device has been developed for residual stress
depth distribution. Measurements on HFMI treated specimens showed that residual stresses are
compressive to a depth of about 1 mm.
In situ monitoring of stresses using Barkhausen Noise (BN) technique was completed on HFMI
treated samples. Although there was quite a large scatter in the BN results, it is believed that the
method shows promising possibilities for future monitoring of critical welded joints.

Fatigue testing was conducted under both constant amplitude cyclic loading (CAL) and variable
amplitude spectrum loading (VAL), with the objective to establish the levels of improvement in
lives of the welded joints.

87
The main conclusions drawn from this programme of work are as follows:

Constant Amplitude Loading:

- Under HFMI, other than stress ratio there was no other consistent effect of either of the
two variable investigated, plate thickness and HFMI condition.
- For laser re-melting it was found that a more focused spot size with less laser power but
more power density is preferable to higher power, larger focus size and less power density.
- For LTT specimens no significant difference was observed for the two wires or plate
thickness investigated in this series of testing.
- A comparison of the techniques investigated suggests that at R=0.1, HFMI and LTT give
the greatest benefit in fatigue strength when compared with the as-welded condition, by a
factor of 2.36 and 2.06 respectively, with laser re-melting giving rise to an increase by a
factor of 1.5. At R=0.5, HFMI again offers the greatest benefit with an increase of 1.69
times versus a factor of 1.08 by LTT.

Variable Amplitude Loading:

- Comparing the performance of the as-welded results against the Class F2 curve in BS7608,
HFMI offers significant improvements in increases in life compared with LTT for the 5mm
thickness S700MC and S960MC material tested. For 10mm thickness S700MC LTT offered
the greatest improvement. No comparative tests were performed by laser re-melting;
however, based on the results it could be assumed that the performance of laser re-melting
under the applied VA spectrum would be similar to that of the LTT technique.

The assessment guidelines for HFMI produced by the IIW are supported by the data produced.

In the modelling work package included in this project, KTH has developed an efficient simulation
approach for the prediction of welding residual stresses. This approach was validated on a large
and complex welded structure (i.e. bogie beam, one of the demonstrated selected by VCE).
Qualitatively good agreement was found between experiments and numerically predicted welding
residual stresses.
The work done on residual stress relaxation revealed that compressive residual stresses were
stable for most of the fatigue life of the specimen in constant amplitude testing. However, for the
specimens tested in variable amplitude loading the residual stresses were observed to relax with a
slow rate after consuming 10-20% of their fatigue life until failure.
Additionally, the effect of LTT filler wires on the formation of residual stresses was analysed using
FEA approach. It was possible to predict a reduction in tensile residual stresses at the weld toe by
incorporating expansion during phase transformation. Further investigations are ongoing to refine
the modelling tools to predict the behaviour of the welded joint under different loading conditions.

The objective of guideline development (WP5) has been to combine the experimental and analytical
results produced within this project with other available expertise, data and published information
to develop practical guidance for enhancing the fatigue strength of high strength steel welded
structures. The guidelines were developed in cooperation with Commission XIII of the International
Institute of Welding (IIW).
A guidance document on Improving the fatigue life of high strength steel welded structures by
post weld treatments has been developed within this project. The draft guidance document is
considered to apply to steel plate thickness 5 to 50 mm and for 235 MPa fy 960 MPa.
The design proposal includes proposals for the 1) effect of material strength, 2) special
requirements for low stress concentration weld details, 3) high R-ratio loading conditions and 4)
variable amplitude loading. A yield strength correction method was proposed and verified. A design
recommendation including one fatigue class increase in strength (about 12.5%) for every 200 MPa
increase in static yield strength was proposed and shown to be conservative with respect to all
available data.
A proposal for procedures and quality assurance for HFMI treated welded joints in steel has also
been presented. It was developed based on discussions, presentations and experimental evidence
published within Commission XIII of the IIW.

Dissemination of the obtained project results was made through several scientific papers.
Furthermore, the achievements made were presented during the Fatigue Design 2013 conference,
held in Senlis (France), as well as during several IIW meetings.

88
3.4 Exploitation and impact of the research results

With the results obtained in this project, many economic sectors of industry (more specifically
European steel industry and their customers) will benefit from the know-how built in terms of
experimental data, modelling tools and guideline development to improve fatigue life of HSS
welded structures. Mainly for load carrying welded structures in several types of heavy equipment
(construction machinery, agricultural machinery, etc) the main benefit will be in the critical zones
of these constructions with an import impact on durability and life cycle cost.

More specifically, the knowledge and experience gathered within the project will be implemented
by the end user involved (VCE) in future projects. The analysis and re-design procedure, as well as
the use of HSS and HFMI treatment for welded joints will be considered for new products. For
instance, the optimisation result obtained for the bogie beam is considered as a valuable solution
for the next generation products.

LTT welding consumables offer the potential to achieve improved fatigue strength without time and
cost extensive post weld treatments. The results can thus be utilised in a variety of steel
processing industries where fatigue stressed components are made of high strength steel.
Additionally, LTT fillers open up the possibility to produce favourable residual stresses even in
areas of a welded joint which are difficult to access. Moreover, volume-like compressive residual
stress fields in the weld metal and HAZ may be produced contrary to conventional treatment
methods which are mainly limited to surface areas. The freedom of design is increased as the
accessibility for weld treatment methods may be neglected during manufacturing. The adjustment
of the residual stresses (level and distribution) is highly reproducible compared to strongly user-
dependent (manual) weld treatment methods. This allows the economical and efficient use of high-
strength steels. The results can thus be utilised in a variety of steel processing industries where
components subjected to fatigue loading, are made of HSS.

Dissemination of the project results will continue through the partners (steel makers, welding
consumable manufacturer, ...) involved in this project. The obtained achievements are currently
being used in presentations to customers. Furthermore, several partners are IIW members, so
project results and developed guidelines will be widely used and transferred to standardisation
bodies.

89
4 List of References

[1] FATWELDHSS, Mid-term report - 2012.

[2] FATWELDHSS, Technical report n3 - 2013.

[3] EN ISO 15614-1, Specification and qualification of welding procedures for metallic materials -
Welding procedure test - Part 1 Arc and gas welding of steels.

[4] EN ISO 15609-1, Specification and qualification of welding procedures for metallic materials -
Welding procedure specification - Part 1 Arc welding.

[5] Gerritsen C., Vanrostenberghe S., Dor M., Laser weld toe re-melting as a means of fatigue
strength improvement, Procedia engineering, vol. 66, pp. 171-180, 2013.

[6] Kromm A., Kannengiesser T., van der Mee V., FATWELDHSS - WP2.3 LTT filler wire
development - research report.

[7] A. Hobbacher, Recommendations for fatigue design of welded joints and components, IIW
document no. XIII-2151-07, December, 2008.

[8] BS7609 - Fatigue design and assessment of steel structures, British Standards Institute,
London, 1993.

[9] Dor M., Maddox S., FATWELDHSS - WP3 Fatigue Testing - research report, 2014.

[10] Marquis G., Mikkola E., Yildirim H. and Barsoum Z., Fatigue strength improvement of steel
structures by HFMI: Proposed fatigue assessment guidelines, International Institute of Welding,
Commission XIII, 2013.

[11] G. Booth, Improving the fatigue strength of welded joints by grinding - techniques and benefits,
Metal Construct, vol. 18, nr. 7, pp. 432-437, 1986.

[12] Haagensen P. and Maddox S., IIW recommendations on post weld improvement of steel and
aluminium structures, IIW document no. XIII-2200r4-07, International Institute of Welding, 2010.

[13] Marquis G., Bjork T., Koskimaki M. and Kuoppala J., Fatigue strength of improved high strength
welds subject to variable amplitude loading, Second international conference on Material and
component performance under variable amplitude loading, Proceedings volume II, pp. 1043-1051,
2009.

[14] Bhatti A., Barsoum Z., Development of efficient 3D welding simulation approach for residual
stress estimation in different welded joints, Journal of Strain Analysis for Engineering Design, vol.
47, nr. 8, pp. 539 - 552, 2012.

[15] Bhatti A., Barsoum Z., Khurshid M., Development of FE simulation framework for prediction of
residual stresses in large welded structures, Computers & Structures, vol. 133, pp. 1 - 11, 2014.

[16] Kim You-Chul et al, Modeling of Laser-Arc Hybrid Welding considered phase transformation, IIW
doc XV-1380-11, International Institute of Welding.

[17] M. Leitner, Local fatigue assessment of welded and high frequency mechanical impact treated
joints, PhD Thesis, Chair of Mechanical Engineering, Montan universitt Leoben, Austria, 2013.

[18] D. Simunek, Implementierung der Scheissnahtnachbehandlung mittels einem hochfrequenten

Hmmerverfahren in die numerische Lebensdauerabschtzung, Diploma thesis, Montan
universitt, Leoben, 2013.

90
[19] Khurshid M., Barsoum Z., Marquis G., Behaviour of Compressive Residual Stresses in High
Strength Steel Welds induced by High Frequency Mechanical Impact Treatment, in ASME
Pressure Vessel and Piping Conference, Paris, 2013.

[20] S. Maddox, Key Developments in the fatigue design of welded constructions, IIW Portvin lecture,
International Institute of Welding, 2003.

[21] A. Bignonnet, Improving the fatigue strength of welded steel structures, in 3rd International
ECSC Offshore Conference, 1987.

[22] P. Haagensen, State of the art and guidelines for improved high strength steel welds, in Intl
Symp on Integrated Design and Manufacturing of Welded Structures, Eskilstuna, Sweden, 2007.

[23] Yildirim H., Marquis G., Fatigue strength improvement factors for high strength steel welded
joints treated by high frequency mechanical impact, International Journal of Fatigue, vol. 44, pp.
168-176, 2012.

[24] G. Marquis, Failure modes and fatigue strength of improved HSS welds, Engineering Fracture
Mechanics, vol. 77, nr. 11, pp. 2051-2062, 2010.

[25] SSAB: Domex 420 MC D Hot roled, high strength, cold forming steel, DATASHEET: 11-02-03
B8415 DOMEX.

[26] SSAB: Domex 700 MC Hot rolled, extra high strength, cold forming steel., DATA SHEET: 11-02-
03 GB8421 DOMEX.

[27] SSAB: Domex 960 Structural strip steel sheet., DATA SHEET: 11-02-16 GB8435 DOMEX.

[28] A. Hobbacher, IIW Recommendations for Fatigue Design of Welded Joints and Components,
WRC Bulletin 520, The Welding Research Council, New York, 2009.

[29] W. Fricke, IIW Recommendations for the fatigue assessment by notch stress analysis for welded
structures, IIW Doc no XIII-2240r2-08, Paris, 2010.

[30] Marquis G., Mikkola E., Yildirim H., Barsoum Z., Fatigue strength improvement of steel structures
by high frequency mechanical impact: proposed fatigue assessment guidelines, Welding in the
World, vol. 57, nr. 6, pp. 803-822, 2013.

[31] Niemi E., Fricke W., Maddox S., Fatigue analysis of welded joints - Designer's guide to the
structural hot-spot stress approach, Cambridge: Woodhead, 2006.

[32] Yildirim H., Marquis G., Barsoum Z., Fatigue assessment of high frequency mechanical impact
(HFMI)-improved fillet welds by local approaches, International Journal of Fatigue, vol. 52, pp.
57-67, 2013.

[33] Haagensen P., Maddox, IIW Recommendations on Post Weld Fatigue Life Improvement of Steel
and Aluminium Structures, Cambridge: Woodhead , 2001.

[34] Weich I. et al, Fatigue behaviour of welded high strength steels after high frequency mechanical
postweld treatments, Welding in the World, vol. 53, nr. 11/12, 2009.

[35] REFRESH-Extension of the fatigue life of existing and new welded steel structures, FOSTA-
Research Association for Steel Applications, Research Projects P 702, Verlag und
Vertriebsgesellschaft GmbH, Dusseldorf, 2011.

[36] Weidner P., Weich I., Ummenhofer T., High frequency hammer peening of LCF-stressed ultra-
high strength steels, IIW Document XIII-2341-10, International Institute of Welding, Paris.

91
[37] E. Niemi, Random loading behavior of welded components, in IIW International Conference on
Performance of Dynamically Loaded Welded Structures, San Francisco, 1997.

[38] Yildirim H., Marquis G., A Round Robin study of high frequency mechanical impact treated
welded joints subjected to variable amplitude loading, Welding in the World, vol. 57, nr. 3, pp.
437-447, 2013.

[39] M. G., Long-life spectrum fatigue of carbon and stainless steel welds, Fatigue and Fracture of
Engineering Materials and Structures, vol. 19, pp. 739-753, 1996.

[40] C. Sonsino, Effect of residual stresses on the fatigue behavior of welded joints depending on
loading conditions and weld geometry, International Journal of Fatigue, vol. 31, pp. 88-101, 2009.

[41] C. Sonsino, Fatigue Testing under Variable Amplitude Loading, International Journal of Fatigue,
vol. 29, pp. 1080-1089, 2007.

[42] G. Marquis, Fracture and fatigue of Welded Joints and Structures (Chapter 8), Cambridge:
Woodhead, 2011.

[43] G. Marquis, Failure modes and fatigue strength of improved HSS welds, Engineering Fracture
Mechanics, vol. 77, nr. 11, pp. 2051-2062, 2010.

[44] Marquis G., Barsoum Z., Fatigue strength improvement of steel structures by high-frequency
mechanical impact: proposed procedures and quality assurance guidelines, Welding in the
World, vol. 58, nr. 1, pp. 19-28, 2014.

[45] ISO 5817:2006; Welding Fusion-welded joints in steel, nickel, titanium and their alloys (beam
welding excluded) Quality levels for imperfections, 2006.

[46] AASHTO. American association of state highway and transportation officials: LFRD bridge design
specifications. (4th ed.), Washington, D.C., 2008.

[47] Tilly G., Jackson P., Maddox S., Henderson R., Proceedings of the ICE - Bridge Engineering (pp.
147-152), 2010.

[48] Kudryavtsev, Y. et al, Fatigue life improvement of welded elements by ultrasonic peening, IIW
doc no XIII-2010-04, International Institute of Welding, 2004.

[49] Statnikov E., Korostel V., Vekshin N., Marquis G., Development of Esonix ultrasonic impact
treatment techniques, IIW doc no XIII-2098-06, International Institute of Welding, 2006.

[50] M. Neher, HiFIT Presentation to the IIW Commission XIII intermediate meeting, Espoo, Finland,
2012.

[51] Lopez Martinez L., Haagensen P., Life extension of Class F and Class F2 details using ultrasonic
peening, IIW Document no XIII-2143-06, International Institute of Welding, 2006.

[52] Jonsson B., Shin Y., Implementing High Frequency Mechanical Impact in industrial components:
A case study, Proceedia Engineering, vol. 66, pp. 202-215, 2013.

[53] Bhatti A., Barsoum Z., van der Mee V., Kromm A., Kannengiesser , Fatigue strength
improvement of welded structures using new low transformation temperature filler materials,
Proceedia engineering, vol. 66, pp. 192 - 201, 2013.

[54] Suominen L., Khurshid M., Parantainen J., Residual stresses in welded components following
post-weld treatment methods, Proceedia engineering, vol. 66, pp. 181 - 191, 2013.

[55] FATWELDHSS, Technical report n2 - 2011.

92
[56] Marquis G. and Barsoum Z, Fatigue strength improvement of steel structures by HFMI: Proposed
procedures and quality assurance guidelines, International Institute of Welding, Commission XIII,
2013.

[57] van Hoecke D., De Geest A., FATWELDHSS - WP2.5 Material characterisation - test report,
2012.

[58] B. Droesbeke, FATWELDHSS - WP1A Welding Procedure Specifications - Data sheets.

[59] Y. Shin, FATWELDHSS - WP1B Case Studies - research report.

[60] Bhatti A., Barsoum Z., Khurshid M., FATWELDHSS - WP4.2 Predicting residual stress using LTT
filler wires - research report, 2013.

[61] Yildirim H., Marquis G., FATWELDHSS - WP5.4 verification of guidance documents -
presentation, 2013.

93
5 Appendices
Table 23: Results of regression analysis of test results from as-welded specimens tested under
CAL

Standard Fatigue strength @ Fatigue strength

Free deviation 2M cycles for free @ 2M cycles for
Material slope of logN slope (MPa) slope of 3 (MPa)
S700MC, R=0.1, 5mm 2.92 0.083 66 67

S700MC, R=0.1, 10mm 2.74 0.047 69 74

S690QL, R=0.1, 10mm 3.20 0.081 104 99

S690QL, R=0.1, 20mm 2.74 0.111 81 86

S960MC, R=0.1, 5mm 2.87 0.030 68 71

S960QL, R=0.1, 10mm 2.64 0.040 77 86

S960QL, R=0.1, 15mm 2.66 0.048 74 81

S700MC, R=0.5, 5mm 2.89 0.019 73 75

S700MC, R=0.5, 10mm 2.91 0.056 71 73

S690QL, R=0.5, 20mm 3.00 0.075 87 87

S960MC, R=0.5, 5mm 2.80 0.030 62 66

S960QL, R=0.5, 15mm 2.85 0.042 80 84

RR, R=0.1, 5mm 2.64 0.089 60 67

RR, R=0.5, 5mm 2.88 0.030 64 65

All 2.78 0.157 74 78

Table 24: Results of regression analysis of results obtained from laser re-melted specimens under
CAL

Standard Fatigue strength @ 2M Fatigue strength @ 2M

Free deviation of cycles for free slope cycles for slope of 3
Material slope logN (MPa) (MPa)
S690QL C1 2.77 0.097 87 85
S690QL C2 2.99 0.548 146 146
S690QL C3 3.60 0.051 126 115
S690QL C4 3.68 0.179 137 123
S690QL C5 3.63 - 142 120
S960QL C1 2.59 0.075 75 85
S960QL C2 3.43 0.222 134 124
S960QL C3 3.43 0.136 122 113
S960QL C4 4.30 0.146 131 105
S960QL C5 2.72 0.204 96 106
All 2.85 0.279 106 109

94
Table 25: Results of regression analysis of data obtained from HFMI treated specimens under CAL

Standard Fatigue strength Fatigue strength Fatigue strength

Free deviation @ 2M cycles for @ 2M cycles for @ 2M cycles for
Material slope of logN free slope (MPa) slope of 3 (MPa) slope of 5 (MPa)
S700MC, R=0.1
6.34 0.350 155 88 135
5mm C1
S700MC R=0.1
3.60 0.638 178 166 196
10mm C1
S690QL R=0.1
2.21 0.167 113 150 205
10mm C2
S960MC R=0.1 5mm
4.17 0.194 158 135 170
C1
S690QL R=0.1
3.39 0.296 182 167 224
20mm C4
S960QL R=0.1
5.60 0.370 205 159 198
10mm C4
S960MC R=0.5 5mm
3.71 0.250 103 90 121
C1
S960QL R=0.5
3.75 0.156 121 106 137
10mm C1
S960QL R=0.5
4.56 0.098 133 92 141
15mm C1
S700MC R=0.5 5mm
5.70 0.257 121 74 112
C2
S960QL R=0.5
3.99 0.034 122 100 138
10mm C2
S960QL R=0.5
6.73 0.513 174 113 155
15mm C2
PIT v UIT S700MC
8.23 0.079 240 152 203
R=0.1 10mm
All 2.96 0.474 113 114 155

All S690, R=0.1 1.63 0.238 94 - 203

All S690, R=0.5 3.57 0.157 109 - 125

All S700, R=0.1 5.04 0.518 181 - 180

All S700, R=0.5 5.57 0.191 118 - 111

All S960, R=0.1 4.49 0.325 176 - 183

All S960, R=0.5 3.91 0.317 120 - 137

All steels, R=0.1 3.72 0.410 159 - 184

All steels, R=0.5 4.23 0.318 120 - 132

95
Table 26: Results of regression analysis of fatigue data obtained from LTT treated specimens
under CAL

Standard Fatigue strength Fatigue strength Fatigue strength

Free deviation @ 2M cycles for @ 2M cycles for @ 2M cycles for
Material slope of logN free slope (MPa) slope of 3 (MPa) slope of 5 (MPa)
S960QL LTT C
3.48 0.397 169 158 192
R=0.1 15mm
S960QL LTT C
2.75 0.057 88 96 135
R=0.5 10mm
S960QL LTT S
6.75 0.257 188 143 174
R=0.1 10mm
S960QL LTT S
3.95 0.104 114 91 132
R=0.5 15mm
S700MC LTT C
3.22 0.172 84 79 115
R=0.5 5mm
S700MC LTT C
6.23 0.473 171 127 159
R=0.1 5mm
S690QL LTT C
3.97 0.010 146 126 162
R=0.1 10mm
S690QL LTT C
4.13 0.085 113 84 132
R=0.5 10mm
S690QL LTT S
5.86 0.059 177 118 165
R=0.1 10mm
S690QL LTT S
3.31 0.416 94 87 127
R=0.5 10mm
S960QL LTT C
4.60 0.145 151 122 156
R=0.1 10mm
S960QL LTT C
3.28 0.014 89 82 123
R=0.5 15mm
S960QL LTT S
4.10 0.055 139 114 153
R=0.1 15mm
S960QL LTT S
3.10 0.035 79 77 113
R=0.5 10mm
S960MC LTT S
3.73 0.019 108 87 135
R=0.1 5mm
S960MC LTT C
3.17 0.048 86 85 97
R=0.5 5mm
S960MC LTT S
3.42 0.085 85 73 121
R=0.5 5mm
All 3.27 0.395 107 100 138
All S690QL,
4.39 0.122 153 - 163
R=0.1
All S690QL,
3.41 0.141 93 82 -
R=0.5
All S700MC,
5.87 0.417 169 - 159
R=0.1
All S700MC,
3.92 0.265 101 84 -
R=0.5
All S960QL,
3.98 0.338 143 - 160
R=0.1
All S960QL,
3.20 0.150 89 84 -
R=0.5
All steels, R=0.1 4.42 0.318 152 - 161
All steels, R=0.5 3.22 0.151 88 82 -

96
 Table 27 Results obtained from as-welded specimens under variable amplitude loading

% of Yield Equivalent n/N for n/N at failure based Average n/N at failure n/N at failure Average n/N at
Steel Max stress range in stress in stress range, Cycles at spectrum at on mean constant based on mean constant based on Class failure based on Class
Grade Thickness, mm spectrum, MPa tension MPa failure applied load amplitude S-N curve amplitude S-N curve F2 mean curve F2 mean curve
560 40 135 951,931 0.20 2.91 1.90
630 45 152 730,290 0.28 3.19 2.08
S700MC 5 700 50 169 421,844 0.39 2.53 2.69 1.65 1.76
386,495 3.09 2.02
770 55 186 0.52
215,591 1.73 1.13
485,974 1.49 0.97
560 40 135 0.20
510,776 1.56 1.02
362,292 1.58 1.03
630 45 152 345,533 0.28 1.51 0.99
359,435 1.57 1.03
218,619 1.31 0.86
S700MC 10 1.47 0.96
700 50 169 247,532 0.39 1.49 0.97
310,777 1.87 1.22
118,619 0.95 0.62
151,427 1.21 0.79
770 55 186 0.52
183,407 1.47 0.96
210,777 1.69 1.10
1,052,327 8.29 5.41
1,144,814 9.02 5.89
768 40 185 0.52
3,814,908 30.04 19.62
1,410,777 11.11 7.26

97
410,776 4.60 3.00
864 45 208 774,069 0.73 8.66 5.66
S960MC 5 8.29 5.42
625,930 7.01 4.58
346,031 5.31 3.46
960 50 231 337,843 1.01 5.18 3.38
257,159 3.94 2.58
157,162 3.20 2.09
1056 55 254 1.34
154,819 3.16 2.06
3,013,146 6.16 4.02
3,062,757 6.26 4.09
4,030,550 8.24 5.38
S700MC
3,023,433 6.18 4.04
(Round 5 490 35 118 0.13 6.05 3.95
1,944,494 3.97 2.60
Robin)
1,859,211 3.80 2.48
2,973,837 6.08 3.97
3,766,568 7.70 5.03
 Table 28: Results obtained from HFMI treated specimens under variable amplitude loading

% of Yield Equivalent n/N for n/N at failure based Average n/N at failure n/N at failure Average n/N at
Steel Thickness, Max stress range in stress in stress Cycles at spectrum at on mean constant based on mean constant based on Class failure based on
Grade Condition mm spectrum, MPa tension range, MPa failure applied load amplitude S-N curve amplitude S-N curve F2 mean curve Class F2 mean curve
751,607 1.73 3.93
770 55 186 0.52
1,813,968 4.18 9.48
1,396,797 4.13 9.35
S700MC C1 5 840 60 202 0.67 3.17 7.18
1,345,144 3.97 9.01
814,838 3.07 6.95
910 65 219 0.86
513,856 1.93 4.38
830,290 1.91 4.34
770 55 186 0.52
1,035,908 2.39 5.42
410,777 1.21 2.75
S700MC C1 10 840 60 202 0.67 1.41 3.20
285,995 0.84 1.91
226,932 0.85 1.94
910 65 219 0.86
331,495 1.25 2.83
483 35 116 10,000,000 0.19 >5.59 >12.68
690 50 166 10,000,000 0.37 >16.40 >37.16
S690QL C1 10 784,369 2.25 5.01 5.10 11.35
828 60 200 0.65
3,389,205 9.72 22.03
897 65 216 847,008 0.82 3.06 6.93
691 36 - 3,605,973 - - -
864 45 208 2,918,619 0.73 9.41 21.34
725,264 3.20 7.26
S960MC C3 5 960 50 231 1.01 4.14 9.38

98
1,110,776 4.91 11.12
1056 55 254 262,412 1.34 1.54 3.49
1152 60 278 210,776 1.74 1.62 3.68
3,148,731 18.49 41.92
1056 55 254 1.34
2,810,777 16.51 37.42
910,777 7.01 15.90
S960QL C1 10 1152 60 278 1.74 12.12 27.48
1,710,776 13.17 29.86
1,380,510 13.49 30.58
1248 65 301 2.21
414,582 4.05 9.18
 Table 29: Results obtained from laser re-melted specimens under variable amplitude loading

Max stress % of Yield Equivalent n/N for n/N at failure based Average n/N at failure n/N at failure Average n/N at
Steel Thickness, range in stress in stress Cycles at spectrum at on mean constant based on mean constant based on Class failure based on
Grade Condition mm spectrum, MPa tension range, MPa failure applied load amplitude S-N curve amplitude S-N curve F2 mean curve Class F2 mean curve
845,532 1.98 4.21
759 55 183 0.50
519,607 1.22 2.59
450,760 1.38 2.93
S690QL C2 10 828 60 200 1,043,873 0.65 3.19 2.37 6.78 5.05
492,120 1.50 3.20
362,260 1.39 2.97
897 65 216 0.82
1,548,729 5.96 12.68
672 35 162 5,176,949 0.52 8.40 17.88
1,810,777 4.38 9.31
768 40 185 0.52
918,619 2.22 4.72
S960QL C2 10 864 45 208 657,158 0.73 2.26 3.65 4.80 7.78
960 50 231 506,100 1.01 2.38 5.07
1056 55 254 264,845 1.34 1.66 3.53
1152 60 278 522,863 1.74 4.29 9.13

99
 Table 30: Results obtained from LTT treated specimens under variable amplitude loading

% of Yield Equivalent n/N for n/N at failure based on Average n/N at n/N at failure Average n/N at
Steel Thickness, Max stress range stress in stress range, Cycles at spectrum at mean constant amplitude failure based on based on Class failure based on
Grade Condition mm in spectrum, MPa tension MPa failure applied load S-N curve mean constant F2 mean curve Class F2 mean
770 55 186 510,776 0.52 1.65 2.67
410,792 0.67 1.70 2.75
840 60 202
S700MC LTT C 5 352,560 0.67 1.46 1.46 2.36 2.36
221,320 0.86 1.17 1.89
910 65 219
248,689 0.86 1.31 2.12
1,214,582 0.52 3.93 6.35
770 55 186
1,256,536 0.52 4.06 6.57
710,800 0.67 2.94 4.76
840 60 202
S700MC LTT C 10 311,700 0.67 1.29 2.49 2.09 4.02
165,214 0.86 0.87 1.41
910 65 219 279,997 0.86 1.48 2.39
537,600 0.86 2.84 4.59
2,210,776 0.50 6.81 11.01
759 55 183
861,972 0.50 2.65 4.29
1,871,000 0.65 7.52 12.16
S690QL LTT C 10 3.15 5.10
828 60 200 522,900 - - -
326,692 0.65 1.31 2.12
897 65 216 221,552 0.82 1.12 1.81
864 45 208 1,417,410 0.73 6.41 10.36
210,780 1.01 1.31 2.11
960 50 231
S960MC LTT C 5 215,506 1.01 1.33 2.62 2.16 4.23

100
248,732 1.34 2.05 3.31
1056 55 254
241,050 1.34 1.98 3.21
2,448,729 0.73 11.07 17.90
864 45 208
2,122,863 0.73 9.60 15.52
440,345 1.01 2.73 4.41
960 50 231
S960QL LTT C 10 497,769 1.01 3.08 6.42 4.98 10.37
1,196,259 1.34 9.85 15.92
1056 55 254
514,582 1.34 4.24 6.85
1152 60 278 402,140 1.74 4.34 7.02
Table 31: Experimental R=0.1 constant amplitude axial fatigue data for HFMI treated longitudinal
welds

Ref. steel type fy treatment plate thickness number of specimens

(MPa) method (mm) in series

8 S700 7002 UP/ UIT 8 16

9 S690QL 7861 UIT 16 16

9 S690QL 7861 HiFIT 16 15

10 16Mn 3902 UP/UPT 8 6

11 S350 3981 UP/UPT 12 5

11 S700 7801 UP/UPT 12 7

11 S900 9001 TIG+UP 12 10

12 SS800 7001 UP/UPT 8 8

12 16Mn 3901 UP/UPT 8 6

12 Q235B 2671 UP/UPT 8 7

13 S355 3552 UIT 8 10

1
measured fy
2
nominal fy

Table 32: Experimental R=0.1 constant amplitude axial fatigue data for HFMI treated transverse
stiffener welded joints

Ref. steel type fy treatment plate thickness number of specimens

(MPa) method (mm) in series

14
S355J2 3981 UIT 12 7

14
S355J2 3981 UIT 12 4

ee14 S460ML 5041 UIT 12 5

14 S460ML 5041 UIT 12 5

15 S260 2602 UIT 20 9

15 S355J2 4771 PIT 12 8

15 S690QL 7811 PIT 12 7

16 AH36 3921 UIT 20 3

16 AH36 3921 UIT 20 3

1
measured fy
2
nominal fy

101
Table 33: Experimental R=0.1 constant amplitude axial fatigue data for HFMI treated butt welded
joints

Ref. steel type fy treatment plate number of specimens

(MPa) method thickness(mm) in series

9 S355J2 4221 UIT 16 14

9 S355J2 4221 HiFIT 16 18

9 S690QL 7861 UIT 16 18

9 S690QL 7861 HiFIT 16 12

18 E690 7631 UP 9.5 8

19 S960 9602 PIT 5 7

1
measured fy
2
nominal fy

Table 34: FAT values for welded joints evaluated in this study

Specimen Type FAT2 (MPa) Sm,A3(MPa) N

Longitudinal
711 97 0.206
attachment

Transverse
80 110 0.206
stiffener joint

Butt joint 90 123 0.206

1
Note that FAT 71 is for 50 < l < 150 mm (l = attachment length). Larger or smaller FAT
values are seen as l changes. For simplicity FAT 71 was assumed for all longitudinal attachments.
2
Typical values used by the International Institute of Welding
3
Mean value corresponding to 50% survival probability at 2106 cycles to failure in the as-
welded condition

102
 HOW TO OBTAIN EU PUBLICATIONS

Free publications:
one copy:
via EU Bookshop (http://bookshop.europa.eu);
more than one copy or posters/maps:
from the European Unions representations (http://ec.europa.eu/represent_en.htm);
from the delegations in non-EU countries (http://eeas.europa.eu/delegations/index_en.htm);
by contacting the Europe Direct service (http://europa.eu/europedirect/index_en.htm) or
calling 00 800 6 7 8 9 10 11 (freephone number from anywhere in the EU) (*).
(*) The information given is free, as are most calls (though some operators, phone boxes or hotels may charge you).

Priced publications:
via EU Bookshop (http://bookshop.europa.eu).

Priced subscriptions:
via one of the sales agents of the Publications Office of the European Union
(http://publications.europa.eu/others/agents/index_en.htm).
 KI-NA-26929-EN-N
The objective of the FATWELDHSS project was to study post-weld treatment
techniques and their effect on the fatigue life of MAG welded attachments
in High Strength Steel (HSS). Fatigue cracks in steel structures often occur at
welded joints, where stress concentrations due to the joint geometry and tensile
residual stresses are relatively high. Fatigue life improvement techniques, which
rely on improving the stress field and/or the surface geometry around the
welded joints, are generally known to be beneficial.
Therefore, within the framework of this project, the following were examined:
diode laser weld toe re-melting;
High Frequency Mechanical Impact (HFMI) treatment;
Low Transformation Temperature (LTT) filler wires
Laser diode re-melting was used to improve the surface profile at the weld toe
and thus reduce stress concentrations.
HFMI treatment involving high frequency hammering of the weld toe is another
technique that can produce a smooth weld toe profile but, more significantly,
which also can introduce compressive residual stresses.
Lastly, two new LTT filler wires were developed within the project as these can
decrease or even remove tensile residual stresses resulting from weld zone
shrinkage.
An extensive fatigue testing programme was set up to establish the levels
of improvement in the fatigue lives of the welded attachments achieved by
application of the selected improvement techniques.
Furthermore, two industrial demonstrators were selected that could show the
project achievements in terms of facilitating the introduction of high strength
steels by overcoming the limitations posed by the fatigue properties of the
welded joints. In addition, modelling tools were developed to predict the residual
stresses at the welded joint.
Finally, practical guidelines were developed for enhancing the fatigue strength
of HSS welded structures.

Studies and reports

ISBN 978-92-79-44077-9
doi:10.2777/705266