PREDICTIVE MODEL FOR PEKKA

N EVA S M A A
THE PREVENTION OF WELD
METAL HYDROGEN CRACKING Department of Mechanical Engineering,
University of Oulu
IN HIGH-STRENGTH
MULTIPASS WELDS

OULU 2003
PEKKA NEVASMAA

PREDICTIVE MODEL FOR

THE PREVENTION OF WELD METAL
HYDROGEN CRACKING IN HIGH-
STRENGTH MULTIPASS WELDS

Academic Dissertation to be presented with the assent of

the Faculty of Technology, University of Oulu, for public
discussion in Raahensali (Auditorium L10), Linnanmaa, on
November 15th, 2003, at 12 noon.

O U L U N Y L I O P I S TO, O U L U 2 0 0 3
Copyright © 2003
University of Oulu, 2003

Supervised by
Professor Pentti Karjalainen

Reviewed by
Professor Horst H. Cerjak
Doctor David A. Porter

ISBN 951-42-7181-5 (URL: http://herkules.oulu.fi/isbn9514271815/)

ALSO AVAILABLE IN PRINTED FORMAT

Acta Univ. Oul. C 191, 2003
ISBN 951-42-7180-7
ISSN 0355-3213 (URL: http://herkules.oulu.fi/issn03553213/)

OULU UNIVERSITY PRESS

OULU 2003
Nevasmaa, Pekka, Predictive model for the prevention of weld metal hydrogen
cracking in high-strength multipass welds
Department of Mechanical Engineering, University of Oulu, P.O.Box 4200, FIN-90014
University of Oulu, Finland
Oulu, Finland
2003

Abstract
This thesis studies controlling factors that govern transverse hydrogen cracking in high-strength
multipass weld metal (WM). The experiments were concerned with heavy-restraint Y- and U-Groove
multipass cracking tests of shielded-metal arc (SMAW) and submerged-arc (SAW) weld metals.
Results of tensile tests, hardness surveys, weld residual stress measurements and microstructural
investigations are discussed. The analytical phase comprised numerical calculations for analysing the
interactions between crack-controlling factors. The objectives were: (i) the assessment of WM
hydrogen cracking risk by defining the Crack-No Crack boundary conditions in terms of 'safe line'
description giving the desired lower-bound estimates, and (ii) to derive predictive equations capable
of giving reliable estimates of the required preheat/interpass temperature T0/Ti for the avoidance of
cracking.
Hydrogen cracking occurred predominantly in high strength weld metals of Rp0.2 ≈ 580-900 MPa.
At intermediate strengths of Rp0.2 ≈ 500-550 MPa, cracking took place in the cases where the holding
time from welding to NDT inspection was prolonged to 7 days. Low strength WMs of Rp0.2 ≤ 480
MPa did not exhibit cracking under any conditions examined. Cracking occurrence was, above all,
governed by WM tensile strength, weld diffusible hydrogen and weld residual stresses amounting to
the yield strength. The appearance of cracking vanished when transferring from 40 to 6 mm thick
welds. The implications of the holding time were more significant than anticipated previously. A
period of 16 hrs in accordance with SFS-EN 1011 appeared much too short for thick multipass welds.
Interpass time and heat input showed no measurable effect on cracking sensitivity, hence being of
secondary importance. Equations were derived to assess the weld critical hydrogen content Hcr
corresponding to the Crack-No Crack conditions as a function of either weld metal Pcm, yield strength
Rp0.2 or weld metal maximum hardness HV5(max). For the calculation of safe T0/Ti estimates, a
formula incorporating: (i) WM strength as a linear function of either weld carbon equivalent CET or
weld HV5(max), (ii) weld build-up thickness aw in the form of tanh expression and (iii) weld diffusible
hydrogen HD in terms of a combined [ln / power law] expression was found descriptive.

Keywords: cold cracking, high-strength steels, hydrogen cracking, multipass welding, weld
metal cracking
 Preface
This work was principally carried out in two laboratories: VTT Industrial Systems
(formerly: VTT Manufacturing Technology) and the University of Oulu during 1999–
2002. Specifically and additionally, experimental cracking tests were performed at VTT,
Oy ESAB and Mäntyluoto Works Oy, Finland, ESAB U.K Ltd, United Kingdom and
Nippon Steel Corporation, Japan. Weld residual stress measurements were made in
Helsinki University of Technology, Finland, whilst the University of Oulu was
responsible for the permeation tests of hydrogen diffusivity, weld metal tensile tests and
metallographic studies of welds. The non-destructive examination of the cracking test
welds was performed jointly by Rautaruukki Oyj and VTT, Finland.
I would like to express my sincere gratitude to Professor Pentti Karjalainen, Dr. David
Widgery and Dr. David Porter for their encouraging attitude, experimental work and
valuable scientific comments concerning the analytical work. I am deeply indebted to
Professor Nobutaka Yurioka for the opportunity to work in Nippon Steel Research
Centre, Japan, as well as for the fruitful theoretical lectures and many discussions. I am
very grateful to Lic.Tech. Mari-Selina Kantanen and PhD Longxiu Pan for their
contribution as co-workers to the experimental part concerning metallographic studies of
weld metal microstructures and measurements of hydrogen diffusion. The persistency
and collaboration of Professor Risto Karppi regarding the arrangements of my working
period in Japan, and the project management, is gratefully acknowledged. I have the
pleasure in thanking MScTech Juha Lukkari for providing me with many rewarding
research papers from recent works in this field.
I wish to thank all the colleagues and staff for their contribution to the various parts of
the work during its lifetime, without whom this all would not have been possible. These
thanks are, above all, due to Esko Kallinen, Esa Penttilä, Aarne Kanerva, Pentti Pajunen,
S. Highland, A. Elmes and M. Andrews for the welding experiments, as well as to David
Widgery, Juha Lukkari, Mika Siren, Janne Mononen, Janne Tamminen and Ari
Joutsenvirta for supervising the welding work. Anssi Brederholm, Hans Gripenberg and
Iikka Virkkunen who were conducting the residual stress measurements are gratefully
acknowledged. It is my pleasure to thank MScTech Anssi Laukkanen for the many
valuable and far-reaching discussions on numerical modelling of hydrogen diffusion and
the associated effects on fracture resistance. I would also like to thank Mrs Hilkka
Hänninen, Mr Tuomo Hokkanen and Mr Erkki Makkonen for their patience in illustrating
the figures and photos, as well as Arja Grahn, Erja Mörsky and Merja Asikainen for their
secretarial work.
This work belonged to the Finnish R & D project “Control of high-strength steel weld
metal hydrogen cracking” co-ordinated by VTT Industrial Systems. Financial support by
the Graduate School of Metallurgy within the University of Oulu, ESAB Group Ltd,
Rautaruukki Oyj, National Technology Agency (Tekes) and VTT Industrial Systems is
gratefully acknowledged. Support in the form of labour as countless number of man-
hours by Mäntyluoto Works Oyj and PPTH Steel Ltd is highly appreciated.
Finally, my dearest thanks are due to my wife Anu, my son Ilmari and my daughters
Anniina and Leena for bearing their invidious share of the burden along the way, as well
as for all the borrowed time I cannot pay back.
"After all this struggle, agony and pain – while now reaching finally my very
destination,
Can't help feeling like a freight train – lost'n' lonely, as if I was abandoned in the
station".

Espoo, November 2003 Pekka Nevasmaa

 List of symbols and abbreviations
α : ferrite
a : activity of the diffusing mass
aw : weld build-up thickness (mm)
ab : height of an individual weld bead, bead size (mm)
AF : acicular ferrite
BM : base material, parent material
B/A : weld (bead) overlap ratio according to bead-on-plate test
BS : bainite transformation (start) temperature
c : concentration of the diffusing mass, e.g., hydrogen
CEIIW : IIW carbon equivalent (%)
CET : German carbon equivalent based on chemical composition (%)
d : weld bead penetration/overlap (mm)
d/hw : weld (bead) overlap ratio
D : diffusion coefficient – general expression
D(T) : diffusion coefficient as a function of temperature (cm2/ sec)
Di : diffusion coefficient at a given interpass-temperature (cm2/ sec)
DH : diffusion coefficient for hydrogen (cm2/ sec)
E : arc energy (in welding) (kJ/mm)
F : weld groove shape factor
FCAW : flux-cored (tubular wire) arc welding
FDM : finite difference method
FEM : finite element method
Fe3C : cementite
FI : specimen/groove severity parameter in terms of stress concentration
FS : side-plate ferrite, lath-like ferrite side-plates resembling upper-bainite
FS(A) : ferrite side-plates with aligned second-phase
FS(NA) : ferrite side-plates with non-aligned second phase
γ : activity coefficient of the mass
γ : austenite
GBF : pro-eutectoid grain boundary ferrite
GMAW : gas-metal arc welding (i.e., MIG/MAG welding)
HAZ : heat-affected zone
HD : weld diffusible hydrogen content per deposited weld metal (DM)
conforming to ISO/IIW 3690-1977 (ml/100 g DM IIW)
HF : weld diffusible hydrogen content per fused metal (ml/100 g FM IIW)
H0 : initial diffusible hydrogen content in the weld (at high temperatures,
immediately after solidification before cooling down to ambient
temperature) according to e.g. ISO 3690 (in which case H0 = HD)
(ml/100 g DM IIW)
Hmax : maximum hydrogen content/concentration in weld metal – general
expression (ml/100 g)
HRmax : final local maximum hydrogen concentration in multipass weld metal
(ml/100 g DM IIW)
HR100 : remaining weld diffusible hydrogen content at 100°C (ml/100 g DM IIW)
HR100max : maximum remaining weld diffusible hydrogen content at 100°C
(ml/100 g DM IIW)
HRmax(d=0) : maximum remaining weld diffusible hydrogen content under conditions
of no dilution or weld overlap (ml/100 g DM IIW)
HR100max(d=0) : maximum remaining weld diffusible hydrogen content at 100°C under
conditions of no dilution or weld overlap (ml/100 g DM IIW)
HRmax(d > 0) : local final (maximum) hydrogen concentration in weld metal affected by
dilution and weld overlap (ml/100 g DM IIW)
Hcr : weld critical hydrogen content with respect to cracking (ml/100 g DM IIW)
HV : Vickers pyramid hardness (HV)
HVave : average value of Vickers hardness (HV)
HVmax : maximum value of Vickers hardness (HV)
h : plate thickness (mm)
hw : thickness of an individual weld bead layer (mm or cm)
J : flux of the diffusing mass
k : thermal efficiency of welding (acc. to EN 1011-1:1998)
l : number of weld bead layers
µ : chemical potential of a substance in the phase
M : martensite
M-A : martensite-austenite constituent
MAC : martensite-austenite-carbide (aligned) layers between ferrite laths
MS : martensite transformation (start) temperature (°C)
N : number of individual weld beads
Pcm : Japanese welding index based on chemical composition (%)
PD–D : integrated hydrogen diffusion-distance parameter
PF : polygonal ferrite
PWHT : post-weld heat treatment
Q : heat input (in welding) (kJ/mm)
R : the gas constant (cal/mol)
RF : intensity of restraint against transverse shrinkage (MPa/mm∗mm)
Re : yield strength – general expression (MPa)
Re(WM) : yield strength of weld metal (MPa)
Rp0.2 : yield strength at 0.2 mm elongation (MPa)
RM : ultimate tensile strength (MPa)
σres : welding residual stress (MPa)
σresL : weld longitudinal residual stress (MPa)
σnet : transverse net stress across the weld throat (MPa)
ΣD∆t : thermal factor of hydrogen diffusion (cm2)
ΣD∆t(100) : estimate for the thermal parameter of hydrogen diffusion ΣD∆t from 500
to 100°C
SAW : submerged-arc welding
SMAW : shielded metal-arc welding (manual metal-arc welding)
τ : thermal parameter of hydrogen diffusion (cm2)
t : time, in general (sec)
t8/5 : weld cooling time between 800 to 500°C (sec)
t500–100 : weld cooling time between 500 to 100°C (sec)
t15/2 : weld cooling time from peak temperature to 200°C (sec)
t200 : weld cooling time from peak temperature to 200°C (sec)
t15/1.5 : weld cooling time from peak temperature to 150°C (sec)
t150 : weld cooling time from peak temperature to 150°C (sec)
t100 : weld cooling time from peak temperature to 100°C (sec)
ti : welding interpass time (min, sec)
Σti : total interpass time during welding (min, sec)
∆t : time integrated over the temperature range from Tm to Ti (sec)
T : temperature, in the connection with diffusion equations (K)
θ : temperature, in the connection with diffusion equations (°C)
T0 : preheat temperature, in general (°C)
Tcr : critical required preheat/interpass temperature for the avoidance of
cracking (°C)
Ti : weld interpass temperature (°C)
Tm : melting temperature corresponding to approx. 1500°C (°C)
WM : weld metal
 Contents
Abstract
Preface
List of symbols and abbreviations
Contents
1 Introduction: Current status of weld metal hydrogen cold cracking .......................... 15
1.1 Hydrogen-induced cracking in weldments – general........................................ 15
1.1.1 Hydrogen cracking in the HAZ ............................................................ 17
1.1.2 Hydrogen cracking in weld metal......................................................... 18
1.1.3 Characteristic appearance of hydrogen cracks ..................................... 19
1.2 Principal controlling factors of hydrogen induced cracking in weld metals ..... 20
1.2.1 Hydrogen.............................................................................................. 20
1.2.2 Microstructure ...................................................................................... 26
1.2.3 Stresses................................................................................................. 29
1.2.4 Welding process ................................................................................... 32
1.3 Mechanisms of hydrogen induced and assisted cracking ................................. 32
1.3.1 Hydrogen enhanced decohesion ........................................................... 33
1.3.2 Hydrogen enhanced localised plasticity ............................................... 34
1.3.3 Adsorption induced dislocation emission............................................. 34
1.4 Hydrogen diffusion and its role in cracking of multiple-pass weld metals....... 36
1.4.1 General equation for the diffusion of mass .......................................... 36
1.4.2 Effects of temperature and continuity conditions................................. 38
1.4.3 Activity coefficient of the diffusing mass ............................................ 39
1.4.4 Potential of hydrogen in structural steel weld metal ............................ 40
1.4.5 Determination of the diffusion coefficient of hydrogen ....................... 41
1.4.6 Determination of the interactive and generalised activity coefficients. 43
1.4.7 Numerical analysis of hydrogen diffusion and accumulation effects... 45
1.5 Specific design factors affecting transverse hydrogen cracking in multipass
weld metal......................................................................................................... 47
1.5.1 Welding residual stresses and stress concentration .............................. 48
1.5.2 Plate thickness/weld build-up thickness and bead layer thickness ....... 50
 1.5.3 Weld penetration and bead overlap – effects on final hydrogen
concentration ........................................................................................ 54
1.5.4 Interpass time and temperature............................................................. 57
1.5.5 MS temperature and hydrogen distribution........................................... 58
1.5.6 Welding heat input ............................................................................... 60
1.6 Preheat assessment methods for the avoidance of hydrogen cracking in
multipass weld metal ........................................................................................ 62
1.6.1 Method according to Okuda et al. ........................................................ 62
1.6.2 Method according to Yurioka et al. ...................................................... 63
1.6.3 Small-scale cracking tests and their usability....................................... 64
1.7 Practical experiences of hydrogen cracking in multipass welds – overview .... 67
2 Aims of the current work ........................................................................................... 70
3 Experimental and analytical procedures .................................................................... 72
3.1 Materials ........................................................................................................... 72
3.2 Experiments and testing methods ..................................................................... 74
3.2.1 Welding processes and procedures....................................................... 74
3.2.2 Weld hydrogen measurements ............................................................. 75
3.2.3 Weld cracking test programme............................................................. 75
3.2.4 Analysis of weld chemical composition............................................... 77
3.2.5 Tensile tests .......................................................................................... 77
3.2.6 Hardness survey ................................................................................... 77
3.2.7 Weld residual stress measurements ...................................................... 78
3.2.8 Measurements of hydrogen diffusion coefficient ................................. 78
3.2.9 Dilatometric measurements .................................................................. 79
3.2.10 Weld bead size measurements.............................................................. 79
3.2.11 Metallographic examinations ............................................................... 80
3.3 Analytical calculations...................................................................................... 80
3.3.1 Effects of thermal and geometrical factors on final local hydrogen
concentration ........................................................................................ 80
3.3.2 Derivation of values for hydrogen diffusion coefficient ...................... 81
3.3.3 Calculation of safe preheat/interpass temperature estimates for
multipass weld metal ............................................................................ 82
4 Results ....................................................................................................................... 85
4.1 Multipass weld metal cracking tests ................................................................. 85
4.1.1 Cracking tests for SMAW thick plate weld metals .............................. 85
4.1.1.1 Small-scale Y-Groove tests.................................................... 86
4.1.1.2 Small-scale U-Groove tests ................................................... 88
4.1.1.3 Large-scale double-welded Y-Groove tests ........................... 91
4.1.1.4 Metallographic examination of crack morphologies.............. 94
4.1.2 Cracking tests for SAW thick plate weld metals .................................. 96
4.1.2.1 Small-scale U-Groove tests ................................................... 96
4.1.2.2 Metallographic examination of crack morphologies.............. 99
4.1.3 Comparative cracking tests for welds in thin plate............................. 102
4.2 All-weld metal tensile properties .................................................................... 103
4.3 Weld metal hardness and chemical composition ............................................ 104
4.4 Residual stresses in weld metal ...................................................................... 107
 4.4.1 Ring-Core measurements for thick plate multipass welds.................. 107
4.4.2 Hole drilling measurements for thin plate welds................................ 108
4.5 Hydrogen diffusion measurements ................................................................. 108
4.5.1 Hydrogen permeation tests for parent steel ........................................ 109
4.5.2 Hydrogen permeation tests for SMAW multipass weld metals.......... 110
4.5.3 Hydrogen permeation tests for SAW and FCAW multipass weld
metals ................................................................................................. 112
4.6 Phase transformation temperature................................................................... 114
4.7 Weld bead size................................................................................................ 115
4.8 Microstructures of multiple-pass weld metals ................................................ 116
5 Discussion................................................................................................................ 118
5.1 Weld metal chemical composition – hardness relationships........................... 118
5.2 Weld metal hardness – strength dependence .................................................. 124
5.3 Effect of weld thermal cycle, bead size and overlap on final hydrogen
concentration in multipass welds according to analytic calculations.............. 128
5.3.1 Thermal cycle and bead size effects – weld overlap excluded ........... 130
5.3.2 Thermal cycle and bead size effects under the influence of
weld overlap ...................................................................................... 131
5.3.3 Differences in final hydrogen concentration between SMAW and
SAW welds......................................................................................... 135
5.4 Derivation of the 'safe-line' for critical hydrogen content in multipass welds 137
5.4.1 Dependence of weld critical hydrogen content on longitudinal
residual stress ..................................................................................... 137
5.4.2 Dependence of weld critical hydrogen content on weld chemical
composition ........................................................................................ 142
5.4.3 Dependence of weld critical hydrogen content on weld
metal hardness ................................................................................... 146
5.5 Comparison of the descriptive potential of approaches based on stress,
composition and hardness related estimates ................................................... 150
5.5.1 Descriptive potential for the SMAW weld metals.............................. 150
5.5.2 Descriptive potential for the SAW weld metals ................................. 152
5.5.3 Descriptive potential for the combined data sets of SMAW and
SAW welds......................................................................................... 153
5.6 Role of weld build-up thickness and residual stress in weld metal hydrogen
cracking .......................................................................................................... 155
5.7 Effect of weld bead size on hydrogen diffusion distance and cracking risk ... 157
5.8 Effect of weld metal chemical composition and local microstructural
mismatch on cracking sensitivity.................................................................... 159
5.9 Comparison of hydrogen diffusivity between SMAW, SAW and FCAW
multipass weld metals..................................................................................... 162
5.10 Comparison of cracking sensitivity between SMAW and SAW multipass
weld metals ..................................................................................................... 163
5.11 Differences in final hydrogen content between single-pass and multiple-
pass welds ....................................................................................................... 165
5.12 Derivation of the prediction formulae for the estimation of safe
preheat/interpass temperatures for multipass weld metals.............................. 169
 5.12.1 Re-analysis of the NSC Y-Groove data sets....................................... 169
5.12.2 Comparison of the NSC Y-Groove and VTT U-Groove data sets in
view of required preheat/interpass temperatures ................................ 172
5.12.3 Analysis of the first SAW U-Groove dataset for optimised
prediction of preheat/interpass temperature estimates for
multipass welds .................................................................................. 174
5.12.3.1 Optimisation of theT0/Ti prediction with respect to weld
metal strength ...................................................................... 175
5.12.3.2 Role of heat input and interpass time in optimisation of
theT0/Ti prediction ............................................................... 176
5.12.3.3 Optimisation of theT0/Ti prediction with respect to
plate/weld build-up thickness .............................................. 177
5.12.3.4 Optimisation of theT0/Ti prediction with respect to weld
hydrogen content ................................................................. 179
5.12.4 Final analysis of the SAW U-Groove data for optimised prediction
of preheat/interpass temperature estimates for multipass welds ........... 181
5.12.5 Postulated mechanism for transverse hydrogen cracking in extra-
high strength multipass weld metals................................................... 185
5.12.6 Proposed procedure as to provide the necessary precautions for the
avoidance of hydrogen cracking in multipass weld metal.................. 188
5.12.7 Areas of non-applicability of the Procedure....................................... 191
6 Conclusive remarks ................................................................................................. 193
7 Summary of final conclusions ................................................................................. 198
8 Future work.............................................................................................................. 200
References ...................................................................................................................... 202
Appendices 1–11
 1 Introduction:
Current status of weld metal hydrogen cold cracking
Arc welding typically leads to the occlusion of hydrogen gas in the arc atmosphere into
the solidifying weld metal, from which sc. diffusible hydrogen can diffuse into the
various regions of the weldment as it cools down. Depending on metal's microstructure,
concentration of weld diffusible hydrogen and the level of weld residual stress, the risk of
hydrogen-induced cold- cracking in ferritic steel can arise when the weld cools down to
below the 150–100°C temperature region. Generally, hydrogen cracking is of a delayed
nature, i.e., cracks can appear several days after welding was completed. Usually,
hydrogen cracks are sited either in the parent steel heat-affected zone or in the weld metal
itself. One of the most effective precautions against weld hydrogen cracking is to use
preheating, i.e., to heat up a sufficiently great volume of the weld area of a structural
member prior to welding or, in the case of multipass welding, to apply elevated interpass
temperature throughout the welding operation.
Recent developments in advanced steel processing routes have considerably improved
base material quality by lowering carbon and impurities, thereby enhancing the resistance
of the weld heat-affected zone (HAZ) to hydrogen cold cracking. From the end user's
viewpoint, these developments have expanded the application boundaries for welding
filler materials, challenging the consumable manufacturers to keep pace with the
development of steel products. Presently, the industrial benefits brought by advanced
steel are too often offset by the shift of hydrogen cold cracking problems to the weld
metal (WM). Consequently, with increasing strength of parent plate and weld metal,
required preheat may be dictated by the cracking sensitivity of the weld metal instead of
the parent steel HAZ.

1.1 Hydrogen-induced cracking in weldments – general

Weld hydrogen cracking is attributable to three main factors: microstructure, hydrogen

and stress. The causal factors governing the occurrence of hydrogen-induced cold
 16

cracking in welded joints of ferritic structural steels are: (i) crack-sensitive, hardened
microstructure containing martensitic and/or bainitic transformation products, (ii)
sufficiently high, local weld hydrogen concentration in terms of weld diffusible hydrogen
content and (iii) elevated stress caused by high structural restraint that is determined by
structural rigidity, i.e., plate thickness and weld bead height (or weld build-up thickness).
Hydrogen cracking sensitivity of weldments is a combination of these three primary
factors[40].
The greatest risk of weld hydrogen cracking occurs when the weld cools down and
temperatures of around 150–100°C are reached; above this temperature range, cracking is
unlikely to initiate in ferritic structural steels[40]. A characteristic feature of hydrogen
cold-cracking is its delayed nature, that is, crack initiation and especially propagation
may take place several hours, or sometimes even days or weeks, after welding has been
completed. The risk of cracking becomes apparent and raises as strength of parent steel
and/or weld metal increases, and when welding thick plate sections that often employ
multipass welding techniques[40].
Overall, there is very little information on the risk of weld metal hydrogen cracking in
multipass welds, from which guidance on safe welding conditions can be derived[8]. Thus,
welding guidelines in the current standards are quite limited in addressing the issue of
avoiding hydrogen cracking in weld metal[8, 76]. AWS D1.1:1988[104] advises to conduct
testing where it is feared there may be a risk. BS 5135:1984[105], in turn, proposes two
approaches: (i) the use of consumables giving less than 2.5 ml/100 g DM hydrogen, or
(ii) the application of 200°C post heat for 2–3 h. Whilst the former approach is outside
the capabilities of most SAW and FCAW consumables, the latter approach has
undesirable implications for, e.g., welding costs. Even the very recent European standards
EN 1011-1:1998[1a] and EN 1011-2:2001[1b] do not give reliable guidance on how to
assess the need of preheat for weld metal quantitatively. These standards do consider the
possibility of weld metal cracking under certain conditions, however, they do not provide
the user with any unified, scientifically validated methodology for the calculation of the
actual level of ‘safe’ preheat temperature.
Simple, small-scale oblique y-groove Tekken, CTS and Implant tests are currently
used for the assessment of HAZ hydrogen cracking of structural steels. The principal
experiments employed for single-pass weld metals are the (symmetrical) Y-Groove
Tekken test, the tensile restraint cracking (TRC) test, the gapped bead on plate (G-BOP)
test and the Welding Institute of Canada (WIC) test. Applying only small-scale single-
pass tests for weld metal, one cannot be convinced whether the results are really
descriptive of the multiple pass welded structure. Testing full-scale products, on the other
hand, indicates only the cracking susceptibility of the particular structure under certain
welding conditions, but the results do not necessarily have any general relevance to other
cases. Consequently, there are substantial gaps to be filled between the outcome of
various cracking tests and a derivation of a unified prediction model for assessment of
hydrogen cracking susceptibility in weld metal.
Of the calculation formulae predicting necessary preheat for weld metals, only
few[2, 3, 12] are systematically based on experiments that actually involve multipass
welding, the others being either modifications of previously existing formulae derived
originally for the HAZ cracking cases[1b], or are solely based on the results of single pass
cracking tests[7]. Single pass weld tests are of little assistance in the provision of
 17

procedural guidance for multipass welds, principally because of the hydrogen retention
characteristics of a much larger volume of weld metal in multipass welds, and because
such factors as interpass time and -temperature are not encountered with single pass
welds[8]. Therefore, suitability of data gained from single pass welding tests to reliable
assessment of the hydrogen cracking risk in multipass weld metals can be questioned.
Every now and then, fabricators are still experiencing cases of hydrogen cold cracking
in the welding of high strength steels and, increasingly, in the weld metal[76]. It is obvious
that the status of knowledge on WM hydrogen cracking is not nearly as well advanced as
that for the HAZ cracking. Even the few calculation formulae[2, 3, 11] that do exist for the
assessment of hydrogen cracking risk in multipass weld metal can give greatly varying
predictions, the differences in preheat temperature estimates becoming confusingly great.
Presently, there is no consensus about the relevant primary parameters that should be
included into such estimation procedures. For instance, different views exist on whether
plate thickness should be considered, or not[2, 3, 76], and do increasing heat inputs actually
result in beneficial or adverse effects with respect to WM cracking risk[8, 9, 11, 76].
Whether cracking takes place in the WM or HAZ depends on the actual chemical
composition of the weld metal in relation to parent steel, as well as on the strength level
in question. Generally, higher strength and greater contents of alloying elements in the
weld metal tend to favour WM cracking, at the expense of HAZ cracking[40].

1.1.1 Hydrogen cracking in the HAZ

For CMn and microalloy/low-alloy high-strength steels, the heat affected zone (HAZ)
cracking is generally the major form induced by hydrogen[13, 24, 25]. In practice, this covers
steels with their yield strengths ranging from 350 to about 600 MPa. The most frequent
types of HAZ cracks associated with common welding processes are root crack, toe
crack and underbead crack.
Of these, underbead cracks are generally associated with shorter delay times and
higher hydrogen levels than toe and root cracks, provided other factors are kept
constant[41, 43]. Thus, lowering the hydrogen content effectively prevents underbead
cracks in both fillet and butt welds. In multipass welding, however, underbead cracking
may become a problem due to the effect of angular distortion (in the case it has not been
prevented), this has been attributed to certain welding sequences[44]. As far as the HAZ
cracking is concerned, the main problem in both single- and multipass welding of butt
joints is usually root and/or underbead cracking. The restraint stress is highest after the
first pass being laid, provided the weld is allowed to cool down to ambient
temperatures[48]. The fact that the stress concentration is generally higher at the root than
at the toe, further accentuates the appearance of root cracking at the expense of toe
cracking[45].
The formation of HAZ underbead/root cracks is essentially governed by the transverse
net stress σnet across the weld and against transverse shrinkage, which, in turn, is
determined by structural restraint and, to some extent, the weld metal strength. Since
structural rigidity in thin plates is far too low to cause structural restraint and transverse
 18

stresses high enough to promote cracking in the HAZ[1, 3], root and underbead cracking
represent a potential risk merely in high-strength steel plates with thickness exceeding,
say, 30 mm. This applies roughly to steel with yield strengths in the range of 355–460
MPa[34, 52], whereas for extra-high strength steel and with yield strengths exceeding
600 MPa, the critical thickness for the occurrence of the HAZ underbead/root cracking is
lowered to 20–25 mm[52, 53]. Another application field where underbead/root cracks in the
HAZ are often faced, are pipeline girth welds.
As the role of transverse stress σnet is so decisive in the case of the HAZ
underbead/root cracking, and the stress concentration at the root is relatively simple to
model numerically and/or calculate analytically from the groove geometry, many of the
current numerical methods[1b, 13] can predict safe preheat temperatures for the avoidance
of cracking quite reliably.

1.1.2 Hydrogen cracking in weld metal

For extra-high strength steels of yield strength greater than 600 MPa and with weld
metals of matching or overmatching strength, weld metal (WM) hydrogen cracking starts
gradually to become the predominant form of hydrogen cracking[2, 3, 25]. Cracking in weld
metal occurs either transverse or longitudinal to the weld direction, the orientation of
cracks depending on the presence of gaps and notches and the direction of the controlling
stress[3, 54]. In general, the susceptibility of weld metal to hydrogen cracking appears to
increase with the strength and diffusible hydrogen of the weld metal. Consideration needs
to be given to the risk of weld metal cracking, in particular, whenever making relatively
thick restrained multipass butt welds[8].
In single-pass welds, the root gap or root bevel/face preparation provides a stress
concentration with respect to stresses transverse to the weld. This leads to longitudinal
hydrogen cracks in the weld metal[54], which is often the predominant form of cracking in
high-strength pipelines where cracks can be encountered, not only in single-pass welds
but, also, in the root run of multipass welds[26]. Many of the hydrogen-cracking tests, like
the (oblique) y-groove Tekken test and Implant test, are designed to assess this kind of
hydrogen cracking susceptibility and, often, predominantly in the HAZ instead of the
weld metal.
In the case of high-strength multipass welds and relatively thick restrained plates,
hydrogen cracking will generally occur transverse to the weld direction, growing either
normal to the weld surface or at an orientation of ≈ 45° angle in the weld thickness
direction. This latter morphology is called Chevron cracking[17b, 54]. In this form of
cracking, the high longitudinal tensile stress causes slip bands to form at 45° angle in the
weld thickness direction and the concentration of plastic strain in the intergranular
proeutectoid grain-boundary ferrite GBF, coupled with transport of hydrogen into these
regions, leads to crack initiation, c.f. Fig. 3. These cracks then link up by a ductile shear
mechanism assisted by hydrogen[54]. Chevron cracking is often recognised in multipass
submerged-arc (SAW) weld metals and, sometimes, in electroslag welds[54].
 19

Most of the small-scale cracking tests for weld metals, e.g., the G-BOP, Y-Tekken and
WIC test, are designed for single-pass welds[54]. Suitable cracking tests for multipass
weld metals are sparse and not standardised within Europe. In Japan, Japanese Welding
Association (JWA) has published a standard WES 1105-1985[19] entitled "Cracking Test
for Single-Bevel Groove Multi-layer Welds". This standard specifies a heavily restrained
small-scale specimen of a size 300 x 600 mm, as well as defines quantitative crack size
criterion. Despite this, different researchers and institutes, even in Japan, are still keenly
devoted to different cracking test specimens and/or groove geometries. For example,
Okuda et al[2] use a 500 x 400 mm U-grooved specimen, whereas Yurioka et al[3, 12] apply
a symmetrically Y-grooved specimen of 300 x 300 mm size. At the TWI, U.K., an 800 x
400 mm size V-grooved specimen configuration is applied[8, 9]. Some of these
experiments and applied test specimens are introduced in Chapter 1.6.
Because of the crucial role of longitudinal stress and accumulation of hydrogen – and
in the absence of any groove geometry-dependent, geometrically defined stress
concentration, the risk of WM transverse cracking in the filling runs is rather complicated
to assess reliably either using small-scale test specimens or applying analytical
calculation methods.

1.1.3 Characteristic appearance of hydrogen cracks

Fractographic examinations have revealed that hydrogen cracking is not associated with
any unified micromechanism, but that cracking can occur by various modes, i.e.,
cleavage, quasi-cleavage, microvoid coalescence or in an intergranular manner along
prior-austenite grain boundaries[42, 46, 54]. According to the microplasticity theory[50, 51],
hydrogen in the lattice ahead of the crack tip assists and promotes whatever microscopic
deformation and crack growth process the particular microstructure will inherently allow.
The fracture mode in question will thereby depend on (i) weldment microstructure, (ii)
the crack tip stress intensity level and (iii) the concentration of hydrogen[50, 51].
A characteristic feature of hydrogen cracking is that the crack propagation phase
usually includes both transgranular and intergranular fracture morphologies[3, 54]. The
highest susceptibility to hydrogen cracking is traditionally associated with intergranular
cracking, and the lowest with microvoid coalescence, i.e., dimple formation. With
increasing stress intensity, a transition from intergranular to cleavage/quasi-cleavage to
microvoid coalescence occurs in the early stages of fracture[42], i.e., a transition from
failure mechanisms involving negligible plasticity to those associated with high levels of
plastic deformation. Investigations on fracture surfaces of cracked weld specimens have
revealed[117] that the amount of crystalline area (resembling cleavage/quasi-cleavage
fracture) increases with the carbon equivalent of weld metal and, hence, its strength.
Hydrogen cracks may vary in length, usually from a few microns to several mm:s[40].
Small cracks easily escape detection by normal NDT methods, such as radiography,
magnetic particle inspection and ultrasonic inspection. In the case of extra-high strength
steels and weld metals, microcracks of the size of a few tenths of a mm to a few mm:s
may already become critical from the standpoint of structural integrity of a component[47].
 20

Hydrogen cracks in weldments can act also as potential nuclei for later failures in
manufacturing or service, e.g., by lamellar tearing, brittle fracture or fatigue.

1.2 Principal controlling factors of hydrogen induced cracking in

weld metals

According to the present knowledge[2–7, 16–18, 21–23, 26, 37], the principal causal factors of
weld metal hydrogen cold cracking have been identified as:
(i) sufficiently high weld hydrogen concentration, generally expressed in terms of weld
initial diffusible hydrogen content, HD or H0
(ii) susceptible microstructure, described usually in terms of microstructural phases,
GBF, M, and/or weld metal hardness HV or ultimate tensile strength RM
(iii) level (and tri-axiality) of weld residual tensile stress, often assumed equivalent to
the actual weld metal yield strength Re(WM).
It has been shown[3, 21] that hydrogen (micro)cracks appear in multipass weld metals
when the weld hydrogen content exceeds a certain critical level that depends on the weld
metal strength, plate/weld thickness and/or weld residual stress. These microcracks tend
to locate mostly beneath the final weld bead layer, i.e., in the 2nd or 3rd last layer where
weld hydrogen is likely to accumulate and remain the most[3]. These microcracks may
gradually extend in length and grow further into characteristic transverse WM
macrocracks[3b, 12].
The role of each factor is discussed separately in the following, with the emphasis on
multipass welds, in particular.

1.2.1 Hydrogen

As the weld cools down, much of the hydrogen absorbed by the weld pool escapes from
the solidified bead by diffusion. In general, the redistribution of hydrogen during cooling
of the weld depends on the thermal history and hence the diffusion coefficient D(T) for
hydrogen, as well as on the stress, especially, the triaxial states of stress close to notches
and cracks[21, 22, 25]. In the case of transverse hydrogen cracking in multipass weld metal,
the role of the latter factor becomes somewhat complicated, since there are no artificial
gaps or bevelled faces inside the weld metal that could act as stress raisers.
In a certain temperature range (200–300°C) during cooling, part of the hydrogen in
steel weldments looses its ability to move by diffusion. This part is called "residual
hydrogen" in contrast to "diffusible hydrogen"[57]. The formation of residual hydrogen
can be attributed to permanent trapping of hydrogen in sc. irreversible traps, for example
voids with oxide layer on their surfaces and different chemical bonding mechanisms have
been identified as sources for irreversible traps[58]. It is the diffusible hydrogen that is
generally considered the predominant fraction of hydrogen in hydrogen cold-cracking of
 21

welds[59, 60]. This has been explained by the fact that stress induced diffusion of hydrogen
plays an important role in hydrogen crack initiation[61, 62], as well as by the observation
that post heating even at relatively low temperatures (i.e., 100–150°C) already is effective
in reducing the cracking risk[40].
It has been widely accepted[3b, 13, 24, 25, 40, 59, 63] that due to the removal of diffusible
hydrogen from a weld during continuous cooling, it is not the initial diffusible hydrogen
content H0 according to ISO/IIW 3690[57], but the remaining diffusible hydrogen content
at cold cracking temperatures, HR100, that determines the cracking risk from the
viewpoint of hydrogen factor. The choice of 100°C as a reference temperature is in
agreement with hydrogen cracking temperatures[3, 24, 25]. It is known[3] that hydrogen
diffusion and evolution from the weld is still quite active at around 150°C. However, the
amount of decrease of diffusible hydrogen content in the bead on plate weld is found[3b]
only 10–15% at temperatures below 100°C for 1 to 2 hours after welding. It is therefore
reasonable to assume that the value of hydrogen content that is directly related to
hydrogen cracking is satisfactorily represented by the value at 100°C[3b].
Consequently, the remaining weld diffusible hydrogen content HR100 that is
responsible for the hydrogen cold-cracking in welds becomes always lower than the
initial diffusible hydrogen H0 absorbed by the weld pool. As far as hydrogen cracking in
single-pass welds is concerned, e.g., HAZ underbead/root cracking, the description of
hydrogen removal from a single bead during continuous post weld cooling applying the
HR100 concept has been proved by the number of earlier works[3b, 13, 24, 25, 40, 59, 63]. The
relationship between HR100 and H0 depends essentially on the weld thermal cycle and
diffusion coefficient D(T) of hydrogen. A general description for the HR100/H0 ratio can
be given using two alternative expressions[13, 63]:
(HR100 / H0) = e(–A ∗ ΣD∆t) (1a)

(HR100 / H0) = e(–β0 ∗ τ) (1b)

where HR100 is the remaining weld diffusible hydrogen content at 100°C, H0 is the initial
weld hydrogen content, ΣD∆t (or τ) is a thermal factor, sc. integrated thermal factor of
hydrogen diffusion in ferritic steel, and A and β0 are material and geometry dependent
constants, respectively[13, 22, 23, 63]. It can be seen that the HR100/H0 ratio decreases
exponentially as the weld thermal content becomes greater, i.e., the weld cooling time is
prolonged.
Measurements of HR100, coupled with plots of the HR100/H0 ratio against the thermal
diffusion factor ΣD∆t, have shown[69] that the slope varies with the weld metal strength,
see Fig. 1. This was ascribed to differences in hydrogen diffusion rate as the WM
composition changes. Consequently, values for constant A were derived[69] as 41, 46, 69,
and 83 for RM = 1800, 1300, 800 and 600 MPa welds, respectively. For basic electrodes,
a general value of A = 70 has been proposed[13, 63].
 22

Fig. 1. Relationship between ΣD∆t and the ratio of the remaining (residual) weld diffusible
hydrogen content at 100°C to the initial hydrogen content, HR100/H0, in HT60, HT80, HY130
and HY180 steels[69].

For Eq. (1), the complete weld thermal cycle is expressed in terms of integrated
thermal factor of hydrogen diffusion ΣD∆t, or τ. The ΣD∆t incorporates the diffusion
coefficient of hydrogen, D, at a certain temperature and the time interval ∆t to which the
D value applies. Theoretically, ΣD∆t should be determined by integrating over the
complete weld thermal history, i.e., from Tm to Ti. In theoretical form, ΣD∆t can be
written as[21–23]:
ΣD∆t = τ = ∫ D(t)dt (2a)

ΣD∆t = ΣD(θj)∆t = ∫ D(θ) (dt/dθ) d(θ) (2b)

In order to avoid complex and tedious integration of time increments, and provided it
is sufficient to consider the cracking conditions at a constant temperature, i.e., 100°C, the
description of hydrogen diffusion over the weld thermal history from Tm to Ti can be
simplified. This way, ΣD∆t can be expressed using an approximative thermal factor of
hydrogen diffusion, ΣD∆t(100) that has been shown[13, 23–25] to approximate Eq. (2)
comparatively well. The advantage of this simple 'engineering' factor is that for its
numerical determination, only one parameter, namely, weld cooling time t is needed.
Why cooling in the low temperature regime becomes important for HR100 stems from the
fact that despite the decrease in the D values, the time available for hydrogen removal
 23

increases considerably. This, in turn, is a direct consequence of the decaying exponential

form of the post-weld cooling curve.
The use of ΣD∆t(100) in place of theoretical ΣD∆t integrated over the complete weld
thermal history is the possibility to calculate the ΣD∆t(100) factor within engineering
accuracy directly from the recorded weld thermal history using e.g. t100 data. For this,
three alternative empirical formulae have been proposed[13, 23, 25, 63, 77] (and validated) for
calculating the estimates of ΣD∆t(100) on the basis of the weld cooling data:
ΣD∆t(100) = (4.2 t200 + 2.73 t150 – 13) * 10–5 (3a)
ΣD∆t(100) = 6.87 * 10 *–4 2
e[0.412 (log t100) – 0.101 (log t100)] (3b)
ΣD∆t(100) = τ (TR = 100°C) = (0.76 t100 + 6.3 t200) * 10 –5
(3c)
where t200, t150 and t100 are the corresponding weld cooling times from peak temperature
to 200, 150 and 100°C, respectively.
The methodology based on HR100 being valid for single pass welds, the situation
becomes more complex when multipass welds are concerned. Due to a number of
repetitive thermal cycles of successive passes, part of the already solidified weld metal
becomes re-melted, as welding proceeds. This re-melting of weld metal, in conjunction
with the thermal effect elevating the temperature of those parts of the weld metal
remaining in the solid state, cause part of the hydrogen present in the previous passes to
become active again[3, 20–23]. There is good reason to believe that part of the
thermodynamically stabilised residual hydrogen of the previous passes also becomes
activated and may hence transform into diffusible state. Therefore, analysis of the
dynamic behaviour of diffusible hydrogen in welds is even more important.
Previous work have shown[3, 20–23, 100] that in multipass welds the critical local (final)
hydrogen concentration, HRmax, is governed by hydrogen diffusion and accumulation
throughout the welding until finishing of the final layer. Therefore, HRmax in the filling
runs where cracks usually locate, can differ not only from the initial weld diffusible
hydrogen content H0 according to ISO/IIW 3690[57], but also from the remaining weld
diffusible hydrogen content HR100max according to Eq. (1) and characteristic of single pass
welds.
The major practical difficulty is how to measure hydrogen distribution and a local
maximum concentration in multiple-pass welds. Many attempts have been
made[22,23,84,96,97] using several techniques, albeit only few have worked. Most methods
are considered to fail because they are capable of measuring only hydrogen that is
spontaneously evolved from welds. Since the final hydrogen distribution comprises also
trapped hydrogen that is thermodynamically stabilised, forced extraction techniques are
needed in order to determine a complete distribution of accumulated hydrogen[22].
Microsectioning of different weld regions at low temperature followed by hydrogen
extraction[20, 22, 96, 97] and spot fusion of a weld cross section by laser beam followed by
mass-spectrometric analysis of evaporated hydrogen[22, 84], are given as the most
promising techniques. However, since both of these methods are technically very
demanding and time consuming, they are not well suited to continual hydrogen control in
production welding fabrication.
The earlier works[3, 20–22, 97, 100] have demonstrated that in multipass welding the
hydrogen distribution across the weld thickness is non-uniform. Numerical FE and FD
 24

modelling, as well as experimental techniques, in Japan reveal accordingly that the local
hydrogen concentration HRmax in multipass welds increases in weldment thickness
direction towards the filling runs as a result of diffusion and accumulation of
hydrogen[19–23,100]. The results[20,100] demonstrate that in multiple-pass welds, HRmax
reaches its maximum value at the height of 0.75–0.90 of the plate thickness (from the
plate bottom), independently of the actual plate thickness, as shown in Fig. 2a.

Fig. 2a. Examples of the measured/calculated hydrogen distributions in through-thickness

direction of multipass weld[20].
 25

This agrees with the findings that the crack density increases with the hydrogen
concentration and towards the filling layers[76, 100, 106], as shown in Fig. 2b, as well as with
the observation that most transverse hydrogen cracks tend to locate predominantly
beneath the weld surface under the 2nd or 3rd last bead layer[3, 9, 12, 19–23, 76]. As this location
also coincides with the peak value of weld longitudinal residual stress in weld thickness
direction[19], it has been concluded that transverse cracking in multiple-pass weld metals
actually results from the critical combination of high local hydrogen concentration and
peak in weld longitudinal tensile residual stress[2,3,19,23]. Continuous hydrogen
accumulation caused by repetitive weld thermal cycling and weld bead overlapping
determines the local (final) hydrogen concentration HRmax[3, 20–22], which is the
predominant hydrogen factor in hydrogen cold-cracking of multipass weld metals.

Fig. 2b. Crack distribution expressed in terms of crack density (CD) through the 50 mm thick
FCAW multipass weld[76, 100].

It has been demonstrated[8, 9, 19–23] that in addition to the initial hydrogen content H0,
the local final hydrogen concentration HRmax in multipass welds depends on complete
weld thermal history, bead size and weld overlap due to penetration of successive passes.
As a consequence, the relationship between HR100max and H0 that characterises the
removal of hydrogen in single-pass welds and obeys Eq. (1), no longer prevails for
multipass welding conditions. Factors affecting HRmax in multiple-pass welding, as well
as methods for its determination, are dealt with in Section 1.5.3.
Furthermore, it has been suggested[37] that the difference in the martensite start
temperature MS between the weld metal and the parent steel HAZ influences the final
 26

hydrogen distribution in the weldment. This, in turn, affects the hydrogen concentration
and whether the cracks locate in the weld metal or in the HAZ. Since the MS factor has
both hydrogen and microstructure related characteristics, it is dealt separately with in
Section 1.5.5.

1.2.2 Microstructure

Whist the general effects of hydrogen and hydrogen diffusion on weld metal cracking are
relatively well established and understood, the same does not apply when it comes to the
influence of WM chemical composition and microstructure[54].
In the case of the HAZ cracking, it is usually sufficient to know the maximum HAZ
hardness, which is determined by the steel's chemical composition (mainly carbon) and
the weld t8/5 cooling time. Generally, the harder the microstructure, the greater the
cracking susceptibility[13, 24, 25, 28–30]. For the weld metal the situation is more complicated,
particularly in multipass welding. In general, the susceptibility of weld metal to hydrogen
cracking appears to increase as the strength increases[2, 3, 12, 54, 74, 76, 100, 106], that is, the
proportion of microstructures exhibiting greater hardenability, is increasing. Unlike for
the HAZ, there is no consensus about any unified ‘critical hardness’ in the case of weld
metal, although in the early works[56] one can find suggestions that WM cracking might
occur when its hardness exceeds that of the HAZ. It was also claimed[54] that WM
hardness has not proved to be a reliable indicator of weld metal cracking susceptibility,
even though correlations do exist among WM hardness, strength and its alloy content.
There exist also evidence[8, 9] that WM hydrogen cracking in multipass welds can occur in
microstructures that are not essentially martensitic.
Previous experience has shown[3, 54, 55] that hydrogen cracking in multipass weld metal
is often accompanied by an apparent high degree of plastic strain. Therefore, cracks tend
to locate in such microstructural regions where intense plastic straining accumulates. A
common feature is the presence of cracks in pro-eutectoid grain-boundary ferrite, GBF,
at prior-austenite grain boundaries and solidification cell boundaries[54]. Cracking will
generally occur in the filling layers and transverse to the welding direction, then growing
either normal to the weld surface or at an angle orientation in the weld thickness
direction. In the latter case, i.e., Chevron cracking, the high longitudinal tensile stress
causes slip bands to form at 45° angle in the weld thickness direction. Microscopic
observations have revealed[54, 98. 99] that concentration of plastic strain in the intergranular
GBF, coupled with transport of hydrogen into these regions, leads to crack initiation.
These staggered cracks then link up by a ductile shear mechanism assisted by
hydrogen[54], see Fig. 3. Microcracks are seen to often grow into several directions to
further link up to form transverse macrocracks[3b]. Apart from linking microcracks,
sometimes parts of the crack do not link, but form a "staircase" pattern.
Examinations on WM crack locations had revealed[3b] that transverse cracks in
multipass welds tend to appear in the as-welded dendritic microstructures, i.e., in the
regions that remained unaffected by the reheat of the following passes and therefore
usually contain proeutectoid GBF ferrite. Those parts of the Chevron crack locating in
 27

the GBF have been found[54] to often show quasi-cleavage fracture surface, whilst cracks
in other regions exhibit more ductile fracture appearance. Both intergranular and
transgranular cracking modes have been recognised, the former usually running through
the GBF phase[54]. Initiation of cracking in the as-deposited microstructures was clearly
associated with the GBF, the internal microstructure being apparently of secondary
importance[70]. Likewise, another study[121] reports that cracks appeared to be initiated
essentially from the weld metal areas having columnar grain structure and composed of
grain boundaries decorated with GBF, whilst equiaxed (reheated) grains provided better
resistance to cracking. In reheated WM regions, the crack appears to have formed first,
then widened as a result of re-heating due to successive passes[54]. A proposed[54, 98, 99]
mechanism of Chevron crack formation in multipass welds is shown in Fig. 3.

Fig. 3. Proposed mechanism of Chevron crack formation associated with dislocation shear
bands: (a) microcrack nucleation, (b) blunting and arrest, (c) crack formation within the
shear band, (d) macrocrack formation[54, 98, 99].

The harmful role of GBF phase is supported also by the recent findings[16–18]
demonstrating a positive correlation between the volume fraction of GBF and WM
hydrogen cracking susceptibility. The results implied that the cracking sensitivity of
multipass WMs may, at low hydrogen levels less than 4.5 ml/100 g DM (IIW),
essentially depend on the weld metal microstructure, particularly, the volume fraction of
GBF. At high hydrogen contents the conditions seem to change, so that the hardness and
tensile strength of weld metal will become increasingly important in controlling the
hydrogen cracking behaviour[16–18].
 28

With extra-high strength weld metals of Rp0.2 ≥ 690 MPa, hydrogen cracks have been
observed, accordingly, within regions of intense plastic strain[54, 55]. Instead of GBF,
cracking appeared at isolated locations in the weld metal and along slip bands parallel to
laths of martensite, M, as well as along slip bands across M laths[55]. Microscopy (SEM)
revealed fracture surface described as quasi-cleavage mode with several tiny shallow
dimples. The later stages of crack growth showed wavy slip bands, again characteristic of
intense plastic strain, and dimpled fracture surfaces. It was believed that the absence of
intergranular cracking in this case can be explained by the very low weld metal oxygen
level[54, 55], which is a characteristic feature of extra-high strength weld metals. Thus,
cracking was usually transgranular and only occurred along solidification cell boundaries
when cells were oriented normal to the strain. These were considered as initiation sites.
When cells were at an angle to the strain, a mixed mode fracture along and across cells
was observed.
In addition to weld metal strength, microphases and inclusion structure have been
reported[21, 22] to have importance in determining the sensitivity of WM to hydrogen
cracking. Microphases and inclusions affect the diffusion rate of hydrogen, available
trapping sites, crack initiation conditions, all of which are considered as controlling
factors in terms of WM cracking. As the morphology of inclusions affects the hydrogen
diffusivity and trapping, it follows, for instance, that different types of weld metals can,
in principle, possess different responses to cracking, even though they would exhibit
relatively similar strength and hydrogen content[8, 9, 16–18].
It is known[54] that weld metal chemical composition has an effect also on the level of
residual (retained) hydrogen. This could account in part for the higher susceptibility of
higher strength weld metals to hydrogen cold-cracking[54], especially in multipass
welding where part of retained hydrogen may become re-activated by the thermal effects
of successive welding runs.
The attempts on defining the effects of specific alloying elements on WM cracking
susceptibility have not been very revealing and do exist mainly for single-pass welds. A
study[64] based on G-BOP tests reports, Nb levels of 0.06% in the weld metal increased
the crack susceptibility significantly, whereas V (0.05%), Mo (0.2%) and Ni (1.2%) had
minimal effect, and Mn appeared to have a mildly beneficial effect. The crack
susceptibility was not related to the hardness and the effects of these alloying elements
could not be characterised by any existing carbon equivalent formula[64].
Reported findings on the relationship between WM hardness and its hydrogen
cracking susceptibility are somewhat conflicting and inconsistent. In terms of Implant
rupture stress and at a given level of hydrogen, the susceptibility of WM to hydrogen
cracking was found[71] to increase linearly with the hardness. As a general trend, this is
consistent also with the findings made elsewhere[2, 3], although results of the Implant test
are likely to apply to reheated weld metal, as the notch of the specimen inevitably locates
there. With regard to hydrogen embrittlement, the reheated weld metal has been
claimed[70] to behave in a similar manner to the parent steel HAZ. Indeed, an early
study[56] has already suggested that WM cracking might occur when its hardness exceeds
that of the HAZ. This, however, contrasts with the later findings[6, 7] demonstrating that in
multipass welds with a low carbon content of 0.06%, the ‘critical hardness’ above which
cracking tended to occur actually was different in the case of the weld metal and the
HAZ. Whilst approximately 300 HV was found[6, 7] ‘safe’ for the HAZ according to the y-
 29

groove Tekken test, cracking in the corresponding FCAW weld metals occurred, under
similar welding conditions, at hardness levels of around 250–280 HV. Similar findings
have been reported in another study[8] on SAW weldments revealing transverse hydrogen
cracks in the weld metals of hardness no more than about 200 HV. A recent study[121]
concluded that cracking in extra high-strength HSLA-100 welds was controlled by the
weld overall strength and the local variation of WM microstructure, but was little affected
by the microhardness variation.
As a consequence, the relationship between WM chemical composition,
microstructure and cracking susceptibility seems to be less straightforward than for the
HAZ. It is not only the hardness, but also the expression for the integrated hydrogen
diffusion parameter-hydrogen dependence, i.e., βτ + logHD, or, HR100/H0 = f(ΣD∆t), that
is found[72] to be a function of weld chemical composition. Whereas some studies have
shown reasonable fits using the conventional IIW carbon equivalent formula CEIIW[72],
others seem to rely more on the Pcm formula[54]. Furthermore, WM hardness and
microstructure are both influenced also by the number and nature of e.g. non-metallic
inclusions that are not accounted for in any common carbon equivalent formula.
From the standpoint of prediction method, it hence appears essential to investigate to
which extent hardness and strength are sufficient factors to describe the weld metal
cracking sensitivity. Any predictive system will become complicated once it has to
consider microstructural factors, such as weld metal oxygen and impurity content,
inclusion morphology, volume fraction of GBF and/or retained austenite / M-A
constituent.

1.2.3 Stresses

Stresses produced during the welding fabrication are either (i) residual welding stresses
due to internal restraint or (ii) reaction stresses caused by external restraint[65]. Of these,
the residual stresses are a result of the non-uniform temperature distribution. These
stresses are in a short range equilibrium in the vicinity of the weld. Reaction stresses, in
turn, are a consequence of the clamping effect, i.e., other structural members somehow
prevent the free thermal shrinkage[65]. These long range stresses are deleterious in that the
net stress is not reduced with cracking. Hydrogen cracking in welds is generally
associated with residual welding stresses[68].
Phase transformations have shown a remarkable beneficial effect on residual and/or
reaction stress development when the γ to α volume change is large and the γ–α phase
transformation takes place at a comparatively low temperature[66]. This stress reducing
effect, prominent in single-pass welding, is considered to attenuate in multipass welding
due to the repetitive thermal cycles, consequently the effect would prevail only when
welding the final weld run[67a].
Experience has shown[65], in single-pass welds the residual stresses play only a
secondary role in hydrogen cracking of butt joints, whilst the reaction stresses
predominate the risk of cracking. This does not, however, apply to multipass welds where
 30

the residual and reaction stresses form a complicated system whose accurate solutions
usually require thermal elastoplastic FEM analyses.
The effect of the geometry of a structural detail on residual stress (actually, reaction
stress) development in butt welds under (external) restraint is described[66, 73] using a
parameter named "restraint intensity", RF. It is simply defined as the force per unit of
weld length caused by the elastic change of a unit in root gap, uniform along a butt weld
joint. In practice, RF increases with the plate thickness h. Predominant stress component
for hydrogen cold cracking is the one perpendicular to the crack orientation[2, 3]. In the
case of HAZ underbead root cracking, what counts is the transverse net stress, σnet,
across the weld throat[13, 25, 63]. In the weld root these stresses can reach the critical level
and finally, as plate thickness and hence restraint intensity, RF, increase, even the level of
the weld metal yield strength, Re(WM)[13].
Whereas the transverse stresses are of decisive importance for the HAZ
underbead/root cracking in butt welds, the situation with regard to the cracking in filling
runs of multipass welds is just the reverse. Transverse hydrogen cold cracking in
multipass WMs is essentially controlled by the longitudinal tensile residual stress in the
welding direction, σresL, which is intensified in the presence of transverse notches or
imperfections, such as an unspliced backing bar or a gap[54]. Unlike longitudinal
hydrogen cracks, transverse cracks have been found to locate predominantly below the
final weld layer[76, 100] or in the 2nd or 3rd last weld layer[3, 12], i.e., in the filling runs in the
area where both longitudinal residual stress and hydrogen accumulation will concentrate
the most[2, 3, 12, 19, 74, 76, 100].
A characteristic finding with respect to multipass welds is that the longitudinal tensile
stress, while first remaining at a comparatively low level in the weld root, increases soon
towards the filling layers and reaches its maximum, i.e., the true yield strength of weld
metal[19, 23, 67b]. This can occur already after some 2 to 3 bead layers have been deposited,
irrespective of the value of RF[19, 23]. Any further increase in plate thickness therefore does
not increase the level of weld longitudinal tensile stress in the filling runs anymore[19], as
shown in Fig. 4.
As a result, intensity of restraint RF due to structural rigidity is considered[19, 23] to
contribute to the multipass WM cracking susceptibility to a lesser extent than is the case
with HAZ root/underbead cracking. As far as heavy plates are concerned, the longitudinal
residual stress in the filling layers of multipass weldment are expected practically
equivalent to the weld metal true yield strength, provided the weld is long enough –
preferably 300–500 mm or more[19, 23, 67b].
Apart from this, however, structural rigidity may influence the precise location of the
cracks. There are indications[19, 20, 74] that hydrogen cracks in highly rigid structures and
thick multipass welds may appear already in the mid-thickness of the weldment, whereas
in small-scale tests[3, 12] they generally tend to locate under the 2nd–3rd latest filling bead
layer closer to the weld surface.
 31

Fig. 4. Effect of number of welding passes on residual stresses in the toe of multipass weld[19].

It has been shown[23] that the critical hydrogen content Hcr for the occurrence of weld
metal cracking decreases exponentially as the weld longitudinal residual tensile stress
σresL elevates. This dependence obeys a decaying exponential function of a general
form[23]:
Hcr = A ∗ 10(–B ∗ σresL) (4)
where: A and B are constants.
As a consequence of Eq. (4), the use of higher strength weld metals yields a drastic
decrease in the critical hydrogen content capable of causing cracks. The increase in the
risk of WM hydrogen cracking with strength is thereby a result of (i) crack sensitive
microstructure, in conjunction with (ii) remarkably elevated stress state. This is in
agreement with the previous experimental findings[3, 12] that have documented transverse
hydrogen cracks in weld metals with no more than ≈ 4 ml/100 g DM (IIW) diffusible
hydrogen in the cases where WM yield strengths were around 700 MPa or more.
The means to overcome the greater cracking risk in high-strength weld metals may
differ between root pass and filling layer welding in the sense how to relax the stress rise
via consumable selection. When welding root runs, it is often acceptable to use filler
metals of undermatching strength, which results in desired reduction of the welding
residual stress in the highly stressed root area[13, 24, 25]. In contrast to root runs, however,
 32

the use of filler metals producing undermatching weld metal strength is often
unacceptable for filling runs and surface layers.
Besides on weld metal strength, the longitudinal residual stresses have been found[74]
to depend on the applied interpass temperature Ti during welding. In a previous work,
residual stress measurements with strain gauge method revealed that while the tensile
residual stresses reached the level of the WM yield strength under low Ti conditions
(< 50°C), increasing Ti (> 120°C) was accompanied by a reduction of σresL so that the
residual stresses remained less, about 75–85% of the WM yield strength[74].
The role of residual stress development and stress concentration aspects in multipass
welds from the design viewpoint are dealt with in more detail in Section 1.5.1.

1.2.4 Welding process

The results of single-pass cracking tests, such as the G-BOP test, generally
demonstrate[54, 72] that hydrogen cracking susceptibility is independent of welding
process. This is usually the case when for example plots of diffusion integrated
parameter, e.g., βτ, against simple composition-hydrogen formulae of the form:
A (CEIIW) + B (log HF) are made[72].
In the case of multipass weld cracking tests, the findings are not always as consistent.
There exist evidence[8, 9] that, other factors being equivalent, hydrogen cracking risk can
be greater for submerged-arc (SAW) weld metals than for shielded metal arc (SMAW) or
flux-cored arc (FCAW) welds. This has been attributed[8, 9, 19–21] to differences in bead
shape and size which, in turn, affect the hydrogen diffusion distances and, finally, the
local hydrogen concentration HRmax. Even with equivalent levels of heat input, different
welding methods tend to yield weld beads of different geometrical shape and size[8, 9].
Typically, SAW is known to produce comparatively narrow weld beads with high depth-
to-width ratio[8, 32, 33], whilst beads with less penetration and wider shape are characteristic
of FCAW[9].
The bead shape and size effects, especially from the standpoint of weld local hydrogen
concentration, are dealt with in more detail in Sections 1.5.2 and 1.5.3. The effects of
heat input on weld hydrogen content are discussed in Section 1.5.6.

1.3 Mechanisms of hydrogen induced and assisted cracking

Unlike many other damage mechanisms, such as e.g. creep failure at high temperatures,
or brittle cleavage fracture at low temperatures, hydrogen cracking is not associated with
any unified micromechanism[24, 25, 36, 42, 46, 50, 51, 54]. Besides in terms of fracture
micromechanisms, the initiation and propagation events of hydrogen cracking can be
investigated and elucidated via observations on those microscale processes that emerge at
the crack tip and fracture process zone under the presence of hydrogen concentration.
This approach puts the focus on surface effects, as well as on the hydrogen-dislocation
 33

interactions. Some of these mechanisms are common to both hydrogen-induced and

hydrogen-assisted cracking. According to this classification[75], the following three
mechanisms are generally defined: (i) hydrogen enhanced decohesion (HEDE), (ii)
hydrogen enhanced localised plasticity (HELP) and (iii) adsorption induced dislocation
emission (AIDE).
All these processes involve a complex combination of a number of events: (i)
hydrogen dissociation, adsorption and transport to the crack tip, (ii) dislocation emission
and egress, (iii) hydrogen diffusion and movement of vacancies, (iv) transport of
hydrogen to dislocations, (v) hydrogen effects on dislocation mobility and hydrogen
trapping at particle-matrix interfaces[75]. Thus, controversies regarding the features how
to distinguish between these three mechanisms, are not surprising. It has been
postulated[75] that the appearance of a certain mechanism may depend also on the
operating temperature in such a way that AIDE and HEDE mechanisms would dominate
at lower temperatures, whereas HELP occurs predominantly at higher temperatures.

1.3.1 Hydrogen enhanced decohesion

Classical hydrogen enhanced decohesion (HEDE) was first discovered by Troiano in the
1960's, then subsequently re-discovered by Oriani in the 1970's and Gerberich et al. in the
1980's. This mechanism involves weakening of the interatomic bonds in the atomically
sharp crack tip and the surrounding process zone as a consequence of absorbed
hydrogen[75]. Decreases in electron charge density between metal atoms occurs due to (i)
absorbed hydrogen at the atomically sharp crack tip, followed by diffusion and presence
of hydrogen in (ii) interstitial sites and (iii) at the particle-matrix interfaces. These
phenomena then result in final decohesion in terms of tensile separation of atoms[75].
Unlike the other two mechanisms, hydrogen enhanced decohesion (HEDE) is
characterised by limited dislocation activity[75]. In the cases where HEDE predominates,
fractographic investigation can reveal atomically flat fracture surfaces, which is
considered a demonstration of negligible plasticity involved in this mechanism. These
findings are consistent with the observations that HEDE is generally the predominant
mechanism for brittle-like intergranular fracture[75].
Earlier work[7, 34] on hydrogen cracking in weldments demonstrates that intergranular
fracture becomes a predominant form as the weld diffusible hydrogen content increases
and/or segregation of impurities into prior-austenite grain boundaries is intensified. The
latter occurs typically when welding conventional steel using comparatively high heat
inputs. Accordingly, proportion of intergranular fracture is found[39] to diminish at the
expense of transgranular fracture in the case of modern, low-impurity steels and weld
metals.
 34

1.3.2 Hydrogen enhanced localised plasticity

Hydrogen enhanced localised plasticity (HELP) was first introduced by Beachem in 1972
and re-introduced by Birnbaum et al. in the 1980's. This mechanism is characterised by a
decrease in repulsive interactions between dislocations and obstacles due to
reconfiguration of their hydrogen atmospheres. Hydrogen accumulates and concentrates
around the voids over a certain volume ahead of the crack tip. As hydrogen concentration
becomes highly localised, so is also the case with plasticity. This mechanism only
operates in certain temperature and strain rate regime where hydrogen atmospheres can
keep up with dislocations[75].
Direct evidence for HELP being a characteristic embrittlement mechanism has been
reported[75], for example, from in-situ TEM observations. When fracture appearances in
thin foils strained in a vacuum have been compared to those strained in hydrogen
atmosphere, the presence of hydrogen was found to result in an increase in dislocation
activity and facilitate fracture by localised shear mechanism. Localised microvoid
coalescence coupled with enhanced dislocation activity in terms of extensive slip on
planes intersecting cracks have been observed[75] directly by TEM when examining thin
foils.
A distinct difference between HEDE and HELP is therefore the increased dislocation
activity in the latter mechanism. Another difference is the amount of plasticity involved in
the fracture process; whilst atomically flat fracture surfaces were associate with HEDE,
observations of dimpled fracture surfaces in the regions of highly localised strains have
been reported in the case of HELP. The latter findings are regarded as characteristic of
ductile crack growth under the presence of substantial external stress which then transfers
to internal localised strain. Owing to its nature, HELP is considered to contribute
especially to slip-band (induced) fractures and hence promote fracture by localised
microvoid coalescence[75].
Earlier work[6, 7, 53] has demonstrated that especially Implant specimens extracted from
the weld metals can exhibit substantial plasticity in the hydrogen-induced fracture
process. This usually manifests itself as a ductile fracture surface with large plastically
deformed areas despite the presence of hydrogen, which totally contrasts to the typical
brittle fracture surface characteristic to the hydrogen embrittled HAZ Implant specimens.

1.3.3 Adsorption induced dislocation emission

Adsorption induced dislocation emission (AIDE) has been proposed by Lynch at 1977.
This mechanism encompasses similarities to both HEDE and HELP and hence
incorporates some features from already existing theories and models.
In accordance with HEDE, the mechanism involves weakening of the interatomic
bonds at the crack tip as a result of absorbed hydrogen. Crack surfaces (and sub-surfaces)
are recognised to act as strong trapping sites for hydrogen. The presence of hydrogen at
these surfaces leads to weakening of interatomic bonds (presumably metal-metal bonds,
while the hydrogen-metal bonds are inherently weak) at crack tips, which further
 35

enhances emission of dislocations. Dislocation emission, in turn, promotes and facilitates

linking of cracks with voids ahead of the cracks and over a characteristic distance, ∆a.
This linking is recognised to occur along some low-index planes or grain boundaries,
leaving shallow dimples on fracture surfaces as the fracture propagates further[75].
The decisive role of dislocation emission in the AIDE mechanism is, in turn, similar to
HELP, except that strains can be even more localised than those for microvoid
coalescence associated with HELP. As evidence supporting AIDE and the dislocation
emission theory, this mechanism has been recognised[75] also under conditions where
there was no time for any significant hydrogen diffusion ahead of cracks. Thus, it is not a
prerequisite of AIDE for hydrogen atmospheres to keep up with dislocations, such as was
the case with HELP[75].
In general, dislocations emit from the plastic zone ahead of the cracks, egressing at
crack-tip surfaces. The influence of embrittling environment, such as absorbed hydrogen,
is seen[75] particularly in that dislocations tend to emit predominantly from crack tips,
which then results in crack advance and crack opening. Consequently, increasing
proportion of dislocation emission from the crack tips manifests itself as greater crack
growth ∆a for a given crack opening displacement, i.e., COD. Accordingly, the strains
required for fracture should lessen due to the embrittling effect of absorbed hydrogen at
the crack tip. It has been reported[75] that in the case of cleavage or quasi-cleavage
fracture, cracking occurs on {100} planes in <110> directions. This suggests that slip on
planes intersecting the crack tips plays a critical role in the AIDE mechanism.
The characteristic features of AIDE that can be identified using electron microscopy
are considered[75] as: (i) dimpled fracture surfaces due to high localised strains, (ii)
extensive slip on planes intersecting cracks, and (iii) formation of localised microvoid
coalescence coupled with intensified dislocation activity. In comparison to HEDE or
HELP, the importance of crack surface effects in hydrogen cracking is regarded[75] more
pronounced in AIDE and is, hence, given more attention. The findings supporting the role
of surface effects in AIDE are given[75] as: (i) a high hydrogen concentration on and just
beneath the crack tip surface, (ii) substantial effect of hydrogen adsorption on atomic
bonding, (iii) cracking at very high velocities and (iv) abrupt ductile-to-brittle transitions
on e.g. changing of temperature[75].
Owing to its complex nature, AIDE seems capable of contributing to cleavage like
fracture, as well as to dimple formation associated with ductile fracture.
According to AIDE, degree of hydrogen embrittlement should increase with (i)
increasing surface coverage of hydrogen on crack surfaces, (ii) increasing filling of sub-
surface sites with hydrogen and (iii) increase in hydrogen absorption at internal crack or
void tips[75]. Quantitative predictions are often problematic, because for instance the
difficulties in modelling dislocation-emission process.
 36

1.4 Hydrogen diffusion and its role in cracking of multiple-pass

weld metals
Earlier work[2, 3, 21–23] have shown that hydrogen cold cracking in multiple-pass weld
metals occurs when accumulating hydrogen exceeds a certain critical level, Hcr. This
hydrogen accumulation via diffusion is thermally activated, continuing process that takes
place throughout the welding until finishing of the final bead. For the analysis of the
dynamic behaviour of diffusible hydrogen, diffusion needs to be described in terms of the
mass diffusion equation, with the introduction of the activity coefficient of the diffusing
mass into the diffusion equation[21, 22]. This enables diffusion and accumulation of
hydrogen to be determined under conditions where the properties of the diffusion
medium vary locally and under defined distributions of stresses and strains.
Diffusion of mass in a uniform medium is governed by the Fick's second law that
relates the rate of composition with time to the concentration profile. A steel weld cannot,
however, be regarded as uniform medium for hydrogen diffusion, since the weldment
consists of regions composed of quite different microstructures. Whereas the HAZ of
ferritic steel comprises mostly hardened, bainitic-martensitic transformation products
with relatively high dislocation density, weld and base metals are often less hardened in
nature, consisting of bainitic or ferritic microstructures[22]. Besides dislocation density,
the number and morphology of inclusions and microphases acting as traps for diffusible
hydrogen, are also different among the HAZ and weld metal[21, 22]. Hence, hydrogen
diffusion is influenced by the diffusive media of a non-uniform nature, where hydrogen
diffusion due to temperature dominates over mass diffusion[79].
A weld root with an acute notch is subjected to triaxial stresses (in the as-welded
condition) that also cause lattice expansion. Hydrogen diffusion in welds is therefore not
only affected by the gradient of microstructures and dislocation densities, but also the
fields of stress and strain in the weld[22]. As, for instance, plastic strain and microstructure
influence the diffusion coefficient of hydrogen, these interactions need to be
evaluated[21, 22].
The equations for the analysis of dynamic behaviour of hydrogen in multipass welds
are given.

1.4.1 General equation for the diffusion of mass

The diffusion of mass in a uniform medium is governed by the Fick's second law[22, 78]
that relates the rate of composition with time to the concentration profile c(x), as:
(∂c/∂t) = D(∂2c/∂x2) = D ∇2c (5)
where c is the concentration of the diffusing mass, t is the time and D is the diffusion
coefficient.
Hydrogen diffusion in steel welds has been analytically solved mainly using Eq. (5)
under the given initial condition and boundary conditions[22]. For these analyses,
hydrogen diffusion in a butt and fillet weld is often approximated with diffusion in an
infinite plate and an infinite cylindrical rod, respectively[40].
 37

The general equation of mass diffusion in a heterogeneous medium is defined[21, 22] as:
(∂c / ∂t) = – ∇ J (6)
where c is the concentration of the diffusing mass, t is time and J is the mass flux. The
flux of the diffusing mass J can further be defined in terms of a chemical potential of
diffusing mass µ, as[21, 22]:
J = – [D ∗ c /(RT)] ∇ µ (7)
where D, R and T are the diffusion coefficient, gas constant and the absolute temperature,
respectively. The chemical potential µ applied in thermodynamics is defined as[21, 78]:
µ = [∂G' / (∂n)] = G0 + RT ln(x) = µ0 + RT ln (a) (8)
a=γc (9)
where G' is the Gibbs free energy in a system, dn is the change in the quantity of atoms
of a substance in a system, a is the activity of a diffusing mass, γ is the activity
coefficient of the mass, and c is the concentration of the diffusing mass.
Combining Eqs (6), (7), (8) and (9), the general equation of mass diffusion
becomes[21]:
(∂c / ∂t) = ∇ ∗ {[(D / γ) ∇ (γc)] + [(Dc / T) (ln (γc)) ∇ T]} (10)
In a welded joint at its cooling stage, mass transfer at a certain local point is assessed
at a certain moment of time. Thus, the change in temperature over an increment of time
can be considered very small in relation to the absolute temperature. Under these
conditions, it can be assumed that the temperature gradient is very small, i.e., ∇ T ≈ 0).
This approximation allows for Eq. (10) to be simplified even further, as[21]:
(∂c / ∂t) = ∇ ∗ [(D / γ) ∇ (γc)] (11a)
[21]
or, alternatively, as :
(∂c / ∂t) = ∇ ∗ (D ∇ c) + ∇ ∗ [(Dc / γ) ∇ γ] (11b)
If it is assumed that the diffusion coefficient D and the generalised activity coefficient
γ are independent of the location, as is the case with a homogeneous diffusion medium, it
follows[22] that: ∇ D = 0 and ∇ γ = 0. Considering these, and rewriting Eq. (11a) in terms
of the activity of the diffusing mass a according to Eq. (9), leads to the following
form[21]:
(∂a / ∂t) = (a / γ) ∗ (∂γ / ∂t) + γ ∇ ∗ [(D / γ) ∇ a] (12)
At a certain moment of time t, any change in the temperature T leads to corresponding
change in the concentration of the diffusing mass c and, hence, also its activity a.
Assuming, however, that the activity coefficient γ is itself independent of time t, it
follows that: ∂γ / ∂t = 0. Consequently, Eq. (12) reduces simply into the following
form[21]:
(∂a / ∂t) = γ ∇ ∗ [(D / γ) ∇ a] (13)
 38

This equation expresses the mass diffusion in a system in terms of the activity gradient
of the diffusing mass ∇a, in conjunction with the activity coefficient of the mass γ and the
diffusion coefficient D. Of these, the diffusion and activity coefficients are strongly
dependent on the temperature, and microstructure & stress, respectively, as discussed
later in Section 1.4.3.

1.4.2 Effects of temperature and continuity conditions

Unlike general diffusion, hydrogen diffusion is influenced by the diffusive media of a

non-uniform nature[79]. Therefore, Fick's second law in the form of Eq. (5) cannot
satisfactorily describe hydrogen diffusion in welds. Owing to its chemical nature,
hydrogen diffusion due to the abrupt temperature changes clearly dominates over mass
diffusion[79]. Consequently, the local accumulation of hydrogen and delayed phenomena
cannot be solved satisfactorily by conducting an analysis of hydrogen diffusion in a
homogeneous medium in accordance with Eq. (5)[22].
On this basis, Eq. (5) can be modified to account for the effect of temperature T on the
diffusion of hydrogen. Thereby, we obtain[79]:
(∂T / ∂t) = κ (∂2T / ∂x2) (14)
κ = λ / ρC (15)
where κ is the thermal diffusion coefficient, λ is thermal conductivity and ρC is the
volume of specific heat of the material. For calculations, values of 0.14 (cm2/sec), 0.12
(cal/cm sec°C), 8.0 (g/cm3) and 0.10 (cal/g) have been proposed[79] for κ, λ, ρ and C,
respectively.
Combining Eqs (14) and (15), the general equation of diffusion due to temperature
becomes:
ρC (∂T / ∂t) = λ (∂2T / ∂x2) (16)
In practice, it is particularly the gradient of potential µ that governs the diffusion of
hydrogen at any medium[79]. The total chemical composition in a system remains always
the same, only the temperature T and the local concentrations ci do change[79]. Thus, the
continuity condition over the A–B boundary between the two separate concentrations, c1
and c2, can be expressed in terms of the chemical potential of hydrogen µ, as[79]:
µA = µB => µ0 + RT ln (γαc1) = µ0 + RT ln (γβc2) (17a)
It thereby follows that:
γαc1 = γβc2 => γα = (c2 / c1) γβ => γβ = (c1 / c2) γα (17b)
If one now assumes that for example atom B is very active in solid α, which applies
e.g., to hydrogen diffused in solid α iron, the activity coefficient of diluted solute can be
much higher than unity. The gradient of its chemical potential, ∇µ, c.f. Eq. (7), thereby
essentially determines the mass diffusion of hydrogen.
 39

1.4.3 Activity coefficient of the diffusing mass

The flux of the mass J that diffuses in any system under the conditions of locally varying
pressure P and chemical potential µ', can be expressed, in accordance with Eq. (7), as[21]:
J = – [D ∗ c / (RT)] ∇ (–PV* + µ') (18)
*
where V is the partial molar volume of the diffusing mass (i.e., hydrogen), and P is the
hydrostatic pressure considered as a mean of negative stresses in the triaxial directions in
solids, as[22]:
P = – (σx + σy + σz) / 3 (19)
The chemical potential µ that is normally applied in thermodynamics [78]
, is given by
Eqs (8) and (9) and can be re-written as:
µ = µ0 + RT ln (a) = µ0 + RT ln (γmc) (20)
where γm is the activity coefficient determined by the type of a given microstructure and
the amount of plastic strain[22].
Comparing Eqs (7) and (18) gives the following relation that incorporates the effects
of locally varying pressure and chemical potential into the flux of the diffusing mass.
This relation is defined[22, 80] by introducing a new term: the generalised potential of
diffusing mass, Φ, being:
Φ = PV* + µ (21)
Combining Eqs (20) and (21), the equation for the generalised potential Φ becomes:
Φ = µ0 + RT ln (γmc) + PV* = µ0 + RT ln (γmc) + RT ln [exp(PV*/ RT)] (22a)
Let us now set the last term of Eq. (22a) as:
exp (PV*/ RT) = γP (22b)
Consequently, re-combining Eqs (22a) and (22b), the final equation of the generalised
potential Φ can be expressed as[22]:
Φ = µ0 + RT ln (γm ∗ γP ∗ c) (23)
Comparing Eqs (23) and (20), the activity coefficient can be written in the form of a
generalised activity coefficient γ of the diffusing mass defined as[22]:
γ = γm ∗ γP (24)
where γm and γP are the interactive activity coefficients due to microstructure (and plastic
strain) and stress, respectively[21, 22], c.f., Section 1.4.6.
When the activity coefficient γ in a system is locally decreasing, the potential or
activity of the diffusing mass, a, becomes also reduced[22] in accordance with Eq. (9). As
a consequence, a kind of potential difference, or "well", is formed. The diffusing mass
thereby diffuses gradually into this potential well, resulting in local accumulation of the
mass[22]. The triaxial stress effect[81, 82], c.f. Eqs (19) and (21) further enhances the
accumulation of hydrogen.
 40

1.4.4 Potential of hydrogen in structural steel weld metal

The activity coefficient γ and the diffusion coefficient D of hydrogen are both expected to
vary sharply in welds, as the microstructures of the HAZ and the weld metal experience
abrupt changes due to the temperature changes caused by successive thermal cycles[21, 22].
In addition, the stresses and plastic strains will vary greatly, especially near those
weldment areas where the stresses concentrate, such as at the weld root or weld toe. Thus,
the potential for hydrogen, µ, must be accounted for at any such boundaries where
discontinuity exists[21].
The theories[21, 22, 79, 80] defining the hydrogen potential µ usually assume that the
hydrogen responsible for cold cracking, i.e., the diffusible (reversible) hydrogen is in a
temporarily trapped state (apart from residual hydrogen) and that no chemical reactions
occur when lattice-diffusing hydrogen diffuses into a trapping site. Under these
assumptions, two conditions can be defined[21] that must prevail at the α–β boundary
between a lattice site and a trapping site: (i) continuity condition of the hydrogen
potential µ meaning that the variation of µ across the boundary must be smooth, and (ii)
conservation condition of the mass flux of hydrogen meaning that the mass of hydrogen J
is conserved at the boundary.
According to the continuity condition of the hydrogen potential µ, it follows that[21]:
µα = µβ (25)
where subscripts α and β denote to the lattice site and the trapping site of the
boundary, respectively.
Combining Eq. (25) with Eqs (8) and (9), the equation for the continuity condition
becomes[21]:
γα c α = γβ c β (26)
In addition, the condition that the hydrogen potential is smoothly continuous across the
α-β boundary gives the following relation[21]:
∇µα = ∇µβ (27)
According to the condition of hydrogen conservation at the α-β boundary, it follows
that[21]:
Jα = Jβ (28)
where J is the mass flux of hydrogen, and subscripts α and β denote to the lattice site and
the trapping site of the boundary, respectively.
Combining Eqs (28) and (7), the equation for the conservation condition becomes[21]:
[(Dα cα) / (RT)] ∇ µα = [(Dβ cβ) / (RT)] ∇ µβ (29)
[21]
Substituting Eq. (27) into Eq. (29), the following relation can be written :
Dα cα = Dβ cβ (30a)
Combining Eqs (30a) and (26) allows an alternative expression to Eq. (30a) to be
written as[21]:
 41

Dα / γ α = Dβ / γβ (30b)
Under the conditions defined by Eqs (26), (27) and (29), the diffusion coefficient of
hydrogen in the trapping site of the α–β boundary, Dβ, depends on the ratio of the local
equilibrium concentration (cα/cβ) or, alternatively, the inverse ratio of the activity
coefficient [(1/γα) / (1/γβ)], at each site of the α–β boundary. These relations have been
verified[21] by demonstrating a satisfactory correspondence between the estimates of an
apparent diffusion coefficient predicted according to the equations and the experimental
results. Specifically, this correspondence has also involved[83] the temperature and plastic
deformation dependencies of the diffusion coefficient of hydrogen in iron and steel.
Apart from the conditions defined by Eqs (25)–(30), an alternative approach[84]
determines the hydrogen potential Φ on the basis of the hydrogen concentration c
according to Sievert's law, as:
Φ = (PH2 / P0)0.5 = c / k (31)
where PH2 is the partial pressure of hydrogen, P0 is the total pressure, c is the
concentration of hydrogen and k is the hydrogen solubility in steel calculated as a
function of temperature T and obeying Eq. (32) of the form:
log k = 1.677 – 1420 / T (32)
Consequently, applying Eqs (31) and (32), the hydrogen diffusion in a weldment can
be solved by the variation calculations using the hydrogen potential Φ [22, 84].

1.4.5 Determination of the diffusion coefficient of hydrogen

For the analytic (or numerical) solutions of hydrogen diffusion, it is necessary to

determine the diffusion coefficient of hydrogen, DH, in steel as a function of
temperature[21, 22]. A number of reports[21, 40, 59, 84–88] have been published on the
determination of apparent hydrogen diffusion coefficients applying techniques such as
gas evaporation and electrolytic penetration. Selection of an appropriate DH value is a
complicated problem, since published values show wide scatter amounting to even three
orders of magnitude in ferritic steels at room temperature, as shown in Fig. 5 for
example[40]. The scatter in apparent DH values has been attributed[21, 22, 25, 40, 59] to factors
such as microstructure, composition, reversible hydrogen traps and surface conditions of
a specimen used for the determination. Reviews[108] considering the scatter bands for
hydrogen diffusion coefficients have been published recently.
Generally, different analytic equations are given for the calculation of the apparent
hydrogen diffusion coefficient DH at different temperature regions[21, 22, 85–87]. At higher
temperatures of about 200°C and above, DH is shown[21, 22, 40] to lie quite close to the
coefficient for lattice diffusion, Dα, whereas DH at lower temperatures is significantly
lower and shifted towards that for the trapped hydrogen as the temperature decreases[22].
 42

Fig. 5. Variation of apparent hydrogen diffusivity with temperature in steels[40].

According to the literature[21, 22, 85–87], the following equations have been proposed for
the determination of the hydrogen diffusion coefficient DH at a given temperature range θ
(°C):
DH = 1.51 ∗ 10–2 ∗ exp[–11970 / (RT)] ; θ ≥ 500°C (33a)
Dα = 1.4 ∗ 10 ∗ exp[–3200 / (RT)]
–3
; 200 ≤ θ < 500°C (33b)
DH = 0.12 ∗ exp[–7820 / (RT)] ; θ < 200°C (33c)
2
where DH (or Dα) is the apparent diffusion coefficient for hydrogen in steel (cm /sec),
R is a gas constant (= 2 cal/mol ≈ 8.4 J/mol), T is the temperature (K) and θ is the
temperature (°C).
Sometimes, hydrogen diffusion equations are used as an alternative to hydrogen
potential based approach[22]. For instance, diffusion equations can be formulated[88] as
sub-equations according to the degree of trapping, i.e., (i) hydrogen of lattice diffusion,
 43

(ii) hydrogen trapped in voids, and (iii) hydrogen trapped permanently or irreversible
hydrogen. To perform these kind of analyses, one does not need to introduce hydrogen
potential to derive solutions for hydrogen diffusion in a heterogeneous medium[22].
Another approach[22, 84] to calculate hydrogen diffusion and concentration in welds
directly without the use of the hydrogen potential, is to derive microstructure-dependent
hydrogen diffusion coefficients for each weld region. For the temperature region from 20
to 200°C, Eqs (34a)–(34c) have been proposed[84] for parent steel, the HAZ and weld
metal, respectively:
DH = 0.89 ∗ exp[–8856 / (RT)] ; parent steel (34a)
DH = 0.31 ∗ exp[–7990 / (RT)] ; HAZ (34b)
DH = 1.22 ∗10 ∗ exp[–5200 / (RT)]
–2
; weld metal (34c)
where DH, R and T are as for Eqs 33a–33c.
This way, the apparent diffusion coefficients can be varied in their analysis depending
on the particular local region of the weldment[22].
The consistency of the resulting hydrogen concentrations and the respective locations
in a weld are dealt with in Section 1.4.7. Anyway, it can already be seen that differences
do exist between the DH estimates according to Eq. (33c) and Eqs (34a)–(34c), for
instance. Nevertheless, the DH values according to Eqs (34a)–(34c) seem to decrease as
the locus shifts from the parent steel to the HAZ and weld metal, as expected.

1.4.6 Determination of the interactive and generalised activity

coefficients

Experience[21, 22, 89–92] on the effects of plastic deformation and alloying elements on
hydrogen occlusion have demonstrated that introduction of interactive activity
coefficients enables to incorporate the effects of plastic strain and microstructure, and
stress, into the analysis of hydrogen diffusion and accumulation.
Previous work[89, 90] that compare the recorded hydrogen occlusion with the hydrogen
solubility in annealed pure α iron, allows for the determination of the interactive activity
coefficient with respect to strain and microstructure, γm. This, coupled with results[21, 90]
on plastically deformed steel leads to the derivation of the interactive activity coefficient
γm of the form[21]:
(1 / γm) = 1 + 0.001228 ∗ vpore ∗ exp[ 6733 / (RT)] (35)
where γm is the interactive activity coefficient due to microstructure and strain, R is the
gas constant (= 2 cal/mol ≈ 8.4 J/mol), T is the temperature (K), and vpore is a parameter
for microvoid volume (%) relating to the available trapping sites for lattice hydrogen[91].
Values of vpore have been determined[21] for iron, steel and weld metals applying cold-
strained tensile bar specimens, some of which have been heat-treated to obtain
microstructures characteristic to the HAZ. The diffusion and interactive activity
coefficients of hydrogen, DH, γm, have been determined experimentally from these bar
 44

specimens by using gas-chromatography measurements of transient effusion of hydrogen

with details described elsewhere[21]. Values of γm have then been calculated[21] applying
Eq. (30b) that compares the measured diffusion coefficient Dβ with that of unstrained and
pure α-iron, Dα. Values of vpore, in turn, have been determined[21] by substituting the
calculated values of γm into Eq. (35).
The interactive activity coefficient with respect to hydrostatic pressure or compressive
triaxial stress, γP, is generally determined[21, 22] as a function of pressure, P, gas constant
R (= 2 cal/mol ≈ 8.4 J/mol), temperature T (K) and the partial molar volume of hydrogen,
V*, in accordance with Eq. (22b). The partial molar volume of hydrogen, V*, represents
the extent of the effect of pressure on hydrogen activity. Values of V* = 2 ml/mole have
been reported[92] according to the hydrogen penetration experiments for steel and iron
under tensile stress.
Table 1 shows an example of the database[21] of collated values of DH and vpore for pure
α iron, structural steel in different cold-strained and heat-treated conditions (simulating
the HAZ microstructures), and weld metal.
Table 1. Hydrogen occlusion and diffusion coefficient DH in steel at 20°C [21].
Steel specimen & condition Total effusion Diffusion vpore
coefficient DH
(ml/100 g) (cm2/sec) (%)
–
Pure α iron 5.74 ∗ 10–6 0
Structural steel SM50B
– as-delivered 0.24 1.8 ∗ 10–6 0.0169
– 2% strained 0.26 7.6 ∗ 10–7 0.0506
– 4% strained 0.43 5.6 ∗ 10–7 0.0715
– 6% strained 0.50 4.4 ∗ 10–7 0.0932
– 7% strained 0.78 4.0 ∗ 10–7 0.1032
– restraint-free, heat-treated*) – 4.2 ∗ 10–7 0.0982
– restrained & heat treated**) – 3.1 ∗ 10–7 0.1350

Weld metal 0.22 1.6 ∗ 10–7 0.0213

*) [21]
: used for Vpore when considering only microstructural change in the analysis
**)
: used for Vpore when both microstructural change and strain distribution in the weld are
considered[21].
An alternative approach[22] to analyse the hydrogen diffusion is to apply a generalised
activity coefficient, γg, instead of interactive activity coefficients. The generalised activity
coefficient is defined as[22]:
γg = {exp[PVH*/ (RT)]} / {1 + 0.001228 ∗ vpore ∗ exp[6733 / (RT)]} (36a)
vpore (%) = v0 + 0.0123 ∗ εp (36b)
where εp is the equivalent plastic strain and v0 is a microstructure dependent parameter
associated with a particular region of the weld. Values of v0 = 0.0169, 0.0982 and
0.0213% have been suggested[22] for the parent steel, the HAZ and the weld metal,
respectively.
 45

Distributions of the equivalent plastic strain εp of single- and multiple-pass weldments

have been calculated[21], for instance, by a finite element method (FEM). As the hydrogen
trapping effects are already incorporated into the generalised activity coefficient γg
according to Eq. (36a), the corresponding diffusion coefficient of hydrogen used in
numerical analyses should be that for the lattice diffusion, Dα [22].

1.4.7 Numerical analysis of hydrogen diffusion and accumulation

effects

Numerical analysis enables solutions of hydrogen diffusion in welds having complex

geometries and under the conditions where the diffusion coefficient of hydrogen varies
depending on time, temperature and location[22]. For the numerical modelling of hydrogen
diffusion, a two-dimensional (2D) finite difference method (FDM) is often applied[93–95],
with hydrogen diffusion in a butt weld being approximated with diffusion in an infinite
plate[40]. The 2D FDM analysis that uses a partial differential equation obeying the law of
mass conservation is applicable for assessing hydrogen diffusion in both single and
multiple-pass welds. For the analysis, Eq. (12) that gives the activity of the diffusing
mass, is formulated and converted into a finite difference equation[21, 22]. This equation
expresses the mass diffusion in a system in terms of the activity gradient of the diffusing
mass ∇a, in conjunction with the activity coefficient of the mass γ and the diffusion
coefficient of hydrogen D, with incorporated temperature dependencies.
Traditionally, numerical analyses are mostly concerned with single-pass welds and
root runs. Accomplished FDM analyses[93–95] have shown, for instance, that hydrogen
evolves more rapidly from a pipe girth weld than from the root beads of Lehigh or
Tekken cold-cracking tests, simply because the root surface exposed to the atmosphere is
wider in a pipe girth weld than in the cracking test welds[93]. Another FDM analysis on
the distribution of hydrogen concentration in a multipass weld has revealed[94] that
hydrogen concentrates beneath the final weld layer. Moreover, the maximum values of
the local hydrogen concentration calculated by FDM were found consistent with the
results of gas chromatographic measurements of hydrogen evolved from specimens
sectioned at various thicknesses of a weld[94]. Accordingly, the calculations and
experiments both demonstrated that this maximum value in multipass weld is close to the
weld hydrogen content determined using a quenching method for a single-pass weld[94, 95].
Regarding hydrogen accumulation, a steel weld cannot be considered as a uniform
medium for hydrogen diffusion, since the weld comprises a microstructural gradient with
differences in dislocation densities and inclusions acting as traps[22]. Any geometrical
discontinuities, such as weld root in a single-bevel groove, also cause stress concentration
and/or the development of triaxial stresses[22]. For the analysis of hydrogen accumulation
effects, additional factors therefore need to be encountered, namely, (i) the differences in
microstructures in different weld regions, (ii) plastic strain distributions, as well as (iii)
stresses and stress concentrations due to groove geometry and notch effects[21, 22]. A
current practice thereby is that the stress and strain distributions, and sometimes also the
heat conduction calculations, are solved by FEM[21, 22], in conjunction with the analysis of
 46

the hydrogen diffusion patterns applying FDM. Alternatively, one can conduct a
complete FEM analysis of all the relevant factors.
The documented FDM and/or FEM analyses[20–22, 100] providing quantitative data on
hydrogen accumulation effects are mostly concerned with single-pass welds and root
runs, only. The effects of (i) hydrogen diffusion, (ii) activity coefficients of hydrogen in
strained and/or heat-treated steel and (iii) the stress and strain fields being incorporated,
the results show that diffusible hydrogen accumulates at the weld root and the HAZ[21].
The transient hydrogen concentration at the root reaches its maximum some time after
completion of welding[22]. Among microstructure, plastic strain and hydrostatic stress, the
most influential factor on hydrogen accumulation was found the microstructural change
associated with the γ–α phase transformation and the gradient of microstructures between
parent steel, the HAZ and the weld metal, followed by the effects of equivalent strain and
hydrostatic stress[21]. Accounting for all these factors, the maximum hydrogen
accumulation was recorded[21] nearly five times as high as that resulting from hydrogen
diffusion due to a concentration gradient, alone.
With respect to the HAZ in the middle of the weld throat depth, no stress nor strain
concentration exist there, consequently, hydrogen accumulation is influenced essentially
by the weld thermal cycle and the resulting microstructural changes[21]. This, in turn,
manifests itself as a prolongation of time required for hydrogen diffusion. Whilst
hydrogen concentration at the weld root was found to reach its maximum within 2 hours
after welding[21, 84], the corresponding maximum in the hydrogen concentration in the
middle of the throat depth was recognised nearly 5–7 hours after welding[21]. As far as the
absence of stress concentration is concerned, these conditions apply also to multiple-pass
weld metal. Consequently, hydrogen cracking in the filling layers of multipass weld
metals is expected to require longer times to occur, than is the case with root/underbead
cracking in the HAZ of single-pass welds.
In the case of single-run welds, applying preheating has been shown[13, 21, 25, 63, 77] very
effective in reducing the maximum level of hydrogen accumulation. For example,
preheating a weld up to 200°C, the maximum content of hydrogen accumulated at the
weld root can become reduced to about one fifth of that attained without preheating[21].
This is essentially a consequence of substantial prolongation in the weld cooling time at
low temperatures, e.g., t100, due to preheating, with a subsequent increase in the thermal
parameter of hydrogen diffusion, ΣD∆t, resulting in a remarkable reduction of the
HR100/H0 ratio in accordance with Eq. (1)[13, 25, 63, 77].
With regard to multiple pass welding, the advantages of elevated preheat/interpass
temperatures in reducing the concentration of accumulated hydrogen are documented to a
lesser extent[8, 20], and are sometimes conflicting[8]. For instance, a Japanese work[20]
reports a relatively small decrease in the HRmax/H0 ratio with increasing preheat/interpass
temperature θi, compared to that recorded for single-pass welds[13, 25, 63, 77], see Fig. 6.
This is particularly true with higher heat inputs (and lesser beads), presumably because of
accentuated weld bead penetration, overlapping and greater diffusion distances for
hydrogen[20]. This contrasts to the single-pass welds that, from the viewpoint of hydrogen
effusion, actually benefit from elevated heat inputs, c.f., Eq. (1).
The effects of preheating on hydrogen concentration in single and multiple-pass welds
are dealt with in more detail in Section 1.5.3.
 47

Fig. 6. The ratio of the final maximum hydrogen concentration to the initial hydrogen
content, Hmax/H0, as a function of preheat/interpass temperature θi and for two different weld
heat inputs and number of bead layers l – multipass welding[20].

1.5 Specific design factors affecting transverse hydrogen cracking in

multipass weld metal

Earlier work[2–9, 12, 19–23, 26, 37] on hydrogen cracking in multipass weld metals have
identified (i) factors whose influence differs from that traditionally experienced with
HAZ root/underbead cracking in single-pass welds, as well as (ii) factors which needed
not to be taken into account at all when assessing the HAZ cracking, but appear to
become relevant for the weld metal cracking situations. These factors have been
found[3, 19–23] to affect hydrogen accumulation towards the filling runs, as welding
proceeds. Hydrogen accumulation, in turn, determines the local (final) hydrogen
concentration HRmax in welds, which is the predominant hydrogen factor in hydrogen
cold cracking of multipass weld metals.
The specific factors are: (i) the absence of stress concentration due to geometric
discontinuities inside the weld metal, (ii) the influence of individual bead layer thickness
hw and weld build-up thickness aw on hydrogen diffusion and accumulation, (iii) the
influence of weld bead penetration and the resulting weld bead overlapping d/hw on
hydrogen accumulation and HRmax in the filling layers, and (iv) the influence of interpass
time ti and interpass temperature Ti on HRmax in welds. Additionally, the role of (v) the
γ-α phase transformation and martensite start (Ms) temperature, as well as of (vi) welding
heat input Q are given.
 48

1.5.1 Welding residual stresses and stress concentration

Both numerical modelling[19–22] and direct measurements[20, 96, 97] using small sections
extracted from actual welds have proven, in multipass welding the local hydrogen
concentration, HRmax, increases in weldment thickness direction towards the filling
runs as a result of diffusion and accumulation of hydrogen. Earlier work in
Japan[19–23, 94, 95, 100], for example, clearly demonstrate that the local hydrogen
concentration in multipass welds always tends to reach its maximum at the height of
0.75–0.90 of plate thickness from the plate bottom, irrespective of the actual plate
thickness, c.f. Fig. 2. This coincides with the observation[3, 12, 20] that transverse hydrogen
cracks tend to locate mostly beneath the weld surface below the 1st or 2nd last layer, where
also the weld longitudinal residual tensile stress has been shown[19] to reach its maximum
value, c.f. Fig. 4. This has led to the conclusion that transverse hydrogen cracking in
multipass WM is a direct consequence of the combination of two essential factors: (i) a
sufficiently high local (final) hydrogen concentration HRmax and (ii) the maximum peak
in the weld longitudinal tensile residual stress σresL.
Experiments on multiple-pass welds carried out by WES[19] and NSC[3, 12, 20–23] have
demonstrated that the weld longitudinal tensile stress, while being first reduced in the
weld root as subsequent bead layers are being deposited, rise subsequently towards the
filling layers as the number of subsequent passes increase and will finally reach its
maximum, i.e., the level of weld metal true yield strength. The maximum of σresL under
the final layer has been found to remain almost unchanged soon after about 2 bead layers
have been deposited[19], as shown in Figs 4 and 7. The prerequisite to the development of
the maximum σresL value is that the weld in a specimen is long enough, preferably at least
300–500 mm or more[19, 67]. Being the weld/specimen too small, the results, in terms of
residual stress build up, cannot be expected to be descriptive of the real welded
structure[67].
To attain the maximum peak in hydrogen accumulation, in turn, about 4 to 5 layers are
generally required[19, 20], as shown in Fig. 9. As far as the longitudinal stress is concerned,
and provided a crack-sensitive microstructure and sufficient amount of hydrogen are
present, transverse WM hydrogen cracking is principally possible in a welded joint with
no more than 2 bead layers.
There exist evidence[13, 24, 25, 66, 67b] that in single-pass welds, low temperature phase
transformations may cause reduced residual stresses in localised regions at welds in low
alloy steels, such as e.g. quenched & tempered extra-high strength Ni-Cr-Mo grades. This
effect is accentuated further by applying sufficiently high preheating[67b]. In the case of
multipass welds, however, this only concerns the final weld bead, since the stresses in
previous bead layers have been shown to become elevated due to cycling to temperatures
below the phase transformation temperature during subsequent weld passes[67b]. This
thermal cycling of successive beads gives a "saw-tooth" stress distribution. With respect
to the residual stress, the beneficial effects of low phase transformation temperature and
preheat in low alloy steel welds hence only apply to the final weld bead[67b].
 49

Fig. 7. Through-thickness distribution of residual stresses in multipass weld metals[19].

 50

It is known[13, 21, 22, 25, 37] that diffusion of hydrogen in the weld root region is promoted
and intensified by the tri-axial stress state due to geometrical discontinuity and the
resulting stress concentration. This kind of stress-concentration is very effective in
elevating the HAZ root cracking sensitivity of single-bevel groove welds, for
instance[19, 21, 22]. Unlike the HAZ root cracking in single-pass welds, there is no such
stress concentration arising from e.g. geometrical effects that could possibly act as local
stress raiser inside the filling beads of multipass weld metal. Because of the absence of
stress concentration, such cracking risk assessment methods that are based on the stress
field or groove severity parameters[13, 25], the former being similar to K in fracture
mechanics, have been traditionally considered difficult to apply.
As there is no apparent stress concentration, the accumulation of hydrogen in the weld
metal is primarily governed by the weld thermal cycle and the associated microstructural
change resulting from the γ-α phase transformation[21, 22]. Because of this, hydrogen
cracking in the WMs may sometimes appear after substantial time passed after the
completion of welding. The time of crack appearance is often 3 to 4 times longer than
what is encountered with the HAZ root/underbead cracking[21, 22]. This may explain why
unexpected WM hydrogen cracks have sometimes been recognised[76, 117] after several
weeks (sometimes nearly 5 weeks) of the completion of welding.
As it comes to predictive systems assessing weld metal cracking risk, one has to
content with estimates of overall level of weld longitudinal residual stress, in place of
parameters such as the stress field/intensity parameter ΣIcr, or the groove/specimen
severity parameter, FI[2, 3, 13, 25]. Thus, the effect of residual stress on the cracking risk of
multipass weld metals is usually encountered using causal factors, i.e., critical threshold
level of stress and applying regression analyses[2, 3], instead of sophisticated local
approach-based numerical modelling capable of determining intrinsic local stresses and
distributions at and near a crack tip[38].

1.5.2 Plate thickness/weld build-up thickness and bead layer thickness

Earlier work[19–22, 96, 97] have consistently demonstrated that the final local hydrogen
concentration HRmax in multipass welds is governed by hydrogen diffusion and
accumulation towards the filling runs throughout the welding until finishing of the final
filling layer. Therefore, the HRmax in the filling runs can differ from both the remaining
weld diffusible hydrogen content HR100max characteristic of single-pass welds, and the
initial hydrogen content H0 determined using a single-pass test according to IIW/ISO
3690[57]. It has been shown[8, 9, 19–23] that besides H0, HRmax depends on weld thermal
history and weld bead geometry and size, i.e., on (i) the complete weld thermal cycle that
determines the hydrogen injection, accumulation and the conditions of subsequent
hydrogen effusion, (ii) weld bead size affecting the diffusion distances of hydrogen, as
well as on (iii) weld bead overlapping due to penetration of successive passes, which
influences the volume of re-activated hydrogen available for subsequent accumulation.
The weld thermal cycle, weld bead size and weld bead overlap are defined[19–23] in
terms of the thermal factor of hydrogen diffusion, ΣD∆t, an individual bead layer
 51

thickness, hw and weld bead overlap ratio, d/hw, respectively. Of these, the overlapping
effects are dealt with in Section 1.5.3. The thermal and geometrical conditions for
hydrogen diffusion can be incorporated by describing them using a single parameter, the
integrated diffusion-distance parameter, PD–D, introduced[19, 20] according to Eq. (37) as:
PD–D = ΣD∆t / hw2 (37)
where hw is the thickness of an individual weld bead layer (cm) and ΣD∆t is the thermal
factor of hydrogen diffusion (cm2).
Theoretically, ΣD∆t incorporates the diffusion coefficient of hydrogen, D, at a certain
temperature and the time interval ∆t to which the corresponding D value applies[19, 21–23].
Numerical solutions for ΣD∆t are determined by integrating over the complete weld
thermal history, i.e., from Tm to Ti acc. to Eqs (2a) and (2b). For analytic calculations,
ΣD∆t is usually written in an approximative form according to Eqs (3a)–(3c)[19, 21–23].
The effect of weld bead size on hydrogen diffusion is encountered[19, 20] via individual
bead layer thickness, hw. In Eq. (37), hw2 is merely proportional to the square of the
diffusion distance from the middle of the bead to its outer surfaces. Alternatively, hw/2
can be regarded as the distance that needs to be covered by the diffusing hydrogen. It has
been shown[19–23] that the approximation of hydrogen diffusion through an ‘element
surface area’, as one may picture the term hw2 to mean, by applying the square of the
individual bead layer thickness, provides a realistic description of the effect of bead size
on the diffusion of hydrogen under the thermal cycle of welding. The ΣD∆t/hw2 ratio as
defined in Eq. (37), is thereby indicative of the loss of hydrogen from the element in the
time available.
According to Eq. (37), the shorter the weld thermal cycle (i.e., fast cooling rate, short
interpass time) and/or the greater the bead size, the smaller the corresponding PD–D value,
meaning that hydrogen diffusion becomes more difficult. Thus, the smaller the PD–D, the
greater is the final hydrogen concentration HRmax and hence the cracking risk, as shown
in Figs 8 and 10. The thermal conditions for hydrogen diffusion are essentially enhanced
with increasing the weld cooling time towards low temperatures, e.g., t100, that is, by
applying higher thermal inputs that yield an increase in ΣD∆t in accordance with Eqs
(3a)–(3c). An increase in bead size hw, in turn, leads to greater diffusion distances, which
further manifests itself as a reduction of the integrated diffusion-distance parameter PD–D
throughout the thermal input range.
It has been proven[19–23] that the maximum value of the remaining weld diffusible
hydrogen content in the case of both single- and multiple-pass welds, denoted as HR100max
and HRmax, respectively, depends essentially on diffusion conditions of hydrogen as
defined by Eq. (37). This is comprehensively demonstrated in Fig. 8 that shows the ratio
between the maximum remaining weld diffusible hydrogen and the weld initial hydrogen,
HR100max/H0, plotted against the hydrogen diffusion conditions, i.e., the integrated
diffusion-distance parameter, ΣD∆t/hw2. Primary data in Fig. 8 originate from Osaka
University (OU) and Nippon Steel Corp. (NSC), Japan[20]; these data do not account for
overlapping effects. Assuming that no weld overlapping occurs, i.e., d = 0, HR100max
corresponds to Hmax(d=0).
 52

Fig. 8. The HR100max/H0 ratio as a function of the integrated diffusion-distance parameter

ΣD∆t/hw2 – overlap not included. Primary data according to Osaka University and NSC[20],
Nos 1, 3, 5 and 7: separate trials.

According to the data[20] in Fig. 8 and looking one pass at a time, Hmax(d=0) (that
corresponds to HR100max when assuming no weld overlap, i.e., d = 0) remains high and
almost at the level of H0 as long as ΣD∆t/hw2 is comparatively small, then decreasing as
ΣD∆t/hw2 starts to increase. The data in Fig. 8 suggests that the relationship between the
HR100max/H0 ratio and the ΣD∆t/hw2 can be expressed using a decaying sigmoidal function
incorporating positive first degree and negative second-degree polynomic terms. A
general form of this expression is derived according to the primary data in Fig. 8, as:
PD–D
)]} + C ∗ P
D–D – D ∗ (PD–D)
2
HR100max(d=0) /H0 = {e[–A PD–D / (B (38)
where A, B, C and D are constants and PD–D is the integrated diffusion-distance parameter
according to Eq. (37).
According to Fig. 9, with any given value of PD–D the local hydrogen concentration
Hmax (that corresponds to HR100max when assuming no weld overlap, i.e., d = 0) seems to
reach a constant, maximum level after a certain number of weld bead layers. As expected,
under conditions allowing only limited diffusion of hydrogen (i.e., PD–D values < 0.1) this
maximum level is reached sooner, after about 2 to 4 layers[19, 20], whereas under
favourable diffusion conditions (i.e., PD–D values > 0.5), not until some 5 to 6 layers were
deposited[19, 20]. This, in turn, coincides with the build-up of weld longitudinal residual
stress σresL in multipass welds, c.f., Fig. 4.
The results of various types of multipass weld metal cracking tests[2, 3, 8, 9, 12, 76], as well
as those of numerical modelling[20–23, 100] are consistently demonstrating that the
occurrence of weld metal cracking appears to increase with the local maximum hydrogen
concentration Hmax. According to the modelling data[19–23], the influence of weld thermal
history and bead size on diffusion and accumulation of hydrogen are satisfactorily
described applying Eqs (37) and (38).
 53

Fig. 9. Hmax(d=0) /H0 ratio as a function of the number of bead layers l (l = aw/hw), at different
levels of the diffusion-distance parameter ΣD∆t/hw2 [20] – overlap not included (d = 0).

Whilst the increase in Hmax or hw has been shown to affect adversely on cracking
occurrence in multipass weld metals, the role of plate thickness h or weld build-up
thickness aw is less clear. Whilst some studies[2, 20] reveal practically no influence of plate
thickness on required preheat/interpass temperature Tcr, others[3, 12, 76] show a significant
trend of increasing Tcr as thickness increases, until the effect is usually offset above 50–
60 mm. In general, the detrimental effects of increasing h and aw have been attributed[3]
either to elevated stress level or an increase in hydrogen accumulation, or both.
On the other hand, contrasting findings showing practically no effect of either h or aw
on weld metal cracking susceptibility have been documented[2]. In their experiments on
extra-high strength SAW weld metals of RM ≈ 755–1080 MPa, Okuda et al.[2] found the
influence of weld build-up thickness aw quite minimal and hence insignificant in the
thickness range of 20 to 80 mm. Accordingly, a Japanese study[20] on SMAW weld
metals reports, at the RM ≈ 800 MPa strength level a HRmax value of 3.5 ppm was high
enough to cause cracking in a wide thickness range from 25 to 150 mm. This implies,
being a given level of weld metal strength (and hardenability) high enough to cause
hydrogen cracking, the critical HRmax level appears to be fairly constant regardless of
plate thickness[20].
Contrary to this, Yurioka et al. reported[3, 12] an approximately logarithmic dependence
of the hydrogen cracking susceptibility on weld build-up thickness aw in the thickness
range from 20 to 50 mm, above which the effect gradually starts to level off as thickness
increases further. Their study was concerned with SMAW and GMAW weld metals of
RM ≈ 590–885 MPa. In line with Yurioka's work, an Australian study[76] on high-strength
basic and rutile FCAW weld metals of RM ≈ 820–855 MPa reports that, other factors
being constant, the occurrence of weld metal cracking in terms of number and density of
cracks progressively decreased as plate/weld build-up thickness was reduced from 50 to
30 mm. Finally, no cracking was found in the 20 mm thick welded section[76].
 54

A recent German work[11] suggests that the 'limit thickness', above which further
increase in thickness does not lead to a subsequent rise in the required preheat
temperature, is not constant but varies depending on the weld hydrogen content. With
low weld hydrogen contents of 2–4 ml/100 g DM (IIW), the effect of thickness was
recognised to offset not until above 65–70 mm, whereas with higher weld hydrogen
contents of 8–10 ml/100 g DM (IIW), the effect of thickness seemed to become levelled
off already above 45 mm. The experimental data of NSC[3, 12] is recognised to fall within
these ranges of hydrogen and thickness.
As a result, presently there is no consensus about the effect of plate or weld build-up
thickness on the hydrogen cracking susceptibility of multipass weld metals. Different
views exist[2, 3, 11, 12, 76] on whether h or aw should be considered as a relevant parameter
into the predictive systems or not, and if so, what is the most appropriate mathematical
description of their influence. Moreover, a possibility that plate or weld build-up
thickness affects the hydrogen cracking risk differently, depending e.g., on the applied
groove geometry of a test specimen, thermal history of welding, weld metal strength and
conditions for hydrogen effusion and escape, must be taken into account.

1.5.3 Weld penetration and bead overlap – effects on final hydrogen

concentration

Besides the thermal factor of hydrogen diffusion, ΣD∆t, and bead size in terms of
individual bead layer thickness hw, weld bead penetration and the resulting bead
overlapping have been shown[19, 20] to affect hydrogen accumulation and hence the final
local hydrogen concentration HRmax in multipass welds. This effect, characteristic to
multiple-pass welds, is described[19, 20] by weld bead overlapping, d, or weld overlap
ratio, d/hw. Numerical modelling have shown[20] that increasing bead overlap leads to
intensified hydrogen accumulation and hence elevated HRmax concentration.
As a result, the final local hydrogen concentration HRmax, here denoted as HRmax(d>0)
accounting for weld bead overlapping (i.e., d > 0) in multipass weld metals, depends
essentially on (i) weld thermal cycle described using the thermal factor of diffusion,
ΣD∆t, (ii) individual bead layer thickness hw and (iii) bead penetration in terms of weld
overlap ratio d/hw, as shown in Fig. 10.
A direct consequence of overlapping is that the conditions of hydrogen accumulation
will inevitably become different between single-pass and multiple-pass welds. In single-
pass welds, the remaining maximum diffusible hydrogen content HR100max is always
lower than the initial diffusible hydrogen content H0[13, 25, 59, 63, 69]. Because of a decaying
exponential dependence[13, 25, 63] of the HR100max/H0 ratio on ΣD∆t, as defined in Eq. (1a),
HR100max decreases progressively as soon as ΣD∆t starts to increase as a result of slowed
cooling e.g., due to preheating. With prolonged cooling times in terms of e.g., t100,
HR100max can reduce down to values being remarkably less than H0[101]. This is
particularly the reason why preheating is so effective in reducing the root/underbead
hydrogen cracking risk in the HAZ of single-pass welds[13, 25, 101].
 55

Fig. 10. Effect of the overlap ratio, d/hw, on the maximum final hydrogen concentration-
maximum remaining diffusible hydrogen content -ratio, HRmax(d>0)/HRmax(d=0), at different
levels of the integrated diffusion-distance parameter, ΣD∆t/hw2 [20] – overlap included (d > 0).

In multiple-pass welding, in turn, HRmax(d=0) corresponding to the maximum remaining

diffusible hydrogen content after the first pass being laid, has been shown[19, 20] to
subsequently increase from the initial level of HR100max towards the filling layers, as
welding proceeds. This is explained[19–23] by additional supply of hydrogen released from
previously solidified passes as they are re-melted by subsequent pass as a result of weld
penetration and bead overlapping. Released hydrogen is then transported further towards
the filling layers due to enhanced diffusion caused by the repetitive thermal cycles of
successive passes. Being accumulated into filling layers, this additional supply of
hydrogen may compensate some of the hydrogen loss due to effusion from the weld
during its cooling and interpass intervals.
A Japanese work[19, 20] have demonstrated that in the case where overlapping in terms
of the d/hw ratio is substantial and the conditions for diffusion of hydrogen are favourable
(i.e., at high PD–D values), the local hydrogen concentration in the filling runs, HRmax(d>0),
may be elevated nearly 3 times higher than in the absence of any overlap, HRmax(d=0), as
shown in Fig. 10. To obtain a realistic description of the HRmax/H0 ratio in the case of
multipass welds, the effect of weld overlap in terms of the weld bead overlap ratio, d/hw,
must therefore be included into the analysis. This requires HRmax(d=0) to be replaced by
HRmax(d>0) in the applied calculation formulae.
 56

Fig. 10 shows the effect of d/hw ratio on the HRmax(d>0) /HRmax(d=0) ratio at different
values of the integrated diffusion-distance parameter of hydrogen, ΣD∆t/hw2. The
primary data in Fig. 10 is based on numerical modelling and weld-sectioning
experiments, having its origin in the earlier work[20] made at Osaka University and
Nippon Steel Corp., Japan. The HRmax(d>0)/HRmax(d=0) ratio in Fig. 10 expresses the local
final hydrogen concentration under the influence of weld bead overlapping, in relation to
the hydrogen content unaffected by any overlap. Here, HRmax(d=0) denotes to hydrogen
present in a weld before any overlapping occurs, which is the situation e.g. in the first
pass of multiple-pass weldment. Thus, with respect to hydrogen level and other welding
procedural factors being equivalent, HRmax(d=0) in a first pass of a multiple-pass weld can
be regarded to correspond to HR100max of a single-pass weld.
For the application of the diagrams in Fig. 10 to the assessment of hydrogen
accumulation, the effects of weld thermal cycle, bead size and overlap on HRmax are
expressed in terms of ΣD∆t, hw2 and d/hw, respectively. Under these premises, the data[20]
in Fig. 10 can be applied to describe the overlap effect using a relationship between the
HRmax(d>0) /HRmax(d=0) ratio and the d/hw ratio at a given level of ΣD∆t/hw2.
According to Fig. 10, a general form[20] of the overlap equation thereby becomes:
HRmax(d>0) / HRmax(d=0) = 1 + A (d/hw)m (39)
where A and m are constants that depend on the particular ΣD∆ t/hw2 value in question.
How steeply the HRmax(d>0) /HRmax(d=0) ratio rises with greater d/hw values, is seen to
depend essentially on ΣD∆t/hw2 in question. Fig. 10 shows that at high ΣD∆t/hw2 values,
the behaviour of HRmax(d>0) no longer obeys power-law dependence, but becomes nearly a
linear function of the d/hw ratio. For the cases where ΣD∆t/hw2 >> 0.8, constants A and m
in Eq. (39) can hence be fixed as: A = 2 and m = 1.
In the case of single-pass welding, enhanced diffusion conditions of hydrogen in terms
of greater ΣD∆t are known[3b, 13, 25, 59, 63, 69] unambiguously beneficial in allowing for
greater hydrogen effusion and escape as the weld cools down. Under the conditions of
increasing weld bead overlap d/hw and multipass welding, as presented in Fig. 10,
intensive hydrogen diffusion in terms of greater ΣD∆t/hw2 appears[20] to turn detrimental
in that it allows for a rapid increase of HRmax(d>0) in the filling runs. According to Fig. 10,
small overlap in terms of d/hw only yields a 10–20% increase in the HRmax(d>0), thus
nearly corresponding to situation of no overlap. Contrary to this, high d/hw ratios may,
under favourable diffusion conditions, cause remarkably high local hydrogen
concentrations of nearly 3 times higher than under conditions of no overlap[20].
Whether HRmax(d>0) towards the filling runs can, under excessive overlap and
intensified diffusion conditions, become even greater than the initial diffusible hydrogen
content H0 according to a single pass hydrogen test, has not been demonstrated
experimentally. According to the modelling data[20] in Fig. 10, HRmax(d>0) values
exceeding H0 seem possible, in theory, under conditions comprising small bead layer
thickness hw and provided that excessive weld bead penetration, d, takes place[20]. The
theoretical and experimental work at Leeds University[96, 97] and TWI[31], however, give
no reason to believe that H0 could be exceeded in multipass welding under any
circumstances. Presently, no experimental data exists[31, 96, 97] to verify that HRmax(d>0)
actually could ever exceed H0, even though HRmax(d>0) can well exceed the level of
 57

remaining weld diffusible hydrogen content HR100max characteristic to single-pass

welds[19, 20, 100].
In practice, inclusion of the overlap effect in terms of d/hw into predictive systems
requires some kind of metallographic examination. This can be done either from a
polished macrosection of the actual weld or from a cross-section of a separate bead-on-
plate trial weld macrosection according to Fig. 11 and welded with the same parameters
to be later used in actual production welding.

Fig. 11. A simple approximative methodology for determining the d/hw ratio from a bead-on-
plate test weld: d/hw ≈ B/A [20].

Ignoring the effect of overlap d/hw is likely to raise a possibility of underestimating the
actual local hydrogen concentration HRmax(d>0) in the filling passes of multiple-pass weld
metals. Even with equivalent levels of heat input, different welding methods tend to yield
weld beads of different shape. Typically, SAW produces narrow weld beads with quite
high depth-to-width ratios[8, 32, 33], whilst beads with less penetration and wider shape are
characteristic with FCAW[9]. Welding procedures that yield excessive penetration and
overlapping may exhibit hydrogen cracking sensitivity greater than that concluded on the
basis of single-pass cracking tests. In such cases, need of preheating may be
underestimated, which results in potential danger of cracking in actual welding
fabrication.
At present, there is no unified predictive system that would account for the effect of
weld bead penetration and the resulting bead overlap on final (local) hydrogen
concentration.

1.5.4 Interpass time and temperature

Because accumulation of hydrogen in multipass weld metals is a continuous, diffusion-

controlled process, interpass temperature Ti and interpass time ti are considered more
important as controlling parameters, than in the case of the HAZ root/underbead cracking
in single-pass welding. However, detailed experimental data on the combined effects of
Ti and ti in affecting WM hydrogen cracking in multipass welds are not available in the
literature.
Multipass V-Groove cracking tests conducted previously at TWI[8, 9] implied that
whilst higher Ti was needed as the weld diffusible hydrogen content and arc energy
 58

increased, cracking can be prevented by raising the interpass temperature Ti also

whenever the total welding time and/or the interpass time ti remained too short[8, 9]. This
was particularly the case with greater plate thickness. In practice, ti can remain too short
when welding e.g., short tack welds, at a high welding speed, or in the cases where weld
cooling is deliberately intensified to shorten the total interpass time. The latter technique
is sometimes used to compensate excessively long welding times resulting from the use
of higher arc energies.
Consequently, other welding procedural factors being constant, longer interpass time
in terms of ti and Σti, or higher interpass temperature, are expected to lessen the risk of
cracking in multipass weld metals, whereas shortening these time factors seems to induce
an opposite effect[8, 9].
In practice welding fabrication, welding is usually not continued until a maximum
allowed Ti has been reached. Thus, the higher the arc energy, the higher the interpass
time ti. Those TWI experiments[8] that employed intensified cooling to maintain constant
interpass temperature as arc energy was raised, had therefore presumably overemphasised
the severity of the applied test, as well as the effect of higher arc energies apparently
increasing the cracking risk, in relation to the practice welding conditions.
Presently, there is no generally accepted expression for the effect of interpass time on
the cracking risk in multiple-pass weld metals. Therefore, it is deemed necessary to
somehow include a parameter describing the effect of interpass time into predictive
systems[8, 102].

1.5.5 MS temperature and hydrogen distribution

Traditionally, martensite start temperature MS has been used as an indicator of the alloy's
hardenability and austenite stability. For the higher strength steels, MS can also indicate a
transition during cooling between the slow and fast diffusion rates of hydrogen. In
general, the higher the MS, the larger will be the temperature range available for rapid
hydrogen transport in the ferrite or martensite phase[37]. Because of the significant
differences in the hydrogen diffusion coefficient and solubility in austenite, γ, and
ferrite/martensite, α, the temperature at which the steel transforms from austenite to
ferrite or martensite will affect the degree of hydrogen transport. Thus, the MS
temperature is both a measure of (i) the microstructure evolution and (ii) the ability to
have available a phase (i.e., ferrite or martensite) for rapid hydrogen transport[37].
With respect to weldments, there exists evidence[37] that the difference in the MS
temperature between the weld metal and the parent steel HAZ influences the hydrogen
distribution in multipass welds. This further affects the final hydrogen concentration and
whether cracks will finally locate in the weld metal or in the HAZ.
During welding, the concentration of diffusible hydrogen in the weld/HAZ interface
increases as a function of time. The driving force for the hydrogen accumulation and the
resulting non-uniform distribution of hydrogen is the thermal diffusion process activated
by the successive overlapping thermal cycles due to multiple-pass welding and assisted
by localised stress gradients in the weldment due to tri-axial stress state or geometrical
 59

discontinuities[21, 22, 37]. During weld cooling, the HAZ that has just experienced the γ–α
phase transformation is capable of transporting hydrogen a significant distance into a
parent metal due to high diffusivity of hydrogen in ferrite. The hydrogen transport
cannot, however, proceed extensively, until the weld metal also transforms to ferrite. As
austenitic, weld metal stores high hydrogen contents (due to high solubility of hydrogen
in austenite) but can not move it fast enough to the fusion boundary due to slow diffusion
rate of hydrogen in austenite[37].
One can determine the MS temperature either experimentally by dilatometric
measurements, or numerically using some of the existing formulae[37, 104] in the literature.
Although there are formulae derived for both parent steels and weld metals[104], most of
them suffer the shortcoming of disregarding the weld cooling rate, which makes their use
somewhat approximative for welds.
Experimental measurements have shown[37] that when martensite transformation in the
HAZ occurs at a higher temperature than in the WM, the resulting hydrogen
concentration in WM is always higher than in the cases where the weld metal transforms
before the HAZ. It follows, being the MS temperature for the parent plate HAZ higher
than that for the WM, excess hydrogen concentration may result in the WM instead of
hydrogen diffusing into the HAZ, as shown in Fig. 12. This situation can promote WM
hydrogen cracking. Its is known that current development in steel processing techniques
have enabled high strength parent plate to be manufactured with considerably leaner
alloying than is the case with the WMs. Thus, the way the MS temperatures tend to differ
between the parent plate and the weld metal is expected to favour cracking in WM, in
particular.

Fig. 12. Hydrogen distribution across the weldment fusion boundary for MS(Weld) > MS(HAZ)
and MS(Weld) < MS(HAZ)[37].
 60

The recent experimental findings[26] based on WIC cracking tests and using different
hydrogen levels have shown a relationship between the MS temperature of the weld metal
and its hydrogen cracking sensitivity. In this study, extra-high strength welds of RM ≈
890–1000 MPa were examined[26]. At a given hydrogen level, there existed a ‘critical MS
temperature’ below which cracking was systematically recorded[26, 37]. The higher the
hydrogen content, the higher the critical MS temperature, see Fig. 13.

Fig. 13. Diagram of hydrogen cracking/non-cracking zones as a function of weld hydrogen

content and weld metal MS temperature[37].

The results[37] imply that weld metal MS temperature, or the difference in MS between
WM and parent steel HAZ, ∆MS, could serve as a kind of composition based hydrogen-
cracking index. In many cases in practice, however, WM and HAZ microstructures are
not essentially martensitic, which leads to thinking that temperatures such as Ar3 or BS
might be more relevant temperatures than MS. Such systematic comparisons of the
different temperatures were, however, not found in the published literature. An MS
expression for weld metal, which incorporates the influence of the WM oxygen content,
has been suggested[37] as a further improvement to better describe the effect of oxide
inclusion on austenite decomposition kinetics.

1.5.6 Welding heat input

In order to reduce the risk of hydrogen cracking in the HAZ, increasing the heat input Q
is unambiguously beneficial. This is ascribed to a decrease in weld cooling rate and hence
 61

the hardenability effect, which reduces the volume fraction of martensitic transformation
products, thus manifesting itself as lower HAZ hardness[6, 7, 13, 24, 25, 27–30, 34, 52, 53, 101].
In the case of multipass weld metals, there exist evidence that the effect of heat input
on cracking can appear differently from the common experience for HAZ cracking.
Experiments on CMn SAW weld metals of Rp0.2 ≈ 355–450 MPa, for instance, have
demonstrated[8] an increased risk of weld metal cracking with increasing arc energy in the
range 1–5 kJ/mm. Thus, raising the heat input can actually worsen the conditions for
cracking in multipass welds, which is the reverse of common experience for HAZ
cracking. This was attributed to greater diffusion distances as bead size in terms of hw
increases, which further results in elevation in the local final hydrogen concentration
HRmax, more hydrogen being retained by larger weld beads[8]. Another consequence of
elevated heat input is an increased weld penetration and weld bead overlap in terms of
the d/hw ratio, which intensifies the hydrogen accumulation towards the filling runs[19, 20].
Contradictory findings have also been reported. A previous work[53] on extra-high
strength (Rp0.2 ≈ 700 MPa) multipass SAW welds found no difference in the weld metal
cracking appearance between the 2.0 and 4.5 kJ/mm heat input welds. Regardless of heat
input, the required preheat, about 180°C, was dictated by cracking sensitivity of the weld
metal instead of the HAZ. Another set of experiments[5] made for extra-high strength
(Rp0.2 ≈ 745 MPa) multipass SMAW weld metals failed to reveal any effect of arc energy
on the cracking occurrence within the 1.5–5.0 kJ/ mm range. This was even though the
bead size in terms of both ab and hw was found to increase with the arc energy. Since
cracking did not appear in any of these weld metals, the lower than expected cracking
sensitivity may have also prevented the appearance of arc energy effects, even if such
existed. It is noteworthy that no intensified weld cooling was applied in any of these
experiments, consequently, with higher arc energies the interpass time also tended to
become higher, hence presumably accentuating the hydrogen effusion.
The effect of heat input on cracking risk of multipass weld metals has usually been
explained[8, 19–21] by its influence on the integrated diffusion-distance parameter
ΣD∆t/hw2 in accordance with Eq. (37). Even though the thermal factor of diffusion,
ΣD∆t, increases with higher heat inputs[13, 25,19,20,63], the fact that bead size and hence the
diffusion distance increase simultaneously in relation to the square of bead size, hw, in the
dominator[8, 19, 20], may result in an overall reduction of the final ΣD∆t/hw2 at higher heat
inputs.
Another fact accounting for the deleterious effect of high heat inputs on weld metal
cracking is that hardness of the weld metal is usually much less affected by the changes
in heat input or weld cooling time, than in the case of the HAZ[6, 7]. Any attempts to
reduce weld metal cracking risk via lowering its hardness by raising the heat input are
therefore likely to be far less effective than what have been recognised e.g., for the HAZ
root cracking. What remains is hence the effect of heat input on weld bead geometry in
terms of hw, weld penetration and bead overlap in terms of d/hw and, consequently, on the
conditions for hydrogen diffusion and accumulation described in terms of ΣD∆t/hw2.
This implies that inclusion of a heat input term into a predictive system should be
accomplished via incorporating it to ‘hydrogen retention’ and the resulting final hydrogen
concentration HRmax, rather than to the weld t8/5 cooling time or weld hardness HV.
 62

1.6 Preheat assessment methods for the avoidance of hydrogen

cracking in multipass weld metal

Only two of the reviewed[1–8, 11, 12, 23] calculation methods for predicting the necessary
preheat for weld metals were based on experimental tests that actually involve multipass
welding. All the other methods were mostly derived from the results of various types of
single pass tests. The two prediction methods are those according to Okuda et al.[2] and
Yurioka et al.[3].

1.6.1 Method according to Okuda et al.

Okuda‘s prediction method[2] is based on experiments using multipass SAW of 50 mm

thick, heavily restrained 500 x 400 mm specimens with a single U-groove preparation
welded from one side and having a 25° groove angle, no root face and a groove depth of
40 mm. This specimen geometry is reported[2] to have been chosen because the existing
evidence[106] showed that the increase in welding residual stress in weld metal with
increasing plate thickness tends to level off to a constant value with plate thickness of
about 50 mm and above. Different consumables (i.e., electrodes and fluxes) were used to
obtain various weld metal strength and hydrogen levels[2].
Most of the weld metal cracks observed in the tests were transverse cracks within a
depth of 5 to 15 mm below the plate surface, i.e., the location where residual stresses and
local hydrogen accumulation are both shown[2, 3, 19–22, 54, 76, 94, 95, 100] to reach their
maximum in multipass welding. It is worth noticing that only extra-high strength weld
metals (i.e., RM = 755–1080 MPa) were used in these experiments[2]. The illustration of
the test specimen preparation according to Okuda et al. is shown in Fig. 14.
Okuda found[2] that the required preheat temperature to prevent hydrogen cracking in
weld metal depends on the weld metal ultimate tensile strength, RM, and the logarithm of
weld diffusible hydrogen, log(HD), measured with the gas chromatography method.
Despite the nearly constant level of residual stress above 50 mm plate thickness, the
cracking susceptibility seemed to increase still further with the increase in weld build-up
thickness aw, as local hydrogen continued to accumulate. The influence of weld build-up
thickness on WM hydrogen cracking susceptibility was found[2] quite minimal and hence
insignificant in the 20–80 mm thickness range. Therefore, Okuda’s model does not
consider plate/weld build-up thickness as a variable, see Appendix 1. Anyway, Okuda’s
model is stated[2] to be applicable irrespective of plate thickness, provided it does not
exceed 40 and 80 mm in the case of single-side and double-side groove joints,
respectively.
 63

Fig. 14. Multipass weld metal U-Groove cracking test according to Okuda et al.[2a].

1.6.2 Method according to Yurioka et al.

Yurioka’s prediction method[3] is based on experiments comprising multipass SMAW

and GMAW of high-strength HT80 class plates of various weld build-up thicknesses and
applying a 300 x 300 mm specimen. The specimen uses a symmetrical Y-groove
preparation welded from one side and having a 60°-groove angle, small root gap (few
mms) and a relatively large root face. In addition to weld metal tensile strength RM and
diffusible hydrogen, log(HD), inclusion of the weld build-up thickness aw as an input
parameter can be regarded as an extension to Okuda’s method. The illustration of the test
specimen configuration according to Yurioka et al. is shown in Fig. 15.
In Yurioka’s experiments, welds of somewhat lower strength (RM = 580–885 MPa)
were used[3], in comparison to Okuda’s work[2]. Similarly to Okuda’s findings, the
observed WM cracks were mostly transverse to the weld direction and situated in the 2nd
or 3rd last weld filling layer[3,12], i.e., the location where the longitudinal residual stress
and local hydrogen in multipass welding are known[2, 3, 19–22, 54, 76, 94, 95, 100] to accumulate
the most.
 64

Fig. 15. Multipass weld metal Y-Groove cracking test according to Yurioka et al.[3b, 12].

Yurioka found[3,12] that the required preheat temperature to prevent hydrogen cracking
in weld metal depends on the weld metal tensile strength RM, the logarithm of the weld
diffusible hydrogen, log(HD) or log(H0), measured with the JIS glycerine method, and the
weld build-up thickness aw. Instead of a single equation, Yurioka’s method uses three
sub-equations in accordance with the aw range in question. These sub-equations are
applicable for weld aw ranges of 15–30 mm, 30–50 mm and over 50 mm, as shown in
Appendix 1. In accordance with Okuda, Yurioka also considers the influence of aw
constant with plate thickness above 50 mm[3].
Yurioka et al.[3] regarded the diffusible hydrogen content the most important parameter
governing cracking. Unfortunately, their hydrogen measurements are based on the
glycerine method, which induces some inaccuracy into the prediction, as one has to use
approximative equations to convert the JIS hydrogen values to the IIW values. A
comparatively large number of conversion formulae can be found in the literature, each
of them giving somewhat different result, see Appendix 1.

1.6.3 Small-scale cracking tests and their usability

Most of the small-scale hydrogen cracking tests for weld metals are designed for single-
pass welds[54]. The principal experiments employed for single-pass weld metals are the
(symmetrical) Y-Groove Tekken test, the tensile restraint cracking (TRC) test, the gapped
bead on plate (G-BOP) test and the Welding Institute of Canada (WIC) test[8, 26]. Single
pass weld tests, however, are of little assistance in the provision of procedural guidance
for multipass welds, principally because of the hydrogen retention characteristics of a
much larger volume of weld metal in multipass welds, and because such factors as
interpass time and -temperature are not encountered with single pass welds[8].
With regard to multipass weld metal cracking tests, different researchers and institutes
have been traditionally devoted to different types of cracking test specimens and/or
 65

groove geometries. In their experiments, Okuda et al[2] used a 500 x 400 mm size U-
Groove specimen of 40 mm groove depth and 50 mm plate thickness, c.f. Fig. 14.
Yurioka et al.[3, 12], in turn, apply a 300 x 300 mm size symmetrical Y-Groove specimen
where the plate thickness and weld build-up thickness can be varied independently on
each other, c.f. Fig. 15. Both of these test specimens are heavily restrained by using
welded strong-backs on the reverse side, to ensure adequate structural rigidity. At The
Welding Institute, U.K. (TWI), an 800 x 400 mm size V-Groove full-thickness specimen
configuration is applied, with restraining anchor welds on both ends[8, 9], as shown in
Fig. 16.

Fig. 16. Specimen configuration of the V-Groove cracking test for multipass weld metals
applied by TWI[8].

According to the earlier work in Japan[19], the U-shaped groove appeared ideal
particularly for the occurrence of weld metal transverse cracks in multiple-pass welding.
The outcome of the tests is interpreted differently among the tests. Okuda et al.[2a]
propose a safety margin of 15%, principally because the average tensile properties are
assumed used in the analysis instead of ‘worst case’ maximum values. Thus, his
description of the ‘safe’ Crack – No Crack line derived from the experimental data[2] is
multiplied with a safety factor of 1.15. Yurioka et al.[3], in turn, use the sub-equations
derived from the Y-Groove experiments directly without any external safety factors.
Since every single test is accounted for in the analysis and designated as cracked, if
 66

cracks were found, his description is a kind of ‘lower bound’ safe line. Having compared
the results of oblique y-groove and symmetrical Y-Groove tests for single-pass and
multiple-pass welds, respectively, Yurioka concludes[3b] that hydrogen cracking in weld
metal can be avoided by adopting a preheat temperature of about 25°C lower than that for
preventing root cracking in the JIS y-groove Tekken test. This view, however, has not
been fully shared by other researchers[107].
Overall, suitable cracking tests for multipass weld metals are sparse and not
standardised within Europe. In Japan, Japanese Welding Association (JWA) has
published a standard WES 1105-1985[19] entitled "Cracking Test for Single-Bevel Groove
Multi-layer Welds". This standard specifies a small-scale 300 x 600 x t mm3 (where t =
plate thickness and t ≥ 38 mm) test specimen with single-bevel (i.e., ½-V) groove
preparation having 45° groove angle, root face and restraining welds on both ends of the
specimen, as shown in Fig. 17. The standard is intended primarily for SMAW welding,
although also SAW and GMAW welding can be applied. The adoption of 45° single-
bevel (½-V) groove is justified by stating that it is suitable for examining parent steel
HAZ root cracks, underbead cracks and toe cracks, as well as enables the occurrence of
transverse weld metal cracks to be confirmed[19]. The standard defines appropriate ranges
of interpass temperatures, in relation to the applied preheat temperature. The standard
also provides defined quantitative criteria for the acceptable crack size: it judges the test
result as "cracked" in the case any cracks having a length of 0.5 mm or more are being
detected[19].
For the assessment of the risk of hydrogen cracking in multipass welds, it is essential
to gain information on the complete thermal history as the weld cools down to
temperatures of 200–100°C, usually in terms of t200, t150 or t100[13, 19, 23, 25, 101]. This way,
one can assure that a given specimen is large enough to be descriptive of real structures,
as far as weld cooling is concerned. With respect to the t100 weld cooling time, there exist
evidence[6, 7] demonstrating that the values measured from small-scale specimens, e.g. the
JIS y-groove Tekken test, correlate reasonably well to the values measured from large-
scale welded sections. Experiments carried out by JWA, Japan[19], ended up to a similar
conclusion when comparing the t8/5 and t150 cooling times between small-scale specimens
and larger test panels corresponding to an actual structure. The outcome was that the heat
cycle during multiple-pass welding of the 300 x 600 x t mm3 (t ≥ 38 mm) test specimen
conforming to WES 1105-1985 (NSC)[19] differed very little from that of larger plates and
panels corresponding to those used in real welded structures. According to the VTT
experiments[7], at similar levels of arc energy and preheat/interpass temperature, there
were no significant differences in the t100 values between butt welds and fillet welds,
either.
Presently, there exists no systematic investigation on whether the differences in
preheat estimates given by different cracking tests can be traced back to the differences in
specimen groove geometries or specimen size.
 67

Fig. 17. Specimen configuration of the single-bevel (½-V) Groove multiple-pass weld cracking
test according to WES 1105-1985 (JWA)[19].

1.7 Practical experiences of hydrogen cracking in multipass welds –

overview

According to the previous work[6, 7] made at VTT, the risk of hydrogen cracking in
multipass weld metals must be taken into account already in 460 MPa nominal yield
strength grades when using consumables producing an overmatching weld metal strength.
The risk was found[6, 7] evident for a 50 mm plate thickness in high rigidity structures
containing rutile FCAW welds with the weld metal Rp0.2 ≈ 600 MPa, RM ≈ 680 MPa and
the weld diffusible hydrogen content of ≈ 9 ml/100 g DM (IIW).
 68

Earlier work at VTT[52, 53] also demonstrated that the required preheat in a 690 MPa
nominal yield strength steel may be dominated by the cracking sensitivity of the multiple-
pass weld metal instead of the HAZ. According to the weld metal Implant tests,
instrumented Restraint tests and analytic calculations, in order to avoid weld metal
cracking in a 40 mm thick 690 MPa yield strength DQT steel, a preheat of 180°C –
higher than that for the HAZ – was found necessary[52, 53]. This was confirmed by the
results of SAW multipass welding tests where WM transverse cracking in the filling runs
could be eliminated only after raising the minimum preheat/interpass temperature up to
150°C. At 2.0 and 4.5 kJ/mm heat input, the occurrence of WM cracking appeared
independent of the heat input[52, 53].
Apart from extra-high strength weld metal exhibiting predominantly martensitic
microstructures, the possibility of hydrogen cracking in rutile FCAW weld metals of
hardness no more than 250–280 HV was confirmed at VTT[6, 7], as well, at least in the
case of thick plates and rigid structures. Therefore, as shown also by TWI[8, 9], a
maximum hardness level in a given steel proved acceptable from the HAZ cracking
viewpoint, say, 350 HV, is not necessarily ‘safe’ in terms of weld metal cracking. Despite
of this, standards such as SFS-EN 288-3:1992[10a] and its very recent successor, Draft
prEN ISO 15614-1[10b], categorically specify a ‘permitted maximum hardness’ of the
welded joint according to steel group (i.e., strength class) and whether heat-treated or not,
hence without distinguishing between the HAZ and the weld metal.
According to the earlier VTT data[6, 7], a single-pass K-Groove Tekken test is expected
to give more reliable preheat estimates than an oblique y-groove JIS Tekken test for such
weldment configurations and structural applications where weld metal root cracking is
the predominant form of cracking. This is often the case with e.g. pipeline girth welds.
The results[6, 7] of VTT are, in this sense, compatible with the earlier findings[107] made at
TNO, The Netherlands. With respect to transverse cracking in multipass weld metals, the
results[6, 7] do not provide sufficient evidence on the overall applicability of the K-Groove
Tekken test in assessing hydrogen cracking susceptibility of multipass weld metals
irrespective of the groove geometry (and FI value) in question.
The analyses made at VTT[6, 7] at that time suggested, whenever transverse weld metal
cracking in the filling layers is the predominant mode of cracking, both Okuda’s and
Yurioka’s prediction methods can be applied. Unfortunately, the two methods were
shown[4–7] to give greatly varying predictions. Okuda’s model[2] considers low hydrogen
more effective in lowering the required preheat/interpass temperature, thus putting more
weight on penalising the increase in hydrogen. Yurioka[3], in turn, considers the weld
build-up thickness aw to affect substantially the preheat estimate, whereas this factor is
completely omitted from Okuda's prediction. Neither of the two methods[2, 3] includes the
effect of heat input. For the cases where there is a possibility of root cracking to become a
predominant mode, e.g., line-pipe girth welds, neither of the models is suggested[4–7, 23] to
be applied, but specific assessments instead.
A recent German work[11] reported good agreement between the EN 1011-2:2001
Method B estimates and the results of multipass V-Groove welding experiments. The
work concludes that EN 1011-2:2001 Method B[1b] originally based on SEW 088
Merkblatt and derived from the results of single-pass HAZ cracking tests, e.g., y-groove
Tekken tests, can be applied safely also for estimating relevant preheat temperatures for
 69

multipass welds. One only has to replace the CET carbon equivalent of the parent steel
with the corresponding value of the weld metal increased by 0.03%[1b].
This, however, contrasts with the earlier NSC data[12] on preheat estimates based on
multipass Y-Groove tests, which differ substantially from those according to EN 1011-2 :
2001 Method B. The former NSC data[12] indicate that weld metal hydrogen cracking can
occur with weld hydrogen as low as ≈ 4 ml/100 g DM (IIW), provided weld metal
strength is high enough, i.e., Rp0.2 ≥ 730–750 MPa and RM in excess of ≈ 820 MPa. It
was recognised[12] that a preheat/interpass temperature of 150–175°C was necessary to
prevent cracking in RM ≈ 820 MPa weld metals with 4 ml/100 g DM (IIW) hydrogen.
According to EN 1011-2 : 2001 Method B[1b], preheat temperatures considerably lower
than those actually needed to eliminate cracking in the multipass Y-Groove tests would
have been predicted as ‘safe’. A comparison of T0 predictions according to the Y-Groove
cracking tests[12] and EN 1011-2 : 2001 Method B[1b] is presented in Appendix 2.
Overall, it is obvious that contrasting findings exist on the necessary preheat levels
predicted by the currently available methods[1b, 2, 3, 11, 12] intended for the cracking risk
assessments in multipass weld metals.
 2 Aims of the current work
The overall aims of the Doctoral Thesis were (i) to identify causal factors that govern
transverse hydrogen cracking in filling runs of high-strength multipass weld metals
(WM), as well as (ii) to define the primary parameters to describe the cracking behaviour.
This was to enable the development of an experimentally verified engineering
methodology that allows reliable predictions of the necessary precautions for the
prevention of WM hydrogen cracking.
The experimental working phase within the present work focused on multiple-pass
weld metal cracking in shielded-metal arc (SMAW) and submerged-arc (SAW) welds. As
a first stage of experiments, SMAW was chosen due to its feasibility in obtaining
different levels of weld diffusible hydrogen in a controlled manner by re-drying and
moisturing the electrodes. The use of SMAW to perform various types of Y- and U-
Groove cracking tests hence allowed the necessary scaling of the critical limits of Crack –
No Crack conditions according to weld metal strength and weld hydrogen. The work
continued with sets of SAW multipass cracking tests and using a similar U-Groove test
specimen configuration as was applied to the SMAW cracking tests. The data obtained
for these two welding processes were then compared, in order to reveal any possible
differences that may appear in the cracking susceptibility between SMAW and SAW
weld metals of otherwise similar strength and initial hydrogen content.
Based on the experimental data, numerical calculations were made for analysing the
interactions between the crack-controlling factors, in order to assess the critical boundary
conditions for weld metal hydrogen cracking in SMAW and SAW multipass weldments.
The interpretation of the findings was supported by the analyses of the results of all-weld
metal tensile tests, all-weld metal hardness surveys, weld residual stress measurements,
weld dilatometric measurements, bead size measurements and microstructural
investigations.
The final deliverables of the Doctoral Thesis were: (i) predictive diagrams and
equations for the assessment of WM hydrogen cracking risk by defining the critical
Crack – No Crack boundary conditions in terms of ‘safe line’ description giving the
desired lower-bound estimates, and (ii) predictive equations capable of giving reliable
estimates of the required preheat/interpass temperature for the avoidance of weld metal
hydrogen cracking in the Crack region.
 71

To accomplish Item (i) it was deemed necessary to define the weld critical hydrogen
content Hcr below which hydrogen cracking in multipass weld metals will not occur even
under the most stringent conditions realistically met in practice welding fabrication. This
was to serve as a ‘first line of defence’ against WM cracking.
Within Item (ii), engineering formulae are to be derived for predicting the required
preheat/interpass temperatures for the avoidance of WM hydrogen cracking, once the
other controlling factors are such that the ‘Crack region’ is inevitably entered. This is
thought to serve as a ’second line of defence’ against WM cracking.
 3 Experimental and analytical procedures
The experimental part of the thesis is concerned with multipass cracking tests on
shielded-metal arc (SMAW) and submerged-arc (SAW) weld metals of various strengths
and hydrogen levels. The outcome of the various cracking tests are analysed in order to
reveal the primary crack controlling factors.
Materials for the cracking tests and the associated experiments comprise SMAW
electrodes and SAW tubular wires & fluxes of different strengths and chemical
compositions. The experiments consisted of (i) multipass WM cracking tests, (ii)
analyses of weld chemical composition, (iii) all-weld metal tensile tests, (iv) weld
hardness measurements, (v) weld residual stress measurements, (vi) measurements on
hydrogen diffusion and diffusivity, (vii) weld dilatometric measurements, and (viii)
metallographic and microscopic studies of the weld bead sizes, weld microstructures and
phase fractions. Based on the experimental data, numerical calculations are made for
analysing the interactions between the crack-controlling factors in order to assess the
critical boundary conditions for hydrogen cold cracking in SMAW and SAW multipass
weld metals.
Within the experimental programme of the present thesis, multipass SMAW cracking
tests were performed at VTT Industrial Systems (formerly, VTT Manufacturing
Technology), Finland, and at Nippon Steel Corp. (NSC), Japan. Multipass SAW cracking
tests were made at ESAB, Finland, ESAB, United Kingdom and Mäntyluoto Works,
Finland. Helsinki University of Technology, Finland, carried out the weld residual stress
measurements, whilst University of Oulu, Finland, was responsible for the hydrogen
permeation tests, diffusivity calculations and derivation of the diffusion coefficients of
hydrogen, as well as for the all-weld metal tensile tests. Hardness surveys and
metallographic studies of weld metals were carried out jointly at University of Oulu and
VTT Industrial Systems.

3.1 Materials

For obtaining a sufficiently wide spectrum of weld metal strengths, electrodes

representing six different classes were used for the cracking experiments of SMAW weld
 73

metals. These were: (i) ESAB OK 48.08 (AWS A/SFA 5.5: E 7018-G; EN 499: E 46 5
1Ni B 32 H5), (ii) OK 74.78 (AWS A/SFA 5.5: E 9018-D1; EN 757: E 55 4 MnMo B
32), (iii) OK 75.75 (AWS A/SFA 5.5: E 11018-G; EN 757: E 69 4 Mn2NiCrMo B 32
H5), (iv) Nittetsu L-80 (IIS Z 3212 D 8016; AWS A5.5 E 11816-G), (v) Atom Arc 12018
(AWS A/SFA 5.5: E12018-M) and (vi) OK 75.78 (DIN 8529: HY 89 86 Mn3NiCrMo B
H5). These electrodes aimed at covering all-weld metal yield strengths in the range of
480–920 MPa.
The ‘average’ chemical compositions of the experimental SMAW weld metals made
using these six electrodes are given in Table 2a. The analysed weld chemical
compositions and the corresponding weld metal strength values are given in Appendix 7
and Table 11, respectively.
As a comparison, SAW tubular flux-cored wires representing four different strength
classes were used for the multipass weld metal cracking experiments. These were: (i)
ESAB OK Tubrod 15.24S (AWS A/SFA 5:23: F5A6-EC-G), (ii) OK Tubrod 15.26S, (iii)
OK Tubrod 15.27S and (iv) OK Tubrod 15.29S. The flux used in all the SAW
experiments was ESAB OK Flux 10.62 (AWS A5.17: F6A4-EM12, F6P5-EM12). These
filler wires & flux combinations aimed at covering all-weld metal yield strengths in the
range of 480–900 MPa.
The analysed chemical compositions of the experimental SAW weld metals made
using these four 4 tubular wires & flux are given in Appendix 7 and Table 2b.

Table 2a. Average chemical composition of the SMAW weld metals made using the
electrodes for the weld metal cracking experiments.
Electrode / C Si Mn Mo Cr Cu V Ni
w-% in weld metal
OK 48.08 0.04 0.35 1.24 0.01 0.04 0.01 0.027 0.97
OK 74.78 0.05 0.31 1.48 0.38 0.03 0.02 0.018 0.04
OK 75.75 0.04 0.37 1.57 0.35 0.34 0.01 0.022 2.24
Nittetsu L-80 0.05 0.50 1.36 0.55 0.32 0.02 0.006 2.58
Atom Arc 12018 0.06 0.33 2.03 0.49 0.88 0.01 0.011 2.14
OK 75.78 0.05 0.25 2.18 0.67 0.42 0.01 0.023 3.10

Table 2b. Average chemical composition of the SAW weld metals made using the tubular
filler wires for the weld metal cracking experiments. Flux: Esab OK Flux 10.62.
Electrode / C Si Mn Mo Cr Cu V Ni
w-% in weld metal
OK Tubrod 15.24S 0.05 0.24 1.79 0.10 0.03 0.03 0.010 0.83
OK Tubrod 15.26S 0.08 0.17 1.80 0.52 0.02 0.03 0.013 0.86
OK Tubrod 15.27S 0.07 0.25 2.03 0.54 0.03 0.03 0.010 2.54
OK Tubrod 15.29S 0.07 0.23 2.00 0.50 0.25 0.03 0.016 3.10
OK Tubrod 15.27S*) 0.08 0.20 1.87 0.33 0.02 0.01 0.011 1.58
*)
: thin plate weld (6 mm) made as a 2-pass weld, see Section 3.2
 74

To ensure a sufficiently wide range of weld hydrogen for the cracking tests, SMAW
electrodes were either re-dried according to the manufacturer's recommendations, or
moistured by holding them in the humidity chamber over a period of time under
controlled levels of moisture and temperature. The procedures are explained in detail in a
previous report[4]. In most cases, humidity chamber conditions of 40°C temperature and
80% relative humidity were found appropriate for obtaining weld diffusible hydrogen
levels of 8–11 ml/100 g DM (IIW). For some electrodes, more aggressive conditions of
40°C / 90% humidity were found necessary for obtaining the highest hydrogen levels of
12–19 ml/100 g DM (IIW).
With respect to SAW, the fluxes were first moistured at 1100°C, following by re-
drying intervals at various temperatures according to a specific procedure designed and
performed by ESAB, Gothenburg, Sweden. A more detailed description is given
elsewhere[4, 5, 112].
Since it has been proven[3, 5a, 19] that transverse WM hydrogen cracking is controlled
essentially by the weld longitudinal residual stress that, in turn, is dictated by the yield
strength of the weld metal, parent plate is not considered as a variable in the present
study. Consequently, fine-grained microalloyed structural steel with nominal Rp0.2 ≈ 460
MPa and 60 mm in thickness was applied throughout the basic testing programme, unless
otherwise noted. Occasionally, a 6 mm thick fine-grained microalloyed thin plate with
nominal Rp0.2 ≈ 650 MPa was used as a reference steel to assess the effect of plate/weld
build-up thickness and weld bead size on the cracking risk.

3.2 Experiments and testing methods

The SMAW weld metals made with the six electrodes were subjected to (i) multipass
weld metal cracking tests applying different Y- and U-groove geometries and various
specimen sizes, (ii) hydrogen diffusion measurements for determining the diffusion
coefficients for different weld metals, (iii) analyses of weld metal chemical composition,
(iv) all-weld metal tensile tests, (v) weld hardness measurements, (vi) weld residual stress
measurements, (vii) weld dilatometric measurements and (viii) metallographic and
microscopic studies of the weld bead size, weld metal microstructures and phase
fractions.
The SAW weld metals made with the four electrodes were subjected to (i) multipass
weld metal cracking tests applying the U-groove configuration and fixed specimen size,
(ii) analyses of weld metal chemical composition, (iii) weld hardness measurements and
(iv) metallographic and microscopic studies of the weld bead size.

3.2.1 Welding processes and procedures

Welding was made as multipass SMAW and SAW using different arc energies depending
on the purpose of the test. In the Y- and U-Groove SMAW cracking experiments made at
VTT Industrial Systems, fixed arc energy and interpass temperature levels of 3.0 kJ/mm
 75

and about 70–100°C, respectively, were constantly applied throughout the cracking test
programme. This was to define lower bound estimates of the boundary curve between
Crack-No Crack regions with respect to the two controlling factors assumed as being of
primary importance[2,3]: strength RM and hydrogen HD. To attain this further necessitated
limiting the number of additional variables.
In the U-Groove SAW experiments made at ESAB, Finland and ESAB, Waltham
Cross, U.K., also the heat input Q was taken as a variable, consequently three levels: 2.0,
3.0 and 4.0 kJ/mm were used. For the Taguchi test matrix prescribed for the latter SAW
tests, two additional variables were adopted besides Q, namely, preheat/interpass
temperature T0/Ti and weld diffusible hydrogen HD. Apart from the test matrix,
complementary SAW U-Groove tests were made at ESAB Waltham Cross, U.K. and
Mäntyluoto Works Oy, Finland. In these experiments, heat input and weld hydrogen were
kept constant and the tests aimed at defining 'safe' interpass temperatures for the two
extra-high strength wires: OK Tubrod 15.27S and OK Tubrod 15.29S.
Additionally, multiple-pass welded SMAW pads were made for obtaining specimens
for tensile tests and hardness measurements. In these cases, the arc energies were
adjusted according to the ratio of the groove shape factor F for bead-on-plate and groove
weld (with bevel preparation), with the intention to obtain equivalent t8/5 weld cooling
times.
The detailed welding parameters are given in the context of each experiment.

3.2.2 Weld hydrogen measurements

With the exception of the SMAW Y-Groove cracking tests made at NSC (c.f., Table 3a),
the weld diffusible hydrogen content HD of welds associated with all the SMAW and
SAW cracking tests in the present thesis was measured using the Osaka University
mercury method. The method has been shown[13, 25] to have a linear correlation to the IIW
mercury method in accordance with ISO/IIW 3690[57].
The individual weld diffusible hydrogen contents measured for each consumable and
condition are presented in the context of the corresponding multipass weld metal cracking
test.

3.2.3 Weld cracking test programme

The cracking test programme comprised SMAW and SAW multipass weld metal
cracking tests applying the Y- and U-Groove specimen configurations. Here, the aim was
to define and design a test specimen that would be small enough to be easily weldable,
but large enough for the cracking conditions to be descriptive of real multipass welded
structures. Specimens of 300 x 300/250 x 50 /60 mm and 500 x 500 x 60/70 mm were
used for the SMAW and SAW processes, respectively. The Y- and U-Groove specimen
configurations are depicted in Figs 15 and 18, respectively.
 76

Fig. 18. The U-Groove multipass weld metal cracking test specimen designed by VTT
Industrial Systems.

The SMAW experiments consisted of (i) symmetrical Y-Groove cracking tests

identical with those made at Nippon Steel Corp. (NSC)[12], (ii) rectangular-shape U-
Groove cracking tests designed by VTT and (iii) sc. double-welded large-scale Y-Groove
cracking tests to verify the small-scale test results.
In these tests, both the interpass temperature and heat input were fixed. The interpass
temperature Ti was constantly controlled down to about 70–100°C by applying forced
cooling of the specimen, with the intention to define, at a given level of weld metal
strength and hydrogen, the lower-bound Crack/No Crack boundary conditions giving the
desired ‘worst-case’ estimates for weldability. Forced cooling conditions enabling the
attainment of low interpass temperatures in conjunction with short interpass time, were
thought to represent the most stringent conditions met in actual welding fabrication of
large welded structural members.
For the SAW experiments, rectangular-shape U-Groove specimen configuration was
employed. Besides strength and hydrogen, here, the interpass temperature Ti and heat
input Q were also taken as variables. Consequently three levels of heat input and
interpass temperature: 2.0, 3.0 and 4.0 kJ/mm and 100, 125 and 150°C, respectively, were
applied in the Taguchi test matrix prescribed for the SAW cracking tests, as explained in
Section 4.1.2.
The objective here was twofold. Firstly, the intention was to reveal any possible
differences in the cracking sensitivity between SAW and SMAW weld metals at a similar
level of strength and hydrogen. Secondly, the aim was to provide data for deriving
mathematical description of the formula enabling the calculation of the required ‘safe’
 77

preheat/interpass temperature as a function of primary influential parameters, once the

Crack-region was entered.
Additionally, some comparative experiments were made for thin V-Groove specimens
of 6 mm in thickness and containing two weld passes, aiming at revealing any effects
plate thickness might have on WM hydrogen cracking. All these specimens were fully
restrained either by bolted clamps or welded strong-backs throughout the welding.

3.2.4 Analysis of weld chemical composition

The chemical composition of SMAW and SAW weld metals in the cracking tests was
determined using optical emission spectroscopy. Analysis was performed for the last
weld bead in the final filling layer, i.e., the location where most of the existing hydrogen
cracks were recorded. In all the cracking test specimens, the last bead to be welded was in
the middle of the final layer and the analyses were made from the centre of the weld bead
in question, in order to avoid any dilution from the parent steel. Consequently, the
analyses results correspond to all-weld metal chemical composition.
These data were used further in weld metal hardenability evaluation that was made to
correlate weld chemical composition in terms of Pcm and CET carbon equivalents to
hardness and, finally, ultimate tensile strength RM.

3.2.5 Tensile tests

Tensile testing was carried out for the SMAW weld metals made using different
electrodes and arc energy levels. The aim was simply to define true yield and tensile
strength, Rp0.2 and RM, of all-weld metals for further analyses of cracking susceptibility
and residual stress profiles, as well as to correlate strength with hardness. Thus, it was felt
adequate to conduct these testing applying SMAW weld metals, only.
All the specimens were extracted longitudinally from the filling layers, aiming at
characterising the weld region most prone to transverse WM hydrogen cracking. Testing
was carried out at room temperature in accordance with SFS-EN 10002-1 using round-
bar specimens (∅ 7 mm) according to SFS 3471.

3.2.6 Hardness survey

Weld hardness profiles were determined as Vickers hardness using HV5 for all the
SMAW and SAW weld metals. Again, measurements were concentrated to the
surroundings of final weld beads in the filling runs.
 78

The intention was primarily to define the average Vickers hardness HVave of as-
welded and re-heated weld metal microstructures, separately. In each case, the
corresponding individual maximum hardness HVmax of the weld metal was also recorded.

3.2.7 Weld residual stress measurements

Weld residual stress measurements were carried out for some of the SMAW weldments
applying a sc. Ring-Core drilling method. The method uses a drilling tool that enables
stresses below weld surface to be recorded in steps of ≈ 5 mm through the whole weld
depth, if necessary. This option was regarded advantageous, since the weld longitudinal
stress that was of primary interest here has been reported[2, 3, 19, 20, 23, 76, 100] to reach its
maximum at the depth of about 10–30% from the top surface of the plate (and the weld).
To reveal any possible effects of specimen (and weld) length, measurements were
made on two 40 mm thick Y-Groove cracking test specimens having (weld) lengths of
300 and 800 mm. The experiments were welded using OK 74.78 electrode with 3.0
kJ/mm arc energy.
Additionally, to investigate the effect of plate thickness on residual stress build-up,
comparative experiments were made using 6 mm thick, 2-pass welded thin plates of 500
mm in length and applying conventional hole-drilling method applicable for stress
measurements in thin plate.
The principles of the Ring-Core test method are presented in Appendix 3.

3.2.8 Measurements of hydrogen diffusion coefficient

Hydrogen diffusion measurements were carried out using the electrochemical hydrogen
permeation test under potentiostatic conditions to measure the hydrogen diffusion
coefficient, DH. A detailed description of the experiments and principles of the applied
methodology are given in separate reports[5b, 36b]. A short introduction is given in
Appendix 4.
The experiments were carried out using a potentiostatic voltage supply EG&G Model
263. With regard to the experiments, two kinds of traps are assumed to be present, deep
and shallow ones[5b]. The double- or multiple-charging technique was therefore adopted
for the hydrogen permeation tests carried out in the present study instead of a single
charging stage, in order to eliminate the effect of deep traps.
A number of series of measurements were performed, with the aim at extending the
number of chargings in each test to 3, 4 or more successive charging cycles. All the
specimens studied here were pre-prepared according to the following procedure:(i)
sawing the specimen down to a thickness of ≈1 mm, (ii) grinding the entry side of the
specimen to 300 grid, (iii) grinding the exit side of the specimen to 1200 grid, polishing
and nickel-coating under 3 mA/cm2 for 3 minutes in Watt's bath (i.e., 250 g/l nickel
 79

sulfate [NiSO4-6H2O], 45 g/l nickel sulphate [NiCl2-6H2O] and 40 g/l boric acid
[H3BO3]), (iv) spot welding a metal wire on both the entry and the exit side, (v) pouring
0.1N NaOH solution into the exit side glass compartment and continuous bubbling with
N2 during the whole time, (vi) prior to charging, pouring 0.1N H2SO4 acid into the enter
side glass compartment, continuous bubbling with N2.
After these treatments, the experiments continued with the actual hydrogen charging
and hydrogen permeation measurements for determining the hydrogen diffusion
coefficients.
The measurements aimed at defining hydrogen diffusion coefficients for various
SMAW, FCAW and SAW weld metals of different chemical compositions. The intention
was to also compare the DH values measured here to the published literature data. The
measured DH values were to be used further in numerical calculations evaluating the weld
local hydrogen concentration HRmax as a function of weld thermal history and weld bead
size.

3.2.9 Dilatometric measurements

Dilatometric measurements were made to determine the γ–α phase transformation

temperatures of the SMAW weld metals, in relation to that of the microalloyed parent
steel. Measurements were made using Gleeble 1500 thermal simulator and applying
various weld t8/5 cooling times to simulate the effect of arc energy on weld cooling and,
hence, on the phase transformation temperature. The peak temperature and holding time
of the thermal cycle were chosen as 1350°C and 1 sec, respectively. Four t8/5 cooling
times of 10, 20, 30 and 40 sec were set, coupled with 3D type cooling curve; these t8/5
values roughly corresponded to the arc energy levels of 1.9, 3.7, 5.6 and 7.5 kJ/mm,
respectively.
Cylinder shaped specimens of 10 mm in diameter and 75 mm in length were used for
the experiments. The specimens were extracted longitudinally from the filling layers of
multipass SMAW weldments.
The intention was to examine, whether differences in the γ–α phase transformation
temperatures between the weld metal and the parent steel HAZ have had any influence on
the cracking tendency of the differently alloyed weld metals. A detailed description of the
measurements is given in a separate report[36a].

3.2.10 Weld bead size measurements

Weld bead size measurements were performed from transverse weld cross sections that
were polished and etched before examining them under optical microscope. Height and
widths of the final weld bead in the filling layer were determined by idealising its shape
into a form of trapezium.
 80

The aim here was to investigate any possible effect a welding process might have on
bead size, especially, whether SMAW and SAW would yield different bead sizes at
similar heat input range.

3.2.11 Metallographic examinations

Metallographic examination was performed as optical microscopy for the SMAW weld
metals made using different electrodes and arc energies. The phase fractions were
calculated from filling runs of the weld specimens and are given as mean values of all the
measurements made from each specimen in question. Microstructures were characterised
and classified to enable the calculation of the phase fractions of different microstructural
constituents. For this purpose, all the specimens were polished to a 3 µm finish and
etched in 4% Nital. Microstructural constituents were classified according to the IIW
recommendations[13, 14] for the classification of weld metal microstructures using optical
microscopy.
The intention here was to find a relationship between the weld metal cracking
sensitivity and microstructure. The details of the examination and classification are
explained in a separate report[36a].

3.3 Analytical calculations

On the basis of the experimental data of this study, numerical calculations were made for
analysing the interactions between the crack-controlling factors, with the aim at assessing
the critical Crack – No Crack boundary conditions for hydrogen cold cracking in SMAW
and SAW multipass weld metals.
These calculations comprise: (i) analyses of the effects of weld thermal cycle in terms
of ΣD∆t, weld bead size hw and bead overlap d/hw on final hydrogen concentration, (ii)
derivation of hydrogen diffusion coefficients DH for various types of weld metals, and
(iii) calculations of preheat/interpass temperature estimates Tcr for the avoidance of
hydrogen cracking in multipass weld metals.

3.3.1 Effects of thermal and geometrical factors on final local

hydrogen concentration

In addition to weld initial diffusible hydrogen H0, the local (final) hydrogen concentration
HRmax in multipass welds has been shown[8, 9, 19–23] to depend on complete weld thermal
history, bead size and weld overlap due to penetration of successive passes. As thermal
gradient, bead size and overlap are welding process related, the present study investigates
 81

their influence on HRmax in the case of arbitrary, yet typical, SMAW and SAW multipass
welds. This is to evaluate whether HRmax could possibly differ between SMAW and SAW
filling runs so significantly that this factor should be included into predictive systems
assessing the weld metal cracking risk. For this purpose, measured data for SMAW welds
are applied for the analysis. With respect to SAW, arbitrary weld geometry was defined
using relevant literature data[32, 33].
For the analysis, the complete weld thermal cycle is expressed in terms of the thermal
factor of diffusion, ΣD∆t, introduced in Section 1.2.1. In theoretical form, ΣD∆t is
defined in Eqs (2a) and (2b)[21–23]. For the comparative calculations made here, it is
sufficient to consider the cracking conditions at a constant temperature, i.e., 100°C,
consequently ΣD∆t can be expressed using another factor, ΣD∆t(100), that has been
shown[13, 23–25] to approximate Eq. (2) comparatively well. The use of ΣD∆t(100) in place
of ΣD∆t, allows calculation of the ΣD∆t(100) factor within engineering accuracy directly
from the recorded weld thermal history using the weld cooling time from the peak
temperature down to 100°C, t100. Therefore, empirical Eq. (3b) allowing for calculation
of estimate of ΣD∆t(100) on the basis of the weld t100 data is applied in the present study.
For this purpose, a set of t100 cooling times were chosen which correspond to the 1.7
kJ/mm arc energy welding of a 50 mm thick plate at various preheat temperatures (i.e., 3-
dimensional heat flow conditions), see Appendix 10.
The effect of weld bead size on hydrogen diffusion distance is described using the
square of the individual bead layer thickness, hw2. The hw2 factor has been shown[19–23] to
provide a realistic description of the effect of bead size on the diffusion of hydrogen in
weld thickness direction and under the thermal cycle of welding, as explained in Section
1.5.2.
The thermal and geometrical conditions for hydrogen diffusion were incorporated by
describing them using the integrated diffusion-distance parameter, ΣD∆t/hw2, as
defined[19, 20] in Eq. (37). The ΣD∆t/hw2 ratio is considered indicative of the loss of
hydrogen from a weld bead in the time available, this time depending on the thermal
conditions of welding. Here, ΣD∆t obeys Eq. (3b) and hw is determined experimentally
from a weld macrosection in question.

3.3.2 Derivation of values for hydrogen diffusion coefficient

A detailed description of the methodology for deriving diffusion coefficients of

hydrogen, DH, is given elsewhere[5b, 36b]. Principles of the applied procedure are explained
in Appendix 4 which shows determination of DH from the potentiostatic curve that
displays current on the exit side of the specimen, Ia, as a function of time t in the
permeation test. Whilst Ic is the constant charging current applied to the entry side of the
specimen, Ia is the recorded current from the reverse (exit) side of the specimen and is
hence proportional to the extent and velocity hydrogen permeates through the specimen
during the test. Thus, Ia is influenced by the hydrogen diffusion coefficient DH of the
given specimen. Consequently, it is Ia that shows correlation with DH, not Ic.
 82

The J/Jmax in the y-axis in Appendix 4 is directly proportional to the charging current
Ic/Ic(max) where Ic corresponds to C0, i.e., the mean value of hydrogen concentration on the
specimen entry side. As long as Ic is kept constant, the recorded Ia–t curve will be an
exponential curve. Different DH values in different materials then manifest themselves as
different shapes of the corresponding Ia–t curve. Before deriving DH from the Ia–t curve,
the sc. onset point must be determined. In the present study, the time at which Ic was
applied to the specimen entry side was set as the onset point. After applying Ic, the
corresponding Ia–t curve can be recorded. The onset point in the Ia–t curve is then defined
as the original zero: t0 = 0, I0 = 0, from which the previous part – being treated as an
diffusion independent background, is omitted from the entire transient curve. Further, the
relaxation time, i.e., the time regarded as being due to finite rate constant for transfer of
hydrogen atoms from the surface into the metal phase[109], approximately equal to 10 sec
on the basis of the present experiments, was applied in the analysis. Usually, the point of
the first maximum Ia peak detected after the original zero was defined as the end point,
see Appendix 4.
The recorded Ia–t curves were processed before they were applied for calculation of
DH according to the requirement of the Fourier's solution. Values of Dt(0.1), Dt(0.2) and
Dt(0.25) were calculated corresponding to the ratio of J/Jmax to 0.1, 0.2 or 0.3,
respectively[110]. These were found[109, 111] to be more reliable than the commonly
determined values for the time lag, breakthrough time or half delay time.
Consequently, the hydrogen diffusion coefficients DH were calculated for Armco iron,
modern fine-grained high-strength structural steel, as well as for the SMAW, FCAW and
SAW weld metals. It was thought that the experiments for Armco iron would serve as a
reference test expected to give DH values characteristic for clean metals practically free
from hydrogen traps and hence exhibiting high hydrogen diffusivity. The DH values for
the modern fine-grained structural steel, in turn, were expected to be somewhat lower
than those for Armco iron but, owing to the low impurity content of the parent steel, still
remarkably higher than the values for the weld metals with a greater number of hydrogen
traps.

3.3.3 Calculation of safe preheat/interpass temperature estimates for

multipass weld metal

Analyses of the calculation methods for predicting safe preheat/interpass temperature

estimates, Tcr, for multipass weld metals comprise to principal phases. These are: (i) re-
analysis of the former data sets from the Y-Groove cracking tests[12] conducted by Nippon
Steel Corp. (NSC), and (ii) derivation of an optimised Tcr prediction on the basis of the
results from the SAW U-Groove experiments made in the present thesis.
The Y-Groove data set[12] from the former experiments made at NSC is given in
Appendix 5. The data consist of SMAW and GMAW multipass experiments with two
weld RM levels of ≈ 820 and 750 MPa, weld build-up thicknesses aw of 20, 30 and 40 mm
and weld diffusible hydrogen in the range of about 2 to 11 ml/100 g DM (IIW). These
 83

data[12] are re-analysed in the present thesis in order to derive alternative expressions that
would enable the avoidance of the use of various sub-equations of the Yurioka-formula,
c.f., Appendix 1.
Since both Okuda et al.[2] and Yurioka et al.[3] have demonstrated the vital role of weld
metal tensile strength RM and diffusible hydrogen HD, RM, and HD were adopted as
primary factors for the re-analysis of the NSC data[12]. As the true value of weld metal
strength is not always known in practice, RM is approximated for the present re-analysis
using the actual weld metal chemical composition in terms of the CET carbon
equivalent[1]. Despite the fact that Okuda et al.[2] – in contrast to Yurioka et al.[3] – found
no significant effect of plate/weld build-up thickness aw on cracking susceptibility, the
inclusion of aw is regarded imperative, because the cracking tests made in the present
thesis comprise also thin plate two-pass welds as reference experiments.
According to present knowledge[1–3, 12, 54, 76, 100, 116, 117] on crack-controlling factors in
multipass weld metals, a first attempt was made to estimate Tcr using the former NSC Y-
Groove data set[12]. These NSC data were fitted by an empirical equation that describes
Tcr as a function of weld metal CET, aw and HD. A general form of this expression is
given as:
Tcr = A ∗ CET + B ∗ (aw)m + C ∗ (HD)n – D (40a)
where A, B, C and D are constants, m and n are exponents (m, n < 1), CET is the carbon
equivalent (%) calculated from the true weld metal chemical composition in accordance
with EN 1011-2:2001[1b], aw is weld build-up thickness (mm) and HD is weld diffusible
hydrogen content (ml/100 g DM) according to ISO/IIW 3690[57].
A second step was to investigate whether the over-conservatism in the Tcr estimates
acc. to Eq. (40a) for moderate strength weld metals investigated here, could be reduced
while still ensuring ’safe’ and ’realistic’ estimates for the extra-high strength welds. This
was made by accentuating the logarithmic effect of hydrogen – a trendline that appears
obvious when analysing the former NSC Y-Groove data set[12], as well as the results of
Okuda et al.[2]. For this purpose, Eq. (40a) is rewritten with respect to the hydrogen term
as:
Tcr = A ∗ CET + B ∗ (aw)m + {[C ∗ ln(HD)] / [D ∗ (HD)n]} – E (40b)
where A, B, C, D and E are constants, whereas exponents m and n and factors CET, aw
and HD are in accordance with Eq. (40a).
Applying a form of Eq. (40b), the experimental data from the SAW U-Groove
cracking tests made in the present study were analysed. These data are thereby fitted by a
revised formula that: (i) puts more weight on the effect of weld metal strength by
adjusting the A ∗ CET factor, and (ii) accentuates the logarithmic effect of hydrogen with
a consequence of rewarding the attainment of low levels of weld hydrogen.
A third step was to evaluate the descriptive potential of parameters that could be
related to weld metal strength, others than CET. It is realised that despite of its
simplicity, any compositional based parameter, such as carbon equivalent, has debility
that its disregards the effect of weld cooling on hardenability. Since the weld ultimate
tensile strength RM has been shown[6, 7, 23] a linear function of weld hardness, HV, and
hardness, on the other hand, incorporates the effects of both the chemical composition
 84

and the weld thermal cycle on hardenability, it was deemed imperative to test alternative
expressions replacing CET in Eq. (40b) by weld metal hardness in terms of HVave and
HVmax.
Consequently, Eq. (40b) is re-formulated as:
Tcr = A ∗ HV + B ∗ (aw)m + {[C ∗ ln(HD)] / [D ∗ (HD)n]} – E (40c)
where HV is the actual weld metal Vickers hardness (HV5) in terms of either average
hardness HVave or the individual maximum hardness HVmax. Constants A, B, C, D and E,
exponents m and n and factors CET, aw and HD are in accordance with Eq. (40b).
The aim here is to pursue optimisation of Tcr predictions in order to (i) accentuate the
effect of weld metal strength on Tcr, (ii) accentuate the logarithmic effect of HD on Tcr at
low hydrogen levels, and (iii) introduce a strength based parameter that is more
convenient to measure than RM, but capable of incorporating the effects of chemical
composition and weld thermal cycle on weld metal hardenability and, hence, its strength.
Finally, the applicability of alternative expressions for aw and HD were investigated.
All the data were therefore re-fitted applying the following expressions for HD and aw, as:
Tcr = f {[C ∗ ln(HD)] / [D ∗ (HD)n]} (41a)
Tcr = f [ C ∗ (HD)n ] (41b)
Tcr = f [ C ∗ log(HD)] (41c)
Tcr = f [ C ∗ tanh(HD / D)] (41d)
Tcr = f [ B ∗ (aw) ]
m
(42a)
Tcr = f [ B ∗ tanh(aw / C)] (42b)
The intention of these final adjustments of the expression for HD and aw was to (i)
describe more realistically the cracking occurrence in the case of thin plate welds and (ii)
avoid overly conservative estimates at high HD levels without a need to sacrifice the
safety at low HD levels. These data assessments were based on the comparisons to the
most relevant literature data[2, 3, 12] and the experimental results of the present thesis.
 4 Results
Results of the experimental programme of the thesis are presented, consisting of (i)
multipass Y- and U-Groove cracking tests for SMAW and SAW weld metals, (ii)
analyses of weld chemical composition, (iii) all-weld metal tensile tests, (iv) weld
hardness measurements, (v) weld residual stress measurements, (vi) measurements on the
diffusion and diffusivity of hydrogen, (vii) weld dilatometric measurements, and (viii)
metallographic and microscopic studies of the weld bead sizes, weld microstructures and
phase fractions.

4.1 Multipass weld metal cracking tests

The cracking test programme comprised (i) small-scale Y-Groove cracking tests on
SMAW weld metals, (ii) large-scale double-welded Y-Groove tests on SMAW weld
metals, (iii) small-scale U-Groove cracking tests on SMAW and SAW weld metals, and
(iv) comparative V-Groove tests on thin plate SMAW and SAW welds.

4.1.1 Cracking tests for SMAW thick plate weld metals

The SMAW cracking test programme on thick plate weld metals consisted of (i) small-
scale Y-Groove cracking tests made at Nippon Steel Corp. (NSC), Japan, and VTT,
Finland, (ii) large-scale double-welded Y-Groove tests made at VTT, Finland, and (iii)
small-scale U-Groove cracking tests made at VTT, Finland.
 86

4.1.1.1 Small-scale Y-Groove tests

The first phase of the experimental programme revolving around SMAW weld metals
composed of six Y-Groove cracking tests made at NSC, Japan, and using extra-high
strength Nittetsu L-80 electrode of RM ≈ 810 MPa. Earlier work[12] has shown, this
electrode to produce weld diffusible hydrogen content HD ≈ 4 ml/100 g DM (IIW) in the
as-received condition. For the electrodes used here, this was confirmed by measurements
made at NSC, as shown in Table 3a. The chemical composition of Nittetsu L-80 weld
metal is given in Table 2a.
The dimensions of the Y-Groove specimen were 300 x 300 x 50 mm3 with 60° -groove
angle, 10 mm root face, 2 mm root gap and welded strong backs in the reverse side to
restrain the specimen against any distortion. The weld build-up thickness aw thereby
becomes 40 mm. The applied specimen configuration hence complies with that
denoted[3, 12] as the "NSC Y-Groove specimen", c.f., Fig. 15. The objective here was to
evaluate the effect of heat input Q on cracking risk of multipass weld metals, since Q is
presently not included either in the Yurioka's[3] or Okuda's[2] prediction method. Three
levels of arc energy E: 1.5, 3.0 and 5.0 kJ/mm were therefore applied. With each arc
energy, welding was made at two levels of interpass temperature Ti of 150 and 175°C
chosen to produce cracked and crack-free weld metals, respectively. The required Ti
values for cracked and crack free welds, at given levels of weld metal strength and
hydrogen, were estimated according to the Yurioka-formula[3b] in Appendix 1.
The results of the Y-Groove cracking tests for SMAW welds made at NSC are shown
in Table 3a. After welding was completed, a period of 48 h was waited before carrying
out the metallographic examination. The possible existence and morphology of the cracks
was confirmed from polished macrosamples of the longitudinally cut, full-thickness weld
sections.
Post-test metallography revealed that none of the six test specimens showed any
characteristic transverse weld metal hydrogen cracks. Thus, the results do not enable one
to draw sound conclusions on the effect of heat input on WM hydrogen cracking
susceptibility in the present case.
Table 3a. Y-Groove cracking tests for Nittetsu L-80 weld metals made at NSC, Japan,
using elevated preheat/interpass temperatures.
Specimen Weld Arc Current Voltage Travel HD Ti ti Total no Crack –
Strength energy speed (ml/100 g of bead No
Rp0.2/RM DM IIW) (min layers Crack
(MPa) (kJ/mm) (A) (V) (cm/min) (°C) ± min)
15-F 790/860 1.5 160 25 15.5 4.0 ≥ 150 6±6 43 NC
15-G 1.5 160 25 14.8 4.0 ≥ 175 5±5 43 NC
30-F 745/845 3.0 160 25 7.7 4.0 ≥ 150 9±4 24 NC
30-G 3.0 160 25 7.6 4.0 ≥ 175 7±2 24 NC
50-F 640/850 5.0 160 25 4.6 4.0 ≥ 150 13 ± 4 15 NC
50-G 5.0 160 25 4.6 4.0 ≥ 175 10 ± 3 15 NC

The second set of SMAW experiments comprised Y-Groove cracking tests made at
VTT and using specimen dimensions identical to those of the NSC specimens[3, 12], c.f.
Fig. 15. Again, the applied electrode was Nittetsu L-80 which, according to the Osaka
 87

University mercury method[13, 25], gave ≈ 3.8 ml/100 g DM (IIW) weld diffusible
hydrogen in the as-received condition. Bead placement and welding sequence followed
that undertaken at NSC. Constant arc energy of 3.0 kJ/mm was applied. The specimens
were bolted tightly onto a table using clamps to ensure absolutely rigid conditions and
negligible deformation during welding, as shown in Fig. 19.

Fig. 19. Principles of specimen fixing and test conditions: bolted clamp fixtures and the
arrangements for intensified cooling during welding using slices of solid CO2.

Since none of the experiments at NSC and given in Table 3a revealed any
characteristic WM hydrogen cracks, the testing conditions for the VTT Y-Groove
cracking tests were deliberately made more severe by employing intensified cooling
during welding. The use of intensified cooling during the time between each and every
bead layer enabled to obtain low interpass temperatures Ti in conjunction with short
interpass times ti. Otherwise, this combination would have been impossible to achieve,
e.g., under 'free' air-cooling conditions. Intensified cooling was done by coupling
continuous blow of compressed air onto the specimen edges with crushes and slices of
solid CO2 placed onto the steel bars on the surface of the specimen. This enabled ti values
of around 3–4 min in the combination with Ti ≈ 70–100°C to be obtained The principles
of fixing by bolted clamps and the cooling arrangements are shown in Fig. 19.
The results of the Y-Groove cracking tests for SMAW welds made at VTT are shown
in Table 3b. After welding was completed, a period of 48 h was waited before carrying
out the metallographic examination. The possible existence and morphology of the cracks
 88

was confirmed from polished macrosamples of the longitudinally cut, full-thickness weld
sections.
It can be seen that the experiments associated with the HD levels of 3.8 and 10.9
ml/100 g DM (IIW) did not reveal any hydrogen cracks. This implies, in contrast to the
earlier NSC data[12], SMAW multipass welds made using the L-80 electrode and at weld
hydrogen content of ≈ 4 ml/ 100 g DM (IIW) are not prone to hydrogen cracking, no
matter how short is the interpass time ti. The fact that N2 did not reveal cracks is regarded
surprising, as the earlier NSC Y-Groove tests have shown[12], the HD contents of ≈ 6.7 and
10.4 ml/100 g DM (IIW) lead to hydrogen cracking of the L-80 weld metal, see
Appendix 5.
Table 3b. Y-Groove cracking tests for Nittetsu L-80 weld metals made at VTT using
intensified cooling.
Specimen L-80 Arc Current Voltage Travel HD Ti ti Total no Crack /
Strength energy speed of bead No
Rp0.2/RM (ml/100 g (min ± layers Crack
(MPa) (kJ/mm) (A) (V) (cm/min) DM IIW) (°C) sec)
N1 745/845 3.0 170 23–24 8 3.8 ≤ 105 4 ± 25" 30 NC
N2 745/845 3.0 170 23–24 8 10.9 ≤ 99 4 ± 31" 31 NC
N3 745/845 3.0 170 23–24 8 13.2 ≤ 94 4 ± 05" 31 C

Next, weld hydrogen was raised further by performing moisturing of Nittetsu L-80
under extreme conditions of 45°C / 95%. This eventually resulted in the appearance of
characteristic transverse WM hydrogen cracks in N3 associated with HD of ≈13 ml/100 g
DM (IIW), see Table 3b. Although being clearly visible in the X-ray examination, these
cracks did not enter the weld surface, as did most of those cracks in the later experiments,
see Sections 4.1.1.2 and 4.1.1.3. The morphology of the cracks was confirmed by
metallographic examination of the longitudinally cut full-thickness weld sections. An
example of transverse hydrogen crack recorded in the N3 weld metal is presented in
Fig. 21.
According to Table 3b, a weld HD content as high as ≈ 13 ml/100 g DM (IIW) was
required, before weld metal hydrogen cracking occurred in Nittetsu L-80 weld metal of
RM ≈ 810 MPa.

4.1.1.2 Small-scale U-Groove tests

For the next stage of small-scale cracking tests, a rectangular-shape U-Groove specimen
was designed at VTT Industrial Systems, acting as a modification of the NSC Y-Groove
test. The intention of the U-Groove configuration was to reduce the total welding time,
while maintaining the inherent severity of the test by ensuring sufficient rigidity of the
specimen. The specimen size was 300 x 300/250 x 60 mm, the groove dimensions
varying to some extent depending on the test. A U-groove having a depth of 25–30 mm
and a width of 40–45 mm was adopted for specimens U1–U11, whilst for specimens
U12–U34 the depth and width both were 40 mm. The groove depth corresponds to the
 89

weld build-up thickness aw and could be varied independently of the specimen thickness.
An example of the U-Groove specimen is shown in Fig. 18. The specimens were bolted
tightly onto a table using clamps to ensure absolutely rigid conditions and negligible
deformation during welding, c.f., Fig. 19.
Similarly to the VTT Y-Groove tests in Section 4.1.1.1, intensified cooling was used
for the U-Groove experiments during the time between each and every bead layer, in
order to deliberately raise severity of testing and to ensure low interpass temperature Ti
and short interpass time ti. Intensified cooling was done by coupling continuous blow of
compressed air onto the specimen edges with crushes and slices of solid CO2 placed onto
the steel bars on the surface of the specimen. This enabled ti values of around 3–4 min in
the combination with Ti ≈ 70–100°C to be obtained The principles of fixing by bolted
clamps and the cooling arrangements are shown in Fig. 19.
Measurements on the weld diffusible hydrogen content, HD, applying the Osaka
University mercury method[13, 25] confirmed that the use of re-dried, as-received and
moistured electrodes produced weld HD values in the range of 3 to19 ml/100 g DM
(IIW).
After welding was completed, a period of 16 h was waited before carrying out the final
NDT inspection to detect possible delayed hydrogen cracks. This time interval was in
accordance with standard EN 1011-2 : 2001[1b]. Both ultrasonic examination and X-ray
inspection were made. In the cases where NDT indications were found, the morphology
of the cracks was confirmed by metallographic examination of the longitudinally cut full-
thickness weld sections.
Additionally, occasional U-Groove cracking tests were conducted for the OK 74.78
weld metals of Rp0.2 ≈ 550 MPa otherwise as previously, but prolonging the period before
the NDT inspection to 7 days (168 h) instead of the 16 h period. Since the cracking
tendency of this particular weld metal appeared surprisingly low in the first phase of the
programme[4], further testing was deemed necessary. This was to now ensure enough time
for complete hydrogen diffusion, as the plate (and weld build-up) thickness were
relatively great and the WM strength close to the lower end of the 'critical' range with
respect to WM cracking risk.
The test matrix and the results of all the U-Groove cracking tests for the five SMAW
weld metals made at VTT are given in Table 4. The U-series finally comprised 34 tests in
total and was focused on OK 48.08, OK 74.78, OK 75.75, Atom Arc 12018 and OK 75.78
electrodes.
It can be seen that of the entire number of test welds in Series U associated with the
16–48 h period, only three sets of specimens exhibited characteristic transverse weld
metal hydrogen cracks. These were: (i) U11, U25, U26 and U29 welded with OK 75.75,
(ii) U13-U17 & U20 welded with OK 75.78 and (iii) U27-U30 welded with Atom Arc
12018. Thus, weld metal hydrogen cracking took place only in the case of the specimens
welded with the electrodes of the highest strengths: OK 75.75, OK 75.78 and Atom Arc
12018 that produced weld metal strengths of Rp0.2 ≈ 680, 880 and 890 MPa, respectively.
 90

Table 4. U-Groove cracking tests for SMAW weld metals made at VTT using intensified
cooling.
Specimen Filler Arc Current Voltage Travel HD Ti ti Total Crack/
No Material energy speed max. or No of No
(ml/100 g range (min ± bead Crack
(kJ/mm) (A) (V) (cm/min) DM IIW) (°C) sec) layers
OK 48.08 : Rp0.2 ≈ 455 MPa, RM ≈ 530 MPa
U6 OK 48.08 3.0 250 24 12 5.1 ≤ 102 4 ± 02" 30 NC
U7 OK 48.08 3.0 250 24 12 6.8 ≤ 97 4 ± 02" 24 NC
U9 OK 48.08 3.0 250 24 12 9.2 ≤ 99 4 ± 01" 24 NC
U19 OK 48.08 3.0 250 24 12 10.8 ≤ 97 3 ± 05" 39 NC
OK 74.78 : Rp0.2 ≈ 550 MPa, RM ≈ 650 MPa
U4 OK 74.78 3.0 250 24 12 5.4 ≤ 108 4 ± 10" 27 NC
U5 OK74.78 3.0 250 24 12 5.5 ≤ 101 4 ± 02" 24 NC
U12 OK74.78 3.0 250 24 12 13.5 ≤ 108 3 ± 05" 39 NC
U8 OK74.78 3.0 250 24 12 15.1 ≤ 100 4 ± 02" 24 NC
U34
*)
OK74.78 2.0 250 23 17 15.5 57–74 4 ± 02" 56 NC
U33
*)
OK74.78 3.0 250 24 12 15.5 90–107 4 ± 05" 37 C
*)

U32
*)
OK74.78 3.0 250 24 12 17.6 88–101 4 ± 05" 37 C
*)

U31
*)
OK74.78 3.0 250 24 12 17.6 91–107 4 ± 10" 39 C
*)

U18 OK74.78 3.0 250 24 12 18.9 ≤ 105 3 ± 05" 39 NC

OK 75.75 : Rp0.2 ≈ 680 MPa, RM ≈ 810 MPa
U1 OK 75.75 3.0 250 24 12 3.2 ≤ 174 4 ± 40" 25 NC
U3 OK 75.75 3.0 250 24 12 3.2 ≤ 105 4 ± 05" 30 NC
U2 OK 75.75 3.0 250 24 12 5.6 ≤ 101 4 ± 15" 27 NC
U10 OK 75.75 3.0 250 24 12 6.2 77–105 10 ± 1" 27 NC
U29 OK 75.75 3.0 250 24 12 10.1 85–92 4 ± 02" 39 C
U11 OK75.75 3.0 250 24 12 12.6 ≤ 97 4 ± 02" 27 C
U25 OK75.75 3.0 250 24 12 14.3 90–100 7 ± 10" 39 C
U26 OK75.75 3.0 250 24 12 14.3 100 15 ± 1" 39 C
OK 75.78 : Rp0.2 ≈ 885 MPa, RM ≈ 965 MPa
U21 OK 75.78 3.0 250 24 12 2.6 98–106 15 ± 1" 39 NC
U24 OK 75.78 3.0 250 24 12 2.8 70–83 10 ± 1" 39 NC
U23 OK 75.78 3.0 250 24 12 2.8 85–95 7 ± 10" 39 NC
U22 OK 75.78 3.0 250 24 12 3.3 101–110 15 ± 1" 39 NC
U20 OK 75.78 3.0 250 24 12 4.3 97–104 15 ± 1" 39 C
U13 OK 75.78 3.0 250 24 12 5.1 ≤ 96 4 ± 02" 39 C
U14 OK 75.78 3.0 250 24 12 5.1 81–93 7 ± 10" 39 C
U15 OK 75.78 3.0 250 24 12 5.1 67–76 10 ± 1" 39 C
U17 OK 75.78 3.0 250 24 12 5.1 98–112 10 ± 1" 39 C
U16 OK 75.78 3.0 250 24 12 5.1 94–103 15 ± 1" 39 C
Atom Arc 12018 : Rp0.2 ≈ 890 MPa, RM ≈ 1020 MPa
U30 Atom 3.0 250 23 12 3.2 95–101 4 + 2" 39 C
Arc
U28b Atom Arc 3.0 250 24 12 3.4 80–94 4 ± 02" 39 NC
U28a Atom 3.0 250 24 12 3.4 80–94 4 ± 02" 39 C
Arc
U27 Atom 3.0 250 24 12 4.4 83–96 4 ± 02" 39 C
Arc
*)
: tests made applying a period of 168 h (7 days) before the final NDT inspection, otherwise 16 h
period in accordance with EN 1011-2 : 2001[1b].
 91

The results demonstrate that once a certain WM strength level is exceeded, the WM
hydrogen cracking risk in multipass SMAW increases quite dramatically. For cracking to
occur at the Rp0.2 ≈ 680–690 MPa strength level, OK 75.75 electrodes producing HD ≥ 10
ml/100 g DM (IIW) were required. To create cracking conditions for the extra-high
strength OK 75.78 and Atom Arc 12018, weld metals of Rp0.2 ≈ 880–890 MPa exhibiting
HD of no more than 3–4 ml/100 g DM (IIW) were crack-sensitive enough. Not even re-
drying of Atom Arc 12018 did always guarantee crack-free welds at HD ≈ 3.2 ml/100 g
DM (IIW), although it did so the case of OK 75.78 welds. As an example, a photograph
of characteristic transverse hydrogen crack found from the U11 weld metal is shown in
Fig. 22.
An interesting finding regarding the interpass time ti was that whenever WM cracking
took place, it did so irrespective of the interpass time ti, even though ti was varied over a
range from 4 to 15 min. Accordingly, provided the other factors, i.e., strength and
hydrogen, were not severe enough to provoke WM cracking, shortening ti alone was not
sufficient to create conditions for triggering the crack initiation event.
For the low strength OK 48.08 weld metals of Rp0.2 ≈ 480 MPa, it was impossible to
create conditions severe enough to result in WM cracking with any practical hydrogen
levels in the range of 5–11 ml/100 g DM (IIW).
In the case of OK 74.48 welds of Rp0.2 ≈ 550 MPa, the cracking behaviour was found
to depend essentially on the applied period before the final NDT inspection. Using a
period of 16 h, as required in EN 1011-2 : 2001[1b], it appeared impossible to create
conditions severe enough to lead to WM hydrogen cracking at any of the applied
hydrogen levels ranging from 3 to 19 ml/100 g DM (IIW). Contrary to this, all the
experiments at 3.0 kJ/mm associated with a prolonged period of 168 h (7 days) resulted
in WM Chevron c00racking, although not until at comparatively high weld hydrogen
contents of 15.5–17.6 ml/100 g DM (IIW). Even in this series, however, a single cracking
test, U34, otherwise similar to the others as regards to the electrode and weld hydrogen,
but welded at 2.0 kJ/mm, appeared as crack-free despite a 7 days period and HD of
15.5 ml/100 g DM (IIW), as shown in Table 4.
These plausible findings are perceived as indisputable evidence that the 16 h period,
EN 1011-2 : 2001[1b] currently requires before final NDT inspection, does not always
guarantee crack-free welds in the case of thick multiple-pass welds. In the present case,
the 16 h period has obviously been inadequate to allow for complete hydrogen diffusion
in the 550 MPa yield strength OK 74.78 weld metals, hence resulting in WM hydrogen
cracking after a remarkably long period of time.

4.1.1.3 Large-scale double-welded Y-Groove tests

To confirm the validity of small-scale tests in assessing the cracking risk of larger
structures, a double-welded large-scale Y-Groove test specimen was specifically designed
at VTT, pursuing to verify the small-scale test results. Groove geometry of the specimen
was, as such, in accordance with the NSC Y-Groove experiments c.f. Figs. 15 and 19,
except that the specimen was lengthened up to 800 mm. The weld build-up thickness aw
 92

and the plate thickness were 50 and 60 mm, respectively. Again, heavily bolted clamp
fixtures were used to ensure structural rigidity, heavy restraint conditions and negligible
distortion and deformations during welding.
The use of the YDT specimen allows the assessment of the cracking sensitivity of two
different weld metals (made using two different consumables) in one single test
specimen. To make the multipass test weld, about ½ of the groove length was deposited
with the 1st electrode, whilst the remaining ½ specimen length was filled with another
electrode. Besides the absolute cracking risk, this technique enabled also the assessment
of the relative cracking risk between the two different weld metals under equivalent
thermal and structural conditions. According to the literature[2, 3, 8, 12, 19, 67], the applied 800
mm specimen should provide sufficient weld length for the development of longitudinal
residual stress σresL so as to finally achieve its maximum level, i.e., weld metal yield
strength Rp0.2. The test matrix and the cracking test results for Series YDT are shown in
Table 5.
Table 5. Double-welded Y-Groove cracking tests for SMAW weld metals made at VTT
using intensified cooling.
Specimen Filler Arc Current Voltage Travel HD Ti ti Total Crack /
Material energy (A) (V) speed (ml/100 g (°C) (min No of No
1st (kJ/mm) (cm/min) DM IIW) ± sec) bead Crack
2nd layers
Y-DT1 OK75.75 3.0 250 24 12 13.5 ≤ 90 4 ± 02" 38 C
OK74.78 3.0 250 24 12 13.5 ≤ 90 4 ± 02" NC
Y-DT2 OK74.78 3.0 250 24 12 18.9 ≤ 81 3 ± 07" 38 NC
OK48.08 3.0 250 24 12 8.8 ≤ 81 3 ± 07" NC

According to Table 5, the results of the double-welded Y-Groove cracking tests

confirm the findings of the U-Groove tests shown in Table 4. Under similar weld
hydrogen content and equivalent thermal and structural conditions, the extra-high
strength Ni-Cr-Mo alloyed OK 75.75 weld metal with Rp0.2 ≈ 690 MPa exhibited
remarkably higher WM hydrogen cracking sensitivity than the lower-strength Mo-alloyed
OK 74.78 weld metal with Rp0.2 ≈ 550 MPa. As an example, a photograph of
characteristic transverse hydrogen crack found from the OK 75.75 weld metal in
specimen YDT1 is shown in Fig. 20.
Table 5 shows that even under most stringent conditions of HD = 18.9 ml/100 g DM
(IIW) and ti ≈ 3 min, hydrogen cracking did not occur in the OK 74.78 weld metal,
provided a 16 h period conforming to EN 1011-2 : 2001[1b] was applied. The results of
double-welded large-scale Y-Groove tests are hence in accordance with the small-scale
cracking test data.
Consequently, it is perceived, all the specimen and groove configurations studied in
the present thesis yielded compatible results in terms of WM cracking sensitivity.
 93

Fig. 20. Characteristic transverse weld metal hydrogen cracks found from the OK 75.75 weld
metal side in specimen YDT1 (large-scale Y-Groove test, upper 2/3rds: electrode OK 75.75.,
lower 1/3rd: electrode OK 74.78).
 94

4.1.1.4 Metallographic examination of crack morphologies

Metallographic investigation revealed that the extra-high strength weld metals: Nittetsu
L-80, OK 75.75, OK 75.78 and Atom Arc 12018 with Rp0.2 ≥ 690 MPa exhibited
transverse, or slightly curved, crack morphologies characteristic to hydrogen cold cracks
associated[2, 3, 12, 19, 54, 74, 76, 100, 106] with high-strength multipass weld metals. As an
example, a photograph of weld metal hydrogen cracks found from weld N3 is shown in
Fig. 21.
The cracks were found to having been initiated below the final weld bead layer. In
most cases, they had entered the weld surface. Simultaneously, cracks had been
propagating normal to the weld surface and grown vertically downwards into the
approximately 1/3–1/2-depth of the weld build-up thickness aw. As examples,
photographs of WM hydrogen cracks found from welds U11 and YDT1 are shown in Figs
20 and 22, respectively.
Unlike in extra-high strength SMAW welds, transverse WM hydrogen cracks recorded
from high-strength OK 74.78 weld metals of Rp0.2 ≈ 550 MPa, did not enter the weld
surface, even though they – as those in the extra-high strength WMs – had been initiated
below the final weld layer. Instead of propagating normal to the weld surface in the weld
thickness direction, these cracks tended to propagate at angle of about 45° downwards
into the approximately 1/3–1/2-depth of the weld build-up thickness aw. These cracks
thereby exhibited a characteristic Chevron-type[54] morphology, see Fig. 20. As an
example, a photograph of WM cracks found from weld U31 and associated with the
prolonged period of 168 h (7 days) before the NDT, is shown in Fig. 23.

Fig. 21. Example of the morphology of transverse hydrogen crack recorded from the L-80
weld metal in specimen N3 (Y-Groove test, electrode: Nittetsu L-80) (25 x).
 95

Fig. 22. Characteristic transverse weld metal hydrogen cracks recorded from the OK 75.75
weld metal in specimen U11 (U-Groove test, electrode: OK 75.75).

Fig. 23. Characteristic weld metal Chevron crack recorded from the OK 74.78 weld metal in
specimen U31 associated with the period of 168 h (U-Groove test, electrode: OK 74.78).
 96

4.1.2 Cracking tests for SAW thick plate weld metals

The SAW cracking test programme on thick plate weld metals consisted of (i)
preliminary small-scale U-Groove cracking tests made at ESAB, Finland, (ii) a
comparatively large set of small-scale U-Groove cracking tests made at ESAB, U.K.,
denoted as ‘final test series’ and designed according to the principles of the statistical
Taguchi analysis, and (iii) complementary small-scale U-Groove cracking tests made at
ESAB, U.K. and Mäntyluoto Works, Finland.
Of these, the preliminary SAW cracking tests in Item (i) were made under intensified
cooling conditions similar to those for the SMAW experiments, aiming at defining lower-
bound estimates of the boundary curve between Crack – No Crack regions with respect to
strength and hydrogen. All the other cracking tests according to Items (ii) and (iii), in
turn, employed elevated preheat/interpass temperatures, with the intention to determine
‘safe’ Tcr estimates.

4.1.2.1 Small-scale U-Groove tests

The SMAW cracking test data in Tables 3–5 showed that all groove configurations and
specimen types yielded practically compatible results in terms of the weld metal cracking
sensitivity. One of the applied variants, i.e., the U-Groove test was therefore chosen for
the entire SAW cracking experiments of the present study.
In the cases of preliminary SAW cracking experiments, the final SAW cracking test
series, the complementary tests at ESAB, U.K. and the complementary tests at
Mäntyluoto Works, Finland, the U-Groove specimen sizes were 500 x 500 x 60 mm, 500
x 500 x 70 mm, 500 x 500 x 70 mm and 510 x 800 x 60 mm, respectively, c.f. Fig. 18 on
page 76. A rectangular U-groove having a depth and width of 40 mm was used
constantly, apart from the specimen thickness and size. The fixing arrangements of
specimens, as well as the NDT inspections and routines were similar to those applied for
the SMAW U-Groove tests and described in Section 4.1.1.2. The minimum period before
final NDT inspection was 16 h in accordance with EN 1011-2 : 2001[1b].
Table 6. Preliminary U-Groove cracking tests for SAW weld metals at ESAB, Finland,
using intensified cooling.
Specimen SAW wire/ Arc Current Voltage Travel HD Ti ti Total Crack /
No OK Flux energy (DC+) speed (ml/100 g range (min no of No
10.62 (kJ/mm) (A) (V) (cm/min) DM IIW) (°C) ± sec) layers Crack
Rp0.2 ≈ 530 MPa, RM ≈ 620 MPa
SU1 OK 15.24S 5.0 570 32 22 6.2 125–155 4 ± 05 20 NC
Rp0.2 ≈ 580 MPa, RM ≈ 670 MPa
SU4 OK 15.26S 4.0 610 37.5 34 13.6 93–105 7 ± 05 31 C
Rp0.2 ≈ 760 MPa, RM ≈ 825 MPa
SU2 OK 15.27S 4.0 620 37.5 35 7.2 71–82 7 ± 05 30 C
SU3 OK 15.27S 4.0 620 37.5 35 9.1 71–96 7 ± 05 30 C
 97

The test matrix and results of preliminary SAW U-Groove cracking tests are shown in
Table 6. Of the experiments, specimens SU4, and SU2 and SU3, welded with high-
strength OK Tubrod 15.26S and extra-high strength OK Tubrod 15.27S tubular wires,
respectively, exhibited characteristic transverse WM hydrogen cracks. The cracks in SU4,
SU2 and SU3 were associated with weld diffusible HD of 13.6, 7.2 and 9.1 ml/100 g DM
(IIW), respectively. Contrary to these, specimen SU1 welded with a lower strength OK
Tubrod 15.24S tubular wire was found crack-free.
Test matrix and results of the final SAW U-Groove cracking test series made at ESAB,
U.K., are shown in Table 7. The objective here was to design a test matrix according to
the principles of a statistical Taguchi analysis. Consequently, for each of the three weld
metal strength classes, 3 variables (i) arc energy, (ii) preheat/interpass temperature and
(iii) weld hydrogen content, were taken, all them at three separate levels, as shown in
Table 7.
Table 7. Final U-Groove cracking tests for SAW weld metals at ESAB, U.K. and using
elevated preheat/interpass temperatures.
Specimen SAW wire/ Arc Current Voltage Travel HD Ti ti Total Crack /
No OK Flux energy (DC+) speed (ml/100 g range (min no of No
10.62 (kJ/mm) (A) (V) (cm/min) DM IIW) (°C) ± min) layers Crack
Rp0.2 ≈ 580 MPa, RM ≈ 670 MPa
395 OK 15.26S 2.0 550 30 50 4.9 100–115 5 42 NC
593 OK 15.26S 2.0 550 30 50 4.9 93–109 5±2 46 NC
396 OK 15.26S 3.0 550 30 33 6.1 125–132 5+4 31 NC
397 OK 15.26S 4.7 550 30 21 8.4 133–152 7±1 20 NC
Rp0.2 ≈ 760 MPa, RM ≈ 825 MPa
398 OK 15.27S 2.0 550 30 50 6.1 144–157 5 45 C
399 OK 15.27S 3.4 550 30 29 8.4 82–111 7–2 28 C
400 OK 15.27S 4.0 550 30 25 4.9 124–135 7+1 21 C
594 OK 15.27S 5.0 550 30 20 4.9 121–138 7+6 20 C
Rp0.2 ≈ 890 MPa, RM ≈ 960 MPa
401 OK 15.29S 2.0 550 30 50 8.4 105–129 5+1 44 C
402 OK 15.29S 3.0 550 30 33 4.9 143–160 5+4 28 C
403 OK 15.29S 4.0 550 30 25 6.1 93–116 7+2 20 C

According to Table 7, two essential observations are perceived. Firstly, the risk of
transverse weld metal hydrogen cracking is seen to increase dramatically with WM
strength, particularly for welds made with OK Tubrod 15.27S and OK Tubrod 15.29S
exhibiting nominal yield strengths of Rp0.2 ≥ 700 MPa. This is in line with the results of
the SMAW U-Groove and preliminary SAW U-Groove experiments, compare Tables 4,
6 and 7. Overall, the results demonstrate that for the extra-high strength weld metals of
Rp0.2 ≥ 700 MPa, HD levels of about 5 ml/100 g DM (IIW) and above are high enough to
provoke cracking, even though moderate interpass temperatures of 125–150°C are
employed. At lower WM yield strengths of Rp0.2 ≤ 580 MPa, weld HD levels above 10
ml/100 g DM (IIW) seem to be necessary in order to create critical conditions for
inducing WM cracking.
What is more surprising is that whilst all the 7 specimens welded with OK Tubrod
15.27S and OK Tubrod 15.29S displayed characteristic WM transverse hydrogen cracks,
none of the 4 specimens welded with OK Tubrod 15.26S revealed any cracks.
Consequently, whether WM cracking did, or did not, occur, both these phenomena took
 98

place totally irrespective of the other parameters: weld hydrogen, arc energy and
preheat/interpass temperature.
For quantitative assessment of the cracking tendency, sc. ‘severity grading’ was
defined for each weldment on the basis of NDT magnetic inspection, ultrasonic
inspection and X-ray examination, all performed at Rautaruukki Oyj, Finland. These
grades are given in Table 8.
According to Table 8, ‘cracking severity grading’ seems not to have any unified
relation to the WM cracking conditions in terms of weld hydrogen or preheat/interpass
temperature. For instance, test No 5 (weld 399) with the highest HD content and the
lowest Ti level among the OK Tubrod 15.27S weld metals, did not display the
correspondingly high severity grading, but rather the opposite. The only noticeable trend
is that, in general, OK Tubrod 15.29S weld metals show higher severity grading than OK
15.27S weld metals. Thus, the WM strength is, again, found to affect the cracking
susceptibility, whilst the severity grading seemed practically insensitive to hydrogen,
interpass temperature and heat input.
Table 8. Final SAW U-Groove cracking tests at ESAB, U.K., using intensified cooling.
Test Weld SAW Filler Wire / Weld hydrogen Arc energy Interpass Crack – Cracking
No OK Flux 10.62 level*) temperature No Crack Severity
(ml/100 g DM IIW) (kJ/mm) (°C) Grading**)
1 U395 OK Tubrod 15.26S 4.9 (7) 2.0 100 NC 0
2 U396 OK Tubrod 15.26S 6.1 (10) 3.0 125 NC 0
3 U397 OK Tubrod 15.26S 8.4 (15) 4.0 150 NC 0
4 U398 OK Tubrod 15.27S 6.1 (10) 2.0 150 C 5;8
5 U399 OK Tubrod 15.27S 8.4 (15) 3.0 100 C 4;5
6 U400 OK Tubrod 15.27S 4.9 (7) 4.0 125 C 2;8
7 U401 OK Tubrod 15.29S 8.4 (15) 2.0 125 C 6;8
8 U402 OK Tubrod 15.29S 4.9 (7) 3.0 150 C 7;9
9 U403 OK Tubrod 15.29S 6.1 (10) 4.0 100 C 3;9
10 U593 OK Tubrod 15.26S 4.9 (7) 2.0 100 NC 0
11 U594 OK Tubrod 15.27S 4.9 (7) 4.0 125 C 5;9

*)
: measured values confirmed; target levels defined according to flux humidity and given in
brackets ()
**)
: 1st value: defect density index, 2nd value: defect depth index; scale from 1 (least severe) to 10
(most severe)
Table 9 shows the results of complementary SAW U-Groove cracking tests for OK
Tubrod 15.27S and Tubrod 15.29S weld metals made at ESAB, U.K. and Mäntyluoto
Works, Finland, respectively. As the cracking sensitivity of these two weld metals
appeared particularly intense, the complementary experiments aimed at finding ‘safe’
preheat/interpass temperatures for both grades at a constant weld hydrogen level. Here, it
was felt realistic to choose hydrogen levels commonly faced with SAW in practice
welding fabrication, i.e., ≈ 4–7 ml/100 g DM (IIW).
 99

Table 9. Complementary U-Groove cracking tests for SAW weld metals at ESAB, U.K. &
Mäntyluoto Works, Finland, and using elevated preheat/interpass temperatures.
Specimen SAW wire/ Arc Current Voltage Travel HD Ti ti Total Crack –
No OK Flux energy (DC+) speed (ml/100 g range (min no of No
10.62 (kJ/mm) (A) (V) (cm/min) DM IIW) (°C) ± min) layers Crack
Rp0.2 ≈ 760 MPa, RM ≈ 825 MPa
U907 OK 15.27S 2.0 550 30 50 6.5 114–132 5 + 0.5 46 C
U908 OK 15.27S 2.0 550 30 50 6.5 168–183 5 + 2 42 C
U905 OK 15.27S 2.0 550 30 50 6.5 189–203 4 ± 2 45 C
U906 OK 15.27S 2.0 550 30 50 6.5 218–233 4 + 0.5 45 C
U144 OK 15.27S 2.0 550 30 50 6.5 243–254 3 + 1 45 C
Rp0.2 ≈ 890 MPa, RM ≈ 960 MPa
MW1 OK 15.29S 4.0 550 30 25 6.0 250–253 7±2 23 C
MW3 OK 15.29S 4.0 550 30 25 6.0 275–278 6±3 25 C
MW4 OK 15.29S 4.0 550 30 25 6.0 285–288 6±3 25 NC
MW2 OK15.29S 4.0 550 30 25 6.0 300–305 6±3 24 NC

It can be seen that, at the applied weld HD of 6.0–6.5 ml/100 g DM (IIW), the levels of
‘safe’ preheat/interpass temperature for the OK Tubrod 15.27S and Tubrod 15.29S weld
metals lie above 250°C and around 285–300°C, respectively. Even considering the
comparatively high strength of these two SAW weld metals, the recorded
preheat/interpass temperatures required for crack-free welds appeared somewhat higher
than anticipated previously.

4.1.2.2 Metallographic examination of crack morphologies

In accordance with SMAW, metallographic investigation revealed that the extra-high

strength SAW weld metals: OK Tubrod 15.27S and OK Tubrod 15.29S with Rp0.2 ≥ 760
MPa exhibited transverse, or slightly curved, crack morphologies characteristic to
hydrogen cold cracks associated[2, 3, 12, 19, 54, 74, 76, 100, 106] with high-strength multipass weld
metals.
As in the case of SMAW, the cracks in the SAW weld metals were found to initiate
below the final weld bead layer. In the majority of the examined SAW welds that
revealed hydrogen cracks, they had grown towards, and were finally entering, the weld
surface. Simultaneously, these cracks had been propagating normal to the weld surface in
the weld thickness direction and finally grown vertically into the approximately ½-depth
of the weld build-up thickness aw.
For the welds associated with the lowest HD of 4.9 ml/100 g DM (IIW), however, this
was not the case and the cracks tended to be sub-surface cracks, either remaining
immediately beneath the final filling layer, or growing downwards into the approximately
½-depth of the weld build-up thickness aw.
As examples, photographs of transverse WM hydrogen cracks exhibiting the crack
orientation of ‘straight vertical propagation in the weld thickness direction’ are shown for
OK Tubrod 15.27S and OK Tubrod 15.29S weld metals in Figs 24a and 24b,
 100

respectively. Example photographs of transverse WM hydrogen cracks exhibiting a

‘staircase patter’ along its path of propagation into the weld thickness direction are shown
for OK Tubrod 15.27S and OK Tubrod 15.29S weld metals in Figs 25a and 25b,
respectively. Example photographs of transverse WM hydrogen cracks locating
immediately below the final filling layer as sub-surface cracks and associated with low
weld hydrogen contents are shown for OK Tubrod 15.27S and OK Tubrod 15.29S weld
metals in Figs 26a and 26b, respectively.

a) b)

Fig. 24. Examples of characteristic transverse weld metal hydrogen cracks exhibiting the
morphology of ‘straight vertical propagation into weld thickness direction’, recorded from
(a) OK Tubrod 15.27S and (b) OK Tubrod 15.29S weld metals in specimens U399 and U402,
respectively (U-Groove test).

a) b)

Fig. 25. Examples of characteristic transverse weld metal hydrogen cracks exhibiting the
morphology of a ‘staircase patter’ along its path of propagation into weld thickness direction,
recorded from (a) OK Tubrod 15.27S and (b) OK Tubrod 15.29S weld metals in specimens
U398 and U403, respectively (U-Groove test).
 101

a) b)

Fig. 26. Examples of transverse weld metal hydrogen cracks exhibiting a sub-surface
appearance, i.e., cracks locating immediately below the final filling layer as sub-surface
cracks and associated with low weld hydrogen contents, recorded from (a) OK Tubrod 15.27S
and (b) OK Tubrod 15.29S weld metals in specimens U594 and U402, respectively (U-Groove
test).

In one case, transverse WM hydrogen cracks were recorded also from the ‘moderate
strength’ OK Tubrod 15.26S weld metals of Rp0.2 ≈ 580 MPa, however, this was
associated with high weld HD of 13.6 ml/100 g DM (IIW) in specimen SU4, see Table 6.
Unlike in extra-high strength SAW welds, cracks in the OK Tubrod 15.26 weld metal in
SU4 did not enter the weld surface, although they were revealed to having been initiated
below the final filling layer. Instead of propagating normal to the weld surface, these
cracks had propagated at angle of 45° in the weld thickness direction and into the
approximately ½-depth of the weld build-up thickness aw. These cracks were hence
similar to those found from the SMAW in the case of OK 74.78 weld metal having Rp0.2
≈ 550 MPa, and exhibited a Chevron-type[54] morphology. As an example, photographs of
WM cracks found from the weld SU4 are shown in Figs 27a and 27b.
 102

b)

a)

Fig. 27. Characteristic weld metal Chevron crack recorded from the OK Tubrod 15.26S weld
metal in specimen SU4 and associated with high weld hydrogen (a) macrograph,
(b) microstructure (U-Groove test).

4.1.3 Comparative cracking tests for welds in thin plate

As comparative tests to heavy plate multiple-pass welds, a separate set of V-Groove

cracking tests were conducted for two-pass weld metals in a thin plate. The experiments
were made on SMAW and SAW weld metals using OK Tubrod 15.27S tubular wire and
OK 75.75 electrode, respectively, in accordance with the consumable characteristics in
Table 2 on p. 73.
For the V-Groove cracking tests, 6 mm thick plate specimens were applied. The
specimen size was 500 x 500 x 6 mm. The V-Groove preparation employed here had 60°-
groove angle and no root face, which made the adoption of a separate root backing
necessary. Welding was done as one-side welding using 2 passes, one pass per layer.
Intensified cooling similar to that in the case of thick plate experiments was applied for
the SMAW experiments, however, in the case of SAW experiments the hydraulic fixtures
of the SAW equipment prevented the appropriate placement of any slices of CO2 ice onto
test specimens.
The intention here was to examine the dependency of the WM cracking risk on plate
and weld build-up thickness aw under equivalent strength and hydrogen in relation to
 103

thick multipass welds. The results of the V-Groove cracking experiments are summarised
in Table 10.
Table 10. Comparative SAW and SMAW V-Groove cracking tests for 6 mm thick
specimens made at VTT and HUT, using intensified cooling and air cooling in the case of
SMAW and SAW experiments, respectively.
Specimen SAW wire / Arc Current Voltage Travel HD Ti ti Total Crack –
No OK Flux energy speed (ml/100 g max (min no of No
10.62 (kJ/mm) (A) (V) (cm/min) DM IIW) (°C) ± sec) layers Crack
SMAW Rp0.2 ≈ 680 MPa, RM ≈ 810 MPa
T-1 OK 75.75 1.3 108 23 12 14.4 68–74 4 ± 02" 2 NC
T-2 OK 75.75 1.3 108 23 12 9.2 69–76 4 ± 02" 2 NC
SAW Rp0.2 ≈ 700 MPa, RM ≈ 820 MPa
S-1 OK 15.27S 1.9 580 28 50 16.0 60–65 4 ± 15" 2 NC
S-2 OK 15.27S 1.9 580 28 50 16.0 63–66 4 ± 15" 2 NC

It can be seen that none of the SMAW or SAW experiments made using thin plate
two-pass welded V-Groove specimens resulted in transverse WM hydrogen cracks at any
of the applied weld hydrogen contents ranging from HD ≈ 9.2–16.0 ml/100 g DM (IIW).
In this respect, the risk of WM hydrogen cracking, at a comparable level of weld metal
strength and weld hydrogen, is found significantly less for the 6 mm weld build-up
thickness in thin plate than in the case of a 40 mm weld build-up thickness in heavy plate
multiple-pass welds, compare Table 10 to Tables 2–9.

4.2 All-weld metal tensile properties

Tensile testing was carried out for the SMAW multipass weld metals at room temperature
in accordance with SFS-EN 10002-1. The intention was to measure the all-weld metal
properties; consequently, round-bar specimens (∅ 7 mm) according to SFS 3471 were
used. The specimens were extracted from the final weld beads of the filling layers.
Summary data of the tensile test results for all the SMAW specimens are given in
Table 11.
All welds were made using an ‘equivalent’ arc energy level of 3.0 kJ/mm[4], meaning
that the actual arc energy values for multiple-pass weld metal pads and bevel groove butt
welds were adjusted to yield equable weld cooling conditions in terms of weld t8/5. Thus,
2.4 kJ/mm arc energy in the OU experiments accrues from the bead-on-plate like
conditions for the weld metal pads, whereas 3.0 kJ/mm was applied in the VTT
experiments that were made using a bevel groove preparation. A lower arc energy was
used to compensate a somewhat lower weld cooling rate t8/5 in bead-on-plate like welding
(shape factor: F ≈ 1.0) of multiple-pass pads, compared with welding into a bevel V-
groove (shape factor: F ≈ 0.8). All the test results are therefore regarded comparable in
this sense. A more detailed description has been given elsewhere[4].
 104

Table 11. Summary of the tensile test data for the experimental SMAW multipass weld
metals.
Specimen Filler material Arc energy Yield Tensile Elongation
code E strength strength A5
(Electrode) (kJ/mm) Rp0.2 RM (%)
(MPa) (MPa)
OU1 OK 48.08 (Ni) 2.4 455 525 27.9
OU1.1 2.4 456 530 33.4
VTT3-5 2.4–3.2 503 578 30.5
VTT1 OK 74.78 (Mo) 3.0 550 650 18.1
OU2 2.4 470 590 22.3
OU2.1 2.4 541 624 30.1
VTT6-8 2.4–3.2 552 640 30.1
VTT2 OK 75.75 (Ni-Cr-Mo) 3.0 680 810 14.6
OU3 2.4 680 780 17.4
OU3.1 2.4 718 818 22.3
NSC15 Nittetsu L-80 (Ni-Cr-Mo) 1.5 790 860 17.7
NSC30 3.0 745 845 19.9
NSC50 5.0 640 850 17.5
OU4 OK 75.78 (Ni-Cr-Mo) 2.4 883 963 19.6
VTT9-11 Atom Arc 12018 (Ni-Cr-Mo) 1.6–3.2 890 1022 19.2

4.3 Weld metal hardness and chemical composition

Weld metal hardness survey was carried out for (i) SMAW weld metals representing the
cracking test welds, (ii) a set of SMAW multiple-pass weld metal pads made with
different arc energies, (iii) preliminary SAW cracking test welds associated with high arc
energies, and (iv) the final SAW cracking test series made using different arc energies
and elevated preheat/interpass temperatures.
The NSC and VTT specimens[4, 5b] were extracted from the actual SMAW Y- and U-
groove cracking test welds, respectively, made at 3.0 kJ/mm arc energy. The OU
specimens[4, 5b] were made as bead-on-plate like SMAW multiple-pass pads (F ≈ 1.0)
using 2.4 kJ/mm arc energy adjusted to yield weld t8/5 cooling time equivalent to that
associated with the bevel groove preparation (F ≈ 0.8) in the NSC and VTT
experiments[4, 5b]. In addition, SMAW pads were made using three arc energy levels: 1.6,
2.4 and 3.2 kJ/mm, to reveal the influence of arc energy on weld metal hardness.
The SU and U specimens were extracted from the preliminary and final SAW U-
groove multipass cracking tests welds, respectively, made at various arc energies and, in
the case of the latter test series, using also different preheat/interpass temperatures. The
preliminary SAW cracking test welds were associated with high arc energies of 4.0–5.0
kJ/mm and were made applying a constant preheat/interpass temperature, see Table 6.
The final SAW U-groove cracking test welds, in turn, were associated with three
different levels of arc energy: 2.0, 3.0 and 4.0 kJ/mm and under preheat/interpass
temperatures of 100, 125 and 150°C.
 105

Table 12. Summary of weldability indexes, Pcm, CET, and weld metal Vickers hardness
values HV5 for the SMAW weld metals at different arc energy levels.
Specimen SMAW Arc energy Pcm CET HVave HVmax
No Electrode as-welded/
(kJ/mm) (%) (%) reheated (HV)
NSC15 Nittetsu L-80 1.5 0.233 0.325 317 342
VTT1.6 1.6 0.201 0.294 286 / 296 303
VTT2.4 2.4 – – 269 / 270 290
NSC30 3.0 0.235 0.325 304 319
VTT3.2 3.2 0.215 0.313 261 / 260 268
NSC50 5.0 0.232 0.321 291 301
VTT1/1.6 OK 74.78 1.6 0.156 0.232 235 / 216 239
VTT1/2.4 2.4 0.155 0.232 234 / 218 237
VTT-U20 3.0 0.160 0.225 – –
OU2 2.4 0.165 0.237 215 230
VTT1 3.0 0.163 0.238 217 232
VTT1/3.2 3.2 0.155 0.232 219 / 191 225
VTT2/1.6 OK 75.75 1.6 0.222 0.324 295 / 287 308
VTT2/2.4 2.4 0.221 0.322 292 / 281 299
OU3 2.4 0.222 0.321 271 281
VTT2 3.0 0.214 0.308 269 279
VTT2/3.2 3.2 0.216 0.316 276 / 273 280
T-1 1.3 0.180 0.254 273 280
VTT3/1.6 OK 48.08 1.6 0.136 0.192 208 / 185 216
VTT3/2.4 2.4 0.131 0.186 215 / 185 221
VTT-U7 3.0 0.129 0.182 – –
OU1 2.4 0.132 0.183 189 209
VTT3 3.0 – – – –
VTT3/3.2 3.2 0.127 0.182 198 / 179 204
VTT4/1.6 Atom Arc 1.6 0.286 0.410 353 / 353 362
VTT4/2.4 12018 2.4 0.281 0.402 337 / 358 365
VTT4 3.0 0.294 0.416 – –
VTT4/3.2 3.2 0.275 0.399 325 / 326 338
VTT5/1.6 OK75.78 1.6 0.287 0.434 (256) / 330 351
VTT5/2.4 2.4 0.277 0.411 335 / 342 351
OU4 3.0 0.273 0.408 – –
VTT5/3.2 3.2 0.256 0.390 326 / 332 338

* specimens VTT1-VTT5 : VTT Y-groove cracking tests: 3.0 kJ/mm

* specimens VTT-U7, VTT-U20 : VTT U-groove cracking tests: 3.0 kJ/mm
* specimens OU1-OU5 : OU multiple-pass welded pads: 2.4 kJ/mm
* specimens NSC15, NSC30 and NSC50 : NSC Y-groove cracking tests: 1.5, 3.0 and 5.0 kJ/mm
* specimens VTT1/1.6-VTT5/1.6 : VTT multiple-pass welded pads: 1.6 kJ/mm
* specimens VTT1/2.4-VTT5/2.4 : VTT multiple-pass welded pads: 2.4 kJ/mm
* specimens VTT1/3.2-VTT5/3.2 : VTT multiple-pass welded pads: 3.2 kJ/mm
* specimen T-1 : VTT V-groove cracking tests: 2-pass welded thin plate
Tables 12 and 13 summarise the results of Vickers hardness traverses (HV5) and
associated weldability indexes (Pcm, CET) for the SMAW and SAW multipass weld
metals, respectively. The Vickers hardness profiles were measured from the filling layers
 106

of the SMAW and SAW welds. Both as-welded and re-heated weld metal microstructures
were examined in order to reveal possible differences in their hardenability. The chemical
composition of all-weld metals was determined using optical emission spectroscopy.
Using these analysis data, Pcm and CET were calculated for the SMAW and SAW weld
metals applying the following formulae:
Pcm = C + Si/30 + (Mn + Cu + Cr)/20 + Ni/60 + Mo/15 + V/10 + 5 (43a)
CET = C + (Mn + Mo)/10 + (Cr + Cu)/20 + Ni/40 (43b)

Table 13. Summary of weldability indexes, Pcm, CET, and weld metal Vickers hardness
values HV5 for the SAW weld metals at different arc energy levels.
Specimen SAW tubular wire/ Arc Pcm CET HVave (HV) HVave (HV) HVmax
No Flux OK 10.62 energy As-welded/ All regions
(kJ/mm) (%) (%) Re-heated (HV)
SU1 OK Tubrod 15.24S 5.0 0.172 0.263 225 225 235
SU4 OK Tubrod 15.26S 4.0 0.205 0.302 237 / 235 234 241
SU2 OK Tubrod 15.27S 4.0 0.256 0.382 – 313 334
SU3 OK Tubrod 15.27S 4.0 0.258 0.383 – 312 335

U395 OK Tubrod 15.26S 2.0 0.227 0.333 273 / 258 265 277
U593 OK Tubrod 15.26S 2.0 0.238 0.345 265 / 262 264 277
U396 OK Tubrod 15.26S 3.0 0.226 0.334 252 / 256 251 262
U397 OK Tubrod 15.26S 4.7 0.216 0.322 251 / 254 249 262

U398 OK Tubrod 15.27S 2.0 0.264 0.396 313 / 310 315 336
U399 OK Tubrod 15.27S 3.4 0.258 0.388 328 / 333 313 367
U400 OK Tubrod 15.27S 4.0 0.271 0.402 303 / 326 315 345
U594 OK Tubrod 15.27S 5.0 0.265 0.387 334 / 353 348 362
S-1 OK Tubrod 15.27S 1.9 0.210 0.341 – 257 266

U401 OK Tubrod 15.29S 2.0 0.271 0.399 348 / 341 345 367
U402 OK Tubrod 15.29S 3.0 0.273 0.402 347 / 331 339 362
U403 OK Tubrod 15.29S 4.0 0.283 0.419 336 / 373 354 376

* specimens No SU1-SU4 : preliminary SAW U-groove cracking tests at ESAB, Finland

* specimens No U395-U403 and U593-U594 : final SAW U-Groove cracking tests at ESAB, U.K.
* specimen No S-1 : SAW V-Groove cracking tests: 2-pass welded thin plate made at HUT &
VTT, Finland
Comparing the data in Tables 12 and 13 reveals that the SAW Tubrod 15.27S weld
metals having nominal yield strength Rp0.2 ≈ 690–700 MPa exhibit somewhat higher
hardness, as well as greater Pcm and CET values than, for instance, the SMAW OK 75.75
weld metals of true Rp0.2 ≈ 690 MPa. Presumably, this explains why, at the similar level
of apparent strength, the SAW weld metals are seen to exhibit WM hydrogen cracking at
considerably lower weld hydrogen contents than the corresponding SMAW welds, as
shown in Tables 4, 5, 6 and 7.
 107

4.4 Residual stresses in weld metal

Weld residual stress measurements were carried out for multipass SMAW thick welds
applying a Ring-Core method. The method uses a drilling tool that enables stresses below
the weld surface to be recorded in steps of ≈ 5 mm through the whole weld depth, if
necessary. The stresses are measured using strain gauges placed into a drilled hole in a
way that the relaxation of stresses caused by material removal, as the hole is machined, is
recorded as micro-strains and converted to stresses applying mathematical algorithms.
The Ring-Core method was regarded advantageous in view that the weld longitudinal
tensile residual stress that was of primary interest here, have been shown[19,20,23] to reach
its maximum at a depth of about 10–25% from the plate surface.
The principles of the Ring-Core method are shown in Appendix 3. The measurements
were made on two 40 mm thick Y-Groove cracking test specimens HL40 and HL402, c.f.,
Figs 15 and 22, having the weld lengths of 300 and 800 mm, respectively. The specimen
length in the Y-Groove test corresponds to the weld length. Both specimens had the width
of 300 mm, weld thickness aw of 30 mm and were welded at 3.0 kJ/mm using OK 74.78
electrode. The specimens were checked using NDT to ensure they did not contain any
cracks.
To investigate any possible effects plate thickness might have on residual stress
development, comparative measurements were made for 6 mm thick, two-pass welded
thin plates of 500 mm in length. Here, the weld build-up thickness aw was 6 mm. The
SAW and SMAW experiments were welded using OK Tubrod 15.27S tubular wire and
OK 75.75 electrode, respectively. For thin plate specimens, two restraining methods were
applied: bolted clamps similarly to SMAW heavy plates, and welded strong backs.
Because of small plate thickness, conventional hole-drilling method was applied.
The results of the residual stress measurements applying the Ring Core and the hole-
drilling methods are presented in Appendix 3. The residual stress components are given to
the weld longitudinal direction (Sa), weld transverse direction (Sc) and to the angle
orientation of 45° between them (Sb). All stresses are plotted against the actual weld
depth. The stress analyses were made using two calculation functions developed at
Helsinki University of Technology (HUT) and Siemens KWU.

4.4.1 Ring-Core measurements for thick plate multipass welds

In the case of thick multipass welds, the results demonstrate that for both specimens, the
first maximum peak of the weld longitudinal residual stress, σresL ≡ Sa, is recorded at the
depth of about 6–10 mm from the weld surface. This location corresponds comparatively
well to that below the final, or the 2nd last, weld bead layer, which is in line with the
findings reported elsewhere[3, 5, 12, 19, 23, 54, 76].
Comparing the maximum peak stress for the 300 mm long OK 74.78 weld in Appendix
3 to its true yield strength (i.e., Rp0.2 for specimen VTT1) in Table 11, shows that the
recorded maximum σresL of 510–520 MPa is only about 30–40 MPa lower than the weld
 108

metal Rp0.2 of 550 MPa. This implies that even with a specimen (and weld) of no more
than 300 mm in length, structural conditions in terms of longitudinal residual stress
development are fairly representative of that in real welded structures.
A similar comparison between σresL and Rp0.2 in the case of the 800 mm long weld
revealed the σresL level practically equivalent to the weld metal Rp0.2. This is consistent
with the fact that the weld longitudinal residual stress have been shown[19, 23, 67] to
increase towards its theoretical maximum, that is, the true weld metal Rp0.2, as the
specimen and weld length increase enough.

4.4.2 Hole drilling measurements for thin plate welds

The results for thin plates demonstrate that the way the specimen are being fixed and
restrained, affects essentially the residual stress development, see Appendix 3.
In the case of SMAW and SAW specimens attached by bolting them onto the welding
table using clamps, longitudinal residual stress σresL remains considerably lower than that
recorded for thick multipass welds, i.e., only about 0.35–0.56% of the true WM yield
strength. For the SAW specimens restrained by strong-backs welded on the bottom side,
the σresL was found clearly higher than in the case of bolt-attached specimens, i.e., about
0.71–0.94% of the true WM yield strength.
As expected, transverse σres remained in all these cases practically nil, or even
compressive. This can be attributed[25, 49, 65, 66] to low restraint of thin plate against weld
transverse shrinkage.
It is noteworthy that owing to inaccuracies of hole drilling as a method, the true
longitudinal σresL can be about 10–40% lower[5c] than the measurements indicate.
The difference in residual stress level between the two restraining methods can
presumably be explained by differences is deformation pattern of the welded plates. For a
thin two-pass welded plate specimen, the welding residual stresses seem to become partly
relieved when the bolts and clamps have been removed after welding, even though the
plate was rigidly attached during the whole welding operation. In the case of thick
multipass welded specimens, even such a minor stress relaxation could not have occurred
because of higher structural restraint of a thick multipass weldment after welding was
finished.
In contrast to the attachment by bolted clamps, restraining accomplished via welded
strong-backs did not allow for any specimen deformation, neither during nor after
welding.

4.5 Hydrogen diffusion measurements

The aim of electrochemical hydrogen permeation tests carried out[5b, 36b] for the present
study under potentiostatic conditions was to define diffusion coefficients of hydrogen,
 109

DH, for different types of weld metals. The examined SMAW and SAW weld metals
were in accordance with Table 2a and 2b, respectively, whereas the FCAW welds
represented typical high-strength welds obtained with rutile tubular wires. The intention
was to also compare the DH values measured here to those data sets available internally at
NSC, Japan[79], and to openly published data[21, 22].
To eliminate the effect of deep traps, double- or multiple-charging techniques were
adopted[5b, 36b] instead of a single charging stage. The experiments, however,
demonstrated that even deep traps could be filled completely already during the first
charging stage, so they no longer induce any effect in subsequent chargings. As a result,
relatively stable and constant DH values were obtained in the repeated (i.e., parallel) tests.
As hydrogen diffusion during the electrochemical permeation test obeys Fick's law and
the test itself meets the boundary conditions of this law, applied experimental techniques
and analyses of the recorded data[5b] enabled the development of an accurate way for
calculating the hydrogen diffusion coefficients.
Consequently, DH values were determined for Armco iron, modern fine-grained high-
strength structural steel, as well as for the SMAW, FCAW and SAW weld metals. It was
thought, the experiments for Armco iron would serve as a reference test expected to give
DH values characteristic for clean metals practically free from hydrogen traps and hence
exhibiting high hydrogen diffusivity. The DH values for the modern fine-grained
structural steel, in turn, were expected to be somewhat lower than those for Armco iron
but, owing to the low impurity content of the parent steel, still remarkably higher than the
values for the weld metals with a greater number of hydrogen traps.

4.5.1 Hydrogen permeation tests for parent steel

The chemical composition of Armco iron used for the experiments is given in Table 14a.
The specimens were annealed at 950°C for 30 min, then cooled in a furnace to room
temperature before measuring the hydrogen diffusivity.
The mean value of hydrogen diffusion coefficient DH for Armco iron at room
temperature was found approximately 70 (±5) ∗ 10–6 cm2/sec.
Table 14a. The chemical composition of Armco iron (wt-%).
C Mn P S Impurity
0.0012 0.017 0.005 0.025 < 0.1

The chemical composition of the microalloyed high-strength low-impurity structural

steel used for the DH experiments is given in Table 14b.
Table 14b. The chemical composition of the high-strength base material (wt-%).
C Si Mn Ni V S P Nb Al
0.16 0.5 1.7 0.45 0.12 0.01 0.025 0.05 0.02
 110

The measured hydrogen diffusion coefficients DH for the parent steel at room
temperature are presented together with the results for the SMAW weld metals in
Table 17, see Section 4.5.2.

4.5.2 Hydrogen permeation tests for SMAW multipass weld metals

In the first set of experiments, multipass weld metals made with OK 74.78 and OK 75.75
electrodes were used. The chemical composition of the electrodes is given in Table 15.
The applied welding parameters and procedures are given in Table 16.
Table 15. Chemical composition of the SMAW electrodes used for DH experiments.
SMAW electrode/ C Si Mn Mo Cr Ni
Composition wt-%
OK 74.78 0.06 0.35 1.5 0.4 – –
OK 75.75 0.06 0.35 1.5 0.4 0.4 2.3

Assuming, hydrogen diffusion changes only slightly once the deep traps in weld metal
have been filled with hydrogen, no more than, say, 2 or 3 hydrogen charging treatments
in the test are needed to calculate the hydrogen diffusion coefficients. Here, several
cycles were used in each test with an increasing charging current Ic, and the first ones
were neglected in the analysis[5b, 36b]. Tests were repeated a few times (codes: I#–IV#).
The conditions where all deep traps in weld metal become filled with hydrogen before
measuring the actual DH value are thought justified by the fact that these conditions
represent multipass welding quite realistically. In the case of single-pass welds, in turn,
one can expect DH values associated with conditions where traps are not filled yet, to be
more descriptive.
Table 16. Welding procedure data for the SMAW weld metals used for DH experiments.
SMAW electrode / OK 74.78 (∅ 5 mm) OK 75.75 (∅ 5 mm)
Welding procedure data
Electrode condition 70% / 40°C / 18 h 70% / 40°C / 18 h
Weld (initial) diffusible hydrogen content 10.0 9.7
(ml/100 g DM IIW)
Welding current (A) 250 250
Arc voltage (V) 23 23
Travel speed (cm/min) 12 12
Preheat temperature (°C) 100 100
Interpass temperature (°C) 125 125
Number of passes 14 14

The results of DH measurements made at room temperature for the two examined
SMAW weld metals are summarised together with the fine-grained microalloyed parent
steel values in Table 17. The DH values are seen to vary according to the location from
 111

which the specimens were extracted. This stems from the slight differences in chemical
composition and microstructure in different locations within the weld.
Table 17. The values of hydrogen diffusion coefficient DH for the examined SMAW weld
metal specimens (at room temperature, RT).
No of charging / Weld Metal Weld Metal Microalloyed high-
DH value OK 74.78 OK 75.75 strength base material
(cm2/sec)
DH (I#) 5.5 (± 0.5) ∗ 10–6 4.0 (± 0.5) ∗ 10–6 25 (± 1) ∗ 10–6
DH (II#) 10.0 (± 0.5) ∗ 10–6 10.0 (± 0.5) ∗ 10–6 25 (± 5) ∗ 10–6
DH (III#) 8.0 (± 0.5) ∗ 10–6 3.5 (± 0.5) ∗ 10–6 –
DH (IV#) 15.5 (± 0.5) ∗ 10–6 – –

Table 18. The values for hydrogen diffusion coefficients DH collated from literature and
NSC databases.
DH value / Temperature DH : RT DH : 100°C DH : 300°C
(cm2/sec) (cm2/sec) (cm2/sec)
Base Material
1.4E-3 exp (-3200/RT) 5.9 ∗ 10–6 19 ∗ 10–6 86 ∗ 10–6
1.3E-3 exp (-3200/RT) 5.6 ∗ 10–6 18 ∗ 10–6 79 ∗ 10–6
0.12 exp (-7820/RT) 0.19 ∗ 10–6 3.4 ∗ 10–6 –
0.89 exp (-8856/RT) 0.24 ∗ 10–6 6.2 ∗ 10–6 –
NSC experiments 5.7 ∗ 10–6 – –
TWI experiments – 3.3 ∗ 10–6 65 ∗ 10–6
Heat-Affected Zone (HAZ)
0.31 exp (-7990/RT) 0.37 ∗ 10–6 6.9 ∗ 10–6 290 ∗ 10–6
NSC experiments 0.31–0.42 ∗ 10–6 – –
SMAW Weld Metals
1.22E-2 exp (-5200/RT) 1.7 ∗ 10–6 11 ∗ 10–6 130 ∗ 10–6
NSC experiments 0.16 ∗ 10–6 – –

According to Table 17, there are no significant differences in the values of DH between
OK 74.78 and OK 75.75 weld metals. As expected, the DH values for weld metal are
somewhat lower than those for the parent steel, presumably due to greater number of
traps available in the weld metal, compared to relatively clean microalloyed fine-grained
steel.
For the comparison purposes, a number of hydrogen diffusion coefficient values
collated from the literature[21, 22] and NSC databases[79] for weld metals, HAZs and base
metals are presented in Table 18.
The data in Table 18 demonstrate substantial scatter in the individual DH values
collated from different sources. The scatter is, in fact, greater among the values
summarised from various investigations even for the same welded region, than it is
among the values associated with entirely different weld regions (i.e., weld, HAZ or
parent steel).
 112

4.5.3 Hydrogen permeation tests for SAW and FCAW multipass weld
metals

The second stage of experiments comprised electrochemical permeation tests for SAW
and FCAW multipass weld metals. Rutile type FCAW wire Filarc PZ 6138 and basic
type SAW wire & flux combination of OK Autrod 13.27 & OK Flux 10.62 were used to
weld the multipass experiments. The applied welding parameters and procedures are
given in Table 19.
Table 19. Welding procedure data for the FCAW and SAW weld metals used for DH
experiments.
Consumable / Filarc PZ 6138 OK Autrod 13.27
Welding procedure data (∅ 1.2 mm) (∅ 4 mm) /
OK Flux 10.62
Electrode condition as-received as-received
Weld (initial) diffusible hydrogen content not measured not measured
(ml/100 g DM IIW)
Welding current (A) 216 700
Arc voltage (V) 25 30
Travel speed (cm/min) 46 25
Arc energy (kJ/mm) 0.7 5.0
Preheat temperature (°C) 115 115
Interpass temperature (°C) < 150 < 150

For the measurements, each welded sample was cut in thickness direction into three
separate sections representing (i) the weld upper-area (i.e., filling runs), (ii) the mid-
thickness area and (iii) the weld root area. The average values of the hydrogen diffusion
coefficient DH for the examined SAW and FCAW weld metals are presented in Figs 28a
and 28b, respectively. All measurements were made at room temperature.
According to Fig. 28a, there is a systematic tendency in all the examined SAW
specimens S10–S113 for their DH values to increase in weld thickness direction when
going from the weld upper-area towards the weld root area. Possible reasons for this
phenomenon that was not recognised in the case of the FCAW weld metals, c.f., Fig. 28b,
are discussed in Chapter 5.9.
Fig. 28b shows that in contrast to the SAW welds, the examined FCAW weld metals
display no corresponding change in DH as a function of weldment thickness. The
variation of DH in weld thickness direction is found not systematic, but relatively random,
between the examined FCAW specimens F9–F13.
 113

Fig. 28a. The average values of hydrogen diffusion coefficient DH in the filling runs (upper),
mid-thickness layers (middle) and root runs (root) for the SAW specimens.

Fig. 28b. The average values of hydrogen diffusion coefficient DH in the filling runs (upper),
mid-thickness layers (middle) and root runs (root) for the FCAW specimens.

For the comparison purposes, the typical ranges of the DH values for all the examined
SMAW, SAW and FCAW weld metals, as well as those for Armco iron and microalloyed
fine-grained high-strength steel are collated in Table 20.
According to Table 20, the measured DH values all decrease in the order of Armco
iron, fine-grained structural steel and the weld metals, as expected. This can be attributed
to the greater number of hydrogen traps available in the weld metals than in the low-
impurity fine-grained structural steel, or in the annealing heat-treated Armco iron that
represents DH values nearly characteristic of pure α-iron.
With respect to weld metals in Table 20, the DH values do not differ significantly
between SMAW and FCAW welds. The DH values recorded for the SAW filling layers
also lie relatively close to those for FCAW and SMAW weld metals, whereas those for
the SAW mid-thickness and root runs are seen to approach those for the fine-grained
structural steel.
 114

Table 20. Summary of the DH data for SMAW, SAW and FCAW weld metals, together
with the values for Armco iron and fine-grained structural steel (at room temperature).
Weld / Steel / Average DH DH range in weld DH range in weld DH range in weld
Range of DH values range filling layers mid-layers root layers
(at RT) (cm2/sec) (cm2/sec) (cm2/sec) (cm2/sec)
Armco iron 70 (±5) ∗ 10–6 – – –
fine-grained steel 25 (±5) ∗ 10–6 – – –
SMAW OK 74.78 5.5–15.5 ∗ 10–6 – – –
SMAW OK 75.75 3.5–10.0 ∗ 10–6 – – –
FCAW PZ 6138 7.9–17.5 ∗ 10–6 9.4–17.5 ∗ 10–6 10.3–14.4 ∗ 10–6 7.9–14.7 ∗ 10–6
SAW OK Autrod 7.0–38.0 ∗ 10–6 7.0–12.5 ∗ 10–6 17.0–29.0 ∗ 10–6 22.5–38.0 ∗ 10–6
13.27 / OK 10.62

4.6 Phase transformation temperature

Dilatometric measurements were made to determine the γ–α phase transformation

temperatures for the SMAW weld metals and, as a reference, for microalloyed parent
steel. The techniques and results are explained in detail elsewhere[36a].
Measurements were made for OK 48.08, OK 74.78, OK 75.75 and OK 75.78 multipass
weld metals, as well as for the RAEX 460ML parent steel using Gleeble 1500 thermal
simulator and four different thermal cycles in terms of weld t8/5 cooling time. The peak
temperature and holding time of the thermal cycle were chosen as 1350°C and 1 sec,
respectively. Four t8/5 cooling times of 10, 20, 30 and 40 sec were set, coupled with 3D
type cooling curve; these t8/5 values roughly corresponded to the arc energy levels of 1.9,
3.7, 5.6 and 7.5 kJ/mm, respectively.
The summary of the phase transformation experiments subjected to dilatometric
measurements is given in Table 21.
Table 21. Results of the dilatometric measurements with Gleeble 1500 – parent steel
RAEX 460 ML and the SMAW weld metals.
Cooling time t8/5 t8/5 = 10 s t8/5 = 20 s t8/5 = 30 s t8/5 = 40 s
(sec)
Temperature (°C) Ar1 (°C) Ar3 (°C) Ar1 (°C) Ar3 (°C) Ar1 (°C) Ar3 (°C) Ar1 (°C) Ar3 (°C)
/ Specimen
OK 48.08 570 720 590 730 605 740 610 750
OK 74.78 540 690 570 710 580 740 585 750
RAEX 460 ML 440 610 510 630 525 645 540 665
OK 75.75 370 540 440 575 460 580 465 590
OK 75.78 290 455 300 490 360 500 380 505

The results show that with all weld t8/5 cooling times, the γ–α phase transformation
temperature in the higher alloyed Ni-Cr-Mo bearing OK 75.75 and OK 75.78 weld metals
starts and ends at a lower temperature than in the microalloyed parent steel. The phase
 115

transformation temperature in the leaner-alloyed Mo bearing OK 74.78 and Ni bearing

OK 48.08 weld metals, in turn, starts and ends at a higher temperature than in the parent
steel. For all weld metals and the parent steel, the transformation temperature raised as
the t8/5 increased.
The relevance of these findings with a view of explaining the pronounced cracking
tendency of the higher alloyed Mi-Cr-Mo bearing weld metals, in relation to the leaner
alloyed Mo bearing weld metals, is discussed in Section 5.8.

4.7 Weld bead size

Table 22 shows the results of the weld bead size measurements made under optical
microscopy for a range of various SAW and SMAW welds. Both heavy multipass welds
with 40 mm weld build-up thickness and two-pass thin plate welds of 6 mm build-up
thickness were included. All measurements were concerned with the final weld bead of
the last deposited filling layer.
Table 22. Results of the bead size measurements for SAW multipass thick welds –
comparison to SMAW and SAW thin plate two-pass welds.
Weld / SU1 SU2 SMAW thick T1 S1
Dimension SAW thick SAW thick multirun SMAW thin SAW thin
(mm) multirun weld multirun weld welds plate weld plate weld
ab 10.0 10.0 3.4–8.1 5.0 9.5
b2 29.0 24.0 10.2–16.8 10.0 16.5
b1 4.0 6.0 4–8 3 6
bave 20.0 15.0 7.5–15.5 7.0 12.0
Heat input 5.0 4.0 1.2–4.0 1.04 1.9
(kJ/mm)
Arc energy 5.0 4.0 1.5–5.0 1.3 1.9
(kJ/mm)

ab : weld bead height

b2 : weld bead ’maximum’ width close to the weld surface at the upper-part of the bead
b1 : weld bead ’minimum’ width close to the weld root
bave : weld bead average width measured at a location of ½-bead height
The results in Table 22 reveal that the differences in the final bead size are mostly heat
input related and not so much welding process or plate thickness related. Therefore,
within the same welding process and similar heat input range, the bead size is not
expected to differ essentially e.g. between two-pass welded thin plate and multipass
welded heavy plate.
This implies that the bead size effect, alone, may not be responsible for the lower WM
cracking risk in thin plates, compared with thick multipass welds of similar strength and
hydrogen. These findings are discussed in more detail in Chapter 5.
 116

4.8 Microstructures of multiple-pass weld metals

Metallographic studies were carried out for all the SMAW weld metals in Table 23.
Optical microscopy was used to visually examine the transverse sections cut from the
specimens. All samples were polished to a 3 µm finish and etched in 4% Nital. Weld
metal microstructures were classified according to the IIW recommendations[14, 15] for the
classification of weld metal microstructures using optical microscopy. The phase
fractions were calculated from the weld filling runs. In total, about 1000 number of points
were calculated. For each specimen, fractions of different microstructural phases are
given as average values of all the measurements made for a given specimen, without
distinguishing specifically between as-welded and reheated weld metal microstructures.
Since, however, the measurements were directed on filling runs, the presence and
proportion of as-welded microstructures dominate the results in Table 23. Some examples
of typical SMAW weld metal microstructures are shown in Appendix 6.
Table 23 summarises the proportions of each microstructural constituent. It can be
seen that all the examined SMAW weld metals consist predominantly of acicular ferrite,
AF. Among the L-80 weld metals (NSC) where arc energy was varied[4, 5b], the highest
amount of grain-boundary ferrite, GBF, is found to be associated with the 5.0 kJ/mm arc
energy welds. This is in line with the corresponding reduction in hardness with a rise in
the arc energy, see Table 12.

Table 23. Phase fractions of the experimental SMAW weld metals (%).
Specimen Electrode / Arc Acicular Polygona Ferrite with Grain
Code Alloying concept energy ferrite l ferrite second boundary
phase ferrite
(kJ/mm) (AF) (PF) (FS) (GBF)
NSC15 L-80 (Ni-Cr-Mo) 1.5 90 2 3 5
NSC30 L-80 (Ni-Cr-Mo) 3.0 87 1 3 9
NSC50 L-80 (Ni-Cr-Mo) 5.0 85 1 2 12
VTT3 OK 48.08 (Ni) 3.0 66 5 5 24
VTT1 OK 74.78 (Mo) 3.0 75 1 5 19
VTT2 OK 75.75 (Ni-Cr-Mo) 3.0 84 2 6 8
VTT4 Atom Arc 12018 (Ni-Cr-Mo) 3.0 91 2 2 5
OU1 OK 48.08 (Ni) 2.4 63 8 3 26
OU2 OK 74.78 (Mo) 2.4 71 2 5 22
OU3 OK 75.75 (Ni-Cr-Mo) 2.4 81 5 4 10
OU4 OK 75.78 (Ni-Cr-Mo) 2.4 86 3 3 8

Besides arc energy, chemical composition and, hence, strength of a weld metal in
question also affects the amount of GBF. It can be seen that the Ni-Cr-Mo bearing L-80,
OK 75.75, OK 75.78 and Atom Arc 12018 weld metals with the highest strengths have
also the highest AF contents, and much less GBF than, e.g., the Mn-Mo bearing OK
74.78 welds. The amount of GBF is the greatest – and the amount of AF the lowest – in
the Mn-Ni bearing OK 48.08 welds of the lowest strength.
 117

In all welds, the amounts of PF and FS are found very minor and more or less
constant. Only in the OK 48.08 welds, the amounts of FS and GBF are seen to have
increased at the expense of AF.
It is noteworthy that the OK 48.08 and OK 74.78 welds that exhibited the greatest
amounts of GBF did not reveal WM hydrogen cracks in any of the experiments made
applying the standard 16 h period[1b], not even in those associated with the highest weld
HD of ≈ 18.9 ml/100 g DM (IIW), see Tables 4 and 5. In the case of OK 74.78 weld
metal, hydrogen cracks did not appear until the period before the NDT inspection was
prolonged to 168 h (7 days). Furthermore, for the OK 48.08 weld metals it was totally
impossible to create conditions severe enough to result in WM hydrogen cracking.
It has been reported[3, 16–18] that the presence of GBF can be detrimental in that it
promotes WM hydrogen cracking. This has been ascribed to the ability of GBF to act as
a transport path to diffusing hydrogen, as well as to a microstructural mismatch, i.e., low
strength (softness) of GBF in relation to the surrounding matrix. Provided that the
presence of GBF in the examined welds had truly reduced the critical hydrogen content
Hcr for WM hydrogen cracking, this effect was obviously far too small to become notable
in the corresponding U- and Y-Groove cracking tests, see Tables 4 and 5.
 5 Discussion
Based on the experiments of the present thesis, calculations are made to analyse the
interactions between the primary crack-controlling factors in SMAW and SAW weld
metals. These include dependencies between (i) WM chemical composition and hardness,
(ii) WM hardness and strength, and (iii) weld critical hydrogen content-hardness-
chemical composition-residual stress -interactions. Comparative calculations on the
HRmax /H0 relationships between SAW and SMAW were performed, supplemented with
an analysis on the differences in the HRmax /H0 relationship between single-pass and
multiple-pass welds. The role of secondary factors, such as bead size, plate thickness,
heat input, phase transformation temperature and local strength mismatch within
multipass weld metals are discussed.
These analyses firstly aimed at defining the Crack-No Crack boundary conditions for
the weld metal hydrogen cold cracking in multiple-pass welds. This was accomplished
applying the Y- and U-Groove cracking data on SMAW multipass welds, to define the
critical hydrogen content Hcr as a function of weld metal strength.
In the second stage, the former NSC Y-Groove data set[12] was re-analysed to derive a
first-step engineering formula for the estimation of the required critical preheat
temperature Tcr. Then, applying the final and complementary SAW U-Groove data sets,
new formulae were derived for optimised prediction of safe Tcr estimates for the
avoidance of hydrogen cracking in multipass weld metals.

5.1 Weld metal chemical composition – hardness relationships

Figs 29a and 29b show all-weld metal Vickers hardness HV5 for the SMAW multipass
welds plotted as a function of the corresponding all-weld metal chemical composition in
terms of Pcm and CET. The Pcm weldability index and CET carbon equivalent are in
accordance with Eqs (43a) and (43b), respectively. The primary data on weld hardness
and composition, consisting of the Y- and U-Groove cracking test welds and multiple-
pass-welded pads, are given in Table 12.
Fig. 29a expresses the hardness-composition relationships in terms of weld metal
maximum hardness HV5(max) recorded from the cracking tests and plotted as a linear
 119

function of WM chemical composition. This expression hence represents a ‘worst-case’

upper-bound curve.

400

350

300 _ 1155 x Pcm+70

HV5(max) <

250
HV5 (HV)

_ 860 x CET+ 60
HV5(max) <
200

150

100

50

0
0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
CET (%)

0 0.050 0.100 0.150 0.200 0.250 0.300 0.350

Pcm (%)
wm0103b.dsf

Fig. 29a. All-weld metal Vickers hardness HV5(max) as a function of weld Pcm and CET using
the SMAW maximum hardness data.

The upper bound curves for the hardness-composition relationship in Fig. 29a can be
expressed using individual maximum HV5 values. Consequently, the HV5(max) data were
fitted by linear functions of all-weld metal Pcm and CET, as:
HV5(max) = 1155 ∗ Pcm + 70 (44a)
HV5(max) = 860 ∗ CET + 60 (44b)
 120

Fig. 29b, in turn, uses WM average hardness measured from the SMAW multiple-pass
pads and covering an arc energy range from E = 1.6 to 3.2 kJ/mm. For the purposes of
comparison, the preliminary SAW data from the U-Groove cracking test weldments and
given in Table 13 have been added into Fig. 29b. The description of the data in Fig. 29b
has been derived using the SMAW data for the lowest E, i.e., 1.6 kJ/mm, in order to fully
demonstrate any WM hardness increase as a result of rapid cooling cycle. The SMAW
and SAW data enabled a separate hardness adjustment function to be derived to display
the hardness reduction from the 1.6 kJ/mm reference level, with increasing arc energies.

400

** **
350 * *
* **
HV5 = 993 x Pcm + 80
300 * ** * * *
*
* ** SU2
* * (SAW)
250
* *
HV5 (HV)

* ** *
200 * * SU1 HV5 = 687 x CET + 80
(SAW)

Pcm CET
150
1.6 kJ/mm
2.4 kJ/mm SMAW
3.2 kJ/mm
100
2.0 kJ/mm
3.0 kJ/mm SAW
* individual 4.0 kJ/mm
50
maximum
values SAW data: 4.0-5.0 kJ/mm

0
0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
CET (%)

0 0.050 0.100 0.150 0.200 0.250 0.300 0.350

Pcm (%)
wm0202b.dsf

Fig. 29b. All-weld metal Vickers hardness HV5(ave) as a function of weld Pcm and CET using
the SMAW average hardness data fitted according to the 1.6 kJ/mm arc energy data points.
SAW hardness data from the U-Groove tests is added.
 121

The ‘best-estimate’ curves shown in Fig. 29b were derived assuming a constant
hardness of ≈ 70–100 HV that should still remain as Pcm and CET approach to nil. It was
reasonable to adopt a unified, constant hardness value apart from whatever was the
weldability index. This results in linear fittings that use the 1.6 kJ/mm arc energy data
points of HVave data set, and are of the form:
HV5(ave) = 993 ∗ Pcm + 80 (45a)
HV5(ave) = 687 ∗ CET + 80 (45b)
These equations fulfil the theoretical limit conditions of HV → 80 HV when CET → 0
and Pcm → 0. To comply with the currently available data, the validity range for Eqs (44)
and (45) should be set as follows:
195 ≤ HV ≤ 365 (HV5)
0.12 ≤ Pcm ≤ 0.30 (%)
0.18 ≤ CET ≤ 0.45 (%)
1.0 ≤ E ≤ 4.0 (kJ/mm)
Theoretically, the ultimate maximum hardness a microstructure can reach is that of the
100% martensitic structure, whose hardness depends practically entirely on its C content.
Provided the weld cooling cycle is rapid enough to yield maximum hardenability at a
given weld C content, the ‘cut-off ’ value, HVlimit, for the weld maximum hardness HVmax
and corresponding to that of the 100% martensitic structure can be expressed[113, 119] as:
HV ≤ HVmax(1) = 802 ∗ C + 305 (46a)
HV ≤ HVmax(2) = 884 ∗ C ∗ (1 – 0.3 C ) + 294
2
(46b)
HVlimit = max{HVmax(1) , HVmax(2)} (46c)
Fig. 30 shows all-weld metal hardness HV5 of the SMAW welds, c.f., Table 12 as a
function of arc energy E. In the range of E = 1.6 to 3.2 kJ/mm, the overall effect of arc
energy on the HV5 is found relatively small. In most cases, the change in HV5 remained
within the statistical scatter band of Vickers hardness, i.e., ≈ ± 30 HV units. Unlike in the
case of the parent steel HAZ[13, 23–25, 28–30], inclusion of a separate cooling time parameter
into composition-hardness formulae for weld metals was therefore felt unnecessary.
In the case of the two lean-alloyed weld metals, Ni bearing OK 48.08 and Mo bearing
OK 74.78, hardness is seen to remain virtually unchanged in the range of E = 1.6–2.4
kJ/mm. For the high-alloyed Ni-Cr-Mo bearing OK 75.75, Atom Arc 12018 and OK 75.78
welds, a reduction in hardness is found more evident and seems to occur in a relatively
linear manner when the arc energy raises from 1.6 to 3.2 kJ/mm, see Fig. 30. The
influence of arc energy on hardness is, in this respect, very similar between the three Ni-
Cr-Mo alloyed SMAW weld metals.
 122

400

* *
350 * *
*

300 ** *
*
*
*
250
* *
*
HV5 (HV)

* *
200 *

150
OK75.78 Pcm= 0.28
Atom Arc 12018 Pcm= 0.28
OK48.08 Pcm= 0.13
100
OK74.78 Pcm= 0.16
Nittetsu L-80 Pcm= 0.20
OK75.75 Pcm= 0.22
50 * Individual maximum values

0
0 1.0 2.0 3.0 4.0 5.0
E (kJ/mm)
wm0105.dsf

Fig. 30. All-weld metal Vickers hardness HV5 as a function of arc energy E (or, in terms of
heat input Q when replacing E by (Q / k) – for SMAW: k = 0.8 [1a]).

According to Fig. 30, simple linear adjustment formulae can be written to account for
the change in hardness HV5 as a function of heat input Q. The formulae use the 1.6
kJ/mm arc energy as the reference level, which ensures its compatibility with Eq. (45).
Thus, these formulae can be used to obtain heat input adjusted hardness estimates,
provided the weld chemical composition and E, or Q, are known. Due to the different
hardness responses of the leaner- and higher-alloyed weld metals, as the heat input is
varied, the formulae are constructed in a manner that relates them to the WM chemical
composition in question, in terms of Pcm and CET, as follows:
 123

HV5(ave) = 993 ∗ Pcm + 80 – {[11 ∗ ((Q / k) – 1.6)] ∗ (3.1 ∗ Pcm)} (47a)

HV5(ave) = 687 ∗ CET + 80 – {[11 ∗ ((Q / k) – 1.6)] ∗ (2.7 ∗ CET)} (47b)
The validity limits given for Eqs (44) and (45) apply also to Eqs (47a) and (47b). What
comes to hardness adjustment at arc energies lower than the 1.6 kJ/mm reference level, it
should be borne in mind that the HVmax limit values according to Eq. (46) apply to Eqs
(47a) and (47b).
These differences in the hardness responses between the Ni or Mo bearing and Ni-Cr-
Mo bearing SMAW welds can be explained by the differences in their hardenability.
Whether the change in hardness occurs over a certain Pcm or CET range, or not, depends
on the capability of a particular weld metal to form the full spectrum of microstructural
constituents from comparatively soft ferritic structures to hardened martensitic
transformation products through the applied heat input range. This is illuminated in Table
24 that presents a comparison between the measured individual WM maximum hardness
of the SMAW and SAW weld metals together with the theoretical maximum hardness of
100% martensite microstructure.
Table 24. Comparison of hardness HV5 between the measured maximum values of the
SMAW and SAW welds, HVmax, and the theoretical maximum values corresponding to
fully hardened 100% martensite structure, HVmax(M), calculated according to Eqs (46b)
and (46a).
SMAW Electrode / WM Measured WM Calculated WM Difference between
SAW tubular wire & carbon max. hardness max. hardness measured and
OK Flux 10.62 content HVmax HVmax(M) calculated max.
(%) (as-welded) (≡ 100% martensite) hardness HV5(max)
Eqs (46b)–(46a)
OK 48.08 0.04 221 329–337 –108...–116
OK 74.78 0.04 239 329–337 –90...–98
Nittetsu L-80 0.03 303 320–329 –17...–26
(VTT)
OK 75.75 0.04 308 329–337 –21...–29
Atom Arc 12018 0.06 362 347–353 +9
OK 75.78 0.05 347 338–345 +2

OK Tubrod 15.26S 0.08 277 364–369 –87...–92

OK Tubrod 15.27S 0.07 362 356–361 +1...6
OK Tubrod 15.29S 0.07 376 356–361 +15...20

Table 12 clearly reveals how great is the difference between the measured HV5(max)
values for the majority of the examined welds and the maximum hardness this weld metal
can theoretically reach under the most stringent (i.e., rapid) cooling conditions. The
smaller the difference, the greater the hardenability and hence the spectrum of those
microstructural constituents the weld metal can display over the applied arc energy range
from 1.6 to 3.2 kJ/mm and 2.0 to 4.0 kJ/mm for SMAW and SAW, respectively.
It can be seen that the measured HVmax values of the high-alloyed Ni-Cr-Mo bearing
OK 75.75, Atom Arc 12018, OK 75.78, OK Tubrod 15.27S and OK Tubrod 15.29S weld
 124

metals approach to, and finally reach, their theoretical maximum level HVmax(M), i.e., that
of a fully hardened microstructure corresponding to 100% martensite. Contrary to this,
the measured HVmax values for the lean-alloyed OK 48.08, OK 74.78 and OK Tubrod
15.26S welds remain – even with the shortest applied weld cooling time associated with
the lowest E of 1.6 kJ/mm – considerably less than their calculated theoretical HVmax(M).
It follows that, as the arc energy is raised from 1.6 to 3.2 kJ/mm, the Ni-Cr-Mo alloyed
SMAW welds, for example, can display the full spectrum of microstructural phases from
small amounts of martensite plus acicular ferrite to acicular ferrite and, finally, acicular
ferrite plus grain boundary ferrite. The Ni and Mo alloyed welds, in turn, exhibit
predominantly acicular ferrite plus grain-boundary ferrite microstructures over the whole
arc energy range, but no notable amounts of martensite even with the lowest applied arc
energy of 1.6 kJ/mm.
Overall, it can be concluded that variations in heat input have a much lesser influence
on the hardness of the examined multiple-pass weld metals, than is usually the case for
the parent steel HAZ in ferritic high-strength structural steels whose hardness-weld
cooling time dependencies obey[13, 23–25, 28–30] the classical, steeply decaying S-curve.

5.2 Weld metal hardness – strength dependence

Earlier works at VTT[6, 7] and NSC[23, 113] have unambiguously shown that a linear
relationship exists between all-weld metal hardness HV and its ultimate tensile strength
RM in the case of multipass welds. The same does not, however, apply to yield strength
Rp0.2, as the Rp0.2 values for different weld metals showed[6, 7] no compatible correlation
with their hardness. At a given level of weld hardness, rutile weld metals were found[6,7]
to always exhibit higher yield strength than basic welds, presumably due to differences in
their yield-to-tensile ratios and the resulting stress-strain characteristics.
Since these work at VTT[6, 7] and NSC[23, 113] contained data primarily from FCAW and
SAW weld metals, respectively, it was necessary to examine whether the SMAW weld
metals of the present thesis followed these hardness-strength dependencies. For this
purpose, the WM tensile data in Table 11 were plotted against the corresponding WM
average hardness HV5 values given in Table 12. These data are shown in Fig. 31,
together with a linear relationship derived experimentally from the earlier FCAW data
obtained at VTT (dotted line) and expressed[6, 7] as:
RM = 2.5 ∗ HV + 93 (48a)
Fig. 31 shows that the SMAW experiments of the present study accord very well with
the earlier FCAW data[7] obeying Eq. (48a). Unlike yield strength Rp0.2, the ultimate
tensile strength RM can hence be reliably expressed as a function of weld metal hardness
HV, irrespective of the type of weld metal, i.e., whether rutile FCAW or basic SMAW.
 125

1100

1000

900

800

700
WM RM (MPa)

600 RM = 2.47 x HV5 + 93

500

400
RM = 2.83 x HV5
300

200

100

0
0 50 100 150 200 250 300 350 400
HV5 (HV)
wm0104.dsf

Fig. 31. All-weld metal tensile strength RM as a function of weld metal hardness HV5 – the
present SMAW data, together with a dotted line according to the former[7] FCAW data.

If one follows the theory according to which RM ≈ 0 when HV ≈ 0, inclusion of the

present SMAW data into the analysis allows for re-writing of Eq. (48a) into more
theoretically justified form that is expressed according to Eq. (48b) simply as:
RM = 2.85 ∗ HV5 (48b)
This new linear expression obeying Eq. (48b) has been added into Fig. 31 (solid line).
It shows good agreement with the present SMAW data, as well as with the former[7]
FCAW data (dotted line).
The present expression according to Eq. (48b) differs slightly from that proposed by
Yurioka et al.[23] and based on the data[113] from high heat input SAW weld metals. A
 126

comparison shows that at a given weld metal HV level, the RM estimates according to
Yurioka et al.[23] tend to become somewhat higher than those calculated in accordance
with Eq. (48b). In the 100–360 HV range, for example, the difference in RM estimates
given by these two expressions remains no more than about 6–11%. Still, the fact that of
these two formulae, it is Yurioka's expression[23] that gives higher RM estimates, can be
regarded as somewhat surprising. As the Yurioka-formula is based essentially on high
heat input SAW data[113], whereas Eq. (48b) is derived predominantly from SMAW[5a,5b]
and FCAW[6, 7] data associated with low and moderate heat inputs, one would expect the
difference to be the reverse.
Alternatively, an estimate for weld metal RM can be obtained also in the cases where
only weld metal chemical composition is known, by applying Eq. (44) or Eq. (45) in
conjunction with Eq. (48). An example of such an estimation of RM values for extra-high
strength Y-80C, L-80, L-74 and L-60 weld metals from the NSC experiments[12] is given
in Table 25 that compares measured RM data to estimates of RM calculated on the basis of
weld chemical composition in terms of CET.
Table 25. Calculated RM estimates for extra-high strength Y-80C, L-80, L-74 and L-60
weld metals from the former NSC experiments[12] – comparison of the measured NSC
tensile test data[12] with the calculated RM estimates according to Eqs (44b), (45b) and
(48).
Filler Weld metal Weld metal Calculated estimates of weld metal RM
material chemical tensile (MPa)
composition strength ‘Worst case’ ‘Worst case’ ‘Best estimate’ ‘Best estimate’
CET(1) RM (2) RM estimate RM estimate of RM acc. to of RM acc. to
(%) (MPa) acc. to Eqs acc. to Eqs Eqs (45b) and Eqs (45b) and
(44b) and (48b) (44b) and (48a) (48b) (48a)
Y-80C 0.321 852 958 923 856 835

L-80-1 0.312 825 935 904 839 820

L-80-2 0.300 814 906 878 815 800
L-80-3 0.293 807 889 863 802 788

L-74-1 0.278 744 852 831 772 762

L-74-2 0.248 745 778 768 714 711
L-74-3 0.280 733 857 835 776 766

L-60-1 0.223 643 717 714 664 669

L-60-2 0.186 618 626 636 592 606
(1)
calculated acc. to Eq. (43b) applying analyses of WM chemical compositions given in Ref.[12]
(2)
measured NSC tensile test data acc. to Ref.[12]

Analysis of the NSC data set[12] in Table 25 is seen to support the applicability of Eqs
(44), (45) and (48) in providing realistic estimates of weld metal RM even when knowing
only the weld metal chemical composition. Considering that the estimation here was
made in two steps, firstly, applying weld chemical composition to derive estimates of
weld hardness HV and, secondly, using these HV data to derive estimates of weld RM, the
estimates in Table 25 agree surprisingly well with the measured NSC data[12]. The ‘worst
 127

case estimates’ of RM calculated according to Eq. (44b) in conjunction with either Eq.
(48a) or (48b), are seen to overestimate the ‘true’ weld metal strength, as they in fact
should, since these estimates were derived applying a ‘maximum hardness data based fit’
c.f. Eq. (44b) and are therefore expected to yield overly conservative estimates of weld
hardness. However, the ‘best estimates’ of RM calculated using Eq. (48b) in conjunction
with Eq. (45b) that obeys an ‘average hardness data based fit’, are seen to accord very
well with the measured[12] values. Applying Eq. (48a) in place of Eq. (48b) is seen to
yield estimates that slightly underrate the ‘true’ weld strength. Therefore, Eq. (48b) that is
based on a larger data set than Eq. (48a), should be preferred whenever ‘best estimates’
of RM are required and calculated applying Eq. (45b) based on the ‘average hardness data
based fit’.
Consistently, comparison of weld metal Pcm – RM relationship between a recent
investigation[118] on extra-high strength SMAW multipass welds carried out by ELGA
and the outcome of Eqs (45a) and (48b) of the present thesis demonstrates good
agreement between the two approaches, as shown in Appendix 9. At a given weld Pcm,
the present approach yields slightly greater values of weld metal RM and is hence
considered ‘safe’ as it comes to the assessment of WM hydrogen cracking risk. The
difference in RM estimates given by these two approaches is, however, minuscule,
remaining less than about 5%.
Analyses of these separate NSC[12] and ELGA data sets[118] are hence regarded to
validate the applicability of Eq. (48b) in conjunction with Eqs (44b) and (45b) in
providing conservative but ‘safe’, and ‘realistic’, estimates of weld metal RM,
respectively, on the basis of weld metal chemical composition Pcm, or weld metal average
hardness HV5ave.
Overall, the possibility to calculate an estimate to all-weld metal tensile strength
directly from Vickers hardness data HV5 applying Eq. (48b) simplifies the ‘engineering
assessment’ of WM cracking susceptibility, as it is not mandatory to carry out all-weld
metal longitudinal tensile tests as a part of the cracking risk assessment. Sometimes,
standards may require tensile testing of the welded joint anyway, albeit more often
transverse than longitudinal to the weld. Unlike strength, hardness can be measured by
relatively simple means e.g., from weld cross section or, by using appropriate devices,
even from a polished weld metal surface without a need for sectioning or breaking the
actual weldment.
A comparison of weld yield-to-tensile ratios, Rp0.2/RM, between the actual values in
accordance with Rp0.2 and RM measured in the tensile test and nominal values calculated
according to manufacturer's information on nominal weld Rp0.2 and RM is summarised in
Appendix 8. Despite the substantial differences in weld Rp0.2 and RM between measured
and nominal values for a range of differently alloyed weld metals associated with varying
welding procedures, the corresponding Rp0.2/RM ratios are always found very similar,
whether based on the actual or the nominal values of weld Rp0.2 and RM. The Rp0.2/RM
values reported in a recent work[118] on SMAW multipass weld metals of strengths
ranging from RM ≈ 550 to 950 MPa were found to comply relatively well with the results
of the present thesis, the only difference being that the values measured here on extra-
high strength welds tended to be not that high as those reported in Ref.[118].
 128

Overall, a comparatively accurate estimate to weld metal Rp0.2 can be obtained

applying the weld metal Rp0.2/RM ratio that is calculated according to manufacturer's
information on nominal weld Rp0.2 and RM, as shown in Appendix 8.

5.3 Effect of weld thermal cycle, bead size and overlap on final
hydrogen concentration in multipass welds according to analytic
calculations

It is widely reported[2, 3, 8, 9, 12, 19–23, 54, 74, 76, 79, 94, 100, 106] that hydrogen concentration profile
in multipass welds in thickness direction is non-uniform, since the maximum peak of
final local concentration HRmax is governed by hydrogen diffusion and accumulation
throughout the welding until finishing of the final layer. Therefore, HRmax in the filling
runs can differ from both the remaining weld diffusible hydrogen HR100max characteristic
of single-run welds, and the initial weld diffusible hydrogen H0 according to ISO/IIW
3690[57]. Earlier work has shown[8,19–23] that in addition to H0, HRmax depends on complete
weld thermal history, bead size and weld overlap due to penetration of successive passes
resulting in re-melting of previously solidified weld metal.
As thermal gradient, bead size and overlap all are welding process related, an attempt
is made in the present thesis to investigate their influence on HRmax in the case of actual
and arbitrary, yet typical, SMAW and SAW multipass welds. This was to evaluate
whether HRmax could differ between SMAW and SAW filling runs so essentially that this
factor should be included into predictive systems assessing the WM cracking risk. For
this purpose, available data[5a, 5b] for SMAW welds was applied for the analysis. With
respect to SAW, arbitrary weld geometry was defined using relevant literature data[32, 33].
For the present analysis, the (i) weld thermal cycle, (ii) weld bead size and (iii) weld
bead overlap are defined[19–23] in terms of the thermal factor of diffusion, ΣD∆t, an
individual bead layer thickness, hw, and weld bead overlap ratio, d/hw, respectively.
The weld thermal cycle in terms of ΣD∆t incorporates the diffusion coefficient of
hydrogen D at a certain temperature and the time interval ∆t to which D applies.
Integration over the complete weld thermal history from Tm to Ti has been shown[21–23] to
obey Eq. (2), as described in Section 1.2.1.
Assuming it is sufficient to consider the hydrogen cracking conditions at a constant
temperature, the description of hydrogen diffusion over the complete weld thermal
history can be simplified. Consequently, ΣD∆t is expressed here using ΣD∆t(100), a
parameter shown[13, 23–25] to approximate Eq. (2) comparatively well, hence only one
parameter, weld cooling time t100, is needed. This allowed calculation of ΣD∆t directly
from the recorded weld thermal history using the t100 data. One of the empirical formulae
presented in Section 1.2.1, Eq. (3b), was applied for calculating the estimates of ΣD∆t(100)
on the basis of the available weld cooling data at low temperatures, t100, given in
Appendix 10.
For this, a set of t100 cooling times were chosen which corresponded to welding of a 50
mm thick plate at 1.7 kJ/mm arc energy and at different preheat temperatures, all
 129

representing essentially 3-dimensional heat flow conditions, see Appendix 10. These t100
values for the analysis are given in Table 25, together with the corresponding ΣD∆t(100)
values calculated according to Eq. (3b).
It is worth pointing out that of the formulae (3a)–(3c) given in Section 1.2.1 for
calculating ΣD∆t(100) from weld cooling time, Eq. (3a) is stated[25] to generally yield the
most accurate approximations of ΣD∆t. However, as Eq. (3a) requires weld cooling time
data to temperatures also other than 100°C, and since such data is seldom available, it
was concluded that for the purposes of comparisons between different welding processes
and parameters, Eq. (3b) that is based solely on t100 data can be satisfactorily used within
reasonable accuracy, consequently Eq. (3b) was chosen to be applied here.
Table 26. Effect of weld thermal history on the thermal factor of hydrogen diffusion
ΣD∆t(100); conditions*) equivalent to 50 mm plate and 1.7 kJ/mm arc energy at different
levels of preheat temperature T0, in accordance with the diagram given in Appendix 10.
Preheat temperature Weld cooling time Thermal factor of hydrogen diffusion
T0 t100 *) ΣD∆t(100)
(°C) (sec) (cm2)
25 45 0.001793
50 70 0.002318
75 180 0.004447
100 800 0.016505
125 1600 0.034148
150 2200 0.048899
175 2500 0.056732
200 3000 0.070430

The effect of weld bead size on hydrogen diffusion distance was then treated using the
square of the individual bead layer thickness, hw2, i.e., the square of the distance that
needs to be covered by the diffusing hydrogen from the middle of the bead to its outer
surfaces. This was regarded relevant, since it has been widely shown[19–23] that the
approximation of hydrogen diffusion through an ‘element surface area’, as one may also
picture the term hw2 to mean, by applying the square of the individual bead layer
thickness provides a realistic description of the effect of bead size on the diffusion of
hydrogen under the thermal cycle of welding. The ΣD∆t/hw2 ratio is thereby indicative of
the loss of hydrogen from the element in the time available, this time depending on the
nature of the weld thermal cycle.
This way, it was possible to incorporate the thermal and geometrical conditions for
hydrogen diffusion by describing them using a single, integrated diffusion-distance
parameter, ΣD∆t/hw2. This parameter has been described in Section 1.5.2 and
defined[19, 20] in Eq. (37). For the present analysis, the dependence of ΣD∆t on weld
thermal cycle was taken in accordance with Eq. (3b) and hw was determined
experimentally from extracted macrosections of SMAW and SAW welds in question.
According to Table 26, thermal conditions for hydrogen diffusion are essentially
enhanced with increasing weld cooling time t100. The increase in t100 being a consequence
of higher preheat temperatures, or, higher heat inputs, this means that higher thermal
inputs in welding manifest themselves as an increase in the ΣD∆t factor.
 130

5.3.1 Thermal cycle and bead size effects – weld overlap excluded

It has been proven[19–23] that the remaining weld diffusible hydrogen content in the case of
single- and multipass welds, denoted as HR100max and HRmax, respectively, depends
essentially on diffusion conditions for hydrogen. These conditions have been shown[19–23]
to obey Eq. (37), as explained in Section 1.5.2.
This is comprehensively demonstrated in Fig. 32 that shows the ratio between the
maximum remaining weld diffusible hydrogen and the weld initial hydrogen, HR100max/H0,
plotted against hydrogen diffusion conditions in terms of the integrated diffusion-distance
parameter, ΣD∆t/hw2. The primary data in Fig. 32, created[20] at Osaka University and
NSC, Japan, do not account for the weld overlap effects, i.e., d = 0. In the absence of any
weld overlap and looking one pass at a time, HR100max corresponds to Hmax which hence
applies to both single-pass and multiple pass welding situations, i.e., Hmax = Hmax(d=0);
HR100max = HR100max(d=0) and Hmax ≡ HR100max(d=0).

1,3,7: original modelling data

1.0 (Osaka University &
Nippon Steel)
(Hmax)d=0 / H0

VTT approximation

0.5

7
3
1

0
0.01 0.1 1.0
D t / hw2
wm0108.dsf

Fig. 32. The ratio between the maximum remaining weld diffusible hydrogen and weld
initial hydrogen content, HR100max(d=0)/H0, as a function of the diffusion-distance parameter
ΣD∆t/hw – overlap not included. Primary data according to Osaka University & NSC[20],
2

calculated approximation according to VTT.

According to Fig. 32 and looking one pass at a time, Hmax remains high and almost at
a level of H0 as long as ΣD∆t/hw2 is low, then decreasing as ΣD∆t/hw2 starts to increase.
 131

As the data in Fig. 32 implied, the relationship between the HR100max(d=0) /H0 ratio and
ΣD∆t/hw2 can be written using decaying sigmoidal dependence, a general form of this
expression was derived according to the primary data in Fig. 32. The general equation
incorporating positive first degree and negative second-degree polynomic terms of the
form of Eq. (38) was given in Section 1.5.2.
In order to validate the general form of Eq. (38), a numerical fitting into the primary
modelling data in Fig. 32 was made in the present thesis. This fit obeying the form of Eq.
(38) gave an experimental expression of the polynomic form, denoted as („) in Fig. 32,
as:
2
HR100max(d=0) /H0 = {e[(–2.8 ∗ ΣD∆t/hw )/(0.18(ΣD∆t/hw ))]} +
2

+ 0.70 ∗ (ΣD∆t/hw2) – 0.45 ∗ (ΣD∆t/hw2)2 (49)

Fig. 32 shows that in general, Eq. (49) gives estimates that are on the safe side in
relation to the modelling data, that is, Eq. (49) proposes equal to or slightly higher values
of the Hmax/H0 ratio than the modelling data. At high values of ΣD∆t/hw2 approaching
unity, the accuracy of Eq. (49) is likely to weaken to some extent (presumably it would
require a third-order term), however, in this ΣD∆t/hw2 range also the original modelling
data exhibits comparatively great variations, as shown in Fig. 32. Consequently, it is
concluded that Eq. (49) derived in the present thesis is an appropriate description of the
dependence of the Hmax/H0 ratio on diffusion of hydrogen, with the effects of the weld
thermal cycle and bead size being realistically accounted.
Eq. (49) was then applied to calculate the HR100max(d=0) /H0 ratio as a function of weld
thermal conditions, ΣD∆t, and bead size in terms of hw. For the analysis, three different
bead layer thicknesses: 3, 4 and 8 mm were applied, as shown in Table 27. The calculated
values of the HR100max(d=0) /H0 ratio are also given in Table 27. An increase in bead size in
terms of hw leads to longer diffusion distances in accordance with Eq. (37), which will
further manifest itself as a dramatic reduction of ΣD∆t/hw2 throughout the thermal input
range, as shown in Table 27.

5.3.2 Thermal cycle and bead size effects under the influence of
weld overlap

In multipass welding hydrogen accumulation is influenced by the extent a successive pass

penetrates to and re-melts the previous pass(es). This overlapping is described by the
weld bead overlap ratio, d/hw. Section 1.5.3 discusses findings from numerical modelling
showing[19, 20] that increasing overlap leads to intensified hydrogen accumulation and
hence elevated final local hydrogen concentration HRmax. This was explained[19–23] by
additional supply of hydrogen released from previously solidified passes as they are re-
melted by the subsequent pass. Hydrogen is further transported towards the filling runs
due to enhanced diffusion caused by repetitive thermal cycles of successive passes.
For the analysis made here, it was presumed that the final local hydrogen
concentration accounting for overlap effects, HRmax(d>0), depends on weld thermal cycle,
 132

bead size and the weld bead overlap ratio. To obtain a realistic description of the HRmax
/H0 ratio in the case of multiple-pass welds, the effect of weld bead overlap in terms of
the d/hw ratio must therefore be included into the analysis. This requires HRmax(d=0) given
in Eq. (49) to be replaced by HRmax(d>0) in the calculation formulae to be further applied
in the present analysis.
Fig. 33 shows the effect of the weld bead overlap ratio, d/hw, on the maximum final
hydrogen concentration – maximum remaining diffusible hydrogen content ratio,
HRmax(d>0) /HRmax(d=0), for different values of the integrated diffusion-distance parameter
ΣD∆t/hw2. Here, HRmax(d>0) refers to the maximum final local hydrogen concentration in
the filling layers in multipass weld under the influence of the effects of weld bead
overlapping. HRmax(d=0), in turn, refers to the maximum weld diffusible hydrogen content
that is remained in an individual pass after hydrogen effusion as the weld pass has cooled
down to the working temperature, without any overlapping effects and hence no matter
whether the weld bead in question is a single-pass weld, or a last solidified pass of
multipass weld at the moment of the analysis. The primary data in Fig. 33 has its origin
in the earlier work[20] made at Osaka University and NSC, Japan. The HRmax(d>0)
/HRmax(d=0) ratio in Fig. 33 thereby expresses the local (final) hydrogen concentration with
the effects of hydrogen accumulation and weld overlapping included, in relation to the
weld diffusible hydrogen content affected by hydrogen effusion, but unaffected by any
overlap.
Let the effects of weld thermal cycle, bead size and overlapping on HRmax be
expressed in terms of ΣD∆t, hw2 and d/hw, respectively. The data[20] on hydrogen
accumulation in Fig. 33 can then be applied to describe the overlap effect using a
relationship between the HRmax(d>0) /HRmax(d=0) ratio and the d/hw ratio, at a given
ΣD∆t/hw2 level. Eq. (39) in Section 1.5.3 gives a general form of the overlap equation
based on the primary data[20] in Fig. 33.
The results in Table 27 reveal that all the ΣD∆t/hw2 values calculated according to Eq.
(37) remain, in the present case, less than 0.8. Fig. 33, in turn, shows that the dependence
of the HRmax(d>0) /HRmax(d=0) ratio on the weld d/hw ratio obeys a logarithmic, or power-
law, dependence at ΣD∆t/hw2 values less than 0.8. For the analysis made here, four
approximative sub-equations, each one corresponding to a different ΣD∆t/hw2 level, were
hence derived from the primary data in Fig. 33. These fittings resulted in 4 power-law
functions of the form:
HRmax(d>0) /HRmax(d=0) = 1 + 1.70 ∗ (d/hw)0.80 (50a)
HRmax(d>0) /HRmax(d=0) = 1 + 1.25 ∗ (d/hw)0.60 (50b)
HRmax(d>0) /HRmax(d=0) = 1 + 0.75 ∗ (d/hw) 0.40
(50c)
HRmax(d>0) /HRmax(d=0) = 1 + 0.45 ∗ (d/hw)0.35 (50d)
Eqs (50a)–(50d) hence correspond to the ΣD∆ t/hw2
values of 0.8, 0.4, 0.2 and 0.1,
respectively, see Fig. 33. These sub-equations were then applied to calculate values of the
HRmax(d>0)/HRmax(d=0) ratio given in Table 27.
 133

A hw
B d
3.0 bw

hw H0
Lap
d
H D t / hw2 = 0.8
2.5
Hmax / (Hmax) d=0

0.4

2.0 0.3

0.2

1.5

0.1

1.0
0 0.5 1.0 1.5
Lap ratio, d /hw

wm0107.dsf

Fig. 33. Effect of weld bead overlap ratio d/hw on maximum final hydrogen concentration –
maximum remaining diffusible hydrogen content -ratio, HRmax(d>0) /HRmax(d=0), for different
values of diffusion-distance parameter ΣD∆t/hw2 – overlap included. Primary modelling data
acc. to Osaka University & NSC[20].

Table 27 shows the effect of weld thermal cycle in terms of weld t100 cooling time, on
the thermal factor of diffusion, ΣD∆t, and the integrated diffusion-distance parameter
ΣD∆t/hw2, calculated using different arbitrary, yet typical, bead sizes ab of 5, 8 and 16
mm. According to VTT's earlier investigations[24, 25, 34, 53] on SMAW and SAW welds,
these ab values should roughly correspond to the individual weld bead layer thicknesses
of hw ≈ 3, 4 and 8 mm, respectively. Of the hw values, 3 mm corresponds quite well to the
measured SMAW bead size data[5b], as well as to the bead reinforcement data for SAW
reported elsewhere[32, 33]. The latter two, 4 and 8 mm, are thought to represent extreme
 134

values for high-heat input SMAW and SAW weld metals, respectively. Values of ΣD∆t
and ΣD∆t/hw2 have been calculated according to Eqs (3b) and (37), respectively.
Table 27. Effect of integrated diffusion-distance parameter ΣD∆t/hw2 on the remaining
weld diffusible hydrogen content – initial weld hydrogen content -ratio HRmax(d=0) /H0
(assuming no overlap) and on the final local maximum hydrogen concentration – initial
weld hydrogen content -ratio HRmax(d>0)/H0 when accounting for weld bead overlapping.
Weld Welding Weld bead Diffusion – Remaining HRmax(d>0) / Final local
cooling method layer distance diffusible HRmax(d=0) maximum hydrogen
time SMAW/SAW thickness parameter1) hydrogen / initial -hydrogen / initial hydrogen -
t100 hw ΣD∆t/hw2 hydrogen multiplier3) as a ratio
(sec) (mm) -ratio2) function of HRmax(d>0) /H0
HRmax(d=0)/H0 d/hw SMAW SAW
-ratio4)
45 SMAW/SAW 3 0.020 0.96 1.10 / 1.15 1.06 1.10
SMAW 4 0.011 0.98 1.10 1.08 –
SAW 8 0.003 0.99 1.05 – 1.04
800 SMAW/SAW 3 0.183 0.60 1.55 / 1.70 0.93 1.02
SMAW 4 0.103 0.77 1.35 1.04 –
SAW 8 0.026 0.94 1.15 – 1.08
2200 SMAW/SAW 3 0.543 0.28 1.90 / 2.65 0.53 0.74
SMAW 4 0.306 0.41 1.80 0.74 –
SAW 8 0.076 0.83 1.25 – 1.04
3000 SMAW/SAW 3 0.783 0.27 2.10 / 3.20 0.57 0.86
SMAW 4 0.440 0.30 1.90 0.57 –
SAW 8 0.110 0.76 1.40 – 1.06

1)
calculated acc. to Eq. (37), see Table 25
2)
calculated acc. to Eq. (49), or determined from Fig. 32
3)
calculated acc. to Eqs (50a)–(50d) or determined from Fig. 33 using ΣD∆t/hw2 as defined in
Eq. (49)
4)
estimated values for the weld overlap ratio: d/hw = 0.6 for SMAW and d/hw = 1.5 for SAW,
c.f., Refs[32, 33]

For Table 27, equivalent thermal welding conditions in terms of ΣD∆t were adopted,
i.e., it was assumed[5b, 24, 25, 32, 33] that an equivalent thermal heat input results in weld
beads having greater d/hw ratios in the case of SAW than SMAW. Thus, any differences
in hydrogen diffusion are solely due to geometry-associated factors: weld bead
penetration (overlap) and bead size, and not due to thermal cycle itself. According to the
published data, values for the d/hw ratio were set as 1.5 and 0.6 for the SAW[32, 33] and the
SMAW[5b, 24, 25] welds, respectively.
In Table 27, the HR100max(d=0) /H0 ratio that assumes no overlap and hence refers, in
that sense, to single-pass welding conditions, is then multiplied with the corresponding
HRmax(d>0) /HRmax(d=0) ratio which does account for the effect of weld overlapping. This
calculation finally gives, at each level of ΣD∆t/hw2, the desired values of the HRmax(d>0)
/H0, that is, the maximum final local hydrogen concentration in multipass weld, in
relation to the initial weld diffusible hydrogen content measured in a single-pass welding
test. The thereby calculated HRmax(d>0) /H0 values are given in Table 27 over a
 135

comparatively wide ΣD∆t/hw2 range and for the three arbitrary, yet typical, bead layer
thicknesses of hw = 3, 4 and 8 mm.
This allows for comparisons of the critical local hydrogen concentration in multipass
welds in relation to the initial weld diffusible hydrogen, between different welding
processes, as well as for evaluation of the influence of welding procedural factors on
weld hydrogen. These are discussed further in Section 5.3.3.

5.3.3 Differences in final hydrogen concentration between SMAW and

SAW welds

According to Table 27, omitting the effect of weld bead overlap, i.e., d = 0, the final local
hydrogen concentration HRmax both in the case of SMAW and SAW remains always
lower than the initial weld diffusible hydrogen H0. Considering that d = 0 corresponds to
the conditions in single-pass welds, it is not surprising that this finding is in agreement
with the existing knowledge[3b, 13, 24, 25, 40, 59, 63] on the decaying exponential dependence of
the HR100max(d=0) /H0 ratio on ΣD∆t in the case of single-pass welds as defined in Eq. (1a),
see Section 1.2.1.
Inclusion of the overlap effect, i.e., d > 0, allows one to conclude that overlapping will
eventually raise the HRmax in the filling runs of multiple-pass welds, as shown in Fig. 33.
Under these conditions, hydrogen accumulation is found different between the SMAW
and SAW processes, see Table 27.
In SMAW welds, HRmax(d>0) remains equal to or lower than H0, even though the effect
of weld bead overlapping in increasing the hydrogen accumulation is taken into account
in accordance with Eqs (50a)–(50d). Table 27 shows, this accrues from: (i) small bead
layer thickness hw that allows substantial amounts of hydrogen to escape in the weld
cooling stage before subsequent weld beads were laid. This is seen as a drastic reduction
in the HR100max(d=0) /H0 ratio with increasing ΣD∆t/hw2 in Table 27. Another reason
attributes to (ii) low degree of weld overlapping (d/hw = 0.6) characteristic of SMAW
process because of the weld beads exhibiting relatively low depth-to-width ratios. A such
low d/hw in SMAW cannot induce any significant rise in the HRmax(d>0) /H0 values as a
result of overlapping, in contrast to SAW associated with higher d/hw, as shown in Table
27.
The lower the weld cooling rate in SMAW, the greater becomes the difference
between HRmax(d>0) and H0. With prolonged weld t100 cooling times and hence greater
ΣD∆t/hw2 values, HRmax(d>0) is seen to remain considerably lower than H0, see Table 27.
Thus, provided weld cooling can be effectively retarded by means other than raising the
heat input (that would simultaneously increase hw), the risk of WM hydrogen cracking in
SMAW multipass welds should lessen along a decrease in HRmax(d>0), even when H0
remains unchanged. In practice, these means would include e.g. higher preheat/interpass
temperatures, longer interpass times and/or retarded post-weld cooling.
In SAW welds, HRmax(d>0) tends to become equivalent to H0 over a wide heat input
range. According to Table 27, this can be ascribed to high degree of weld overlapping
 136

owing to the d/hw ratios of 1.5–2.0 reported[32,33] typical to SAW process that generally
produces weld beads of high depth-to-width ratios, hence accentuating the re-melting
effect of previously solidified weld beads. Provided that SAW beads are also larger in
size than SMAW beads, as is the case in general[5b, 24, 25, 32, 22], the size effect in
conjunction with prolonged weld t100 cooling time is seen to accentuate the retention of
hydrogen still further. This stems from larger bead layer thickness hw and hence greater
diffusion distances that do not permit any remarkable hydrogen release during the weld
cooling stage before subsequent weld beads are laid. This is finally seen as a relatively
small reduction in the HR100max(d=0)/H0 ratio with increasing ΣD∆t/hw2, c.f., Table 27.
Table 27 shows that irrespective of the weld t100 cooling time or the thermal factor of
diffusion ΣD∆t, the HRmax(d>0)/H0 ratio in SAW tends to always range around unity,
except for very tiny bead size of hw = 3 mm (which are rarely encountered in practice SA
welding). This demonstrates that a combination of (i) large weld beads and (ii) high
degree of weld overlapping in SAW can result in conditions where the effusion and
release of hydrogen during the weld cooling stage actually become counterbalanced by
the subsequent additional hydrogen supply due to pronounced weld overlapping and
associated re-melting of previous beads. Therefore, instead of becoming reduced as in the
case of SMAW, HRmax(d>0) in SAW can eventually reach the level of H0, as shown in
Table 27. For SMAW, these conditions can also prevail, but only in the case of low
thermal heat inputs and hence short t100 cooling times, see Table 27.
Consequently, the analysis made here allows to conclude, under equivalent thermal
conditions and weld initial diffusible hydrogen in terms of ΣD∆t (or t100) and H0,
respectively, the final local hydrogen concentration HRmax can become higher in SAW
multipass welds than is the case with SMAW. This is in line with some of the previously
reported[8] cases where intense occurrence of WM hydrogen cracking was associated with
high heat input SAW, in particular.
It is noteworthy that although low ΣD∆t/hw2 values are seen to yield high HRmax(d=0)
contents during the weld cooling stage, low ΣD∆t/hw2 also retards further accumulation
of hydrogen in subsequent stages of weld overlapping by lowering the ‘thermal driving
force’ ΣD∆t for the diffusion of hydrogen. This is seen in Figs 32 and 33, as well as in
Table 27. That low values of ΣD∆t/hw2 reduce hydrogen accumulation during weld
overlapping stages, leads in most cases to lower multipliers of the HRmax(d>0) /HRmax(d=0)
ratio in the case of SAW than SMAW, as shown in Table 27. This, in turn, will balance to
some extent the differences in the final hydrogen concentration HRmax(d>0) between the
two welding processes.
Overall, according to the analysis in Table 27 it seems unlikely that HRmax could
exceed the H0 by more than ≈ 10%, neither in the case of SAW nor SMAW process. This
is consistent with the views[67a] of the TWI that it should be physically impossible for
HRmax to exceed H0, no matter whether it is some intermediate or the final weld bead
layer one is concerned with. Contrary to this, the Japanese modelling data[20] in Figs 10
and 33 indicate that HRmax(d>0) values exceeding H0 seem possible in principle, under
conditions comprising small bead layer thickness hw and provided that excessive weld
bead penetration, d, takes place[20]. In the absence of sound experimental evidence, and
based on the present calculations for the welds studied here, it feels reasonable to
conclude that HRmax ≈ H0 can be taken as a general approximation for the predictive
 137

formulae assessing the hydrogen cracking risk of multipass weld metals. For SAW welds
in thick plate, setting HRmax ≈ 1.1 ∗ H0 might be more appropriate.

5.4 Derivation of the 'safe-line' for critical hydrogen content in

multipass welds

According to the objectives of the present thesis, the first step was to define the weld
critical hydrogen content Hcr below which hydrogen cracking in multipass weld metals
will not occur even under the most stringent conditions realistically met in practice
welding fabrication. This approach was to serve as a ‘first line of defence’ against WM
cracking.
This chapter provides experimentally verified numerical formulae for calculating the
estimates for the weld critical hydrogen content Hcr with respect to hydrogen cracking in
multiple-pass weld metals. As candidates, those parameters appearing most promising
according to the experiments of the present thesis, as well as to the literature
data[2–9, 11, 12, 16–23, 26, 35, 37, 50, 51, 53, 54, 62, 67, 74, 76, 79, 94, 96, 97, 100, 102, 106], were chosen. These are
based on (i) weld longitudinal residual stress σresL and assumed equivalent to the true
yield strength of the weld metal Rp0.2(WM), (ii) weld chemical composition in terms of
either Pcm or CET, and (iii) weld metal maximum Vickers hardness in terms of HV5max.
The aim here is to form a fundamental basis for the evaluation of the descriptive
potential of these parameters for ‘safe’, yet, not overly conservative, predictions of the
Hcr in the case of multipass welds.

5.4.1 Dependence of weld critical hydrogen content on longitudinal

residual stress

According to the earlier theoretical and experimental work[2, 3, 13, 19, 23–25], the critical
hydrogen content with respect to cracking, Hcr, decreases exponentially as the
longitudinal residual stress σresL in a weldment increases. This can be regarded as
analogous to the conditions in the Implant hydrogen cold-cracking test where the critical
fracture stress σcrit in a notched specimen and under axial tensile loading has been
shown[13, 24, 25, 63] to depend on the microstructural hardness HVmax and the weld diffusible
hydrogen content HD. The general form of this dependence is given[24] according to Eq.
(51), as:
σcrit = A – B ∗ HVmax – C ∗ log(HD) (51)
where A, B and C are constants. From Eq. (51), stress σ is seen to depend on a
negative logarithm of hydrogen H, which is expressed in a general form as:
σ ≡ –C ∗ log(H) (52)
 138

Writing Eq. (52) in another form that expresses the weld critical hydrogen content Hcr
as a function of weld residual stress σres, we obtain:
Hcr = A ∗ 10(–B ∗ σres) (53)
It can be noted that the thereby derived Eq. (53) complies well with a corresponding
expression of Hcr as defined in Eq. (4) and given by Yurioka[23].
The Ring-Core measurements in Section 4.4 confirmed that the weld longitudinal
residual tensile stress σresL in the filling runs of thick multipass weldments is equivalent
to the actual weld metal yield stress Rp0.2, see Appendix 3. This applies to welds that are
long enough, i.e., ≥ 300 mm and under the most stringent restraint conditions that can be
met in rigid large-scale welded constructions. Thus, Eq. (53) is considered to represent
the ‘worst-case’ scenario with respect to weld metal cracking risk. The present results
that prove σresL(max) ≈ Rp0.2(WM) to prevail are regarded to justify the transferability of the
Implant test philosophy into the present WM cracking scheme. Provided an Implant
specimen extracted from all-weld metal is loaded axially up to the level of the all-weld
metal Rp0.2, the test itself should, in this sense, be descriptive of the stress-strain
conditions in heavily restrained multipass welded joint.
Evaluation in Section 5.3 that was based on analytical calculations showed, it is
reasonable to conclude that HRmax ≈ H0 describes the hydrogen factor characteristic of
multipass welds. Since it is particularly HRmax that is considered[19, 20] as the actual local
maximum hydrogen concentration being responsible for the occurrence of WM cracking,
we can conclude that HRmax ≡ Hcr in the case of multipass welds. On the other hand, H0
represents the initial weld diffusible hydrogen content determined using a single-pass test
according to IIW/ISO 3690[57]. It hence follows that for the analyses of the present study,
it is justified to postulate, Hcr = H0 can be adopted to ensure ‘safe’ estimates of Hcr. As H0
can be determined using a single-pass test, this means that a link between the outcome of
the Y- and U-Groove cracking tests given in Section 4.1 and the associated weld
hydrogen content measured using single-pass tests conforming to ISO/IIW 3690[57], can
be established by setting Hcr = H0.
Thus, in the present thesis Hcr denotes to ’standard’ weld initial hydrogen content
determined using single-pass test, i.e., Osaka University mercury method and gas
chromatography method for the SMAW and SAW weld metals, respectively.
The values of Hcr being known on the basis of the Y- and U-Groove test results and the
associated weld hydrogen measurements, and setting σresL = Rp0.2(WM), constants A and B
in Eq. (53) can be determined according to the SMAW cracking test results in Tables 3–
5. Consequently, a decaying exponential function obeying Eq. (53) was fitted into the
experimental SMAW cracking test data available at those weld Rp0.2 strength levels that
revealed WM hydrogen cracks in the corresponding cracking tests.
According to earlier experience[23, 24], the dependence of Hcr on σresL should obey
exponential function of a form of Eq. (53). Fitting according to the SMAW cracking test
data in Tables 3–5 and in Fig. 34 gives the following approximative expression for
calculating best estimates for Hcr as a decaying exponential function of weldment
longitudinal residual stress σresL:
Hcr = 275 ∗ 10(–0.00235 ∗ σresL) (54a)
 139

Fig. 34 shows that Eq. (54a) fitted according to the present SMAW cracking test data,
gives a ‘realistic’ description of the dependence of Hcr on σresL also with respect to the
SAW cracking experiments in Tables 6–7. For deriving a lower-bound of the ‘safe’
description of Hcr, a slightly modified expression is proposed as:
Hcr = 250 ∗ 10(–0.00235 ∗ σresL) (54b)
It should be realised that applying Eq. (54) errors may arise at low σres values, since
for the weld metal Rp0.2 levels below 550 MPa, only No Crack data exists, see Tables 3–
7. Consequently, weld Rp0.2 ≥ 550 MPa should be set as a validity limit to Eq. (54).
Fig. 34 shows the weld critical hydrogen content Hcr as a function of weld longitudinal
residual stress σresL, together with the Crack – No Crack boundary curve defined
according to the multipass SMAW Y- and U-Groove cracking experiments and obeying
Eq. (54a). As a reference, data from the U-Groove cracking tests for the SAW weld
metals c.f. Tables 6 and 7, as well as from the two-pass V-Groove cracking tests for thin
6 mm plates c.f. Table 10, have been added.

OK74.78

Hcr = 275 x 10 (-0.00235 x res)

20 Tubrod 15.26S
Tubrod 15.24S
Nittetsu L-80
(VTT)
S1 S2 ( res = 650 MPa)
Hcr (ml/100g DM)

15 OK75.75
T1
OK48.08
S2 ( res = 650 MPa - 10 % res)

Tubrod 15.27S

10 Tubrod 15.29S
T2

OK75.78

5 Esab
Atom Arc
12018
Nittetsu L-80
(NSC)

300 400 500 600 700 800 900 1000

res (MPa) wm0201e.dsf

Fig. 34. Weld critical hydrogen content Hcr as a function of weld longitudinal residual stress
σresL. The Crack – No Crack safe line fitted according to the SMAW Y- and U-Groove
cracking tests (Ti = 70–100°C). Data from SAW U-Groove cracking tests and thin plate V-
Groove cracking tests added, together with the scatter band associated with hole-drilling
measurements.
 140

According to Fig. 34 it can be explained, for instance, why the cracking sensitivity of
the extra-high strength Rp0.2 ≈ 680 MPa weld metals in thin plate V-Groove cracking
tests was such low as demonstrated in Table 10. For the V-Groove specimens fixed by
bolted clamps, relaxation of the σresL from the level of WM yield strength down to
around 400 MPa that was recorded in hole-drilling measurements, c.f., Appendix 3,
results in a corresponding shift of the data points to the left along the x-axis. The
influence of this stress relaxation in terms of tolerated hydrogen content can then be read
from the y-axis: the required Hcr to cause cracking in this case becomes so high that WM
hydrogen cracking is considered totally impossible in practice welding fabrication. In the
case of V-Groove specimens restrained with welded strong-backs, the level of σresL
would, in principle, have remained sufficiently high to cause WM cracking, as shown in
Fig. 34. That no cracks were recorded in this case either, is regarded as an indirect
evidence that some other factor apart from σresL must have been responsible for the
negligible WM cracking risk in the case of thin plate two-pass welds, see Table 10.
Analysing a recently reported[76] multipass cracking test data set on extra-high
strength FCAW weld metals applying Eq. (54) yields contradictory findings. These
experiments[76] comprise multipass welding tests on fully restrained plate specimens with
a 60° V-groove preparation and using both basic and rutile type tubular wires designated
as E110T-5 K4 and E111-T1 K3, respectively. Values of Rp0.2 for the corresponding
basic and rutile weld metals are reported[76] as 700 and 755 MPa, respectively, whilst
those of RM were 855 and 825 MPa, respectively. It is evident that these welds become
ranked differently with respect to their cracking susceptibility, depending on whether
ranking is based on weld metal’s Rp0.2 or RM. Adopting Rp0.2 as a controlling factor in
accordance with Fig. 34, neither Eq. (54a) nor (54b) would predict an individual test
result[76] with the most critical combination of weld metal Rp0.2 = 700 MPa and weld HD
= 4.3 ml/100 g DM (IIIW) to locate within the Cracking region. This, in turn, contrasts to
the reported[76] NDT indications where hydrogen cracks were recorded from this
particular weld. Setting Rp0.2 as 700 MPa and applying Eq. (54b) gives, in this case, the
Hcr level of 5.6 ml/100 g DM (IIW), whilst cracking was recorded[76] at the 4.3 ml/100 g
weld hydrogen, see Table 28.
Since the RM of these weld metals, rather than Rp0.2, is found to comply with weld
Pcm, these FCAW results[76] are discussed further in the context of the Pcm based
approach in Section 5.4.2.
Table 28 compares the calculated values of weld critical hydrogen content Hcr
applying Eq. (54) to those values derived from the results of the Y- and U-Groove
cracking experiments in Tables 3–7. Values of Hcr are given as a function of weld
residual stress σresL for the examined SMAW and SAW weld metals. Table 28 shows, in
all the cases where cracking did occur in the SAW weld metals, estimation of the Hcr
according to Eq. (54b) fitted by the SMAW data would have indicated these test results to
locate in the Cracking region, c.f., Fig. 34, thereby leading to ‘safe’ Hcr estimates for the
SAW welds, as well. It should be realised, however, that only in the case of OK Tubrod
15.26S welds, both cracked and non cracked data were available, the rest of the SAW
welds made with OK Tubrod 15.24S, OK Tubrod 15.27S and OK Tubrod 15.29S
displaying either only cracked, or only non-cracked, results. This inevitably leaves us
 141

with some uncertainty of the absolute safety of the Hcr estimates even according to Eq.
(54b) in the case of SAW welds.
However, accounting for the presumed difference in the actual HRmax between SAW
and SMAW welds as discussed in Section 5.3.3 and shown in Table 27, the level of ‘true’
HRmax for SAW welds can be assumed 10–40% higher than for SMAW welds.
Considering this elevation, the experimental SAW data in Table 28 and Fig. 34 imply,
the corresponding adjustment of the Hcr values obeying Eq. (54b) would have yielded
‘safe’ Hcr estimates also in the case of the SAW welds in Table 28.
Table 28. Weld critical hydrogen content Hcr at different WM strengths as a function of
weld longitudinal residual stress σresL and assuming weld σresL = Rp0.2.
Weld Weld critical hydrogen Hcr Nature of result: SMAW / SAW
longitudinal (ml/100 g DM IIW) – experimental Filler material
residual stress: calc..(* calc.(** experim.(*** – interpolated Electrode or consumable
σresL – extrapolated
(MPa) – literature
920 1.9 1.7 3.3–4.3 semi-experimental OK 75.78 (nominal)
890 2.2 2.0 < 3.2 experimental Atom Arc 12018 (true)
890 2.2 2.0 << 4.9 experimental OK Tubrod 15.29S / OK
Flux 10.62
885 2.3 2.1 3.3–4.3 experimental OK 75.78 (true)
835 3.0 2.7 < 3.2 semi-experimental Atom Arc 12018
(nominal)
820 3.2 3.0 – interpolated
765 4.4 4.0 < 4.9 experimental OK Tubrod 15.27S / OK
Flux 10.62
755 4.6 4.2 < 8.5 literature[76] FCAW E111T-1 K3
745 4.9 4.3 > 4.0 experimental Nittetsu L-80 (NSC)
700 6.2 5.6 < 4.3 literature[76] FCAW E110T-5 K4
690 6.6 6.0 6.2–10.1 experimental OK 75.75
680 6.9 6.3 10.9–13.2 semi-experimental Nittetsu L-80 (VTT)
620 9.6 8.7 – extrapolated
585 11.6 10.5 8.4–13.6 experimental OK Tubrod 15.26S / OK
Flux 10.62
550 14.0 12.7 15.1–15.5 experimental OK 74.78
530 (15.6) 14.2 > 8.4 experimental OK Tubrod 15.24S / OK
Flux 10.62
480 (20) 18.6 > 10.8 extrapolated OK 48.08
*)
first sub-column: exponential best fit calculated acc. to Eq. (54a)
**)
second sub-column: exponential lower-bound safe fit calculated acc. to Eq. (54b)
***)
third sub-column: range of highest safe value-lowest unsafe value, or the highest individual
safe value (>), or the lowest individual unsafe value (<) according to the experimental cracking
test data

According to Table 28, an individual FCAW test result[76] with the combination of
weld metal Rp0.2 = 700 MPa and weld HD = 4.3 ml/100 g DM (IIW) is the only single test
result whose cracking risk prediction applying Eq. (54b) appeared unconservative. It can
 142

be seen that this particular test result is also the only one that contrasts radically with the
rest of the SMAW, SAW and FCAW data. Anyway, this implies that the σresL = Rp0.2
based approach may not guarantee safe description of the Hcr for weld metals with
comparatively low yield-to-tensile ratios and high strain hardening characteristics.
In Section 1.2.3, results are quoted from a former NSC study[12] where hydrogen
cracking in multipass weld metal L-80 with Rp0.2 ≈ 745 MPa has been reported at weld
HD ≈ 4 ml/100 g DM (IIW). According to Fig. 34, such a situation would appear
unlikely. Table 28 shows, the critical Hcr corresponding to the 745–750 MPa yield
strength level would lie around 4.3 ml/100 g DM (IIW). It should be borne in mind,
however, that the former NSC experiments[12] at that time, exhibited constantly greater
cracking risk than the experiments of the present thesis, as shown in Appendix 2. Also
any attempts to reproduce the former NSC test results in the experimental phase of the
present thesis appeared unsuccessful. This discrepancy has been addressed also in
Chapter 1.7 and possible explanations discussed in more detail in Section 5.12.2.
Microscopic investigation in Section 4.8 revealed that significant amounts of grain-
boundary ferrite GBF were present in the OK 48.08 and OK 74.78 weld metals, as shown
in Table 23. None of these SMAW weld metals, however, revealed any transverse
hydrogen cracks in the Y- or U-Groove cracking tests after a ‘standard’ 16 h period[1]
before the NDT inspection, see Tables 4 and 5. Provided the presence of GBF in these
welds had truly reduced the critical hydrogen content for WM cracking, this effect had
obviously been too small to be noticed in the present cracking tests. Another explanation
is that GBF affects WM cracking sensitivity only at low weld hydrogen levels of less
than ≈ 4 ml/100 g DM (IIW)[16–18], i.e., at levels considerably lower than those associated
with the cracked OK 74.78 welds studied here.
It is therefore obvious that WM hydrogen cracking in the lower strength weld metals
of Rp0.2 ≤ 550 MPa requires some other detrimental factor apart from σresL to provoke
cracking. Indeed, a significant detrimental factor other than σresL was, according to the
data in Table 4, obviously the holding period before the NDT was made, since prolonging
the period up till 168 h eventually resulted in the occurrence of WM hydrogen cracking
in some of the OK 74.78 weld metals. It is noteworthy that some of the cracks associated
with this prolonged holding time appeared at slightly lower weld HD contents than those
still producing sound welds in the case of the 16 h holding time, c.f. Table 4 and Figs 34–
36. Whether the prolonged holding period was necessary for hydrogen to diffuse and
accumulate to sufficient extent to trigger hydrogen-induced crack initiation, or for an
already initiated microcrack to grow further into a critical size, cannot be solely
concluded.

5.4.2 Dependence of weld critical hydrogen content on weld chemical

composition

In order to investigate compositional effects, a comparatively large WM cracking test

data set by Wong et al.[26] was taken as a basis for the assessment of cracking risk of the
 143

SMAW and SAW weld metals studied in the present thesis. These data are presented in
Fig. 35 and comprise results of Modified Cruciform (MC) and WIC cracking tests for
both single- and multiple-pass welds. The data are considered useful for the present
study, since the MC cracking test was described[26] as multipass WM cracking test that
provides the combination of high structural restraint and thermal severity due to specimen
geometry in conjunction with rapid weld cooling cycle.
Fig. 35 presents the ‘No Cracking Region’ of the original diagram (grey area) defined
according to the results of the WIC and MC experiments[26]. The diagram plots weld
critical hydrogen content Hcr against weld metal chemical composition in terms of Pcm.
Besides the severity of the MC test specimen, the experimental data applied to derive the
Cracking – No Cracking boundary was reported[26] to consist of extra high-strength weld
metals with RM ≈ 890–1000 MPa. Thus, the original diagram is, in this respect,
considered as representative of high structural rigidity and residual stress development
characteristic to the Y- and U-Groove tests of the present thesis.

Cracking: VTT Y-and U-groove tests

No cracking: VTT Y-and U-groove tests
Cracking: ESAB SAW U-groove tests
22 No cracking: ESAB SAW U-groove tests
OK74.78
Håkansson SAW
20 FCAW
Tubrod 15.26S Kuebler et al.
18 18
N-L-80 (VTT) Source: Richard J. Wong, 1996
16 16
Hcr (ml/100g DM)

OK75.75
Diffusible Hydrogen (ml/100g)

No cracking (WIC)
14 OK48.08 14 Tubrod 15.26S
Cracking (WIC)
Cracking (25 mm Cruciform)
12 12 Tubrod 15.27S
Tubrod 15.24S Tubrod 15.27S Tubrod 15.29S
Cracking
10 10 region
OK75.78 Atom Arc
12018
8 8

6 6

4 4

2 2 No cracking region
0
0.13 0.15 0.17 0.20 0.25 0.30 0.35
Pcm (%)
wm0206b.dsf

Fig. 35. ‘Weld metal Cracking – No Cracking’ Diagram according to Wong et al.[26]: grey
area. Weld critical hydrogen content Hcr as a function of weld metal Pcm. Safe line according
to VTT’s SMAW Y- and U-Groove cracking tests (Ti = 70–100°C) (spots): cracked: red
(black); non-cracked: yellow (grey). Data from SAW U-Groove cracking tests added
(diamonds & squares).

Next, the results of the SMAW Y- and U-Groove multipass cracking tests in Tables 3–
5 were added into Fig. 35 by plotting them against their Pcm values calculated for the
 144

weld metal using Eq. (43a). For the reasons described in Section 5.4.1, Hcr denotes to
’standard’ weld diffusible hydrogen content according to single-pass test, such as gas
chromatography or IIW/Osaka University mercury method. The red (black) and yellow
(grey) spots denote to Crack and No–Crack results, respectively. Fig 35 shows that in
general, the results of the Y- and U-Groove experiments made at VTT comply quite well
with the original results[26] of the WIC and MC cracking tests.
It can be seen that the original No Crack area according to Wong et al.[26] and given in
Fig. 35 has been expressed in the form of a triangle, thereby implying a decaying linear
dependency of weld Hcr on its Pcm. Therefore, the Y- and U-Groove cracking test data for
the SMAW weld metals of the present study and given in Tables 3–5 are firstly fitted
with a linear expression of the form:
Hcr = 116 ∗ (0.295 – Pcm) (55)
On the basis of the present SMAW cracking test data, Eq. (55) is valid in the Pcm and
Hcr ranges of: 0.13% ≤ Pcm ≤ 0.30% and 3.0 ≤ Hcr ≤ 19.0 ml/100 g DM (IIW).
As a step further, the SAW U-Groove cracking test data from Tables 6 and 7 have
been added into the original diagram in Fig. 35. It is noteworthy that all the U-Groove
cracking test results for the SAW multipass weld metals agree very well with the Y- and
U-Groove cracking test data base consisting of the SMAW welds.
Following the theory on microstructure-hydrogen-stress interrelations according to Eq.
(51), it is relevant to assume that the weld metal Hcr–Pcm relationship should, in fact, be
exponential, just as was the case for the weld Hcr–σresL relationship according to Eq. (53).
The SMAW Y- and U-Groove test results were therefore fitted by a decaying exponential
expression of the form:
Hcr = 175 ∗ 10(–6.460 ∗ Pcm) (56)
The validity limits for Eq. (56) are as those for Eq. (55).
Again, the U-Groove cracking test results for the SAW multipass weld metals are seen
to accord comparatively well with the Crack – No Crack boundary curve obeying Eq.
(56), as shown in Fig. 35. In one case only, an individual test result associated with the
OK Tubrod 15.26S weld metal, would have been predicted as crack-susceptible according
to Eq. (56), although it was actually crack-free as were also the other OK Tubrod 15.26S
weld metals, see Fig. 35. Table 7 shows, this individual SAW cracking test was welded
applying the Ti of 150°C, consequently the conditions were, in this respect, somewhat
less severe than in the SMAW U-Groove cracking experiments c.f., Table 4. Should this
particular SAW experiment made applying Ti ≤ 100°C, as was the case with the SMAW
U-Groove tests where the Ti = 70–100°C was applied, the associated cracking risk would,
accordingly, have been accentuated.
In general, the exponential expression obeying Eq. (56) is seen to shift the Crack–No
Crack boundary line slightly downwards, i.e., towards lower critical Hcr values, compared
to the linear expression according to Eq. (55), see Fig. 35. In this respect, Eq. (56) is
expected to yield safer predictions of weld Hcr than Eq. (55) in the intermediate weld Pcm
range of 0.18–0.28%, in particular. At low Pcm levels of less than 0.16%, Eq. (55) may,
in turn, be safer owing to its linearity, whilst the exponential expression c.f. Eq. (56)
 145

allows somewhat higher Hcr levels. It should be borne in mind, however, that for lower
strength weld metals with Rp0.2 ≤ 500 MPa, there exists only No Crack data.
As a comparison, a recently reported[76] multipass cracking test data set on extra-high
strength FCAW weld metals was analysed using the present diagram in Fig. 35. These
data[76] comprised multipass welding tests on fully restrained, 400 x 370 x 50 mm plate
specimens with 60° V-groove preparation and using basic and rutile type tubular wires
designated as E110T-5 K4 and E111-T1 K3, respectively. The reported[76] values of Rp0.2
for the corresponding basic and rutile FCAW weld metals were 700 and 755 MPa,
respectively, whilst those of RM were 855 and 825 MPa, respectively. For the cracking
tests, two levels of weld hydrogen HD were chosen: 4.3, 2.5 and 12.8, 8.0 ml/100 g DM
(IIW) for the basic and rutile consumables, respectively[76]. All welds were made under a
maximum Ti of 140°C and using comparatively low heat inputs of about 1.6 and 0.95
kJ/mm[76]. Ultrasonic and visual NDT, confirmed with examinations of extracted
macrosections, revealed[76] the occurrence of transverse WM hydrogen cracks in all the
other welds except in one: the basic weld metal associated with the lowest weld HD of 2.5
ml/100 g DM (IIW). It is noteworthy that whilst most cracks did appear already soon
after welding, some of them had appeared only after 5–34 days after the completion of
welding[76].
The reported[76] chemical compositions of the FCAW welds allowed calculation of
their corresponding Pcm values according to Eq. (43a), which resulted in values of 0.29
and 0.24% for the basic and rutile FCAW weld metals, respectively. Plotting these Pcm
values coupled with the corresponding weld HD (i.e., 0.29%: 2.5, 4.3 ml/100 g; 0.24%:
8.5, 12.8 ml/100 g) into Fig. 35 reveals that the application of the present Cracking – No
Cracking diagram would have predicted the cracking behaviour correctly, i.e.,
consistently with that recorded in the NDT examination[76] for all the four weld metals. It
can be seen that, with the exception of the weld associated with the lowest HD content of
2.5 ml/100 g, all the other welds locate clearly in the Cracking region of the diagram in
Fig 35. Even the one with the lowest weld HD appears to locate quite close to the Crack
region and the ‘safe line’, so the applied Ti of 140°C in the cracking tests has presumably
contributed to the crack-free outcome.
A recently published[117] data set consisting of cracking tests on extra-high strength
SAW welds offered another occasion to validate the present diagram in Fig. 35. These
data[117] comprised multipass welding tests on rigidly restrained, 800 x 150 x 50 mm plate
specimens with 70° Y-groove preparation and 28 mm weld build-up thickness. Three
combinations of solid and tubular wires were used, designated as: (i) SPOOLARC 120 +
SPOOLARC 140 (OK Flux 10.63), (ii) OK Tubrod 15.29S (OK Flux 10.47) and (iii) OK
Tubrod 15.29S + SPOOLARC 140 (OK Flux 10.47). The true weld metal strength was not
given, however, reported[117] chemical compositions of the SAW weld metals allowed
calculation of their corresponding Pcm and CET values acc. to Eqs (43a) and (43b),
respectively. This resulted in Pcm and CET values of 0.26, 0.28 and 0.30% and 0.36, 0.40
and 0.42% for the three SAW weld metals, respectively. Each of them displayed an
individual weld hydrogen HD content, which were recorded as: 4.5, 8.3 and 7.0 ml/100 g
DM (IIW), respectively[117]. For each weld metal, the experiments were made under three
different T0/Ti of 80, 140 and 200°C and using moderate heat inputs in the 2.3–2.7 kJ/mm
range[117]. Post test examination of the fracture surfaces revealed[76] that out of the total of
 146

nine experiments, only one test weld did not exhibit any transverse WM hydrogen cracks:
the one with weld HD of 4.5 ml/100 g DM (IIW) and associated with the lowest CET of
0.36% in conjunction with the highest T0/Ti of 200°C. It is, again, noteworthy that in
several welds cracks appeared only after 3 to 5 days after the completion of welding[117].
Elevating T0/Ti, or lowering the weld CET, in particular, were found to result in longer
fracture times[117].
The fact that one of the applied levels of T0/Ti was 80°C enables comparison between
those particular data[117] and the SMAW data of the present thesis and made under Ti of
around 70–100°C, c.f. Table 4. Plotting the Pcm values for the three SAW weld metals[117]
coupled with the corresponding weld HD (i.e., 0.26%: 4.5 ml/100 g; 0.28%: 8.3 ml/100 g;
0.30%: 7.0 ml/100 g), into Fig. 35 reveals that the application of the present Cracking –
No Cracking diagram would have, again, predicted the cracking behaviour correctly for
all the three SAW weld metals. It can be seen that, with the exception of the weld
associated with the lowest HD of 4.5 ml/100 g and Pcm of 0.26%, the other welds locate
clearly in the Cracking region of the diagram in Fig 35. The one with the lowest weld HD
and Pcm appears to locate nearly within the No Cracking region and thereby extremely
close to the ‘safe line’ according to Eq. (55). This implies that the more conservative,
exponential description for the ‘safe line’ obeying Eq. (56) is likely to provide better
assurance against weld metal cracking.
Overall, it can be concluded that the present Hcr–Pcm diagram according to Fig. 35, in
conjunction with Eq. (56), seems a viable ‘engineering tool’ in predicting safe, crack free
welding conditions for different types of multipass weld metals, once the WM’s true Pcm
and weld hydrogen HD are known. The applicability of composition based parameters,
such as Pcm, in indicating weld metal's hydrogen cracking risk has been demonstrated
experimentally also in a recent study[118] on high strength multipass SMAW weld metals.
It is interesting to note that the present Pcm based approach studied here provided a
better description of safe Hcr also for the FCAW welds in Ref.[76], in comparison with the
σresL = Rp0.2 based approach in Section 5.4.1. As for these FCAW welds, Pcm is found to
coincide with weld metal RM and not with Rp0.2, and since RM relates to hardness HV,
this suggests that weld metal HV would be more appropriate as a controlling factor, than
either Pcm or Rp0.2 alone.

5.4.3 Dependence of weld critical hydrogen content on weld

metal hardness

There exist evidence[6–9, 52, 53] showing that transverse hydrogen cracking in multipass
weld metals is often associated with microstructures exhibiting lower hardness than in the
case of the HAZ underbead/root cracking. Thus, any hardness level proven ‘safe’ against
HAZ cracking is not necessarily safe for weld metal. Current standards for specification
and qualification of welding procedures, such as EN 288-3 : 1992[10a], or its recent
successor, Draft prEN ISO 15614-1[10b], however, treat the welded joint as an entity and
do not thereby recognise the possibility of having hydrogen cracks in weld metal at lower
 147

hardness levels than those specified as permitted maximum values for a corresponding
parent steel group.
Earlier findings[27–30] have demonstrated that critical hardness with respect to hydrogen
cracking is not constant, but decreases as steel’s carbon content and hence CEIIW are
reduced. Following this, one may await lower critical hardness for weld metals with
typically no more than 0.04–0.05% C, in comparison with high-strength structural steel
with C ≈ 0.07–0.15%. Furthermore, as the recent cracking test results[76] for extra-high
strength FCAW multipass welds in Section 5.4.1 showed, basic and rutile welds may
exhibit quite different yield-to-tensile characteristics, resulting sometimes into situations
where it is the weld metal RM that seems to coincide with its cracking susceptibility,
rather than Rp0.2. Therefore, it is deemed necessary to investigate whether weld metal’s
cracking susceptibility could be expressed in terms of a relationship between weld metal
hardness HV5 and weld critical hydrogen content Hcr. For the reasons already described
in Section 5.4.1, Hcr denotes to ’standard’ weld diffusible hydrogen content according to
a single-pass test, such as e.g. gas chromatography or IIW/Osaka University mercury
method.
Consequently, the Y- and U-Groove multipass cracking test data for the SMAW weld
metals in Tables 3–5 were analysed in terms of weld diffusible hydrogen HD and weld
metal’s maximum hardness HV5(max) given in Tables 3–5 and 12, respectively. Fig. 36
shows the weld critical hydrogen content Hcr plotted as a function of the associated all-
weld metal maximum hardness HV5(max) for the experimental welds in accordance with
their Crack – No Crack results given in Tables 3–5. The cracked and non-cracked
specimens are denoted as red (black) and yellow (grey) spots, respectively. As
supplementary data, the SAW U-Groove cracking test data from Tables 6 and 7 have
been added into Fig. 35.
Weld critical hydrogen content with respect to cracking, Hcr, has been
shown[2, 3, 13, 19, 23–25] to decrease exponentially in accordance with Eqs (4) and (53) as the
longitudinal residual stress σresL in a weldment increases. The weld residual stress
measurements in Section 4.4, in turn, revealed σresL peak values comparable to weld
metal Rp0.2, as shown in Appendix 3. On the other hand, Fig. 31 shows how the weld
metal hardness HV5 increases with the tensile strength RM and, hence, with Rp0.2. These
findings result in a conclusion that the weld Hcr should, in principle, decrease in a similar
manner to Eq. (53), as the weld metal hardness HVmax increases.
The Implant cold cracking test has been shown[13, 24, 25] to describe the critical fracture
stress σcrit in axial tensile loading of a round-bar specimen as a function of
microstructural hardness HVmax and weld diffusible hydrogen HD in a specimen. Thus,
Eq. (51) giving a general expression of the dependence of σcrit on HVmax and HD, can be
converted into the form:
B ∗ HV ≡ C ∗ [–log(HD)] (57)
Resolving Eq. (57) to obtain a general expression for weld critical hydrogen, gives:
Hcr = C’ ∗ 10(–B ∗ HVmax) (58)
The values of Hcr and HVmax being known on the basis of the SMAW Y- and U-
Groove test results and the associated hydrogen measurements in Tables 3–5, and the
 148

corresponding weld hardness data being available in Table 12, constants C’ and B in Eq.
(58) can be determined according to the SMAW cracking test data c.f. Tables 3–5. The
SMAW Y- and U-Groove cracking test data in Fig. 36 were thereby fitted by a decaying
exponential expression obeying Eq. (58), being of the form:
Hcr = 280 ∗ 10(–0.00585 ∗ HVmax) (59)

OK74.78 Tubrod 15.26S

20
Nittetsu L-80
18 (VTT)
OK75.75

16 Nittetsu L-80
Tubrod (NSC) Tubrod
15.26S 15.29S
14
Hcr (ml/100g DM)

12 OK48.08 Tubrod
15.27S

10 OK75.78

8 Tubrod
15.24S

6
Esab
Cracked Atom Arc
4 Tubrod 12018
Non-cracked 15.26S

0 50 100 150 200 250 300 350 400

HV5 (max) (HV) wm0201f.dsf

Fig. 36. Weld critical hydrogen content Hcr as a function of weld metal maximum hardness
HV5(max). Safe line according to VTT's SMAW Y- and U-Groove cracking tests (Ti = 70–
100°C) (spots): Cracked: red (black); non-cracked: yellow (grey). Data from SAW
experiments added (diamonds).

The Crack – No Crack boundary curve obeying Eq. (59) is presented in Fig. 36. It can
be seen that also in terms of the Hcr–HV5(max) relationship, the present U-Groove test
results for the SAW multipass weld metals in Tables 6 and 7 comply very well with the
SMAW data from the Y- and U-Groove cracking experiments given in Tables 3–5.
According to Fig. 36, in all the cases where cracking did occur in the SAW weld metals,
estimation of the Hcr applying Eq. (59) fitted by the SMAW data would have indicated,
 149

these test results will locate in the Cracking region c.f., Fig. 36, thereby leading to safe
Hcr estimates for the SAW welds, as well. Accordingly, in all the cases where cracking
did not occur in the SAW weld metals, Eq. (59) would also have predicted these test
results as crack-free, i.e., the Hcr for these HVmax–HD combinations were predicted higher
than the actual weld HD content of a given specimen. As supplementary data, the Crack –
No Crack data points of the SAW cracking tests have been added into Fig. 36.
Since the two recent investigations on FCAW[76] and SAW[117] welds did not report
any weld metal hardness data, such validation of the present HVmax based approach that
was accomplished in the case of the Pcm based compositional diagram, could not be made
here. However, a previous investigation[122] concluded that for a weld metal hydrogen
level of HD = 5 ml/100 g, the WM hardness must be kept below 300 HV to avoid
hydrogen cracking. This is very well in line with the outcome of the present Hcr–HV5(max)
diagram c.f. Fig. 36, as it was also pointed out in a recent review[123] on consumable
development and the associated effects on hydrogen cracking.
On the basis of the present cracking test data as a whole, the validity limits to Eq. (59)
should be set approximately as: 210 > HV5(max) ≥ 370 HV and 2.0 ≤ Hcr ≤ 19.0 ml/100 g
DM (IIW).
In Section 1.2.2, results are quoted from a study[8] where hydrogen cracking in SAW
multipass weld metals with a hardness of about 200 HV was recognised. According to the
results of the present thesis, such a situation would not have been expected, until at the
weld HD amounting 18–19 ml/100 g DM (IIW), see Fig. 34. This implies that the
descriptive potential of the weld HVmax as a parameter predicting the WM cracking risk,
may weaken towards lower WM hardness. On the other hand, hardness such as 200 HV
is situated at the lower tail of the whole data set of the present study and is therefore,
strictly speaking, outside the validity limits of Eq. (59). This is by no means to state that
the outcome of the quoted study[8], itself, would be insignificant or irrelevant to consider,
but only to address that factors others than weld hardness may become increasingly
dominant at low hardness welds. Investigation on such factors in low strength weld
metals was not the scope of the present thesis. The applicability aspects are discussed in
more detail in Section 5.12.7.
It feels reasonable to conclude that Eq. (59) is a safe description of the Hcr–HVmax
dependence in the case of multipass WMs, at least for the majority of the present
HV5(max) data in the examined HV range. With respect to the lower tail of the data, i.e.,
HV5(max) ≤ 230 HV, one cannot, however, be fully convinced of the safety of Eq. (59),
since only non-cracked data are available, as shown in Fig. 36. Therefore, the precise
level for the lowest HD value that will be high enough to cause WM cracking at these low
weld metal HV levels, cannot be established.
 150

5.5 Comparison of the descriptive potential of approaches based on

stress, composition and hardness related estimates

Table 29 compares the Hcr estimates for the SMAW and SAW weld metals calculated
according to the weld longitudinal residual stress (= yield stress), weld metal chemical
composition and weld metal hardness based approaches discussed in Sections 5.4.1, 5.4.2
and 5.4.3, respectively.
The estimates according to σresL (= Rp0.2), Pcm and HV5(max) based approaches have
been calculated applying Eqs (54a) & (54b), Eqs (55) & (56) and Eq. (59), respectively.

5.5.1 Descriptive potential for the SMAW weld metals

For the SMAW welds the three approaches are all seen to yield essentially the same
results and similar ranking of the examined weld metals in terms of their cracking
susceptibility, thereby indicating good descriptive potential of the applied parameters:
σresL = Rp0.2, Pcm and HV5(max) in assessing hydrogen cracking risk of multiple-pass weld
metals.
Looking the three approaches individually in the case of SMAW, the Pcm approach
acc. to Eqs (55) and (56) is constantly found to predict somewhat higher Hcr estimates
than the other two approaches; this is particularly so for the linear Pcm fitting acc. to Eq.
(55). The differences in the Hcr estimates between the exponential Pcm based fit acc. to
Eq. (56) and the other two approaches, however, is comparatively small. The linear Pcm
based fit that resembles closely the original treatise of the Hcr–Pcm diagram by Wong et
al.[26] c.f. Fig. 35, is seen to deviate quite a lot from the other two approaches in that the
Hcr estimates tend to become higher than those predicted by the other approaches. This
raises some suspicion that the linear Pcm based fit acc. to Eq. (55), as well as the Wong
diagram[26], may not always yield safe estimates of the weld Hcr content.
The use of the HV5(max) approach acc. to Eq. (59), in turn, is seen to yield in most cases
the lowest Hcr estimates among the three approaches, whereas those predicted by the σresL
= Rp0.2 approach acc. to Eq. (54) seem to fall in between those of the HV5(max) and Pcm
based approaches.
From the standpoint of safety, the outcome of the σresL = Rp0.2, Pcm and HV5(max) based
approaches was evaluated with respect to the range between the highest, but still safe, and
lowest unsafe, weld HD content in the cases of crack free and cracked SMAW Y- and U-
Groove experiments, respectively. It is found that the HV5(max) approach yields safe Hcr
estimates in every single case, as shown in Fig. 36 and Table 29. The lower-bound σresL
= Rp0.2 approach acc. to Eq. (54b) did so in all cases except one, i.e., Nittetsu L80 (NSC)
where the lowest ‘critical’ weld HD content that would cause cracking is not known,
because only one weld hydrogen level was attained, see Table 3a. The Pcm approach with
the exponential fit acc. to Eq. (55) is likely to yield accordingly safe Hcr estimates, as it
would not have predicted any of the cracked SMAW Y- or U-Groove experiments to
 151

locate within the No Cracking region, and neither did any of the three approaches,
compare Figs 34, 35 and 36.
Addressing one example, namely, the relative WM cracking risk between OK 75.78
and Atom Arc 12018 weld metals, illuminates the importance of knowing true weld metal
strength. In view of nominal Rp0.2, OK 75.78 welds should be somewhat more crack-
sensitive and correspondingly exhibit slightly lower weld Hcr content than Atom Arc
12018 welds, see Table 28. Contrary to this, weld HD values in the U-Groove cracking
tests c.f., Table 4, plotted against the weld Pcm, indicate slightly higher cracking
sensitivity for the Atom Arc 12018 than OK 75.78 welds, as shown in Fig. 35. The actual
U-Groove cracking test results also demonstrate this relative ranking between the two
weld metals, see Table 4.
Table 29. Comparison of the Hcr estimates for the examined SMAW and SAW weld metals
calculated according to the weld longitudinal residual stress σresL (= Rp0.2), weld metal
chemical composition Pcm and all-weld metal hardness HV5(max) based approaches.
Consumable / WM WM WM WM Weld critical hydrogen content Hcr
Approach Rp0.2 RM Pcm HVmax (ml/100 g DM (IIW))
(nom) *) (nom) *) (%) (HV5) σresL (=Rp0.2) Pcm HVmax
(MPa) (MPa) Eq. (54a) 1) Eq. (56) 3) Eq. (59) 5)
Eq. (54b) 2) Eq. (55) 4)
OK 48.08 480 530 0.136 210 (20)–(18.6) (23) ; (18.1) (16.5)
OK Tubrod 15.24S 530*) 620*) 0.172 235 (15.6)–(14.2) (13.5) ; (14.3) (11.8)
OK 74.78 550 650 0.165 230 14.0–12.7 15.0 ; 15.1 12.6
OK Tubrod 15.26S 580*) 670*) 0.220 262 11.9–10.8 6.6 ; 9.4 8.2
OK Tubrod 15.26S 580*) 670*) 0.232 277 11.9–10.8 5.6 ; 8.1 6.7
Nittetsu L80 (VTT) 680 810 0.210 270 6.9–6.3 7.7 ; 10.4 7.3
OK 75.75 690 810 0.215 281 6.6–6.0 7.1 ; 9.9 6.4
OK 75.75 690 810 0.222 299 6.6–6.0 6.4 ; 9.1 5.0
Nittetsu L80 (NSC) 745 845 0.235 319 4.9–4.4 5.3 ; 7.8 3.8
OK Tubrod 15.27S 765 825 0.267 340 4.4–4.0 3.3 ; 4.4 2.9
OK Tubrod 15.27S 765 825 0.261 365 4.4–4.0 3.6 ; 5.1 2.1
OK 75.78 880 965 0.275 335 2.4–2.1 2.9 ; 3.6 3.1
OK 75.78 885 965 0.285 350 2.3–2.1 2.5 ; 2.6 2.5
OK Tubrod 15.29S 890*) 960*) 0.272 365 2.2–2.0 3.0 ; 3.9 2.1
OK Tubrod 15.29S 890*) 960*) 0.283 376 2.2–2.0 2.6 ; 2.8 1.8
Atom Arc 12018 890 1020 0.290 365 2.2–2.0 2.3 ; 2.1 2.1

1)
Exponential best fit calculated acc. to Eq. (54a)
2)
Exponential lower-bound safe fit calculated acc. to Eq. (54b) (italic)
3)
Exponential lower-bound safe fit calculated acc. to Eq. (56)
4)
Linear simple fit calculated acc. to Eq. (55) (italic)
5)
Exponential lower-bound safe fit calculated acc. to Eq. (59)
*)
Nominal strength not confirmed by tensile testing
Values in brackets (): outside the validity limits of the particular equation.

This is also the case when plotting the weld HD values from the U-Groove cracking
tests against the measured all-weld metal maximum hardness HV5(max). According to Fig.
36 one can realise why Atom Arc 12018 welds should exhibit higher RM than OK 75.78
weld metals – that is because of the higher hardness HV5(max) recorded for the Atom Arc
 152

12018 weld metal. In terms of nominal strengths of these two SMAW weld metals,
however, an opposite ranking would have been expected. According to Table 11, the
measured all-weld metal strength for Atom Arc 12018 welds appears notably higher than
the nominal values given by the manufacturer. Thus, the actual RM of Atom Arc 12018
welds was, within the arc energy range applied here, indeed higher than that of the OK
75.78 welds. The discrepancies in the relative WM cracking sensitivity between the OK
75.78 and Atom Arc 12018 welds are hence attributed to the different levels of their true
weld metal strength, in relation to the nominal strength.
Overall, analyses in Table 29 demonstrate that the HV5(max) approach will
unambiguously predict safe, yet realistic, Hcr estimates, hence being a potential approach
of guaranteeing crack-free weld metals. The Pcm approach, however, inevitably leaves us
with some uncertainty in those cases the Hcr estimates locate above the highest safe weld
HD content, while the lowest unsafe HD content remaining unknown, see Table 29 and
Fig. 35.

5.5.2 Descriptive potential for the SAW weld metals

As far as the SAW weld metals are concerned, comparison between Figs 35 and 36
reveals that the σresL = Rp0.2, Pcm and HV5(max) based approaches all rank these weld
metals ‘correctly’, i.e., relative to their actual cracking sensitivity recorded in the U-
Groove tests, compare Table 4 to Table 29.
Certain differences in the predicted Hcr estimates between the approaches, however,
are evidently seen. For instance, the Pcm and HV5(max) based approaches acc. to Eqs (56)
and (59), respectively, tend to constantly yield lower Hcr estimates than the σresL = Rp0.2
based approach, as shown in Table 29. The only exception, as in the case of SMAW weld
metals, is the linear Pcm based fit acc. to Eq. (55) that is seen to always predict somewhat
higher Hcr estimates than the exponential Pcm based or the HV5(max) based approaches, see
Table 29. That the Hcr estimates predicted by the σresL = Rp0.2 based approach tend to
become suspiciously high, may accrue from the fact that the strength values of the SAW
weld metals c.f., Table 29 and given by the consumable manufacturer do not accurately
correspond to those in the present U-Groove cracking experiments (here, it is worth
highlighting that tensile testing in the present study was conducted to the SMAW weld
metals only). Being so, the fact that the SAW weld metals studied here seem to be more
crack sensitive according to the Pcm and HV5(max) based approaches than the σresL = Rp0.2
based approach, implies that either (i) the true weld metal Rp0.2 in the SAW U-Groove
experiments have actually been higher than the values in Table 29, or (ii) the tubular
SAW consumables produce weld metals that are more crack sensitive than what would be
awaited according to their strength.
As the SAW welds in the U-Groove experiments were, unlike the SMAW
experiments, welded using three levels of arc energy: 2.0, 3.0 and 4.0 kJ/mm, one would
expect the effect of Q to appear in the cracking test results, as well as in the comparisons
of the cracking occurrence between SMAW and SAW welds. Figs 34, 35 and 36 prove,
however, that irrespective of generally higher heat inputs associated with SAW in
 153

comparison to SMAW, all the SAW U-Groove cracking test results still fell realistically
within the Crack–No Crack regions derived according to the SMAW cracking data.
Thus, the higher heat input in the case of SAW welds did not lead to such low
hardness weld metals, that this would have manifested itself as lower weld Hcr according
to the HV5(max) based than the Pcm based approach. In fact, the situation is rather the
reverse, that is, for the SAW weld metals studied here the weld Hcr is always predicted
lower according to the HV5(max) based than the σresL = Rp0.2 based approach. That the
weld Hcr is predicted lower according to the HV5(max) based than the Pcm based approach,
applies to the extra-high strength OK Tubrod 15.27S and OK Tubrod 15.29S weld metals.
This means that the alloying content of these weld metals had been sufficiently high to
retard any overall WM softening caused by high heat inputs. This is in line with the
hardenability analysis and weld hardness measurements given in Tables 24 and 13,
respectively, which show that these two weld metals have the alloying levels high enough
to attain their maximum hardenability and sustain it throughout the applied Q range. For
the high strength OK Tubrod 15.26S weld metals, the Pcm based approach yields the
lowest Hcr estimates among the three approaches, hence indicating that this leaner alloyed
weld metal has became softened to some extent as a result of high Q level, as shown in
Tables 13 and 24.
This can be regarded a demonstration of a diminutive effect of the arc energy on WM
hardness in the case of extra-high strength SAW multipass weld metals. Whilst the effect
of arc energy on weld cooling time and, hence, hardenability is incorporated, or actually
built-in, into the HV5(max) based approach, the Pcm based approach is a plain description
of weld chemical composition effects. Should the heat input induce a notable effect on
weld hardness, one would expect this effect to appear as notable differences in the SAW
cracking test result plots between the exponentially fitted Pcm and the HV5(max) based
approaches acc. to Eqs (56) and (59) and c.f. Figs 35 and 36. According to Table 29, no
such differences are recognised.

5.5.3 Descriptive potential for the combined data sets of SMAW and
SAW welds

Analysing the SMAW and SAW cracking test results in Tables 3–7 as a combined data
set, certain differences can be found regarding the descriptive potential of the σresL =
Rp0.2, Pcm and HV5(max) based approaches in predicting estimates of the weld Hcr. The
data in Table 29 reveal differences in the relative ranking between the SMAW and SAW
weld metals with respect to their cracking behaviour, depending on the applied approach.
Looking the 890 MPa yield strength level, the predicted Hcr estimates for the SMAW
and SAW weld metals are comparatively similar irrespective of the applied approach.
However, when comparing the 765 MPa yield strength SAW welds to the 880 MPa yield
strength SMAW welds, it can be seen that the weld Hcr estimate according to the σresL =
Rp0.2 based approach is allowed nearly twice as high for the OK Tubrod 15.27S weld
metal (Rp0.2 ≈ 765 MPa) than for the OK 75.78 welds (Rp0.2 ≈ 880–885 MPa), whereas
 154

the HV5(max) based approach predicts almost similar Hcr for both SMAW and SAW weld
metals, regardless of the over 100 MPa difference in their yield strengths.
This controversy between the SAW and SMAW welds applies more or less also to
lower yield strengths in the 580–690 MPa range. For instance, a comparison between the
OK Tubrod 15.26S SAW weld metals (Rp0.2 ≈ 580 MPa) and the Nittetsu L-80 & OK
75.75 SMAW welds (Rp0.2 ≈ 680–690 MPa) reveals, again, that whilst the σresL = Rp0.2
based approach predicts the weld Hcr estimate for the OK 75.75 and Nittetsu L-80 to
remain no more than half of that permitted for the OK Tubrod 15.26S weld metal (Rp0.2 ≈
765 MPa), the Pcm and HV5(max) based approaches predict nearly similar weld Hcr for
both weld metals, regardless of the about 100 MPa difference in their yield strengths.
Should this difference in the yield strengths between the present SAW and SMAW
weld metals appear accordingly also in the U-Groove experiments given in Tables 4, 6
and 7, leads inevitably to the conclusion that the tubular SAW consumables c.f. Table 2b
produce weld metals that, in relation to their strength, are more crack sensitive than the
SMAW weld metals c.f. Table 2a at the corresponding, or even higher, strength.
Therefore, evaluating the descriptive potential of the three approaches for the entire
spectrum of the SMAW and SAW welds studied in the present thesis, the HV5(max) based
approach seems to accommodate the actual cracking susceptibility of different weld
metal grades and strengths most realistically. The HV5(max) based approach is
demonstrated to predict safe, yet realistic, weld Hcr estimates for all the examined cases
of the SMAW and SAW multipass weld metals, as shown in Table 29.
The SAW data also emphasise the danger of obtaining unrealistic, or even
unconservative, weld Hcr estimates in such cases where the σresL = Rp0.2 based approach
is applied to weld metal whose actual strength properties e.g., in the cracking test differ
essentially from the nominal weld metal strength of a particular consumable.
From the fundamental viewpoint, certain differences in physical characteristics of the
investigated parameters, σresL, Pcm and HV, can be endorsed. Theoretically, it can be
stated that the σresL based approach is based on actual physical properties of the material,
i.e., yield strength, and is in this sense both scientifically relevant and theoretically
justified. Weld residual stresses, in particular, relate to the weld metal yield strength, and
not the ultimate tensile strength. The Pcm based approach, in turn, ignores the influence of
weld thermal cycle and weld residual stress on the cracking behaviour. As far as hardness
is concerned, it is considered merely an indication, or indirect measure, of a property
called ‘tensile strength’, as shown in Fig. 31. A practical advantage of using the HVmax
based approach is, however, that hardness combines conveniently the influence of both
weld chemical composition and welding thermal cycle on weld metal's hardenability.
Besides, weld hardness is still relatively simple to measure in practice and these
measurements are required, anyway, in standards for specification and qualification of
welding procedures, such as EN 288-3:1992[10a] and Draft prEN ISO 15614-1[10b]. If
necessary, hardness can be measured even without breaking the specimen by using
devices for surface measurements.
The Pcm and HV5(max) based approaches acc. to Eqs (56) and (59) are, in this sense,
perceived as ‘engineering tools’ which are handy to use in practice, whereas the
difficulties in applying σresL = Rp0.2 based approach acc. to Eq. (54) lie on its accurate
quantification under situations where no directly measured data exist.
 155

5.6 Role of weld build-up thickness and residual stress in weld metal
hydrogen cracking

Presently, different views exist[2, 3, 76] on whether plate or, actually, weld build-up
thickness should be considered, or not, as an influential factor affecting the risk of
hydrogen cracking in multipass weld metals. Okuda et al.[2] found the influence of weld
build-up thickness aw quite minimal and hence insignificant in the 20–80 mm thickness
range. Contrary to this, Yurioka et al. reported[3, 12] an increase of the hydrogen cracking
susceptibility with the weld build-up thickness aw in the 20–50 mm thickness range,
above which the effect gradually starts to level off as thickness increases further.
The V-Groove experiments in Table 10 demonstrate that none of the SMAW or SAW
cracking tests made using thin plate two-pass welded specimens of 6 mm thickness
resulted in transverse WM hydrogen cracks at any of the applied weld hydrogen contents
in the range of HD ≈ 9.2–16.0 ml/100 g DM (IIW), despite the applied extra-high strength
weld metal of Rp0.2 ≈ 680 MPa. In this respect, risk to WM hydrogen cracking, at a
comparable level of weld metal strength and weld hydrogen, is found considerably less
for the 6 mm weld build-up thickness aw in thin plate, than in the case of heavy plate
multipass welds of aw = 40 mm.
Analysis of weld residual stresses in thin plate indicate, for the specimens restrained
by bolted clamps relaxation of longitudinal σresL from the WM yield strength level down
to around 400 MPa took place after the clamps were removed, see Appendix 3. According
to Fig. 34, this relaxation would be enough to shift the cracking test data points from the
Crack-region into the No Crack-region. With respect to the specimens restrained by
welded strong-backs, such stress relaxation did not occur, consequently, some other
factor apart from longitudinal residual stress is responsible for the fact that these
experiments did not reveal any cracks either, see Fig 34.
It is hence concluded that for the majority of thin plate structures, the σresL can be
expected to remain substantially lower than in the case of heavy plates, unless the
structural configuration is so extremely rigid that it can prevent any possible
deformations both during and after welding.
The residual stress measurements for multipass welds in thick plate, in turn,
demonstrated that the weld longitudinal tensile residual stress σresL approaches the level
of weld metal true yield strength Rp0.2(WM), provided the weld is long enough, i.e.,
≥ 300 mm. Raising the weld (and specimen) length up to 800 mm was shown to elevate
the σresL still further, consequently σresL met a level equivalent to the true weld metal
Rp0.2. The absolute rise in σresL with increasing weld length from 300 to 800 mm is,
however, regarded as minuscule, as shown in Appendix 3.
Fig. 37 plots the weld Hcr for thin plate SMAW and SAW welds as a function of weld
metal Pcm. These SMAW and SAW welds were made using OK 75.75 and OK Tubrod
15.27S, respectively, whereas the TM parent steel was of a leaner chemical composition.
Comparing Figs 35 and 37 clearly shows how the welding of thin plate using two passes
has lead to substantial dilution of weld chemical composition, in relation to the
corresponding weld metals in multiple-pass welded thick plate. The corresponding Pcm
values for the OK 75.75 and OK Tubrod 15.27S weld metals in thin plate have descended
 156

from 0.22 to 0.18% and from 0.26 to 0.21%, respectively. In the case of the OK 75.75
weld metal T1, the intensified cooling conditions employed in SMAW have obviously
compensated part of this reduction in hardenability due to leaner alloying, since the WM
hardnesses for thin and thick plate welds were nearly identical. The corresponding HVave
and HVmax values in these cases were ≈ 270 and 280 HV, respectively, see Table 12.
Because of the absence of intensified cooling arrangements in the case of SAW, the
leaner weld metal chemistry coupled with prolonged weld t8/5 cooling time due to a two-
dimensional heat flow in thin plate, have resulted in a marked decrease of WM hardness,
and hence strength, in thin plate weld metal S1 from the values of the corresponding OK
Tubrod 15.27S thick plate welds. The corresponding HVave and HVmax values for the thin
and thick plate OK Tubrod 15.27S welds were ≈ 257 and 266 HV and 313–315 and 336–
367 HV, respectively, Table 13. This, in turn, manifests itself as an additional reduction
in the risk of hydrogen cracking in the SAW thin plate welds studied here. Despite this,
however, those SMAW and SAW welds associated with the highest weld hydrogen
contents still locate within the Crack-region as it is defined for the multipass thick plate
welds, see Fig 37.

22
Cracking: VTT Y-and U-groove tests
OK74.78 No cracking: VTT Y-and U-groove tests
20
Tubrod 15.27S
18 18
N-L-80 (VTT) Source: Richard J. Wong, 1996
S1
16 16
Hcr (ml/100g DM)

T1 OK75.75
Diffusible Hydrogen (ml/100g)

14 14 No cracking (WIC)
OK48.08 Cracking (WIC)
Cracking (25 mm Cruciform)
12 12
Cracking region
10 10
T2 OK75.78 Atom Arc
12018
8 8

6 6

4 4

2 2 No cracking region

0
0.13 0.15 0.17 0.20 0.25 0.30 0.35
Pcm (%)
wm0206a.dsf

Fig. 37. Weld critical hydrogen content Hcr as a function of weld metal Pcm. Safe line
according to VTT’s SMAW experiments (Ti = 70–100°C) (spots): cracked: red (black); non-
cracked: yellow (grey). Data from thin plate SMAW and SAW experiments added (squares).

Structures made of thin plate, but simultaneously rigid enough to prevent any weld
residual stress relaxation via deformation, are regarded extremely rare. Overall, it is
therefore concluded that in the case of thin plate (< 10 mm) transverse WM hydrogen
cracking is precluded, provided weld metal Rp0.2 ≤ 700 MPa and weld HD ≤ 16 ml/100 g
 157

DM (IIW). In relation to heavy plate welds of similar strength and hydrogen, the absence
of transverse WM cracks in thin plate welds is thereby attributed to (i) lower weld
longitudinal residual stress, negligible transverse stress and stress triaxiality, coupled with
(ii) enhanced hydrogen effusion and escape due to less weld metal volume beneath the
final bead.
A recently reported[76] work on multipass V-Groove cracking tests of extra-high
strength FCAW weld metals comprise experiments aimed at determining the effect of
weld build-up thickness aw on the occurrence and presence of weld metal hydrogen
cracks. Multipass weld metal cracking tests were made as 50, 40, 30 and 20 mm thick
welds on fully restrained 400 x 370 x 50 mm plate specimens with 60° V-groove
preparation and using rutile tubular wire designated as E111-T1 K3. Welding was made
at 1.6 kJ/mm arc energy using the T0/Ti of 140°C. The weld diffusible hydrogen was
measured as 8.0 ml/100 g. Subsequent NDT and destructive examinations revealed, a 50
mm thick weld exhibited high crack density of uniformly distributed transverse sub-
surface cracks distributed through the top half of the weld cross section. A 40 mm thick
weld showed substantially less cracking, with lower crack density, whereas the 30 mm
thick weld accordingly revealed cracks, but again the cracking density had decreased.
Finally, no cracking was found in the 20 mm thick weld.
These very recent findings[76], in conjunction with the results of the present study seem
to support the findings of Yurioka et al.[3, 12] in that reducing weld build-up thickness aw
will influence the WM hydrogen cracking risk by progressively reducing it. All the
reviewed studies[2, 3, 12, 76] consistently report that the effect of weld build-up thickness on
cracking is most prominent in the 20–50 mm aw range. In fact, no data was available for
aw below 20 mm[4]. The results of the present thesis demonstrate that the WM cracking
risk is progressively reduced with the aw and becomes insignificant at weld build-up
thicknesses below 10 mm, regardless of high weld metal strength and hydrogen content.

5.7 Effect of weld bead size on hydrogen diffusion distance and

cracking risk

Earlier work[8] on SAW multipass welding of thick plate have shown that contrary to the
HAZ, hydrogen cracking risk in SAW multipass weld metals is accentuated as the heat
input increases. This was ascribed to greater diffusion distances of hydrogen as bead size
increases, which results in higher local concentrations of hydrogen within the weld
metal[8,9].
Microscopic examination of the experiments made in the present thesis revealed that
bead size in the filling layers of the SMAW welds was not essentially different between
the two-pass thin plate welds and thick plate multipass welds, provided they were welded
with comparatively similar heat inputs, see Table 22. Only in the cases where the thick
plate welds were made using clearly higher heat inputs than for the thin plate welds, the
bead size both in terms of bead height, ab, and bead width, b2, b1 and bave, became
accordingly greater.
 158

With respect to SAW, bead height ab appears relatively constant between thick and
thin plate welds, irrespective of the higher heat input that was used consistently for thick
plate welds, see Table 22. Bead width b2, however, tends to become greater in thick
multipass welds as a consequence of higher heat input, compared with the two-pass thin
plate welds. This accrues from relatively similar welding current in SAW welding of both
two-pass thin plate welds and multipass thick welds, as different heat inputs were attained
by adjusting the torch travel speed, compare Tables 7 and 10. As a result, weld bead
penetration and hence ab and hw become relatively similar, regardless of plate thickness.
Contrary to this, arc voltage was higher for the multipass welds, with associated widening
of weld bead width b2, as shown in Table 22.
Consequently, the differences in bead size of the final filling layers are mostly heat
input related and not so much welding process or plate thickness related. Realising that
the weld bead height ab and individual bead layer thickness hw, in particular, relate to
hydrogen diffusion distances in accordance with Eq. (37), one cannot expect significant
differences in the final local hydrogen concentration HRmax between SAW and SMAW
that would only arise from the reason that they are two different welding processes. With
regard to the possibility of having higher weld bead overlap ratio d/hw in the case of
SAW, its effects on hydrogen accumulation and hence on the HRmax cannot be confirmed
by the experiments performed in the present thesis. It is therefore thought that numerical
modelling using FDM and/or FEM would be viable of elucidating this phenomenon.
According to Table 4, in one case two arc energy levels of 3.0 an 2.0 kJ/mm were
adopted for the OK 74.78 electrode at an equivalent weld hydrogen content of HD = 15.5
ml/100 g DM (IIW), the corresponding specimens being U33 and U34, respectively. The
outcome was that U33 exhibited cracks (after 7 days from the completion of welding),
whereas U34 did not. This could be regarded as a demonstration that weld metal in U34
associated with lower arc energy and hence displaying smaller bead size than its
counterpart U33, was less prone to WM cracking due to the bead size effect, in particular.
Table 22 shows that arc energy in the 1.5–5.0 kJ/mm range affects notably the bead size
in the examined SMAW welds. These findings imply, bead size may induce a secondary
effect on cracking occurrence, once the other crucial factors, i.e., strength and hydrogen
are such to place the weld metal in question sufficiently close to its Crack – No Crack
boundary line. This would explain why the bead size effect seems to become
indiscernible for the majority of the data locating essentially either in the No Crack or the
Crack region, c.f. Figs 34–36. Anyway, as only one test per arc energy level is available
at a comparable hydrogen level for this comparison, one cannot draw any generalised
conclusions concerning the bead size effect.
According to the majority of the U-Groove experiments made in the present thesis, the
possibility of bead size affecting weld metal hydrogen cracking susceptibility as an
underlying mechanism, could not be confirmed nor excluded, see Tables 3–7 and Figs 35
and 36. What the cracking experiments indisputably demonstrate is that, eventually, the
SMAW and SAW cracking test results were practically congruent with each other, as
shown in Figs 34, 35 and 36, provided that the results were normalised with the
corresponding values of the true weld metal hardness HV5(max) and the actual weld metal
chemical composition Pcm.
 159

As a result, whatever the effects of bead size on WM cracking sensitivity might have
been – since the bead size in terms of width b2 was unambiguously greater in the case of
SAW – these effects are obviously still of a second order of magnitude, in relation to
weld metal strength and weld hydrogen.

5.8 Effect of weld metal chemical composition and local

microstructural mismatch on cracking sensitivity

Earlier work[37] on high strength welds suggests that those extra-high strength weld
metals which transform from austenite into ferrite at a lower temperature than the
surrounding parent steel HAZ may be more prone to WM hydrogen cracking than weld
metals displaying the reverse transformation sequence in relation to the parent steel HAZ.
Lower γ–α phase transformation temperature in the weld metal means that a greater part
of the weld cooling period will be spent with the weld metal being austenitic. As
austenitic, weld metal has lower hydrogen diffusion coefficients and higher hydrogen
solubilities than in the ferritic state, consequently, an austenitic weld metal can store high
hydrogen contents that cannot move fast enough to the fusion boundary due to slow
diffusion rate of hydrogen in austenite. This eventually results in elevated hydrogen
concentration in the weld metal[37].
Should this be the case, besides weld metal, parent steel grade in question may also
induce an indirect effect on WM cracking sensitivity. According to Table 21, those two
Ni-Cr-Mo bearing SMAW weld metals, OK 75.75 and OK 75.78, that had their γ–α
phase transformation temperatures lower than that of the parent steel were recognised to
exhibit extensive hydrogen cracking, whereas the other two leaner-alloyed OK 74.78 and
OK 48.08 weld metals associated with higher phase transformation temperatures than the
parent steel tended to be crack-free over a wide range of weld hydrogen. Since only one
parent steel grade was used in the present work, one, however, cannot reliably verify
whether this outcome in the cracking tests c.f. Tables 3–5 is a plain coincidence, or a
consequence of differences in the γ–α phase transformation temperature that had further
contributed to cracking behaviour of the examined SMAW weld metals.
In many cases in practice, weld metal and HAZ microstructures are not essentially
martensitic, which leads to thinking that temperatures such as Ar3 or BS might in these
cases be more relevant temperatures than MS. The way one study[37] equates martensite
with ferrite in distinguishing between microstructures with slow (austenite) or high
(ferrite/martensite) hydrogen diffusion rates implies that, as far as diffusion of hydrogen
is concerned, MS temperature would be a relevant description irrespective whether phase
transformation from austenite eventually results in a ferrite, or martensite, microstructure.
Any systematic comparisons of relevant temperatures in such cases, however, were not
reported in the literature. An MS temperature expression for weld metal, which
incorporates the influence of the WM oxygen content, has been proposed[37] as further
improvement to better describe the effect of oxide inclusion on austenite decomposition
 160

kinetics. This suggests that modified MS temperature expressions of that kind would
more realistically take account of weld metals that are not essentially martensitic.
Nevertheless, the rise in the WM cracking occurrence when moving from the 480–580
MPa yield strengths up to the 680–890 MPa yield strengths is found quite dramatic for
the investigated SMAW and SAW weld metals, as shown in Tables 3–7 and Figs 34–36.
Whilst Fig. 34 implies, this phenomenon could be explained solely by the residual stress
σresL effect, the rapid jump from the No Crack region into the Cracking region in the case
of SAW weld metals when transferring from OK Tubrod 15.26S to OK Tubrod 15.27S
leaves some doubt that the lower γ–α phase transformation temperature in the case of the
extra-high strength OK Tubrod 15.27S weld metal could have contributed to its
unambiguously higher crack sensitivity, compared to high strength OK Tubrod 15.26S
weld metal, see Tables 6–9 and Figs 35–36.
Overall, the present experimental results thereby demonstrate that the occurrence of
transverse hydrogen cracking in multipass WMs is associated with the development of
elevated residual stresses approaching the weld metal yield strength. Otherwise, critical
conditions can hardly be met within a multipass weldment that lacks any geometrical
discontinuity, such as a groove bevel acting as stress raiser in the case of single-pass root
or HAZ underbead cracking. Conversely, should the residual stresses become relieved, as
in the case of, e.g., thin plate welds and/or under conditions of low structural rigidity, it is
found nearly impossible to create cracks under any combination of strength and
hydrogen, as shown in Fig. 37 and Table 10.
It is widely demonstrated[2, 3, 16–18, 54, 98, 99, 114, 115] that intensive accumulation of plastic
strain in softer microstructures such as grain-boundary ferrite GBF is often a prerequisite
for WM hydrogen cracking in thick multipass welds. Metallographic examination
conducted in the present thesis, however, revealed that even the presence of high volume
fractions of GBF, alone, did not lead to crack formation. According to Table 23, the
highest amounts of GBF were associated with the OK 74.78 and OK 48.08 weld metals,
i.e., particularly those that tended to be crack-free, unless subjected to very long periods
before the NDT in the case of OK 74.78.
Assuming that the role of GBF acting as a path for diffusing hydrogen is conclusive, it
is not easy to explain why particularly those weld metals with the highest amounts of
GBF either did not crack at all, or required extreme conditions, i.e., the combination of a
7 days period and 15.5–17.5 ml/100 g DM (IIW) hydrogen, to reveal cracks. Meanwhile,
the 16 h period was found perfectly sufficient for cracking to appear in the extra-high
strength weld metals having much less GBF than the lower strength ones, see Table 23.
One explanation may be that all those recent cases where the presence of GBF was found
to indisputably contribute to WM hydrogen cracking, are reportedly[116] associated with
comparatively low weld hydrogen of HD ≤ 4.5 ml/100 g DM (IIW) and with experiments
using externally loaded hydrogen charged specimens[16–18]. In the Y- and U-Groove
cracking experiments of self-restraining nature, being applied in the present thesis, such
low weld HD is found absolutely inadequate to cause weld metal cracking in those 480–
550 MPa yield strength weld metals that contained the largest amounts of GBF.
Therefore, it is concluded that the cracking experiments applied in the present thesis
have been, in any case, unable to asses these microstructural effects due to GBF, whether
such effect existed or not. On the other hand, as most structural conditions in reality are
 161

of self-restraining nature, the occurrence of WM hydrogen cracking provoked by the

combination of HD ≤ 4.5 ml/100 g DM (IIW) and large fractions of GBF in the weld
metal appears somewhat unrealistic in the light of the present experimental data, compare
Tables 3–7 and Table 23.
Hardness traverses summarised in Table 30 reveal that an increase in SAW weld metal
Rp0.2 when transferring from 550–630 to 690–900 MPa yield strength level is
accompanied by a pronounced increase in local hardness mismatch within the multipass
weld metals. This manifests itself as an approx. 50–80 HV difference in hardness
between the individual maximum and minimum value in the case of OK Tubrod 15.27S
weld metal, whereas the corresponding difference for OK Tubrod 15.26S weld metal is
only about 20–25 HV.
Table 30. Local weld hardness mismatch ∆HV (= HVmax–HVmin) in the multipass SAW
weld metals at different levels of strength and arc energy.
Specimen SAW tubular wire / Arc Pcm CET HVmax–HVmin HVave ∆HV
No Flux OK 10.62 energy difference (HV)
(kJ/mm) (%) (%) (HV) All regions (HV)
U395 OK Tubrod 15.26S 2.0 0.227 0.333 251–277 265 26
U396 OK Tubrod 15.26S 3.0 0.226 0.334 241–262 251 21
U397 OK Tubrod 15.26S 4.0 0.216 0.313 241–262 249 21
U593 OK Tubrod 15.26S 2.0 0.238 0.345 257–277 264 20

U398 OK Tubrod 15.27S 2.0 0.264 0.396 289–336 315 47

U399 OK Tubrod 15.27S 3.0 0.258 0.388 286–367 313 81
U400 OK Tubrod 15.27S 4.0 0.271 0.402 295–345 315 50
U594 OK Tubrod 15.27S 4.0 0.265 0.387 280–362 348 82

U401 OK Tubrod 15.29S 2.0 0.271 0.399 303–367 345 64

U402 OK Tubrod 15.29S 3.0 0.273 0.402 299–362 339 63
U403 OK Tubrod 15.29S 4.0 0.283 0.419 306–376 354 70

* specimens No U395–U403 and U593–U594 : final SAW U-Groove cracking tests at ESAB,
U.K.

It should be highlighted that the hardness survey performed in the present thesis did
not comprise any microhardness measurements, consequently the measurements cannot
reveal the nature of very narrow phase interfaces. Hence, there is no direct evidence that
such steep gradient of mismatching strength would have existed also across the phase
interface. The results are therefore considered merely an indication of plausible
differences in local hardness – and hence strength – properties of different
microstructural regions of multipass weld metal, and that these local differences were
emphasised with the increase in WM strength.
That local microstructural variations within multipass weld metal can affect the
occurrence of hydrogen cracking, is demonstrated in a recent study[121] on SAW welding
of extra-high strength HSLA-100 steel. The study reports that most of the appeared
transverse WM cracks were densely populated in the weld region in which the
microhardness increased gradually from the minimum hardness of a tempered band to the
 162

maximum hardness in the recrystallised region. As the former region had essentially
columnar grain structure, whereas the latter composed of equiaxed grains, this means that
intensive cracking took place particularly in locations exhibiting both hardness gradient
and microstructural gradient. The study[121] concluded that the location of hydrogen
cracks formed in the WM was mainly controlled by the local variation in microstructure
but was little affected by the initial microhardness of a single microphase itself. This
means that a steep gradient in microstructures (and hence in hardness) is more
determinant from the standpoint of hydrogen cracking occurrence than an individual
microhardness of local microstructural phase. Consequently, cracking was not necessarily
related to that WM microphase exhibiting the highest microhardness[121]. These findings
are regarded consistent with the present thesis in that steep, local variations in weld
microstructures that, for instance, manifests themselves as marked difference in the
HVmin and HVmax of the given local microstructure, can accentuate the sensitivity of the
weld metal to hydrogen cracking.
Apart from microhardness, it should be stressed that the results of the present thesis, in
accordance with a number of works made elsewhere[2, 3, 12, 23, 26, 52, 53, 76, 107, 117] clearly
demonstrate that WM hydrogen cracking sensitivity, and the need of elevated T0/Ti,
increase abruptly with the increase in WM tensile strength RM, accompanied by an
increase in macrohardness, HV5(ave) or HV5(max), as shown in Figs 36 and 39.
It has been shown[38] that for bimetallic welds, accentuated local strength mismatch
incorporating softened local microstructural regions adjacent to coexisting
microstructures of higher strength, can, under external loading, create conditions that
favour elevation of linear-elastic stress in the high strength microstructures, whilst
accumulation of plastic strain occurs in the softened regions. Should such a steep strength
gradient exist across the phase interface in the present case, while the simultaneous
transport and accumulation of hydrogen happens to result in microcrack formation at the
phase interface, conditions for unstable crack growth may, under high residual tensile
stress, be met at the interface of the two regions, of which one exhibits plastic behaviour
while the other still behaves practically linear-elastically.
Anyway, metallography made here is yet too deficient to draw any firm conclusions
on the mechanism principally described here. Moreover, in practice, overall weld metal
overmatching strength is likely to ‘shield’ the weld against such kind of unstable crack
propagation, anyway.

5.9 Comparison of hydrogen diffusivity between SMAW, SAW and

FCAW multipass weld metals

In general, the measured DH values were found to decrease in the order of Armco iron,
fine-grained structural steel and the weld metal, as one would expected, see Table 20.
This is attributed to the greater number of hydrogen traps available in the weld metal than
in the low-impurity fine-grained structural steel, or in the annealing heat-treated Armco
iron that represents DH values nearly characteristic of pure α-iron.
 163

Regarding weld metal it was found that the DH values do not differ significantly
between SMAW and FCAW welds. The DH values recorded for the SAW filling layers
also lie relatively close to those for the FCAW and SMAW weld metals, whereas those
for the SAW mid-thickness and root runs are seen to approach those for the fine-grained
structural steel, as shown in Table 20.
The measurements of the present thesis revealed a systematic tendency in all the
examined SAW specimens for their DH values to increase in weld thickness direction
when transferring from the weld upper-area to the weld root area. This phenomenon,
however, was not recognised in the case of the FCAW weld metals, compare Figs 28a
and 28b. It seems obvious that the weld chemical composition in the SAW filling layers
and the root runs is not identical, because of the dilution that is always more pronounced
in the weld root and is, overall, a characteristic feature of the SAW process. However,
considering only the effect of chemical composition, it is still difficult to explain the
observed differences in DH between the filling runs and the weld central part, as the
chemical composition in these weld metal regions should be more or less similar.
An explanation[5b, 36b] for the observed differences in the DH values associated with
different SAW weld depths could lie in the nature of weld thermal cycle. The high heat
input Q of 5 kJ/mm in SAW welds results in pronounced re-heating of the previously
solidified weld layers by the subsequent welding runs. This, in turn, causes significant
microstructural changes over a wide distance in the weldment thickness direction. The
higher the Q and the greater the number of successive thermal cycles, the lesser also
becomes the number of hydrogen traps available in the weld metal. It could also be
postulated that the smaller bead size of the FCAW welds allows the inclusion of all types
of microstructures in all the diffusion samples, whereas the SAW bead size means that
the diffusion specimens from near the surface only contain as-welded microstructures as
opposed to more reheated microstructures towards the weldment root.
These phenomena are likely to explain the higher DH values in the root area of the
SAW welds, compared to the filling layers, as well as why the examined FCAW weld
metals, unlike the SAW welds, display no corresponding change in DH as a function of
weldment thickness, see Fig. 28b. The variation of DH in weld thickness direction is
found not systematic, but relatively random between the examined FCAW specimens.
Overall, the scatter in measured DH values appears greater among the values
summarised from various investigations even for the same welded region, than it is
among the values associated with entirely different weld regions, i.e., weld metal, HAZ,
or parent steel, as shown in Table 18.

5.10 Comparison of cracking sensitivity between SMAW and SAW

multipass weld metals

In the first instance and according to nominal yield strength Rp0.2, the SAW OK Tubrod
15.27S weld metals seem to exhibit cracks at lower weld HD contents than the SMAW
OK 75.75 weld metals, compare Tables 4 and 7. This would imply, at similar levels of
 164

nominal strength and hydrogen, SAW process is likely to represent greater risk to weld
metal hydrogen cracking than the SMAW process.
A comparison of the SMAW and SAW U-Groove experiments in terms of their
cracking risk according to Tables 4 and 7 reveals that in the case of the SMAW welds of
Rp0.2 ≈ 680–700 MPa, hydrogen contents of HD ≈ 10 ml/100 g DM (IIW) were necessary
to induce cracks. For cracking to occur in the SAW welds of 700 MPa nominal yield
strength, the corresponding critical HD was only about 5 ml/100 g DM (IIW). These
findings are in line with the preliminary SAW U-Groove test results c.f. Table 6, thereby
suggesting that, at a comparable level of WM yield strength, SAW multipass weld metals
would be more susceptible to WM hydrogen cracking than the SMAW welds. As a result,
one would conclude that the tubular SAW consumables c.f. Table 2b produce weld
metals that, in relation to their strength, are more crack sensitive than the SMAW weld
metals c.f. Table 2a.
Sorting all the examined SMAW and SAW U-Groove cracking tests according to their
actual weld metal chemical composition and measured weld metal hardness, the
differences in WM cracking sensitivity between the SAW and SMAW experiments are
seen to attenuate, as shown in Figs 35 and 36. Consequently, in terms of weld metal Pcm
and HV5(max), all the SAW U-Groove cracking test results are found to agree nicely with
the Crack – No Crack boundaries in Figs 35 and 36, i.e., the boundaries that were
originally derived according to the SMAW Y- and U-Groove data c.f. Tables 3–5.
It follows that both in terms of Hcr–Pcm and Hcr–HV5(max) relationships, the present
SAW U-Groove cracking test results in Tables 6 and 7 comply satisfactorily well with the
corresponding SMAW Y- and U-Groove cracking test data, as shown in Table 29. The
SAW U-Groove test results obtained in the present thesis do not, therefore, assign any
need to modify the original Crack – No Crack boundaries derived from the SMAW
cracking test data and expressed in terms of the Hcr–Pcm and Hcr–HV5(max) relationships in
Figs 35 and 36.
With regard to the Hcr–σresL (= Rp0.2) relationship, the question of general validity is
less clear, as shown in Table 29 and Fig 34. Assuming the WM strength values given for
the examined SAW welds according to Table 29 by the consumable manufacturer
prevailed also in the U-Groove cracking tests, leads to the conclusion that the examined
SAW multipass weld metals would be more crack sensitive than awaited on the basis of
their strength. Nevertheless, Figs 35 and 36 show that the Pcm and HVmax based
approaches provide a realistic description of WM cracking sensitivity of both the SAW
and SMAW multipass weld metals examined in the present thesis.
It is interesting to notify that all the SAW U-Groove cracking test results fell
realistically within the Crack – No Crack regions derived according to the SMAW data,
regardless of the greater heat input range of Q = 2.0–5.0 kJ/mm associated with the SAW
process, in comparison with the 2.4 kJ/mm for the SMAW process. At a comparable level
of WM strength, one would have expected the higher Q in the case of SAW to yield weld
metals of lower hardness than for the SMAW welds. Should this occur, it would have
manifested itself as corresponding differences in the weld Hcr levels between the Hcr–Pcm
and Hcr–HV5(max) based diagrams, since the effects of Q are incorporated into the
HV5(max) based approach, but not accounted for in the Pcm based approach. Figs 35 and
 165

36, however, do not reveal any those kind of systematic differences in the cracking test
results, as shown also in Table 29.
It can hence be concluded that the effect of Q on WM hydrogen cracking occurrence
in multiple-pass welds either (i) does not exist, or simply (ii) was too small to become
notable in the present SAW cracking tests c.f. Tables 6–8. Since higher heat inputs have
been previously reported[8] to accentuate WM cracking risk in multipass welding, the
latter explanation seems more likely. Should, for instance, the local hydrogen
concentration HRmax become higher in SAW than SMAW as a result of pronounced weld
bead overlapping and somewhat greater bead width, could not be fully confirmed in the
experimental cracking tests of the present thesis. The analytic calculations in Sections
5.3.2 and 5.3.3 suggest, this might well have been the case.
Presumably, heat input affects cracking sensitivity of weld metal within only a narrow
region in the immediate vicinity of the Crack – No Crack boundary of a certain weld
metal strength – weld hydrogen -combination. Therefore, the effect of Q seldom becomes
notable and is hence difficult to reveal intentionally in the experimental cracking tests.
According to Figs 34–36, only few of the data points actually lie so very close to the
Crack – No Crack boundary line, whilst the vast majority of the data locates entirely
either deep within the Cracking, or the No Crack, region. The same explanation probably
applies also to interpass time ti. Otherwise, it would be difficult to consider theoretical
reasons why no effect of ti on hydrogen cracking was apparent in the SMAW cracking
experiments c.f. Table 4. A nearly fivefold reduction in ti from 15 to 4 sec should lead to
a corresponding reduction in thermal factor of diffusion ΣD∆t during the time between
two successive passes and throughout the welding operation. That the effect of ti would
be so effete it only becomes notable for those data points locating immediately adjacent
to the Crack – No Crack boundary line, would thereby explain why the changes in ti did
not affect the final outcome of the cracking tests c.f. Table 4.
Nevertheless, the conclusion on the basis of the experiments made here is that, relative
to the WM strength and weld hydrogen, heat input Q and interpass time ti are of a
secondary importance as crack controlling factors, as regards to hydrogen cracking in
multipass weld metals.

5.11 Differences in final hydrogen content between single-pass and

multiple-pass welds

The SAW U-Groove cracking tests for the extra-high strength OK Tubrod 15.27S and OK
Tubrod 15.29S multipass weld metals c.f. Table 9 demonstrated that exceptionally high
interpass temperature Ti was necessary to prevent WM hydrogen cracking, once the
other crack controlling factors: weld metal strength and hydrogen were such that
the Cracking region was indisputably entered. This contrasts with the present
knowledge[3b, 13, 22–25, 40, 59, 63, 69, 77, 101] on single-pass welds that, in this sense, do benefit
remarkably from the use of elevated preheat temperature T0. This has been attributed to
intensified diffusion and effusion of hydrogen that progressively lower the HR100/H0
 166

values as soon as T0 reaches 75–100°C, the consequence being that weld t100 cooling
time and hence ΣD∆t(100) start to increase sharply according to Eqs (1)–(3), respectively,
see Appendix 10.
Therefore, analysis was made to evaluate the differences in the final hydrogen content
HRmax between single-pass and multipass welds in terms of their HR100(d=0)/H0 and
HRmax(d>0) /H0 values, respectively.
With regard to the distribution of hydrogen concentration in multipass weld, it is
known[21, 22, 94] that hydrogen concentrates beneath the final weld bead layer where the
maximum values of the local hydrogen concentration were found[94] consistent with the
results of gas chromatographic measurements of hydrogen evolved from specimens
sectioned at various thicknesses of a weld. The calculations and experiments both
demonstrate[94, 95] that the HRmax in multipass weld is close to the weld hydrogen content
determined using a quenching method for a single-pass weld, that is, the weld initial
diffusible hydrogen, H0.
Assuming the difference in the HRmax/H0 ration between single- and multiple-pass
welds was responsible for the insensitivity of the multipass SAW welds to elevations of
the Ti in the actual U-Groove experiments c.f. Table 9, analytic calculations were made to
derive a set of HR100(d=0)/H0 and HRmax(d>0) /H0 values for single- and multiple-pass welds,
respectively, over a range of weld thermal conditions. These calculations are presented in
Table 31 that defines the weld thermal cycle in terms of T0–t100 combinations
corresponding to 1.7 kJ/mm arc energy welding of a 50 mm thick plate in accordance
with Appendix 10.
In general, Table 31 shows that, at a given level of T0 and t100 which transfer further to
ΣD∆t(100) according to Eq. (3b), single-pass welds are rewarded for applying elevated T0
in that the HRmax/H0 ratio tends to always become lower in single-pass weld than
multipass weld. This trend is seen to accentuate towards higher thermal heat inputs, i.e.,
greater values of ΣD∆t(100), which means that, at an equivalent level of H0, elevating the
T0/Ti is much less beneficial in lowering the HRmax in the case of multiple-pass welds than
for single-pass welds. This finding is in agreement with the outcome from the SAW U-
Groove experiments c.f. Table 9 and may hence explain, why it was necessary to elevate
Ti so remarkably once the Cracking region was entered, despite a comparatively low weld
H0 content. This is demonstrated also in Fig. 39.
It can hence be concluded that for multipass welds, going for strength – hydrogen
combinations readily locating in the No Crack region is much more effective and
desirable for preventing WM hydrogen cracking than the enhancement of hydrogen
effusion by elevating the T0/Ti under conditions that inevitably attribute to Cracking
region at a given weld metal strength – hydrogen combination. In this respect, effective
ways and means to control hydrogen cracking behaviour of multipass welds do differ
from those suffusing single-pass welds.
Looking Table 31 in more detail, it can be seen that for single-pass welds, Eq. (1a) that
is a plain description of the thermal driving force for hydrogen diffusion and hence does
not include any bead size effects, predicts lower HR100(d=0)/H0 values than Eq. (49)
incorporating the bead size and thermal driving force, compare (2) and (3) in Table 31. This
is particularly so with high thermal heat inputs, where HR100(d=0)/H0 acc. to Eq. (1a)
become unrealistically low, compared to the corresponding values acc. to Eq. (49). It is
 167

also evident that large beads (i.e., hw = 8 mm) retain much greater amount of hydrogen
than small beads (hw = 3 mm) do, which further manifests itself as a wide range in the
calculated HR100(d=0)/H0 values in accordance with Eq. (49), see Table 31. Even then, the
HR100(d=0)/H0 values for single-pass welds and associated with hw = 3–8 mm tend to go
under the corresponding range of the HRmax(d>0)/H0 values for multipass welds acc. to Eqs
(49)–(50), compare (3) and (5) in Table 31.
Table 31. Comparison of the HRmax/H0 ratio(*) between single- and multiple-pass welds
over a range of weld thermal conditions.
Preheat Weld cooling Thermal Remaining diffusible Final local maximum hydrogen
temperature time t100 factor of hydrogen / initial hydrogen / initial hydrogen -ratio in
T0 (sec) diffusion (1) -ratio in single-pass weld multipass weld
(°C) ΣD∆t(100) HR100(d=0) /H0 HRmax(d>0) /H0
(cm2) thermal(2) geom. incl.(3) graphic sol.(4) analytical sol.(5)
25 45 0.001793 0.88 0.96–0.99 0.85–0.90 1.04–1.10
50 70 0.002318 0.85 – 0.85–0.90 –
75 180 0.004447 0.73 – 0.80–0.90 –
100 800 0.016505 0.32 0.60–0.94 0.70–0.85 0.93–1.08
125 1600 0.034148 0.10 – 0.65–0.85 –
150 2200 0.048899 0.03 0.28–0.83 0.60–0.80 0.53–1.04
175 2500 0.056732 0.02 – 0.55–0.70 –
200 3000 0.070430 0.01 0.27–0.76 0.50–0.70 0.57–1.06

*)
for single- and multiple-pass welds, HRmax refers to HR100(d=0) and HRmax(d>0), respectively
1)
calculated acc. to Eq. (3b)
2)
calculated acc. to Eq. (1a) excluding the bead size effects, applying ΣD∆t(100) in accordance
with Eq. (3b) and assuming A = 70
3)
calculated acc. to Eq. (49) accounting for bead size effects (hw = 3 and 8 mm correspond to the
lower and higher value of the given range), or determined from Fig. 32
4)
graphic solution acc. to Osaka University & Nippon Steel Corp. c.f. Fig. 6 originally presented
in Ref.[20]
analytical solution c.f. Table 27 and calculated acc. to Eqs. (49)–(50) using ΣD∆t/hw values
5) 2
[32,33]
acc. to Eq. (49) and estimated values for the weld overlap ratio d/hw c.f. Refs (applied
range: bead sizes 3–8 mm)

For multipass welds, the graphic and analytic solutions of the HRmax(d>0)/H0 are seen to
agree comparatively well in that both expressions show accordingly, the HRmax(d>0) /H0
decreases with increasing thermal heat inputs in terms of T0 and t100 and, hence,
ΣD∆t(100), compare (4) and (5) in Table 31. It is noteworthy that the range of variation for
these two expressions originates from different sources. In the case of the analytic
expression (5), the range of variation is caused by the bead size related effects, the
minimum and maximum values corresponding to hw of 3 and 8 mm, respectively,
whereas for the graphic expression (4) the range is attributed[20] merely to the differences
in the number of weld bead layers, l, and heat input Q, as shown in Fig. 6. Anyway, both
these expressions show consistently that the HRmax(d>0)/H0 associated with multipass
welding declines to a lesser degree than the corresponding HR100(d=0) /H0 for single-pass
welds, compare (4) and (5) with (3) in Table 31. This means that hydrogen effusion as a
result of elevated T0 occurs more efficaciously in single-pass welding, whereas multiple-
 168

pass welds, particularly those with greater bead size, tend to retain hydrogen even when
thermal heat inputs increase due to elevated T0/Ti, as shown in Table 31.
This finding is in line with the outcome from the SAW U-Groove experiments c.f.
Table 9 and Fig. 39. It suggests that, besides increasing WM strength, the described
difference in the responses between the HRmax(d>0)/H0 and HR100(d=0)/H0 to the elevation of
T0/Ti might explain, why the need to elevate the Ti temperature for the extra-high strength
OK Tubrod 15.27S and 15.29S weld metals is so abrupt in multipass welding, once the
WM strength and hydrogen are such that the weldment in question inevitably falls within
the Cracking region.

Fig 38. Final maximum hydrogen concentration – initial hydrogen content -ratio Hmax/H0 in
multipass weld as a function of post-weld heat treatment in terms of heating time tp and post-
heating temperature θp[20].

Apart from preheating, or sustaining the elevated interpass temperature during

welding, applying retarded post-weld cooling, or post-weld heat treatment, is known to
effectively reduce hydrogen cracking risk by promoting hydrogen effusion. Examples[20]
of such post-weld treatments for multiple-pass welds are shown in Figs 38a–38c.
According to Fig. 38, post-weld heat treatment at temperatures about 200°C appears,
in the case of multipass welds, somewhat more effective in lowering the Hmax/H0 than the
use of elevated preheat/interpass temperatures[20], compare Fig. 38 and Table 31.
 169

Applying post-weld heat treatments in accordance with Fig 38, the Hmax/H0 is seen to
become depressed to around 0.30–0.50[20]. Table 31, in turn, shows that in order to gain
Hmax/H0 values equal to or less than 0.50 requires the T0/Ti levels higher than 200°C. It
should be highlighted, however, that comparatively long holding times are required in
these cases for the treatment to affect more effectively than preheating. In order to
achieve Hmax/H0 values less than 0.50, for example, around 3–5 hours at the temperature
of 200°C are necessary[20], as shown in Fig 38.

5.12 Derivation of the prediction formulae for the estimation of safe

preheat/interpass temperatures for multipass weld metals

A ’second line of defence’ against WM hydrogen cracking consists of engineering

formulae for predicting the required preheat/interpass temperatures T0/Ti for the
avoidance of cracking, once the other controlling factors: strength and hydrogen are such
that the Cracking region c.f. Figs 34–36 is inevitably entered.
This analysis comprises two steps: (i) re-analysis of the former SMAW and GMAW
Y-Groove data sets of Nippon Steel Corp. (NSC), and (ii) analysis of the SAW U-Groove
cracking test results of the present thesis for optimised prediction of ’safe’ T0/Ti
estimates.

5.12.1 Re-analysis of the NSC Y-Groove data sets

A data set consisting of former Y-Groove cracking tests[12] made at NSC is given in
Appendix 5. These data comprise SMAW and GMAW multipass welds of ultimate tensile
strengths around 820 and 750 MPa, weld build-up thicknesses of 20, 30 and 40 mm and
weld diffusible hydrogen in the range of about 2 to 11 ml/100 g DM (IIW). It was
concluded[12] that the critical preheat/interpass temperature Tcr for the avoidance of WM
hydrogen cracking can be expressed in terms of WM tensile strength RM, weld build-up
thickness aw and weld diffusible hydrogen content HD.
Both Okuda et al.[2] and Yurioka et al.[3] have demonstrated the decisive role of WM
strength and weld hydrogen more or less accordingly. For the present re-analysis, weld
RM is approximated using weld chemical composition in terms of CET as defined in Eq.
(43b). Unlike Yurioka, Okuda found[2] the influence of weld build-up thickness on WM
cracking susceptibility quite minimal and hence insignificant in the 20–80 mm thickness
range. The U- and V-Groove test results of the present thesis, however, demonstrate that
plate thickness does affect the WM cracking occurrence – at least when transferring from
40 mm to 6 mm weld build-up thickness, compare Tables 4 and 10. Inclusion of weld
build-up thickness aw as a parameter into the prediction formulae of Tcr is therefore
regarded imperative.
 170

For the purposes of ‘safe’ predictions of ‘engineering accuracy’, a general expression

of the form of Eq. (40a) was taken as a primitive approximation to describe the crack
controlling factors in multipass welding, as explained in Section 3.3.3. As first attempt to
estimate Tcr, the former NSC Y-Groove data set[12] was fitted by the following empirical
equation obeying a general form of Eq. (40a), as:
Tcr = 700 ∗ CET + 9 ∗ aw0.86 + 134 ∗ HD0.40 – 498 (°C) (60)
where CET is the carbon equivalent (%) according to SFS-EN 1011-2 : 2001 Method
B[1b] calculated from the all-weld metal chemical composition, aw is weld build-up
thickness (mm) and HD is weld diffusible hydrogen content (ml/100 g DM) according to
ISO/IIW 3690-1977[57].
Table 32 summarises the required T0/Ti values for some of the extra-high strength Y-
80C, L-80 and L-74 weld metals from the NSC Y-Groove cracking experiments[12]. These
experimentally determined T0/Ti values are then compared with calculated Tcr estimates
according to the Okuda-formula[2] and Eq. (60) derived in the present thesis. Table 32
reveals that compared to the experimental Y-Groove data[12], the Okuda-formula[2] puts
more weight on weld hydrogen, which is mainly due to the logarithmic dependence of Tcr
on HD, see Appendix 1. This is seen to weaken the accuracy of Tcr predictions towards the
ends of the applied HD range, consequently, Tcr tends to become somewhat
underestimated at very low HD levels, whereas the highest HD levels result in
overestimation of the required Tcr. Furthermore, the Okuda formula[2] disregards the
effect of weld build-up thickness aw, thereby implicitly assuming that the same Tcr
estimate originally[2] derived for the 40 mm weld thickness would apply over the 20–40
mm range. This is undoubtedly a safe presumption; however, Table 32 shows it yields
even unnecessarily conservative estimates of Tcr in the case of 20 and 30 mm weld
thickness.
 171

Table 32. Predicted T0/Ti estimates for extra-high strength Y-80C, L-80 and L-74 weld
metals from the NSC experiments[12] – comparison of T0/Ti estimates according to
Okuda-formula[2] and VTT formulae Eqs (60) and (61) with the required T0/Ti based on
the experimental NSC Y-Groove cracking test data[12].
Filler Weld Weld Weld Weld Preheat / Interpass temperature
material metal metal build-up hydrogen T0/Ti (°C)
strength alloying thickness HD NSC Okuda- VTT VTT
RM 1) CET 2) aw (ml/100 g Y-Groove formula[2] Eq. (60) Eq. (61)
(MPa) (%) (mm) DM IIW) experi-
ments[12]
Y-80C 852 0.321 40 2.0 75 65 118 74
L-80-1 825 0.312 20 4.3 75 155 79 82
30 125 128 131
40 175 175 179
L-80-2 814 0.300 20 6.7 100-125 209 117 121
30 175 166 170
40 > 200 214 217
L-80-3 807 0.293 20 10.4 150 266 167 152
30 200 217 202
40 > 200 264 249
L-74-1 744 0.278 30 4.0 75 95 76–98 78–99
(0.248)
L-74-2 745 0.248 20 6.7 > 50 167 81 85
30 ≤125 130 134
40 175 177 181
L-74-3 733 0.280 20 9.9 > 75 213 129–151 117–140
(0.248) 30 ≤150 178–200 167–189
40 200 226–248 214–236
1)
measured NSC tensile test data acc. to Ref.[12]
2)
calculated acc. to Eq. (43b) applying analyses of WM chemical compositions given in Ref.[12]

According to Table 32, Eq. (60) derived in the present study yields surprisingly
accurate estimates of Tcr, in view of the fact that a single expression is applied to describe
the whole dataset, instead of three sub-equations as the Yurioka-formula[3] does, see
Appendix 1. An obvious shortcoming of Eq. (60) is, however, that it does not allow for
adequate relaxation of Tcr at very low weld HD levels, thereby resulting in an
unnecessarily high Tcr estimate in the case of the Y-80C weld with only 2.0 ml/100 g DM
(IIW) hydrogen, for example.
In order for adjusting Eq. (60) to reward lowering hydrogen, an alternative expression
is derived so that the logarithmic effect of weld hydrogen HD on Tcr would become more
accentuated, particularly at the lower tail of the HD range. Eq. (40b) in Section 3.3.3 gives
a general form of this revised formula. Applying a general formulation in accordance
with Eq. (40b), the NSC Y-Groove datasets[12] in Table 32 were fitted by a revised
formula of the form:
Tcr = 700 ∗ CET + 9 ∗ aw0.86 + {[2800 ∗ ln(HD)] / (12 ∗ HD0.21)} – 505 (°C) (61)
[12]
Table 32 shows that within the examined NSC datasets , Eq. (61) yields a desired
improvement of accuracy of the Tcr predictions associated with both low and high weld
hydrogen contents of HD ≈ 2.0 and 9.9–10.4 ml/100 g DM (IIW), respectively.
 172

It is noteworthy that on account of the available NSC data[12], the following validity
limits for Eqs (60) and (61) should be considered:
20 ≤ Tcr ≤ 200 (°C)
0.24 ≤ CET ≤ 0.33 (%)
20 ≤ aw ≤ 40 (mm)
2.0 ≤ HD ≤ 10.4 (ml/100 g DM (IIW))
Consequently, in view of the analysed NSC Y-Groove datasets[12], Eqs (60) and (61)
provide a realistic description of the required Tcr for extra-high strength multipass welds.
In the case of either very low, or comparatively high, weld HD, Eq. (61) is concluded to
provide ‘realistic’ and ‘safe’ estimates of T0/Ti, consequently Eq. (61) should be preferred
to Eq. (60) for these cases.

5.12.2 Comparison of the NSC Y-Groove and VTT U-Groove data sets
in view of required preheat/interpass temperatures

In the light of the results of the present thesis, Eqs (60) and (61) that both rest entirely on
the NSC Y-Groove data[12], seem to overestimate the ‘true’ T0/Ti that was actually
required to prevent cracking in the SAW U-Groove cracking tests, compare Table 7 and
Table 32. This finding is endorsed by the NSC Y-Groove data[12] on lower strength L-60
weld metals, see Appendix 5.
Contemplating the outcome of all the cracking tests in Tables 3–8 and Table 32, it
looks evident that WM cracking susceptibility according to the present VTT cracking
experiments is lower than anticipated from the NSC Y-Groove data[12], or according to
the formulae of Okuda et al.[2] and Yurioka et al.[3]. This applies particularly to welds of
moderate strength in the RM ≈ 600–700 MPa range.

Table 33. Examples of differences in T0/Ti estimates according to the VTT and NSC
cracking test results for SMAW multipass weld metals.
Electrode Weld metal Weld Weld Preheat / Interpass temperature
strength thickness hydrogen T0/Ti
RM aw HD (°C)
(MPa) (mm) (ml/100 g NSC VTT Y- and
DM IIW) Y-Groove U-Groove
experiments[12] experiments
OK 75.78 965 40 3.3 290 ≈ 100–110
965 40 4.3 310 > 110
Nittetsu L-80 845 40 4.0 175 < 90–100
845 40 10.9 > 200 ≈ 100
OK 74.78 650 50 10.0 150 < 90–100
650 50 18.9 180 > 110
 173

Examples of differences in the cracking test results between the present VTT
experiments and the former ones at NSC[12] are given in Table 33. It shows, whenever
preheating was necessary to prevent cracking, the Y- and U-Groove experiments of the
present thesis indicate lower required Tcr levels than the former NSC Y-Groove tests[3b,12].
The present Y- and U-Groove cracking tests also demonstrate preheating not to become
necessary until at WM strength and weld hydrogen levels notably higher than in the case
of the NSC Y-Groove tests[12], compare Table 4 with Table 32 and Appendix 5. It is worth
highlighting that with respect to Nittetsu L-80 welds, the Y-Groove specimen geometry
and the weld metal strength, thickness and diffusible hydrogen content all were
equivalent between the NSC[12] and the VTT cracking experiments, compare Table 3 and
Table 32.
The reason for this discrepancy between the VTT and NSC cracking test results has
yet remained unexplained. Since the time from the completion of welding to NDT
inspection was not reported for the former NSC experiments[12], differences in the
outcome of the cracking tests at NSC and VTT could, in principle, arise from different
time periods applied in the experiments made at these two institutes. This would also
explain why the attempts made in the present thesis to reproduce the former NSC Y-
Groove test results[12] were unsuccessful, as shown in Tables 3 and 33. The lack of
recorded time data for the former NSC experiments, however, prevents one to draw any
sound conclusions on possible effects of time from welding to NDT in these cases. The
results from a recent study[117] on hydrogen cracking in multipass SAW weld metals
allow one to draw a conclusion that the time factor becomes increasingly important
towards lower strength weld metals, whilst the extra-high strength welds resembling the
L-80 exhibited cracks already within a much shorter period of time. Lowering the
contents of alloying elements in the weld metal was reported[117] to result in longer
fracture times. This is in line with the recognised responses of the different weld metals
investigated in the present thesis to the time period before NDT, as the influence of time
from welding to NDT was recognised particularly in the case of the moderate strength
OK74.78 weld metal of Rp0.2 ≈ 550 MPa, see Table 4 and Figs 34–36.
Another possible explanation may lie on the fact that all the SMAW weld metals
investigated in the present work, except perhaps Atom Arc 12018, had very low impurity
levels, i.e., S < 0.006% and P < 0.013%. Of these, Nittetsu L-80 with 0.001% S and
0.005% P produced the cleanest weld metal, as shown in Appendix 7. Thus, weld metals
may simply have become cleaner over the decades; the trend similar to what had been
recognised[7, 34] in the case of modern steel that nowadays meet extremely low impurity
levels, in comparison to conventional steel.
Once in a while, certain doubts have been expressed[27–30] on that lowering steel's
impurity level inevitably reduces the number of non-metallic inclusions acting as sites for
ferrite nucleation. This, in turn, would manifest itself as increased hardenability that
actually results in elevated hydrogen cracking risk[27–30]. Earlier work[7, 34, 52, 53] at VTT,
however, demonstrated outstandingly high HAZ hydrogen cracking resistance of some
low-carbon modern TMCP steels, in spite of their low impurity contents. It was
concluded[34, 52, 53] that lowering carbon contents of modern steel counterbalance the
adverse effects of increased hardenability due to cleanliness. As a result, the overall
 174

hydrogen cracking risk in modern low-carbon steel is actually reduced, compared to

conventional medium carbon steel.
It is known[40, 75] that not only inclusions, but dislocations, vacancies and prior-
austenite grain boundaries can act as traps to hydrogen in ferritic steel. That HAZ
cracking in clean steels tends, even at high hydrogen contents, to occur predominantly as
transgranular rather than intergranular cracking[39] implies that cleanliness of steel can
impede grain boundary decohesion that is usually promoted[75] by increasing hydrogen
contents. Consequently, at comparable strengths, modern steel with low carbon and
impurities tolerates higher hydrogen contents than conventional steel without the
occurrence of hydrogen cracking[7, 34, 39, 52, 53].
Analyses of the SMAW weld metals of the present thesis, for example, showed[5b, 36]
the presence of substantial amounts of inclusions acting as traps to diffusible hydrogen,
which further manifested itself as remarkably lower hydrogen diffusion coefficients
measured for the weld metals than in the case of the HAZ or parent steel[5b], see Table 20.
Low weld impurity content can therefore be expected to reduce WM hydrogen cracking
sensitivity, analogously to that in clean steels, without concomitant, undesirable increase
in hardenability or disappearance of hydrogen traps.

5.12.3 Analysis of the first SAW U-Groove dataset for optimised

prediction of preheat/interpass temperature estimates for
multipass welds

The U-Groove cracking test results of the first set of SAW experiments of the present
thesis demonstrated that the cracking sensitivity of weld metal is, above all, governed by
its strength, as shown in Tables 7–8. Whilst all the 7 specimens welded with OK Tubrod
15.27S and OK Tubrod 15.29S displayed characteristic WM transverse hydrogen cracks,
none of the 4 specimens welded with OK Tubrod 15.26S revealed any cracks. Whether
WM cracking did, or did not, occur, both these phenomena took place totally irrespective
of the other parameters: weld hydrogen, heat input and preheat/interpass temperature.
This is considered rather obscure and it would certainly be doubtful to interpret or
generalise these results in such a way that the other parameters would be totally
inconsequential in relation to strength.
The apparent insensitivity of parameters, other than strength, to the occurrence of WM
cracking is likely to be a consequence of a deficient hydrogen range in the first SAW test
series. Table 8 shows that the measured weld HD contents not only had fallen from the
target values, but also that the range of the measured values had narrowed substantially
from the intended target levels. As a result, the inherent influence of factors such as
hydrogen, heat input or preheat/interpass temperature on weld metal cracking was
probably impeded, whilst the effect of the most decisive factor from the standpoint of
hydrogen cracking, i.e., strength, remained and became accentuated, as shown in Table 7.
 175

5.12.3.1 Optimisation of theT0/Ti prediction with respect to weld

metal strength

As Table 34 shows, applying Eqs (60) and (61) to predict the required Tcr for the present
SAW U-Groove dataset tends to result in overestimation of the T0/Ti actually needed for
the avoidance of cracking in the case of OK Tubrod 15.26S weld metals. Both these
equations, however, were demonstrated to describe satisfactorily the former NSC Y-
Groove dataset[12], as shown in Table 32. This implies, again, that WM strength asserts an
even greater influence on the cracking occurrence than anticipated previously.
Consequently, Eqs (60) and (61) derived from the NSC data[12] on extra-high strength
welds in the RM ≈ 750–850 MPa range result in overestimation of the required T0/Ti in
the case of lower strength weld metals in the RM ≈ 600–700 MPa range.
Thus, in order to obtain ’realistic’, but still safe, estimates of Tcr for the SAW U-
Groove dataset created in the present thesis, Eq. (61) was re-adjusted with the aim of
pursuing optimisation of Tcr predictions. Applying this re-adjusted formula of the form in
accordance with Eq. (61), the experimental data from the SAW U-Groove multipass
cracking tests in Table 7 were fitted by a formula that puts more weight on the effect of
WM strength via its CET, as:
Tcr = 2310 ∗ CET + 9 ∗ aw0.86 + {[2800 ∗ ln(HD)] / (12 ∗ HD0.21)} – 1145 (°C) (62)
A comparison of T0/Ti values applied in the SAW U-Groove experiments is presented
in Table 34, with the numerical predictions of T0/Ti using the Okuda formula[2] c.f.
Appendix 1, in conjunction with the VTT formulae Eqs (60)–(62). According to Table 34,
the re-adjusted Eq. (62) yields more realistic T0/Ti estimates than the former Eqs (60)–
(61), as far as the SAW U-Groove cracking test results of the present thesis are
concerned. This is particularly the case for weld metals of moderate strength in the RM ≈
600–700 MPa range, such as OK Tubrod 15.26S.
 176

Table 34. Comparison of the applied T0/Ti temperatures in SAW U-Groove cracking tests
with the predictions for T0/Ti using Eqs (60), (61) and (62) for the estimation.
Specimen SAW Arc CET Weld Crack / Applied Prediction for the required
No tubular energy (%) hydrogen No interpass T0/Ti
cored (kJ/ HD crack tempera- (°C)
wire mm) (ml/100 g ture Okuda VTT Eq. VTT Eq. VTT Eq.
OK DM T0/Ti formula[2] (60) (61) (62)
(1)
Tubrod (IIW)) (°C)
U395 15.26S 2.0 0.333 4.9 NC 100 130–167 203 209 105
U396 15.26S 3.0 0.334 6.1 NC 125 135–172 227 232 130
U397 15.26S 4.0 0.332 8.4 NC 150 177–213 263 260 154

U398 15.27S 2.0 0.396 6.1 C 150 248–290 270 276 273
U399 15.27S 3.0 0.388 8.4 C 100 288–330 303 299 284
U400 15.27S 4.0 0.402 4.9 C 125 217–260 251 257 264

U401 15.29S 2.0 0.399 8.4 C 125 344–390 310 307 309
U402 15.29S 3.0 0.402 4.9 C 150 260–304 251 257 264
U403 15.29S 4.0 0.419 6.1 C 100 316–362 287 292 326

(1)
T0/Ti range calculated acc. to Appendix 1 from the weld metal HV data converted into weld RM
estimates using Eq. (48b) and Ref.[119]

Since the No Crack region for the extra-high strength OK Tubrod 15.27S and OK
Tubrod 15.29S weld metals was not reached in the first SAW U-Groove test series c.f.
Table 7, the absolute safety of Eq. (62) for weld metals in the RM ≈ 820–1000 MPa range
cannot be guaranteed. Anyway, Table 34 shows that for these extra-high strength welds,
the T0/Ti estimates according to Eq. (62) are very similar to those given by Eqs (60) and
(61) that were based directly on the former NSC Y-Groove data[12] and can hence be
expected to lie on the safe side for modern high-strength weld metals.

5.12.3.2 Role of heat input and interpass time in optimisation of theT0/Ti

prediction

According to Tables 7 and 34, heat input does not assert any notable effect on WM
cracking susceptibility. What still requires further experiments is the assessment of the
effect of plate/weld build-up thickness, aw; the present data is yet too sparse to allow for
accurate quantification of the thickness factor. Nevertheless, the Tcr–aw dependence
employed in Eq. (62) conforms with the dependence reported[3,12] for the former NSC Y-
Groove experiments[12], thus being compatible with the Yurioka formula[3] c.f.
Appendix 1.
As far as interpass time ti is concerned, the present experimental results imply that it
does not affect the WM cracking boundary conditions to any significant extent either, as
shown in Table 4. According to the SMAW U-Groove test results in Table 4, provided
that WM strength and hydrogen are sufficiently high to cause cracking, it takes place
 177

irrespective of ti in the range of 3 to 15 min. Similarly, in the cases where strength and
hydrogen are too low to cause cracking, crack-free welds are obtained over a wide ti
range from 3 to 15 min. These findings suggest, it is not imperative to include ti as a
separate factor into the predictive formulae assessing the required T0/Ti for the avoidance
of WM cracking in multipass welds.
From Figs 34–36, it is recognised that only few of the recorded cracking test incidents
actually lie very close to the Crack – No Crack boundary line, whilst the vast majority of
the data locates entirely either deep within the Cracking, or the No Crack, region.
Presumably, heat input affects cracking sensitivity of weld metal within only a narrow
region in the immediate vicinity of the Crack – No Crack boundary of a certain WM
strength – hydrogen -combination. Therefore, the effect of Q seldom becomes notable
and is hence difficult to reveal intentionally in the cracking tests. The same explanation
probably applies to interpass time ti. Otherwise, it would be difficult to consider
theoretical reasons why no effect of ti on hydrogen cracking was apparent in the SMAW
cracking experiments c.f. Table 4. A nearly fivefold reduction in ti from 15 to 4 sec
should lead to a corresponding reduction in thermal factor of diffusion ΣD∆t during the
time between two successive passes and throughout the welding operation. That the
effect of ti would be so effete it only becomes notable for those data points locating
immediately adjacent to the Crack – No Crack boundary line, would thereby explain why
the changes in ti did not affect the final outcome of the cracking tests c.f. Table 4.

5.12.3.3 Optimisation of theT0/Ti prediction with respect to

plate/weld build-up thickness

The reviewed studies[2, 3, 12, 76] consistently report that plate/weld build-up thickness aw
affects WM cracking most prominently in the 20–50 mm range. Above 50 mm, the effect
of aw gradually starts to level off as thickness increases further, whilst lowering aw will
progressively reduce the cracking risk. This has been attributed[2, 3, 12, 76] mainly to a
nearly constant level of weld residual stress above 50 mm thickness, whereas local
hydrogen can accumulate still further with the increase in thickness[2]. The results of the
present thesis demonstrate that the cracking risk becomes insignificant at weld build-up
thickness less than 10 mm, regardless of high weld metal strength and hydrogen content,
see Table 10.
Considering this, it can be realised that the power-law function applied to describe the
effect of aw on Tcr in Eqs (60)–(62) is not an optimum expression for the thickness effect.
According to Table 35, a term of the form: Tcr (aw) = {9 ∗ aw0.86} in Eqs (60)–(62) yields
comparatively realistic estimates in the 20–40 mm aw range when compared with the
actual weld thickness effect pertaining to the NSC Y-Groove experiments[12]. With
greater thicknessses, however, this is not true and the power-law expression c.f. Eq. (60)
is seen to exaggerate the effect of aw remarkably, without any abate of the effect, even
when coming up to aw values as great as 100 mm.
 178

Table 35. Assessment of the plate/weld build-up thickness effect – comparison of the
effect of aw on ∆Tcr prediction between the outcome of the experimental NSC data[12] and
the VTT based expressions according to Eqs (60) and (63).
Plate/weld build-up ∆Tcr *) acc. to NSC Thickness Factor acc. Thickness factor acc. to
thickness Y-Groove to Eq. (60) – VTT SAW U-Groove
aw (mm) experiments[12] NSC data fit based data fit c.f. Eq. (63)
(°C) (in°C) (in°C)
6 42 36
10 65 60
15 93 90
20 118 118
∆ ≈ 50 (∆ 50) (∆ 51)
30 168 169
∆ ≈ 50 ∆ 47 (∆ 45)
40 215 214
50 260 250
60 304 280
70 348 302
80 390 319
90 431 332
100 472 342

*)
∆ : shift in the Tcr when moving from smaller weld aw to a greater one (i.e., 20 → 30 and 30 →
40 mm)

Therefore, an alternative expression was derived in the present thesis by replacing the
power-law form c.f. Eqs (60)–(62) with a hyperbolic tanh function of the form:
Tcr (aw) = {367 ∗ tanh (aw / 60)} (63)
As the thickness effect was not systematically assessed in the present thesis, the
former NSC Y-Groove data[12] was applied to derive Eq. (63). This simultaneously
ensured compatible treatment of the thickness effect in Eq. (63) with the outcome of the
NSC data[12] in the aw range of 20–40 mm, i.e., the range where experimental NSC results
were available at the first place, see Table 35.
It can be seen that, in comparison to the expression c.f. Eq. (60), the optimised
expression of the effect of aw obeying Eq. (63) results in (i) the desired gradual abate of
the aw effect at thicknesses greater than about 70–80 mm, (ii) greater reward of lowering
aw at thicknesses less than 15 mm, and (iii) excellent compatibility with the outcome of
the actual NSC Y-Groove experiments[12] in the 20–40 mm aw range.
Furthermore, Eq. (63) is also consistent with the treatment of the plate thickness effect
in the case of predicting Tcr for the avoidance of hydrogen cracking according to EN
1011-2 : 2001[1b].
 179

5.12.3.4 Optimisation of theT0/Ti prediction with respect to weld

hydrogen content

It is widely accepted[1–3, 12, 13, 20–25, 28–30, 63, 69, 77, 101] that the effect of weld hydrogen HD on
Tcr is not linear, but rather of logarithmic nature. Still, different formulations are applied
to describe this logarithmic dependence. For instance, Okuda et al.[2] use direct [a ∗ log
(HD)] expression, whereas Yurioka et al.[3] apply [a ∗ log (HD/b)] type formulation
resulting in somewhat more abating increase of Tcr with HD. Standard EN 1011-2 :
2001[1b], in turn, uses a power-law expression of the form: [a ∗ (HD)b, where b < 1] to
describe the ceasing increase of Tcr with HD.
Therefore, the outcome of different types of formulations for the logarithmic hydrogen
effect on ∆Tcr were compared against the measured values of the ∆Tcr recorded from the
former NSC Y-Groove cracking tests[12], see Table 36. Weld HD levels taken to this
comparison are compatible with the NSC Y-Groove experiments[12], as far as the 2.0–10.4
ml/100 g DM (IIW) range is concerned. The intention for seeking various expressions for
the effect of HD on ∆Tcr was to find a description that (i) would match with the recorded
effect in the NSC Y-Groove experiments[12] in the 2.0–6.7 ml/100 g DM (IIW) range
where data was available, without (ii) simultaneously exaggerating the effect at either end
of the examined HD range. The alternative expressions for the HD effect applied in the
comparison are in accordance with Eqs (60b), (61b), (64), (65) and (66), as:
Tcr (HD) = {134 ∗ (HD)0.40} (60b)
Tcr (HD) = {[2800 ∗ ln(HD)] / (12 ∗ HD0.21)} (61b)
Tcr (HD) = {351 ∗ log(HD)} (64)
Tcr (HD) = {[2310 ∗ ln(HD)] / (12 ∗ HD0.15)} (65)
Tcr (HD) = {323 ∗ tanh (HD / 5.0)} (66)
It can be seen that Eqs (60b) and (61b) that were the first attempted fittings based on
the NSC Y-Groove test data[12], both describe quite satisfactorily the effect of HD on ∆Tcr
in the 4.3–10.4 ml/ 100 g DM (IIW) range. However, owing to its power-law form, Eq.
(60) does not allow for an adequate relaxation of Tcr at very low weld HD levels of about
2 ml/100 g DM (IIW), whereas Eq. (61) applying a combined ln-power-law form is seen
to accentuate the effect of lowering HD on ∆Tcr somewhat too much to be safe, or
realistic, any longer. A simple log-function c.f. Eq. (64), in turn, seems to overstate the
effect of HD on ∆Tcr at low and moderate weld hydrogen levels around 2–8 ml/100 g DM
(IIW), although leading to compatible description with the power-law form c.f. Eq. (60)
at high weld HD contents of around 10–16 ml/100 g DM (IIW).
 180

Table 36. Assessment of the weld hydrogen effect – comparison of the effect of HD on
∆Tcr prediction between the outcome of the experimental NSC data and the VTT based
expressions according to Eqs (61), (64), (65) and (66).
Weld Recorded ∆Tcr Hydrogen Factor acc. to various Equations – influence on ∆Tcr (in°C)
hydrogen *) acc. to NSC NSC NSC VTT based VTT SAW VTT SAW
HD Y-Groove data-based data-based 'log'-fit U-Groove U-Groove
(ml/100 g) experiments[12] 'power-law' fit 'ln-power-law' fit c.f. Eq. (64) data-based, data-based,
(°C) c.f. Eq. (60) c.f. Eq. (61) 'ln-power-law'-fit 'tanh'-fit
c.f. Eq. (65) c.f. Eq. (66)
2.0 177 140 106 121 123
∆ ≈ 100 ∆ ≈ 63 ∆ ≈ 111 ∆ ≈ 116 ∆ ≈ 104 ∆ ≈ 102
4.3 240 251 222 225 225
∆ ≈ 50 ∆ ≈ 47 ∆ ≈ 47 ∆ ≈ 68 ∆ ≈ 50 ∆ ≈ 56
6.7 287 298 290 275 281
∆ ≈ 25–50 ∆ ≈ 55 ∆ ≈ 36 ∆ ≈ 67 ∆ ≈ 42 ∆ ≈ 32
10.4 342 334 357 317 313
– ∆ ≈ 43 ∆ ≈ 20 ∆ ≈ 45 ∆ ≈ 25 ∆≈8
14.0 385 354 402 342 321
– ∆ ≈ 21 ∆≈7 ∆ ≈ 21 ∆ ≈ 10 ∆≈1
16.0 406 361 423 352 322

*)
∆ : shift in the Tcr when moving from lower weld HD to a higher one (i.e., 2.0 → 4.3; 4.3 → 6.7;
6.7 → 10.4; 10.4 → 14.0 and 14.0 → 16.0 ml/100 g)

According to Table 36, both of the two new expressions derived in the present thesis:
the revised, combined ln-power-law form acc. to Eq. (65) and the hyperbolic tanh form
acc. to Eq. (66) result in (i) the desired gradual abate of the HD effect at weld hydrogen
contents greater than about 11–14 ml/100 g DM (IIW), (ii) realistic reward of lowering
HD in the 2–4 ml/100 g DM (IIW) hydrogen range, and (iii) excellent compatibility with
the outcome of the actual NSC Y-Groove experiments[12] in the 4–10 ml/100 g DM (IIW)
HD range.
Of the two expressions, Eq. (66) applying the tanh form is seen to yield an abatement
of the HD effect which, at weld hydrogen levels of around 14 ml/100 g DM (IIW),
already becomes so accentuated that it may no longer be realistic. This may lead to
underrating of the effect of HD on ∆Tcr at high weld hydrogen contents when using Eq.
(66). Thus, the use of Eq. (65) applying a combined, revised ln-power-law form is
presumably more safe in this sense, and is hence advised to be preferred to the other
functions examined in the present thesis.
 181

5.12.4 Final analysis of the SAW U-Groove data for optimised prediction
of preheat/interpass temperature estimates for multipass welds

Table 37 summarises the database comprising the SAW U-Groove cracking tests on thick
plate welds and made under different levels of preheat/interpass temperature T0/Ti. The
objective here was to elevate the applied Ti until crack-free welds were obtained for the
investigated SAW weld metals: OK Tubrod 15.26S, OK Tubrod 15.27S and OK Tubrod
15.29S.
Table 37. Comparison of the T0/Ti estimates derived according to different prediction
formulae against the required T0/Ti temperatures in the SAW U-Groove cracking tests –
thick plate data from experiments made at ESAB U.K. and Mäntyluoto Works.
Specimen SAW Arc CET Weld Crack Applied Prediction for the required
No tubular energy (%) hydrogen / T0/Ti T0/Ti (°C)
cored wire (kJ/mm) HD No (°C) VTT VTT Yurioka Okuda VTT
OK (ml/100 g crack Eq. Eq. formula[3] formula[2] Eq.
3) 3)
Tubrod DM (IIW)) (67a) (67b) (62)
U395 15.26S 2.0 0.333 4.9 NC 100 105 105 143–193 130–167 105
U396 15.26S 3.0 0.334 6.1 NC 125 130 131 128–177 135–172 130
U397 15.26S 4.0 0.332 8.4 NC 150 155 159 149–197 177–213 154

U400 15.27S 4.0 0.402 4.9 C 125 257 257 259–316 217–260 264
U398 15.27S 2.0 0.396 6.1 C 150 267 268 277–333 248–290 273
U399 15.27S 3.0 0.388 8.4 C 100 278 282 297–354 288–330 284
W11) 15.27S 2.0 0.39 6.5 C 175 260 261 280–336 255–297 265
W21) 15.27S 2.0 0.39 6.5 C 200 260 261 280–336 255–297 265
W31) 15.27S 2.0 0.39 6.5 C 225 260 261 280–336 255–297 265

U402 15.29S 3.0 0.402 4.9 C 150 257 257 315–374 260–304 264
U403 15.29S 4.0 0.419 6.1 C 100 317 318 368–429 316–362 326
U401 15.29S 2.0 0.399 8.4 C 125 302 307 371–432 344–390 309
M12) 15.29S 4.0 0.41 6.0 C 250 296 297 347–407 299–345 304
M32) 15.29S 4.0 0.41 6.0 C 275 296 297 347–407 299–345 304
M42) 15.29S 4.0 0.41 6.0 NC 285 296 297 347–407 299–345 304
M22) 15.29S 4.0 0.41 6.0 NC 300 296 297 347–407 299–345 304

1)
weld metal CET estimate calculated as a mean value of the measured values for the OK Tubrod
15.27S welds
2)
weld metal CET estimate calculated as a mean value of the measured values for the OK Tubrod
15.29S welds
3)
T0/Ti range calculated acc. to Appendix 1 from the weld metal HV data converted into weld RM
estimates using Eq. (48b) and Ref.[119]
 182

Table 37 introduces three VTT equations based on weld metal CET to describe the
effect of WM strength on the required T0/Ti. This stems from the outcome of the present
SMAW and SAW cracking tests, which demonstrated that cracking sensitivity of weld
metal is, besides hydrogen, above all governed by its strength, see Tables 3–8. According
to Table 34, WM strength is seen to have an even greater influence on the cracking
occurrence than anticipated previously. This is consistent with the recent findings[117] that
showed, the risk of WM hydrogen cracking increases abruptly when transferring from
intermediate strength weld metals associated with CETs of 0.33–0.36% to extra-high
strength weld metals exhibiting CETs of 0.40–0.42%. This manifested itself as
considerably higher required T0/Ti in the case of the extra-high strength weld metals,
compared with the intermediate strength ones, as shown in Fig. 39.
For the purposes of comparison, two new equations, Eq. (67a) and (67b) were thereby
introduced in addition to former Eq. (62), all of them with slightly different formulations
to account for the effects of weld metal chemical composition CET, weld build-up
thickness aw and weld diffusible hydrogen HD. The two new expressions are of the form:

Tcr = 2195 ∗ CET + 9 ∗ aw0.86 + {[2800 ∗ ln(HD)] / (12 ∗ HD0.21)} – 1106 (°C) (67a)

Tcr = 2195 ∗ CET + 367 ∗ tanh(aw /60) + {[2310 ∗ ln(HD)] / [12 ∗ HD0.150]} – 1081 (°C) (67b)

Table 37 shows, despite their slightly different formulations all the three equations
yield more or less similar estimates of T0/Ti. In the cases where the Crack-No Crack
boundary is known from the results of the U-Groove experiments, as for the OK Tubrod
15.26S and OK Tubrod 15.29S welds, all the three equations are seen to yield safe
predictions of T0/Ti. For the rest of the cases where the Crack-No Crack boundary
remained unknown, the equations predict T0/Ti estimates that accord satisfactorily well
with those given by the Okuda method[2] and, as it comes to the estimates based on weld
metal strength acc. to Eq. (48b), also the Yurioka method[3]. Thus, there is good reason to
believe that all the three VTT equations derived here result in safe predictions of T0/Ti
also for the rest of the cases in Table 37.
Although Eq. (67b) is intentionally adjusted to fit into the former NSC database[12] as
it comes to the effects of aw and HD on the estimated T0/Ti and obeying Eqs (63) and
(65), respectively, it can be seen that for the majority of the cases in Table 37 the use of
Eq. (62) yields slightly higher estimates of T0/Ti than Eq. (67a) or (67b). As a
consequence, among the three equations used to assess the required T0/Ti for the
experiments in Table 37, Eq. (62) is likely to yield the safest T0/Ti prediction for the
entire dataset.
In order to extend the T0/Ti assessment towards thin plate welds and greater hydrogen
contents, additional experimental cracking test data from Tables 10 and 6 were
incorporated into Table 38. These include (i) V-Groove cracking test data for two-pass
SAW and SMAW welds of 6 mm build-up thickness aw, and (ii) preliminary SAW U-
Groove cracking test data up to 13.6 ml/100 g DM (IIW) weld hydrogen, HD. Here, the
effects of aw and HD on T0/Ti were taken as defined in Eqs (63) and (65), respectively,
thus ensuring that their descriptions conform to the former NSC Y-Groove cracking test
database[12], see Tables 35 and 36 and Appendix 5.
 183

Table 38. Comparison of the T0/Ti estimates derived according to VTT prediction
formulae based on weld CET, HV5(ave) and HV5(max) against the required T0/Ti
temperatures in the SAW U-Groove cracking tests – complete database comprising all
the thick and thin plate welds.
Specimen SAW Weld Weld Weld Crack / Applied Prediction for the required
No Tubular CET HV hydrogen No
Cored (%) HD crack T0/Ti T0/Ti
Wire (ml/100 g (°C) (°C)
OK DM (IIW))
Tubrod
ave max Yurioka Okuda Eq. Eq. Eq.
formula[3] formula[2] (67b) (68a) (68b)
4) 4)
CET HVave HVmax
SU1 15.24S 0.263 225 235 6.2 NC 88 70–115 92–126 RT 83 86

U395 15.26S 0.333 265 277 4.9 NC 100 143–193 130–167 105 129 128
U593 15.26S 0.345 264 277 4.9 NC 100 140–191 128–166 131 127 128
U396 15.26S 0.334 251 262 6.1 NC 125 128–177 135–172 131 128 128
U397 15.26S 0.332 249 262 8.4 NC 150 149–197 177–213 159 157 160
SU4 15.26S 0.302 234 241 13.6 C 93 144–191 217–252 136 172 168
–
U400 15.27S 0.402 315 345 4.9 C 125 259–316 217–260 257 219 238
U594 15.27S 0.387 344 362 4.9 C 125 326–386 268–313 224 271 265
U398 15.27S 0.396 315 336 6.1 C 150 277–333 248–290 268 219 248
W11, 3) 15.27S 0.39 322 346 6.5 C 175 280–336 255–297 261 263 271
W21, 3) 15.27S 0.39 322 346 6.5 C 200 280–336 255–297 261 263 271
W31, 3) 15.27S 0.39 322 346 6.5 C 225 280–336 255–297 261 263 271
SU2 15.27S 0.382 313 334 7.2 C 71 285–342 267–309 254 257 262
U399 15.27S 0.388 330 367 8.4 C 100 297–354 288–330 282 303 330
SU3 15.27S 0.383 314 335 9.1 C 71 305–361 301–343 279 281 286
T1 6mm 75.75 0.254 273 280 14.4 NC 68 RT 69 58
S1 6mm 15.27S 0.341 257 266 16.0 NC 60 – – 56 48 44
–
U402 15.29S 0.402 339 362 4.9 C 150 315–374 260–304 257 262 265
M12, 3) 15.29S 0.41 346 368 6.0 C 250 347–407 299–345 297 297 298
M32, 3) 15.29S 0.41 346 368 6.0 C 275 347–407 299–345 297 297 298
M42, 3) 15.29S 0.41 346 368 6.0 NC 285 347–407 299–345 297 297 298
M22, 3) 15.29S 0.41 346 368 6.0 NC 300 347–407 299–345 297 297 298
U403 15.29S 0.419 354 376 6.1 C 100 368–429 316–362 318 314 313
U401 15.29S 0.399 345 367 8.4 C 125 371–432 344–390 307 330 330

1)
weld metal CET estimate calculated as a mean value of measurements for the OK Tubrod 15.27S
welds
2)
weld metal CET estimate calculated as a mean value of measurements for the OK Tubrod 15.29S
welds
3)
weld HVave and HVmax estimates calculated as mean values of measurements for the
corresponding weld metals
4)
T0/Ti range calculated acc. to Appendix 1 from the weld metal HV data converted into weld RM
estimates using Eq. (48b) and Ref.[119]
 184

It is known that the effect of weld metal strength RM on required T0/Ti can be
expressed either via (i) weld metal chemical composition CET hence conforming to SFS-
EN 1011-2[1b], or using (ii) weld metal Vickers hardness HV5 as the results of the present
thesis strongly indicate, see Fig. 36. The advantage of applying the HV based approach
rests on the fact that it incorporates the effects of both chemical composition and weld
thermal cycle. Thus, it was deemed necessary to investigate, whether the use of HV based
estimation of weld metal RM, in place of the CET based estimation, would bring any
extra advantages e.g., in terms of improved accuracy of the T0/Ti prediction. Two
additional VTT equations, Eqs (68a) and (68b), were therefore derived from the
experimental data. These formulae use the measured values of average and maximum
weld metal Vickers hardness HVave and HVmax, respectively, being of the form:
Tcr = 1.80 ∗ HVave + 367 ∗ tanh(aw /60) + {[2310 ∗ ln (HD)]/[12 ∗ (HD)0.150 ]} – 803 (68a)
Tcr = 1.62 ∗ HVmax + 367 ∗ tanh(aw /60) + {[2310 ∗ ln (HD)]/[12 ∗ (HD)0.150 ]} – 776 (68b)
Table 38 shows the comparison of the T0/Ti estimates calculated for the complete
cracking test dataset and applying: (i) the CET based approach obeying Eq. (67b), (ii) the
two new HV based approaches as defined in Eqs (68a) and (68b), and (iii) the estimates
given by the Okuda[2] and Yurioka[3] formulae. Overall, it can be seen that in the cases
where the Crack-No Crack boundary is known as an outcome of the U-Groove
experiments, i.e., OK Tubrod 15.26S and OK Tubrod 15.29S welds, all the three VTT
equations yield safe predictions of T0/Ti. For the rest of the cases where the Crack-No
Crack boundary remained unknown, the equations predict T0/Ti estimates that, again,
accord satisfactorily well with those given by the Okuda-formula[2] and, as it comes to the
estimates based on the weld metal strength acc. to Eq. (48b), also the Yurioka-formula[3].
It is noteworthy that, irrespective of the predictions in Table 38, the use of T0/Ti
temperatures above 300°C may be detrimental due to the possibility of austenite being
retained throughout welding, even if the predictive formulae suggest higher values.
Looking the T0/Ti estimates calculated using the CET, HV5ave and HV5max based
formulae and given in Table 38 in more detail, the following observations, conclusion
and recommendations can be drawn:
(1) In general, the interpass temperature estimation can be realistically based on either
weld metal chemical composition or weld Vickers hardness. The VTT formulae
based on weld metal CET, HVave and HVmax were all proven applicable for the Ti
estimation, hence resulting in realistic description of the required interpass
temperature Ti.
(2) For the majority of the examined cases, the HVmax based formula is found to yield
the highest T0/Ti estimates among the three evaluated VTT formulae, consequently,
the HVmax based Eq. (68b) is likely to result in the most safe T0/Ti predictions for
these cases.
(3) The VTT formulae based on weld metal hardness HV seem plausible in predicting
realistic Ti estimates also in the case of thin plate two-pass welds and to thicknesses
less than 10 mm.
(4a) Should the difference between weld metal average hardness and the individual
maximum HV value be exceptionally great, i.e., equal to or more than 20 HV units,
it is advisable to apply the HVmax based formula instead of the HVave based one.
 185

(4b) Should the difference between weld metal average hardness and the individual
maximum HV value be abnormally small, i.e., less than 10 HV units, it is advisable
to apply the HVave based formula instead of the HVmax based one.
(5) With low CET values less than 0.30%, the two HV based formulae are likely more
safe than the CET based formula. In these cases the use of the CET based formula
may yield unconservative estimates of Ti.
(6) For thin plate (< 10 mm) and essentially 2-dimensional heat flow conditions
resulting in prolonged weld cooling rates, it is advisable to use the CET based
formula instead of the HV based ones, unless CET is very low (< 0.27%) in which
case the use of the HVave based formula should be preferred instead of the HVmax or
CET based formulae.
(7) The Okuda-formula[2] is seen to overestimate the need of Ti towards increasing weld
hydrogen contents, particularly at HD levels exceeding 10 ml/100 g DM (IIW).
(8) The Yurioka-formula[3] is found to overestimate the need of Ti at very high levels of
weld metal strength, i.e., RM > 900 MPa.
(9) In conjunction with Yurioka's strength-hardness formula[119], the methods according
to both Okuda and Yurioka will overrate the need of Ti. This implies that Yurioka's
formula may exaggerate weld tensile strength at a given level of weld metal
hardness HV. Coupled with the VTT strength-hardness description as defined in Eq.
(48b), the Okuda and Yurioka formulae are seen to predict comparatively realistic Ti
estimates that are compatible with those given by the VTT formulae, i.e., Eqs (67b),
(68a) and (68b).
Based on the VTT formulae derived from the data in Tables 35–38 and in Fig. 39, the
proposed procedure for predicting safe T0/Ti temperatures for the prevention of hydrogen
cracking in multipass weld metals is outlined in Section 5.12.6.

5.12.5 Postulated mechanism for transverse hydrogen cracking

in extra-high strength multipass weld metals

The large- and small-scale Y-Groove and the U-Groove cracking tests all demonstrated
that for SMAW and SAW weld metals, at equivalent levels of interpass temperature and
weld hydrogen, the cracking risk of Rp0.2 ≥ 690 MPa welds was significantly higher than
that of Rp0.2 ≈ 550–590 MPa welds. This, in turn, accords with a recent study[117] that
showed how the occurrence of WM hydrogen cracking increased abruptly when
transferring from intermediate strength welds associated with CETs of about 0.33–0.36%
to extra-high strength weld metals exhibiting CETs of 0.40–0.42%. As regards to the
findings of the present study, this is thought to explain the need of using considerably
higher interpass temperatures for the extra-high strength weld metals of Rp0.2 ≥ 750 MPa,
compared with the intermediate strength ones, see Fig. 39.
Nonetheless, the steepness of increase in the required Ti level with the weld metal's
CET value c.f. Fig. 39 still is somewhat peculiar, considering how Ti affects weld
residual stress σres. A recent study[124] concerned with measurements of σres distributions
 186

in multipass weldments reports that the σres values for the specimens welded at the Ti of
100–120°C already remained lower than the specimens welded with the Ti below 30°C.
Thus, the beneficial effects of elevating the interpass temperature should not be restricted
to intensifying hydrogen removal only, but also to reducing weld residual stresses, both
of which lower the risk of hydrogen cracking. This leads back to the conclusion drawn
from the data in Table 31 (p. 167) suggesting that, at an equivalent level of weld H0,
elevating the T0/Ti is less beneficial in lowering the HRmax in the case of multiple-pass
welds than single-pass welds, see Section 5.11.
The Vickers hardness measurements in Table 30 (p. 161) reveal that a rise in WM
strength from Rp0.2 ≈ 550–590 to 750–900 MPa was accompanied by a steep increase in
local macrohardness mismatch within the multipass SAW weld metals. This mismatch
manifested itself as 50–80 HV difference in hardness between an individual maximum
and a minimum HV5 value. Since macrohardness translates to ultimate tensile strength
RM as defined e.g. in Eq. (48), the recorded hardness difference should manifest itself as a
corresponding difference in local strength of the different WM microstructures. As these
were not microhardness measurements and hence cannot reveal microscale properties,
such as nature of the phase interfaces, the results are considered merely an indication of
substantial differences in strength of local microstructural regions, which was found a
characteristic feature of the extra-high strength multipass SAW weld metals in the present
thesis. That the local variation of microstructures actually controls the location of
hydrogen cracks in multipass weld metal, is supported by a recent study[121] on extra-high
strength SAW welds. This study reported[121], most of the cracks were densely populated
in the weld region in which the microhardness increased gradually, but were not
necessarily related to that particular individual microstructural phase exhibiting the
highest microhardness.
Metallographic examination of the present thesis revealed that the presence of larger
amounts of GBF did not directly accentuate WM cracking, although cracks were found to
sometimes grow intergranularly along the GBF-matrix interface. The crack propagation
path revealed mainly transgranular, but occasionally, intergranular cracking modes. This
is consistent with a recent study[124] that reported, the formation of transverse hydrogen
cracks did not follow the GBF phase, rather, they propagated across the grains.
Furthermore, fracture morphology implied[124] that these cracks had occurred in the
highly stressed areas of the weldment. In the extra-high strength WMs of Rp0.2 ≥ 690
MPa studied here, transverse cracks growing normal to the weld surface and
perpendicular to the axis of the weld interface was a predominant mode, whereas in
Rp0.2 ≈ 550 MPa welds Chevron cracking showing a characteristic 'staircase' pattern and
propagating at 45° angle to the surface in weld thickness direction was recognised,
compare Figs 20–22 against Fig. 23 and Figs 24–26 against Fig. 27. This is in line with a
recent study[117] on fracture surface appearance in multipass SAW welds, showing that
the amount of crystalline area resembling cleavage/quasi-cleavage fracture increased with
the WM carbon equivalent and, hence, its strength. In the present thesis, WM hydrogen
cracks were found to initiate in the filling runs beneath the final weld bead layer, and
propagated perpendicularly to the axis of the weld interface into approx. 1/3-thickness of
the weldment. This accords with a recent study[124] on 50 mm thick multipass weldments
 187

where transverse hydrogen cracks were detected at depths of 9–14 mm from the weld
interface.
Analytic calculations showed that in multiple-pass welding, elevation of the
preheat/interpass temperature is less effective in removing weld hydrogen than in the case
of single-pass welds, see Table 31. On the other hand, reported indications[117] on extra-
high strength welds have shown that cracking in multipass welded specimens made
applying comparatively low Ti of around 80°C had frequently occurred soon after
welding and at welding temperatures higher than those traditionally associated with
hydrogen cracking, i.e., close to 300°C, on cooling to the applied Ti. Furthermore, the
amount of ductile area in the fracture surfaces of cracked specimens was reported[117] to
increase with the applied Ti.
Contemplating all these observations leads to conclusion that extra-high strength weld
metals require considerably high Ti temperatures, not only to remove hydrogen, but also
to promote conditions that will favour ductile fracture micromechanism in these weld
metals that due to their strength tend to be inherently more prone to brittle
cleavage/quasi-cleavage fracture.
Considering this against the recent findings made elsewhere[117, 121, 124] suggest, local
variations in microstructure resulting in gradients of hardness and strength may
contribute to the formation of hydrogen cracks at the GBF-matrix interface in extra-high
strength weld metal. High strength of the matrix microstructure allows the build-up of
great level of residual stress, while the presence of hydrogen at the interface is awaited to
weaken the interfacial cohesion. During weld cooling to the applied Ti, the build-up of
weld residual stresses coupled with continuous accumulation of hydrogen towards the
filling layers presumably favour crack initiation at the interface of two regions, of which
one exhibits plastic deformation behaviour while the other still behaves practically linear
elastically. Provided this causes formation of microcrack(s) at the interface, the situation
should be more or less analogous to that in a pre-cracked specimen, in which case it has
been shown[38], a combination of adjacent weld regions with mismatching mechanical
properties, rather than the presence of a single weak region, frequently resulted in
unstable crack growth under external load. Here, a heavily restrained multipass welded
specimen enables elevation of weld residual stress to amount the WM yield strength,
which thereby corresponds, in that sense, well enough to the conditions prevailing in an
externally loaded small specimen.
Overall, the findings suggest that the abrupt increase in multipass WM cracking
sensitivity when transferring from intermediate strength to extra-high strength weld
metals can be explained by the combination of: (i) steeply elevating weld residual stress,
(ii) increased matrix macrohardness of the weld microstructure in conjunction with
pronounced local hardness and, hence, strength mismatch, and (iii) inherent propensity
of the microstructure to brittle cleavage/quasi-cleavage fracture micromechanism.
 188

5.12.6 Proposed procedure as to provide the necessary precautions for

the avoidance of hydrogen cracking in multipass weld metal

The Procedure outlining the necessary precautions against hydrogen cracking in

multipass weld metals comprises two steps. These are: Step 1: Calculation of the 'worst
case' lower-bound estimate of weld critical hydrogen content Hcr to assess the Crack – No
Crack conditions, and Step 2: Calculation of a 'safe' estimate of the preheat/interpass
temperature T0/Ti needed for the avoidance of weld metal hydrogen cracking when
operating within the Crack region (according to Step 1 estimation).
The idea behind the Procedure is that the built-in conservatism decreases through the
sequence: Step 1, Step 2 Level 1, Step 2 Level 2. This way, the procedure penalises the
user with the minimum amount of required data for the assessment, and rewards the user
who is willing to create more than the minimum amount of required data and/or analyses.
Step 1: Calculation of the worst case lower-bound estimate of weld Hcr using the
Crack-No Crack Diagrams
Use one of the three methods to compare between the actual weld diffusible hydrogen
content HD (≡ H0) and the weld critical hydrogen content Hcr.
Method A : Stress based approach that assumes that weld Rp0.2 = σresL and uses the
true yield strength (MPa) of the multipass weld metal in question.
Hcr = 250 ∗ 10(–0.00235 ∗ Rp0.2) [ml/100 g DM (IIW)]
Method B : Weld Pcm based approach that uses the actual chemical composition of the
weld metal in question in terms of the Pcm (%) weldability index as defined in Eq. (43a).
Hcr = 175 ∗ 10(–6.460 ∗ Pcm) [ml/100 g DM (IIW)] , or
Hcr = 116 ∗ (0.295 – Pcm) [ml/100 g DM (IIW)]
The lower value of the two Hcr estimates calculated using the given formulae, is then
applied.
Method C : Weld HVmax based approach that uses the measured individual maximum
hardness value of the weld metal in question in terms of Vickers hardness HV5.
Hcr = 280 ∗ 10(–0.00585 ∗ HVmax) [ml/100 g DM (IIW)]
Choose the method to be applied either: (i) according to the most relevant actual data
that is available, or can be generated, for the case in question or, (ii) provided there is
enough data to apply all the three methods, take the one giving the lowest value of Hcr.
Compare the thereby calculated Hcr value to the actual weld diffusible HD value
measured for the weld metal in question using the single-pass hydrogen test conforming
to ISO/IIW 3690.
If Hcr > HD, there is no risk of transverse WM hydrogen cracking in multipass welds
for the case in question and in conditions referring to normal welding fabrication,
provided Ti ≈ 100°C.
 189

If Hcr ≤ HD, the possibility of having a risk of weld metal hydrogen cracking in
multiple-pass welding must be considered. In this case, one should either: (i) take
protective measures such as adjusting the welding procedure and/or changing to another
filler material to lower the actual weld diffusible hydrogen content HD, or (ii) consider
additional precautions against cracking such as the use of elevated preheat/interpass
temperature in which case one should proceed to Step 2.
Validity limits of the equations in Step 1:
480 ≤ Rp0.2 ≤ 890 (MPa)
210 ≤ HVmax ≤ 370 (HV5)
0.130 ≤ Pcm ≤ 0.300 (%)
0.180 ≤ CET ≤ 0.420 (%)
1.3 ≤ Q ≤ 5.0 (kJ/mm)
2.0 < Hcr < 19.0 (ml/100 g DM (IIW))
80 < Ti ≤ 300 (°C)
Note that until more cracking test results are available for weld metals with Pcm <
0.130%, HVmax < 210 HV or CET < 0.180%, it is advisable for safety reasons to use a
cut-off of the Hcr level for this regime. That is, the same Hcr value corresponding to either
weld metal's Pcm = 0.130%, HVmax = 210 HV or CET = 0.180% is applied also to any
other weld metal having lower Pcm, HVmax or CET value.
Step 2: Calculation of a safe estimate of T0/Ti to prevent weld metal hydrogen
cracking in multipass welds
Step 2 comprises two Levels, of which the latter one provides three Options. The
selection among these depends on the accuracy of the required T0/Ti estimate, as well as
on certain case-related factors such as the actual weld build-up thickness and weld
diffusible hydrogen content.
Level 1: Screening Test: Rough scaling of the required T0/Ti range as a function of
weld metal CET and weld hydrogen HD applying a "Three-Zone Diagram" illustrated in
Fig. 39.
The Diagram in Fig. 39 consists of three zones defined according to the range of weld
metal's CET value in question. The level of safe T0/Ti thereby depends on the range of
weld metal CET and weld HD. The Diagram can be applied provided the CET and HD
values are within the ranges indicated separately for each Zone.
Zone 1: 0.26% ≤ CET ≤ 0.33%
HD ≤ 5.0 ml/100 g DM (IIW) => T0/Ti = 100°C
HD ≤ 6.0 ml/100 g DM (IIW) => T0/Ti = 125°C
HD < 8.5 ml/100 g DM (IIW) => T0/Ti = 150°C
Zone 2: 0.33% < CET ≤ 0.39%
HD ≤ 6.0 ml/100 g DM (IIW) => T0/Ti ≥ 100 + 3340 ∗ (CET – 0.33) (°C)
Zone 3: 0.39% < CET ≤ 0.42%
HD ≤ 6.0 ml/100 g DM (IIW) => T0/Ti = 300°C
 190

0.26 < CET < 0.42%

No Cracks
1 0.26 < CET < 0.33 & Hd < 5.0 ml; Ti = 100°C cracks
300 H < 6.0 ml; T = 125°C
d i Low hydrogen
Hd < 8.5 ml; Ti = 150°C Hd 4.5 - 4.9
Intermediate hydrogen
250 2 0.33 < CET < 0.39; & Hd < 6.0 ml Hd 5.0 - 8.0
<
Ti 100 + 3340 • (CET-0.33) High hydrogen
3 0.39 < CET < 0.42; & Hd < 6.0 ml Hd 8.4 - 13.6
200 Ti = 300°C Preliminary tests
Tcr (°C)

Håkansson's data
150 Low hydrogen
Hd 4.5
Håkansson's data
100
Intermediate hydrogen
Hd 7.0 - 8.5

50

0
0.20 0.25 0.30 0.35 0.40 0.45
CET (%) wm03_1.1.dsf

Fig. 39. Critical preheat/interpass temperature T0/Ti as a function of weld metal CET for
multipass SAW weld metals according to the U-Groove test – Screening of Scale. Data from
Håkansson[117] added. Note: for deriving recommendations as the applied T0/Ti values
referring to Zone 1, also the data in Tables 37–38 have been used. Filled (red) symbols:
cracking; open (yellow) symbols: no cracking.

Should any of the Level 1 estimations yield T0/Ti values greater than 300°C, then set
T0/Ti = 300°C.
Should the validity limits set for the weld CET or HD become violated, one should go
to Level 2.
Level 2: Calculation of a safe T0/Ti estimate of the required Tcr using weld metal
strength in terms of either weld CET or weld metal HVmax, weld build-up thickness aw
and weld diffusible hydrogen HD as controlling factors:
Option 1: If 20 ≤ aw ≤ 40 mm and provided conservative estimates are regarded as
satisfactory:
Tcr = 2310 ∗ CET + 9 ∗ aw0.86 + {[2800 ∗ ln(HD)] / (12 ∗ HD0.21)} – 1145 (°C)
Option 2: When best estimates are required and the estimation is based on weld metal
chemical composition:
Tcr = 2195 ∗ CET + 367 ∗ tanh(aw /60) + {[2310 ∗ ln(HD)] / [12 ∗ HD0.15]} – 1081 (°C)
 191

Option 3: When best estimates are required and the estimation is based on weld metal
hardness:
Tcr = 1.80 ∗ HVave + 367 ∗ tanh(aw /60) + {[2310 ∗ ln (HD)] / [12 ∗ (HD)0.15]} – 803 (°C)

Tcr = 1.62 ∗ HVmax + 367 ∗ tanh(aw /60) + {[2310 ∗ ln (HD)] / [12 ∗ (HD)0.15]} – 776 (°C)
The validated applicability range of the equations in Level 2 Options 2 and 3 are as
follows:
20 ≤ Tcr ≤ 300 (°C)
2.0 ≤ Q ≤ 5.0 (kJ/mm)
0.25 ≤ CET ≤ 0.42 (%)
6 ≤ aw ≤ 40 (mm)
6 ≤ h ≤ 70 (mm)
2.0 ≤ HD ≤ 16.0 (ml/100 g DM (IIW))
The complete Flow-Charts of the Procedure Step 1 and Step 2 are presented in
Appendix 11.
Applying the equations given for Step 1 of the proposed Procedure thereby provides a
worst-case estimate of weld critical hydrogen Hcr as ’first line of defence’ against
transverse hydrogen cracking in multipass weld metal. For the assessment, weld
diffusible hydrogen content HD according to single-pass hydrogen test conforming to
ISO/IIW 3690, and either the true yield strength Rp0.2, chemical composition in terms of
Pcm, or actual maximum hardness HV5max of the multipass weld metal, is required.
Should Step 1 analysis reveal a potential WM cracking risk, Step 2 provides equations
for predicting a safe estimate of the required preheat/interpass temperature T0/Ti as
’second line of defence’ precautions for the avoidance of WM hydrogen cracking. To
conduct this assessment requires (i) weld diffusible hydrogen content HD, (ii) weld build-
up thickness aw and (iii) either weld metal true chemical composition in terms of CET, or
actual weld metal hardness in terms of HV5ave or HV5max, as input parameters.

5.12.7 Areas of non-applicability of the Procedure

The Procedure has its fundamental basis on multipass cracking experiments shown
descriptive of transverse hydrogen cracking occurring in the filling runs. Therefore, it
should not be applied to assess hydrogen cracking susceptibility of single-pass welds.
The experiments of the present thesis demonstrated that weld Hcr levels for weld
metals having their HVmax ≤ 210 HV, Pcm ≤ 0.130% and Rp0.2 ≤ 480 MPa become so
high that cracking is very unlikely in practice welding fabrication. Every now and then,
fabricators and, occasionally, even laboratories[8] are still experiencing cases of hydrogen
cracking in the welding of structural steel and in weld metal, which are associated with
weld strengths lower than those anticipated to provoke cracking. The reasons for the
occurrence of cracking in these cases have often remained unexplained. Until more
 192

cracking test data are generated for these low strength weld metals, it is therefore
advisable, for safety reasons, to use a cut-off of the Hcr level for weld metals with Pcm <
0.130%, HVmax < 210 HV or CET < 0.180%. That is, the same Hcr value corresponding
to either weld metal Pcm = 0.130%, HVmax = 210 HV or CET = 0.180% is applied also to
any other weld metal having lower Pcm, HVmax or CET.
It should be highlighted that the Crack-No Crack diagrams describing the ‘worst-case’
conditions with respect to WM cracking risk in Step 1 are, after all, based on cracking
tests involving Ti = 70–100°C. Any practice welding conditions where the actual Ti
might, at some stage during welding, go under the 70–100°C temperature level, can
therefore represent an acute risk of cracking.
It is noteworthy that the use of T0/Ti temperatures above 300°C may be detrimental
due to the possibility of austenite being retained throughout welding, even if the
predictive formulae applied in Step 2 suggest higher values.
The Procedure rests essentially on the outcome of the SMAW and SAW multipass
cracking tests. The analysis of other cracking test data from literature[76] implies that the
Procedure can presumably be applied to FCAW process, as well. However, such
assessments should be treated with caution. The applicability of the Procedure to welding
processes other than SMAW, SAW and FCAW cannot be guaranteed without performing
additional experiments.
Apart from the experiments of the present thesis, any exceptional welding conditions
and/or configurations may possess additional sources that can provoke cracking. Such
are, for example, welding with cellulose electrodes, welding of highly stressed thick
multipass fillet welds, multiple-pass single/double-bevel T butt joints in thick plate
welded from both sides and, finally, welding at an open air where welds could become
directly exposed to moisture or rain.
 6 Conclusive remarks
This doctoral thesis summarises experimental findings and analytical calculations on
controlling factors that govern hydrogen cold cracking in high-strength multipass weld
metal. The work is concerned with heavily restrained Y- and U-Groove multipass
cracking tests of shielded metal arc (SMAW) and submerged arc (SAW) weld metals of
yield strengths in the range of 455–900 MPa.
Overall, the results demonstrated that transverse hydrogen cracking in multipass welds
occurred predominantly at extra-high strength levels of Rp0.2 ≈ 580–900 MPa. Within this
range, weld metal strength and diffusible hydrogen have a substantial influence on
cracking sensitivity. Whereas at Rp0.2 ≈ 680–690 MPa strength level, weld hydrogen
contents of HD ≈ 7–10 ml/100 g DM(IIW) were necessary to induce cracks, already
≈ 3 ml/100 g DM(IIW) weld HD was sufficient for cracking to occur in the Rp0.2 ≈ 880–900
MPa weld metals.
In the case of low strength SMAW weld metals of Rp0.2 ≤ 480 MPa, it appeared
impossible to create conditions severe enough to result in WM cracking under any
hydrogen levels examined in the range of 5 to 11 ml/100 g DM(IIW).
At intermediate strengths of Rp0.2 ≈ 500–550 MPa, WM cracking occurrence was
found to depend decisively on the time from welding to NDT inspection. Cracking in
SMAW welds appeared in the cases when this holding time was prolonged from the
’standard’ 16 hours[1] to 168 hours (7 days), in conjunction with high weld HD of
≈ 15 ml/100 g DM(IIW). Was the period of 16 h applied, none of the welds showed any
cracks with hydrogen levels in the range of HD ≈ 3–18 ml/100 g DM(IIW). In the case of
SAW welds of Rp0.2 ≈ 585 MPa, cracking took place when moving from the weld HD of
≈9 to ≈13 ml/100 g DM(IIW). The crucial role of holding time that ought to be long enough
to allow for hydrogen crack occurrence is consistent with the recent findings made
elsewhere[117]. These demonstrated, the period of 16 h specified in EN 1011-2[1b] can, in
some cases, be much too short to allow for complete hydrogen diffusion and/or cracks of
critical size to be developed in the case of thick multipass welds.
Weld residual stress measurements made for thick multipass weldments indicated that
the weld longitudinal tensile residual stress σresL approaches the weld metal yield
strength Rp0.2, provided that the weld is long enough, i.e., at least 300 mm but,
preferably, 500–800 mm. Transverse residual stresses were found accordingly high.
 194

These stresses are presumably accompanied by high stress triaxiality. For thin plates,
weld residual stresses in the weld transverse and thickness directions are practically non-
existent. Unlike for heavy plates, the build-up of the σresL depended on how the plate was
fixed: (i) using bolted clamps the stress became partly relieved once the bolts were
removed after welding, consequently, the stress remained remarkably lower than the WM
yield strength; (ii) with welded strong-backs ensuring absolute rigidity of the weldment
throughout the welding and NDT, the σresL approached the WM yield strength.
The results of the double-welded large-scale Y-Groove, small-scale Y-Groove and the
U-Groove cracking tests for thick multipass weldments were all proven compatible.
These experiments employing rigidly bolted specimens subjected to intensified cooling
were shown to maximise the development of σresL and enable low interpass temperature
Ti and short interpass time ti to be obtained simultaneously. According to diffusion
theory, the combination of low Ti and short ti should minimise the effusion of hydrogen
during multiple-pass welding, thereby accentuating the accumulation of hydrogen and
hence the elevation of the local final hydrogen concentration Hmax in the filling runs.
The cracking tests demonstrated that for both SMAW and SAW weld metals, at
equivalent levels of interpass temperature and weld hydrogen, risk of hydrogen cracking
in the Rp0.2 ≥ 690 MPa weld metals was significantly higher than in the Rp0.2 ≈ 550–590
MPa welds. This accords with a recent study[117] that demonstrated how the occurrence of
cracking in multipass weld metal increased abruptly when transferring from intermediate
strength welds associated with CETs of 0.33–0.36% to extra-high strength welds
exhibiting CETs of 0.40–0.42%. Cracking seemed[117] to have frequently occurred at
temperatures higher than those traditionally associated with hydrogen cracking, i.e., close
to 300°C on cooling to lower interpass temperatures, Ti. The amount of ductile area in the
fracture surfaces of cracked specimens was also found[117] to increase with the Ti. As
regards to the present study, these phenomena are thought to explain the need of
considerably higher Ti for the extra-high strength welds of Rp0.2 ≥ 750 MPa, compared
with the intermediate strength ones.
Weld metal Vickers hardness showed good correlation with cracking susceptibility,
provided the actual WM maximum hardness HV5(max) is applied. The measurements
indicated, a rise in WM strength from Rp0.2 ≈ 550–590 to 750–900 MPa was
accompanied not only by a rise in the individual HV5(max), but also by a steep increase in
local hardness mismatch within the weld metal. This manifested itself as a 50–80 HV
difference between an individual maximum and minimum value and is considered merely
an indication of great differences in strength properties of local microstructural regions,
which was found a characteristic feature of the investigated extra-high strength multipass
SAW welds. A recent study[121] on similar SAW welds supports the view that the local
variations of microstructure can control the appearance and location of hydrogen cracks
in multipass weld metal.
Metallography revealed that the presence of larger amounts of grain-boundary ferrite
(GBF) did not accentuate WM cracking, although cracks were found to sometimes grow
intergranularly along the GBF-matrix interface. The crack propagation path consisted of
both transgranular and intergranular cracking modes. This accords with a recent study[124]
reporting that the formation of transverse hydrogen cracks did not follow the GBF phase,
rather, they propagated across the grains and concentrated on the highly stressed areas of
 195

the weldment. In the extra-high strength weld metals of Rp0.2 ≥ 690 MPa studied here,
transverse cracks growing in the weld thickness direction normal to the weld surface and
perpendicular to the axis of the weld interface was a predominant mode, whereas in the
Rp0.2 ≈ 550 MPa welds Chevron cracking showing a 'staircase' pattern and propagating in
the weld thickness direction at 45° angle to the weld surface was recognised. This is in
line with a recent study[117] on fracture surface appearance in multipass SAW welds,
showing that the amount of crystalline area resembling cleavage/quasi-cleavage fracture
increased with the WM carbon equivalent and, hence, its strength. In the present thesis,
WM hydrogen cracks were found to initiate in the filling runs beneath the final weld bead
layer, from where they had propagated perpendicularly to the axis of the weld interface
into ≈1/3-thickness of the weldment. This is in agreement with the experience gained
elsewhere[2, 3, 12, 76, 124].
Sorted out according to their actual weld chemical composition and WM hardness, the
present SAW U-Groove data in the Rp0.2 range of 500–900 MPa was found to comply
well with the Crack-No Crack boundaries derived from the SMAW data. The present
SAW cracking test results do not therefore assign any need to modify the Crack-No
Crack boundaries derived from the SMAW data.
Overall, the present thesis found the cracking risk of multipass weld metal somewhat
lower than expected on the basis of previous experience[2, 3, 12], even under the most
stringent conditions applying the combination of short interpass times and low interpass
temperatures of 3–4 min and 80–90°C, respectively. For instance, any attempts to
reproduce the former NSC Y-Groove test results[12] were unsuccessful. It is thought that
lowered impurity contents of modern welding consumables and hence weld metals can
impede grain boundary decohesion that otherwise tends to occur under the presence of
increasing accumulation of hydrogen. Thus, grain boundaries in clean welds can have
higher tolerance to the presence of hydrogen, which then elevates their resistance to
intergranular hydrogen cracking in a similar manner to that reported[39] for clean steels.
In all the cases where WM hydrogen cracking took place, it did so irrespective of
interpass time ti, even though ti was varied over a range from 4 to 15 min. Thus, ti seems
to have no measurable (significant) influence on WM cracking risk. This was attributed
to the dominant role of weld strength and hydrogen, which presumably overrides any
effect ti might have, unless the cracking test manages to produce a result that lies in the
immediate vicinity of the Crack-No Crack boundary line. Owing to the crude nature of
the applied cracking experiments, this would be considered as a rare event.
According to the SAW U-Groove experiments, heat input Q in the range of 2–5
kJ/mm seems to have no measurable effect on WM cracking susceptibility. Besides the
dominant role of weld strength and hydrogen, this is explained by the finding that heat
input affected the hardness in multipass weld metal to a much lesser extent than expected
on the basis of parent steel HAZ hardenability.
Equations were derived to describe average hardness HV5ave of SMAW and SAW
weld metals as a function of their chemical composition in terms of Pcm and CET. The
effect of arc energy on hardness was incorporated into the equations, although it was
found very modest in the range of 1.6–3.2 kJ/mm. Weld metal ultimate tensile strength
RM was shown to be reliably estimated as a function of HV5ave; a linear expression was
derived and shown to have good agreement with the present SMAW data, as well as with
 196

the earlier FCAW data. In the case of knowing only the weld chemical composition,
estimates of weld HV5 and RM can be calculated using the formulae given here. Weld
metal's Rp0.2 can then be estimated within reasonable accuracy using its yield-to-tensile
ratio based on nominal strength. Analysis of separate data from former works[12,118,119]
showed good agreement with the present results, thereby validating the approach outlined
in the present thesis.
Analytical calculations showed that the local final hydrogen concentration HRmax in
multipass weld becomes considerably higher than the remaining diffusible hydrogen
content HR100max in single-pass weld, even though the initial diffusible hydrogen
contents, H0, were identical. This was attributed to diffusion and accumulation of
hydrogen towards the filling runs during multiple thermal cycles, which counterbalances
the effusion of hydrogen during the cooling stage of each weld pass. Such accumulation
does not occur in single-pass welds, consequently, the HR100max always becomes lower
than H0. Moreover, calculations implied, under equivalent weld thermal cycle and initial
hydrogen in terms of ΣD∆t and H0, respectively, the HRmax could become higher in SAW
than SMAW multipass welds. This was attributed to higher weld bead overlap in the case
of SAW. As greater bead size results in longer diffusion distances, values of the HRmax/H0
close to, and even beyond unity were obtained especially for SAW and over a wide range
of investigated welding conditions. According to calculations, the HRmax did not, even
under extreme conditions, exceed the H0 by more than ≈ 10%. Thus, HRmax = H0 can be
taken as a general approximation for the predictive formulae assessing cracking risk of
multipass weld metal, while HRmax ≈ 1.1 ∗ H0 is considered appropriate for SAW welds
in thick plate. Values of H0 (= HD) can be determined from single-pass weld hydrogen
test conforming to ISO/IIW 3690.
Incorporating all the results showed that WM cracking occurrence was primarily
governed by (i) the weld metal ultimate tensile strength RM, (ii) weld longitudinal
residual stress σresL amounting the WM yield strength Rp0.2 and (iii) weld diffusible
hydrogen HD (= H0). Weld build-up thickness aw was found to affect cracking in that, at
comparable levels of strength and hydrogen, the occurrence of cracking vanished when
transferring from 40 to 6 mm thick multipass welds. The implications of the time from
welding to NDT were much more significant than anticipated previously, whereas the
interpass time ti and heat input Q are regarded as being of secondary importance.
As ‘first line of defence’, experimentally verified numerical formulae defining Crack-
No Crack diagrams are provided for calculating worst-case estimates for the weld critical
hydrogen content Hcr when knowing either: (i) weld metal yield strength Rp0.2, (ii) weld
chemical composition Pcm, or (iii) weld metal maximum hardness HVmax. All the three
approaches yielded essentially the same results and similar ranking of the examined
SMAW and SAW weld metals as individual datasets and in terms of their cracking
susceptibility, thereby indicating good descriptive potential in assessing hydrogen
cracking sensitivity of multipass weld metals. Differences in the relative ranking between
the SMAW and SAW welds were recognised, suggesting tubular SAW filler wires may
produce weld metals that, in relation to their strength, are inherently more crack sensitive
than the SMAW weld metals. In view of the descriptive potential of these approaches for
the entire spectrum of the SMAW and SAW welds, the HV5(max) based approach was
found to accommodate the actual cracking susceptibility of the different WM strength
 197

grades most realistically. That weld Hcr depends on its hardness as defined here is
consistent with a former work[122] on hydrogen cracking in weld metals. Furthermore,
analysis of separate cracking test data[76, 117] showed that, with the exception of one
individual result and the Rp0.2 based approach, applying the cracking diagrams outlined
in the present thesis would have yielded correct predictions of cracking occurrence for
the rest of the data. These are considered as plausible evidences on the validity of the
approaches defined in the present thesis.
An estimate of maximum final hydrogen concentration Hmax in multipass welds can be
calculated as a function of (i) weld initial diffusible hydrogen content H0, (ii) thermal
factor of hydrogen diffusion ΣD∆t approximated using weld t100 data and (iii) individual
weld bead layer thickness hw accounting for the weld overlap effect d/hw, using the given
formulae. Comparing the weld Hcr and the Hmax values then enables the assessment of the
boundary conditions for hydrogen cracking in multipass weld metal.
As ‘second line of defence’ and for the cases where Hcr < Hmax, equations were derived
for the calculation of safe preheat/interpass temperature estimates, T0/Ti, for the
avoidance of cracking. Analysing the whole dataset, a formula incorporating: (i) weld
metal tensile strength as linear functions of either weld metal CET or weld HV5(max), (ii)
weld build-up thickness aw in the form of tanh based expression and (iii) weld diffusible
hydrogen HD in terms of a combined [ln / power law] based expression was found
descriptive.
The results imply that (i) precautions against WM hydrogen cracking in low-strength
multipass welds in thick plate, as well as (ii) the possible dependency of the critical
holding time before NDT on weld metal strength grade, should deserve increasing
attention in the future.
The identification of the primary causal factors governing hydrogen cracking in
multipass weld metals in the present thesis further enables the local approach treatise of
hydrogen effects. The assessment of the hydrogen-crack tip interactions applying local
approach treatise of damage-based parameters according to continuum based mechanics,
coupled with local stresses, actual hydrogen concentration, microstructure and actual
failure mechanism, provide the latitude for performance of analyses utilising the
micromechanical descriptions of hydrogen cracking.
 7 Summary of final conclusions
• Hydrogen cracking occurred predominantly in extra-high strength weld metals of
Rp0.2 ≈ 580–900 MPa. Within this range, cracking depended essentially on weld
strength and hydrogen.
• At intermediate strengths of Rp0.2 ≈ 500–550 MPa, cracking in SMAW welds took
place in the cases where the time from welding to NDT was prolonged to 7 days and
at high weld hydrogen of HD ≈ 15 ml/100 g DM (IIW).
• Low strength weld metals of Rp0.2 ≤ 480 MPa did not exhibit hydrogen cracking
under any conditions examined.
• Results of the large-scale Y-Groove and the small-scale Y- and U-Groove cracking
tests were all compatible in the sense of ranking multipass weld metals according to
their cracking susceptibility. The experiments employing rigidly bolted specimens
subjected to intensified cooling were shown to maximise the development of
longitudinal σresL and enable low interpass temperature and short interpass time to be
obtained simultaneously.
• Despite a range of heat inputs and interpass temperatures, all the SAW cracking test
results, when sorted according to their actual weld composition and hardness, fell
within the Crack-No Crack regions derived from the SMAW experiments.
• According to analytical calculations, Hmax can – with equivalent weld thermal cycle
and initial hydrogen – become higher in SAW than SMAW multipass welds. Setting
Hmax ≈ 1.1 ∗ H0 is considered appropriate for SAW welds in thick plate, while Hmax =
H0 can be taken as a general approximation. Here, H0 = HD and denotes to weld
diffusible hydrogen determined from a single-pass hydrogen test.
• Equations were derived to describe weld HV5 as a function of its chemical
composition in terms of Pcm and CET, as well as to calculate weld RM as a function
of HV5ave. Separate data from previously published works[12, 118, 119] accord well with
the present results, thereby validating the outlined approach.
• As ‘first line of defence’, equations were derived to assess the weld critical hydrogen
content Hcr corresponding to the Crack-No Crack conditions as a function of either
weld metal Pcm, yield strength Rp0.2 or weld hardness HV5(max). All these parameters
were shown to rank the separate datasets of the examined SMAW and SAW weld
metals accordingly with respect to the occurrence of hydrogen cracking. Separately
 199

published cracking test data[76, 117, 122, 123] show good agreement with the present
results.
• As ‘second line of defence’, equations were derived for the calculation of safe T0/Ti
estimates for the avoidance of cracking. Analysing the whole dataset, a formula
incorporating: (i) weld metal strength as a linear function of either weld CET or
HV5max , (ii) weld build-up thickness aw in the form of tanh based expression and (iii)
weld diffusible hydrogen HD in terms of a combined [ln / power law] based
expression was found descriptive.
• In thick multipass welds hydrogen cracking occurrence was, above all, governed by
the weld metal tensile strength, weld diffusible hydrogen and the weld σresL
amounting to the true weld yield strength.
• Plate/weld build-up thickness was found to affect WM cracking risk in that, at
comparable levels of strength and hydrogen, the occurrence of cracking vanished
when transferring from 40 to 6 mm thick multipass welds.
• The implications of time from the completion of welding before NDT were much
more significant than anticipated previously. A period of 16 h from welding to NDT
in accordance with SFS-EN 1011-2 appeared far too short in the case of thick
multipass welds.
• Interpass time and heat input showed no measurable effect on the WM hydrogen
cracking sensitivity, hence turning out to be less important than previously believed.
• Precautions against WM hydrogen cracking in low-strength multipass welds in thick
plate, as well as the possible dependency of the critical holding time before NDT on
strength grade, deserve increasing attention.
• The identification of the primary causal factors governing hydrogen cracking in
multipass weld metals in the present thesis further enables the local approach treatise
of hydrogen effects at the crack tip.
 8 Future work
Whilst the principal causal factors contributing to hydrogen cracking in multipass weld
metals are nonetheless comparatively well understood, investigation of intrinsic factors
as it comes to local conditions prevailing in a microstructure under the presence of
hydrogen and stress, have so far received much less attention. For instance, the effect of
welding residual stress on the weld cracking risk has traditionally been encountered using
robust, causal parameters, i.e., level of longitudinal stress and applying regression
analyses[1–5, 11, 19, 26, 67b], instead of sophisticated local approach based numerical
modelling capable of determining local stresses and distributions at and near the crack
tip, or phase interface. Thus, assessments of hydrogen-interface interactions based on
local approach treatise of damage mechanics parameters using continuum based
mechanics, coupled with local stresses, actual hydrogen concentration, microstructure
and actual failure mechanism, are yet relatively rare and poorly understood. The results
of the analyses of mass diffusion of hydrogen, for example, are difficult to quantify in
terms of cracking risk since no reliable coupling between the local hydrogen
concentration and material damage has been presented.
As an introduction, local approach based numerical analyses of the interactions
between hydrogen diffusion and accumulation and the associated effects on the material's
cracking resistance in the case of multipass weld metals has recently been initiated[120] at
VTT Industrial Systems. Former numerical modelling[21,22] performed at Nippon Steel,
Japan, has been analysed and applied as background data. Numerical analyses enables
solution of hydrogen diffusion in welds having complex geometries. A finite difference
method (FDM) based temperature solution can be used as an input to a finite element
transient mass diffusion analysis. In the mass diffusion problem, a three-dimensional
residual stress field is to be used as input to describe the pressure stress dependency of
the transient diffusion process. The resulting concentration profiles are considered by use
of a novel damage mechanics material model in a finite element analysis (FEA) cell
modelling framework, which links the local concentration to a continuum mechanics
damage description. The use of the damage mechanics constitutive material model will
thereby provide means to evaluate the conditions for hydrogen cracking risk in multipass
welds.
 201

The first results[120] of the mass diffusion analysis seem to be in harmony with the
experimental findings and measured hydrogen concentrations of the present thesis. These
imply, the damage mechanics analysis will provide means for evaluating the rupture
process by the use of continuum mechanics. These modelling work[120] are currently
underway at VTT Industrial Systems. Still, open questions to solved are, for instance,
consideration of time for crack development at different strength levels, as well as how to
treat crack nucleation event. Existing constitutive models, as such, do not set limitations
to model the crack initiation event, however, from the fracture mechanics viewpoint
crack tip considerations are only appropriate to the growth of an existing crack whose
initial size hence needs to be postulated. An obvious difficulty is that there is practically
no experimental evidence whatsoever, from which it could be reliably concluded whether
the incubation period from welding to crack appearance, if being as long as hundreds of
hours, is essentially for (i) the nucleation of a crack, or (ii) growth of an already initiated
crack into critical size.
The results of the present thesis clearly demonstrated that the effect of time from
welding to NDT is very important. In practice, the welding engineer is concerned with
ensuring freedom from hydrogen cracking in his structure for the whole of its lifetime. It
is of little consolation to know that there will be no cracking up to 16 hours after welding,
as specified in SFS-EN 1011[1] if cracks may appear after 7 days or whatever. In real
structures welding residual stresses are not relaxed after, say, 16–48 h, for example. Thus,
welding engineers need to estimate the time under restraint before NDT to be sure that all
possible hydrogen cracks have formed for any given weld.
Therefore, the possible dependency of the critical holding time before the NDT on
weld metal strength grade should deserve increasing attention in the future. An indication
of its magnitude is available from the present results on the lower strength welds,
whereas the question whether the time effect appears similarly also in the high-strength
welds remains yet to be answered. Anyway, separate results from a recent study[117] on
hydrogen cracking in multipass SAW weld metals allow one to conclude that the time
factor becomes increasingly important towards lower strength weld metals, whilst the
high strength welds exhibited cracks already within a much shorter period of time. This is
consistent with the recognised responses of the different weld metals investigated in the
present thesis to the time period before NDT. This would mean that as far as the higher
strength regime is concerned, the present results should ensure freedom from hydrogen
cracking when applying weld Hcr levels given by the formulae and diagrams of the
present thesis.
Still, the scope of the work carried out in the thesis does not allow accurate answers to
be given in these questions. Therefore, the holding time under restraint to ensure that all
cracking is captured during NDT is obviously an area requiring further work. Therefore,
additional quantitative data should be produced to further convince and validate the
conclusions made in the present thesis, as regards to the dependence of holding time on
strength.
 References
[1a] EN 1011-1:1998: ‘Welding – Recommendations for welding of metallic materials
– Part 1: General guidance for arc welding’. CEN/TC 121, CEN 1998. 18 p.
[1b] EN 1011-2:2001: ‘Welding – Recommendations for welding of metallic materials
– Part 2: Arc welding of ferritic steels’. CEN/TC 121, CEN 2001. 113 p.
[2a] Okuda N, Nishikawa Y, Aoki T, Goto A & Abe T: ‘Hydrogen-induced cracking
susceptibility of weld metal’. IIW-Doc. II-1072-86 / II-A- -86. Kobe Steel
Ltd./The International Institute of Welding, Japan, 1986. 16 p.
[2b] Okuda N, Ogata Y, Nishikawa Y, Aoki T, Goto A & Abe T: ‘Hydrogen-induced
cracking susceptibility in high-strength weld metal’. Welding Journal 66 (1987):
May, pp. 141-s – 146-s.
[3a] Yatake T & Yurioka N: ‘Studies of delayed cracking in steel weldments (Report
3)’. Journal of the Japan Welding Society (JWS) 50 (1981) 3, pp. 291–296.
[3b] Suzuki H & Yurioka N: ‘Prevention against cold cracking in welding steels’.
Australian Welding Journal 27 (1982) 1: Autumn 1982. Pp. 9–27.
[4] Nevasmaa P & Karppi R: ‘Implications on controlling factors affecting weld metal
hydrogen cold-cracking in high-strength shielded-metal arc (SMAW) multipass
welds’. Report AVAL64-011043: WM-HICC VTT/3. IIW-Doc. IX-1997-01. VTT
Manufacturing Technology, Espoo 2001. 94 p. + Apps 26 p.
[5a] Nevasmaa P: ‘Controlling factors influencing hydrogen cold cracking in high-
strength multipass weld metals – Status Review report’. Report VAL A:WM-HICC
VTT/1. VTT Manufacturing Technology, Espoo 2000. 51 p. + Apps 4 p.
[5b] Nevasmaa P, Kantanen M-S & Pan L: ‘Hydrogen cold-cracking in high-strength
shielded-metal arc (SMAW) multipass weld metals: experimental findings’.
Report AVAL64-001029:WM-HICC VTT/2. VTT Manufacturing Technology,
Espoo 2000. 66 p. + Apps 23 p.
 203

[5c] Nevasmaa P: ‘Controlling factors affecting hydrogen cold-cracking in high-

strength multipass weld metals: comparison of the cracking test results between
SMAW and SAW welds’. IIW-Doc. No IX-2027-02. VTT Industrial Systems,
Espoo 2002. 46 p. + Apps 14 p.
[6] Nevasmaa P & Karppi R: ‘Avoidance of hydrogen cracking in modern high-
strength steel multipass welds’. Proceedings of the "17th Nordic Welding Meeting
(NSM’97)", Copenhagen, 14–16 May 1997. Copenhagen: Danish Welding
Society, 1997. 17 p.
[7] Nevasmaa P & Karppi R: ‘Welding for the economic and safe use of modern high-
strength steels in order to avoid hydrogen cracking in multipass welds’. Research
Report VAL B 202. VTT Manufacturing Technology, Espoo 1997. Jernkontorets
Forskning TO 40-39. Jernkontoret, Stockholm 1998. 81 p.
[8] Pargeter RJ: ‘Effects of arc energy, plate thickness and preheat on C-Mn steel
weld metal hydrogen cracking’. TWI Research Report (CRP) No. 461/1992. The
Welding Institute, UK, November 1992.
[9] Kinsey A.J: ‘Weld metal hydrogen cracking during welding 450 N/mm2 yield
strength steel using tubular cored electrodes’. TWI Research Report (CRP) No.
655/1998. The Welding Institute, UK, October 1998. 24 p.
[10a] Standard: SFS-EN 288-3:1992: ‘Hitsausohjeet ja niiden hyväksyntä metallisille
materiaaleille’. Osa 3: Terästen kaarihitsauksen menetelmäkokeet’, Suomen
Standardisoimisliitto, 1992. (in Finnish)
[10b] Final Draft Standard: prEN ISO 15614-1: ‘Specification and qualification of
welding procedures for metallic materials – Welding procedure test – Part 1: Arc
and gas welding of steels and arc welding of nickel and nickel alloys’. CEN/TC
121 N 998. CEN/ISO, 2002. 28 p.
[11] Thier H, Eisenbeis C & Winkler R: ‘Untersuchungen zur Absicherung der
Berechnung der Mindestvorwärmtemperatur beim Mehrlagenschweissen’.
Schweissen und Schneiden 49 (1997) 7, pp. 426–430. (in German)
[12] Yurioka N: ‘Test results of cold cracking in multi-pass weld metal’. IIW-Doc. IX-
1903-98. Nippon Steel Corporation, The International Institute of Welding, Japan,
1998. 11 p.
[13] Karppi R, Ruusila J, Toyoda M, Satoh K & Vartiainen K: ‘Predicting safe welding
conditions with hydrogen cracking parameters’. Scandinavian Journal of
Metallurgy 13 (1984), pp. 66–74.
[14] ‘Guide to the light microscope examination of ferritic steel weld metals’. IIW-
Doc. IX-1533-88. The International Institute of Welding, 1988. 5 p.
[15] Tihekari H & Karppi R: ‘Mikrorakenteen vaikutus hienoraeteräksen hitsaus-
liitoksen sitkeyteen arktisissa olosuhteissa’. Technical Research Centre of
Finland. Valtion teknillinen tutkimuskeskus (VTT), Espoo, 1984. 40 p.
(in Finnish)
 204

[16] Wildash C, Cochrane RC, Gee R & Widgery DJ: ‘Microstructural factors
affecting hydrogen induced cold cracking in high strength steel weld metal’.
Proceedings of the "5th International Conference on Trends in Welding Research",
June 1998, Pine Mountain, Georgia, 1–5 June 1998. U.S.A 1998.
[17] Wildash C, Gee R, Cochrane RC & Widgery DJ: ‘The influence of hydrogen and
microstructure on the tensile properties of high strength steel weld metal’.
Proceedings of the "9th International Conference on Joining of Materials",
Helsingör, May 1999. Denmark. Institute for Joining of Materials, 1999. Pp. 335–
340.
[18] Wildash C, Gee R & Cochrane RC: ‘Designing a microstructure to resist HIC in
HS steels’. Welding & Metal Fabrication 68 (2000): April, pp. 15–20.
[19] Japanese Welding Association Standard WES 1105-1985. ‘Cracking test for
single-bevel groove multi-layer welds’. Japanese Welding Association, Japan
1985. 38 p.
[20] Osaka University & Nippon Steel Interim Report on weld metal hydrogen
cracking applying numerical modelling of hydrogen distribution – Interim Report.
(in Japanese)
[21] Yurioka N, Ohshita S, Nakamura H & Asano K: ‘An analysis of microstructure,
strain and stress on the hydrogen accumulation in the weld heat-affected zone’.
IIW-Doc. IX-1161-80. Nippon Steel Corporation, The International Institute of
Welding, Japan, 1980. 18 p.
[22] Yurioka N: ‘A review of numerical analyses on the hydrogen diffusion in welding
of steel’. IIW-Doc. IX-1553-89. Nippon Steel Corporation, The International
Institute of Welding, Japan, July 1989. 15 p.
[23] Yurioka N: ‘Predictive methods for prevention and control of hydrogen assisted
cold cracking’. IIW-Doc. IX-1938-99. Nippon Steel Corporation, The International
Institute of Welding, Japan, 1999. 16 p.
[24] Karppi R: ‘Effect of weld hydrogen content and preheat temperature on the
Implant fracture strength of a few heat resisting pressure vessel steels’.
Publication No 1/1976. Helsinki University of Technology, Laboratory of
Materials Technology, Espoo 1976. 422 p.
[25] Karppi R: ‘A stress field parameter for weld hydrogen cracking‘. VTT
Publications 9. Technical Research Centre of Finland, Espoo 1982. 119 p.
[26] Wong RJ: ‘Hydrogen cracking resistance of high strength steels in single-pass and
multiple-pass weldability tests’. Proceedings of the "Materials Week '95"
Symposium "Welding and Weld Automation in Shipbuilding", Cleveland, Ohio, 29
October – 2 November 1995. Ed. R. SeNale. The Minerals, Metals and Materials
Society, Pennsylvania, USA 1996. Pp. 33–46.
[27] Boothby PJ: ‘Weldability of offshore structural steels. Part 1 & 2’. Metal
Construction 17 (1985): August & September, pp. 2–14.
 205

[28] Hart PHM & Harrison PL: ‘Compositional parameters for HAZ cracking and
hardening in C-Mn steels’. Welding Journal 66 (1987): October, pp. 310–322.
[29] Hart PHM, Matharu IS & Jones AR: ‘The influence of reduced carbon equivalent
on HAZ cracking in structural steels’. Proceedings of the "7th International
Conference on Offshore Mechanics and Arctic Engineering", Houston, Texas, 7–
12 February 1998. Ed. M. Salama et al. U.S.A: The American Society of
Mechanical Engineers (ASME), 1998. Vol. III. Pp. 111–120.
[30] Hart PHM & Harrison PL: ‘Compositional parameters for HAZ cracking and
hardening in C-Mn steels’. Rivista Italiana della Saldatura No 2, 1990. Pp. 27–43.
[31] Private information. Discussion, R Pargeter & P Hart, TWI, in 21st February 2000.
[32] Yang LJ, Chandel RS & Bibby MJ: ‘The effects of process variables on the weld
deposit area of submerged arc welds’. Welding Journal 72 (1993): January, pp.
11-s – 18-s.
[33] Gunaraj V & Murugan N: ‘Prediction and optimization of weld bead volume for
the submerged arc process – Part 1 and 2’. Welding Journal 79 (2000): October &
November, pp. 286-s – 294-s and 331-s – 338-s.
[34] Nevasmaa P, Cederberg M & Vilpas M: ‘Weldability of accelerated-cooled (AcC)
high strength TMCP steel HT50’. VTT Research Notes 1410. Technical Research
Centre of Finland, Espoo 1992. 30 p. + Apps. 8 p.
[35] Private information. Discussion, N Yurioka, Nippon Steel Corp. (NSC), June
2001.
[36a] Kantanen M-S: ‘Controlling Factors Influencing Hydrogen Cold Cracking in
High-Strength Multi-Pass Weld Metal – Microstructure and tensile properties’.
Interim Report, Oulu University, Oulu 2001.
[36b] Pan L: ‘Literature search on hydrogen diffusion data and permeation measure-
ments’. Interim Report, Oulu University, Oulu 1999.
[37] Wang WW, Wong RJ, Liu S & Olson DL: ‘Use of martensite start temperature for
hydrogen control’. Proceedings of the "Materials Week '95" Symposium "Welding
and Weld Automation in Shipbuilding", Cleveland, Ohio, 29 October – 2
November 1995. Ed. R. SeNale. The Minerals, Metals and Materials Society,
Pennsylvania, USA 1996. Pp. 17–31.
[38] Nevasmaa P, Laukkanen A & Ehrnstén U: ‘Fracture resistance and failure
characteristics of AISI 304/ SA508 bimetallic weld in ductile regime’.
Proceedings of the “13th European Conference on Fracture – Fracture
Mechanics: Applications and Challenges (ECF 13)”, San Sebastian, 6–9
September 2000. Eds. M Fuentes & D Eastbury. Spain: Engineering Materials
Advisory Services Ltd., 2000. Paper No 1N.49. 8 p.
 206

[39] Liimatainen J & Martikainen H: ‘New high-purity cast steels for offshore
applications’. Proceedings of the “International Conference Welding-90:
Technology, Materials, Fracture”, Geesthacht, 22–24 October 1990. Ed. M.
Kocak. Germany: Institute for Industrial Technology Transfer, France 1990.
[40] Coe FR: ‘Welding steels without hydrogen cracking’. The Welding Institute,
Cambridge, United Kingdom, 1973. 68 p.
[41] Baillie JG: ‘Underbead and toe cracks’. British Welding Journal 14 (1967) 2, pp.
51–61.
[42] Gooch TG & Cane MWF: ‘Fractographic observations of hydrogen induced
cracking in steel. Practical implications of fracture mechanics’. Institution of
Metallurgists: Spring meeting, Newcastle upon Tyne, 27–29 March 1973.
Institution of Metallurgists, London 1973. Paper No 10. Pp. 95–102.
[43] IIW: ‘Note on the underbead hardness and cold cracking susceptibility’. IIW-Doc.
IX-600-68. The International Institute of Welding, Warsaw 1968. 3 p.
[44] Inagaki M, Nakamura H, Inoue T, Takeshi Y, Minami K & Kanazawa S: ‘Delayed
cracking in high-strength steel weld with angular distortion’. Transactions of the
Japan Welding Society 4 (1973) 1, pp. 19–29.
[45] Ito Y & Iwanaga H: ‘The effects of alloying elements on weldability’. Journal of
the Japan Welding Society (JWS) 36 (1967) 9, pp. 56–64. (in Japanese)
[46] Matsuda F, Nakagawa H, Tsuji T & Tsukamoto M: ‘Fractography of cold crack of
various carbon steels and alloy steels by means of the Implant test’. Transactions
of the Japan Welding Research Institute (JWRI) 7 (1978) 1, pp. 71–85.
[47] Randall MD, Monroe RE & Rieppel PJ: ‘Causes of microcracking and
microporosity in ultra-high strength steel weld metals’. Welding Journal 41 (1962)
5, pp. 193-s – 206-s.
[48] Satoh K, Terai K, Yamada S, Nagano T & Matsumura H: ‘Study on weld cracking
in multiple pass welds of high-strength steel’. Transactions of the Japan Welding
Society 1 (1970) 1, pp. 112–118. IIW-Doc. IX-678-70. The International Institute
of Welding, 1970. 7 p.
[49] Satoh K, Matsui S, Nishimura I, Hyama H & Chiba N: ‘Effect of intensity of
bending restraint on weld cracking in multipass weld’. Transactions of the Japan
Welding Society 8 (1977) 1, pp. 42–49. IIW-Doc. IX-960-76. The International
Institute of Welding, 1976. 13 p.
[50] Gedeon SA & Eagar TW: ‘Assessing hydrogen-assisted cracking fracture modes
in high-strength steel weldments’. Welding Journal 69 (1990): June, pp. 213-s –
219-s.
[51] Beachem CD: ‘A new model for hydrogen-assisted cracking (hydrogen
embrittlement)’. Metallurgical Transactions 3 (1972) 2, pp. 437–451.
 207

[52] Nevasmaa P, Cederberg M & Vilpas M: ‘Weldability of new HT-50 and HT-80
class thermomechanically rolled high strength steels’. Proceedings of the "5th
Conference on Joining of Materials – JOM-5", Helsingör, 10–12 May 1991. Eds.
RL Apps & O Al-Erhayem. The European Institute for the Joining of Materials,
Denmark 1991. Pp. 387–394.
[53] Nevasmaa P, Cederberg M & Vilpas M: ‘Weldability of direct-quenched and
tempered (DQ-T) high strength steel HT80’. VTT Research Notes 1406. Espoo
1992. 32 p. + Apps. 12 p.
[54] Graville BA: ‘A survey review of weld metal hydrogen cracking’. Welding in the
World 24 (1986) 9/10, pp. 190–199.
[55] Matsuda F, Nakagaura H & Matsumoto T, Transactions of the Japan Welding
Research Institute (JWRI) 10 (1981) 1, pp. 81–87.
[56] Interrante CG & Stout RD, Welding Journal 43 (1964): April (Research
Supplement), pp. 145-s – 160-s.
[57] ISO/IIW 3690-1977 E: ‘Determination of hydrogen in deposited weld metal
arising from the use of covered electrodes for welding mild and low-alloy steels’.
International Organisation for Standardisation, 1977. 4 p.
[58] Moreton J & Coe FR: ‘A survey of the sources, distribution and movement of
hydrogen in weld metal’. IIW-Doc. II-512-69. The International Institute of
Welding, 1969. 21 p.
[59] Coe FR & Chano Z: ‘Hydrogen distribution and removal for a single bead welding
during cooling – Part 3’. Welding Research International 5 (1975) 1, pp. 66–80.
[60] Schwarz W & Zitter H: ‘Determination of diffusible hydrogen in the weld metal
deposited by means of coated electrodes’. IIW-Doc. II-A-258-70. University of
Mining and Metallurgy, Leoben 1970. The International Institute of Welding, 16
p.
[61] Gerberich WW, Chen YT & St John C: ‘A short-time diffusion correlation for
hydrogen-induced crack growth kinetics’. Metallurgical Transactions A 6A
(1975) 8, pp. 1485–1498.
[62] Savage WF, Nippes EF & Homma H: ‘Hydrogen cracking in HY-80 steel
weldments’. Welding Journal 55 (1976) 11, pp. 368-s – 380-s.
[63] Terasaki T, Karppi R & Satoh K: ‘Relationship between critical stress of HAZ
cracking and residual diffusible hydrogen content’. Transactions of the Japan
Welding Society 10 (1979) 1, pp. 53–57.
[64] Lazor RB & Graville BA: Canadian Welder & Fabricator 74 (1983) 7: July, pp.
21–23.
[65] Satoh K, Ueda Y & Kihara H: ‘Recent trend of researches on restraint stresses and
strains for weld cracking’. IIW-Doc. IX-788-72, X-659-72, Osaka University,
1972. 39 p. /Welding in the World 11 (1973) 5/6, pp. 133–155.
 208

[66a] Matsui S: ‘Reaction stresses and weld cracking in restrained butt joints’. Doctoral
Thesis. Osaka University, Japan 1967. 109 p. (in Japanese)
[66b] Satoh K & Matsui S: ‘Reaction stress and weld cracking under hindered
contraction’. IIW-Doc. IX-574-68. Osaka University, 1968. 22 p. /Technological
Reprints of the Osaka University 17 (1967) 2, pp. 353–375.
[67a] Private information. Discussion, R Leggatt, P Hart & R Pargeter, TWI,
Cambridge, 21st February, 2000.
[67b] Leggatt RH: ‘Welding residual stresses’. Proceedings of the "5th international
Conference on Residual Stresses", Linköping, 16–18 June 1997 (Reprint). The
Welding Institute (TWI), Cambridge, United Kingdom 1997. 14 p.
[68] St John C & Gerberich WW: ‘The effect of loading mode on hydrogen
embrittlement’. Metallurgical Transactions 4 (1973) 2, pp. 589–601.
[69] Matsuda F, Nakagawa H, Shinozaki K, Tabaka S & Kihara H: Transactions of the
Japan Welding Research Institute (JWRI) 10 (1981) 2, pp. 183–191.
[70] Watkinson F: Welding Journal 48 (1969) 9, pp. 417-s – 424-s.
[71] Hart PHM & Watkinson F: Welding Journal 54 (1975): September, pp. 288-s –
295-s.
[72] McParlan M & Graville BA: IIW-Doc. IX-922-75. The International Institute of
Welding, 1975 /Welding Journal 55 (1976) 4: April, pp. 95-s – 102-s.
[73] Satoh K & Matsui S: ‘Notation system of joint restrain severity’. IIW-Doc. X-910-
78. Osaka University & Kawasaki heavy Industries, 1978. 6 p.
[74] Lee HW, Kang SW, Um DS: ‘A study on transverse weld cracks in thick steel
plate with the FCAW process’. Welding Journal 77 (1998): December, pp. 503-s –
510-s.
[75] Lynch S: ‘Mechanisms & kinetics of environmentally assisted cracking’. Special
Visiting Lecture. Helsinki University of Technology (HUT), Espoo, 30 May 2002.
Aeronautic & Maritime Research Laboratory, Defence Science & Technology
Organisation, Melbourne, Australia, 2002.
[76] Kuebler R, Pitrun M & Pitrun I: ‘The effect of welding parameters and hydrogen
levels on the weldability of high strength Q&T steel welded with FCAW
consumables’. Australian Welding Journal 45 (2000): 1st Quarter 2000 – Welding
Research Supplement, pp. 38–47.
[77] Satoh K, Terasaki T & Ohkuma Y: ‘Relationship between critical stress of HAZ
cracking and residual diffusible hydrogen content’. Journal of the Japan Welding
Society (JWS) 48 (1979) 4. pp. 248–252.
[78] Porter DA & Easterling KE: ‘Phase Transformations in Metals and Alloys’.
Chapman & Hall, London, U.K., 1992. 514 p.
 209

[79] Yurioka N: ‘Diffusion and accumulation of hydrogen in multiple-pass welds’.

Special Visitors Lectures. Nippon Steel Corporation Research Centre (NSC),
Kimitsu, Chiba, 2 July 1999. Nippon Steel Corporation, Chiba, Japan, 1999.
[80] Yurioka N & Nakamura H: ‘Investigation of the mass diffusion equation with
activity as a variable’. Journal of the Japan Welding Society (JWS) 48 (1979) 9,
pp. 726–730.
[81] Steigerwald EA, Schaller FW & Troiano AR: Transactions of the Metallurgical
Society of AIME 218 (1960) 10, p. 382.
[82] Naiki T & Okabayashi H: Journal of the Japan Welding Society (JWS) 40 (1971)
7, p. 675.
[83] Kohira K, Yatake T & Yurioka N: ‘A numerical analysis of the diffusion and
trapping of hydrogen in steels and steel weldments’. IIW-Doc. IX-951-76. The
International Institute of Welding. Nippon Steel Corporation, Chiba, Japan 1976.
[84] Sleptsov OI, Mikhailov VE & Smijan OV: IIW-Doc. IX-B-179-89. The
International Institute of Welding, 1989.
[85] Volkl J & Alfeld G: ‘Hydrogen in Metals’. Ed. G. Alfeld. Vol. 28 (1978), p. 321.
[86] Johnson EW & Hill ML: ‘The diffusivity of hydrogen in alpha iron’. Transactions
of the Metallurgical Society of AIME 218 (1960): December, pp. 1104–1112.
[87] Sykes C: ‘Hydrogen in steel manufacture’. Journal of the Iron & Steel Institute
(1947): June.
[88] Andersson BAB: Transactions of AIME / Journal of Engineering Materials and
Technology 102 (1980) 1, p. 64.
[89] Phragmen G: Jernkontorets Annaler 128 (1944), pp. 537–. (in Swedish)
[90] Lange G & Hoffman W: ‘Relation between hydrogen up-take and porosity in
iron’. Archiv für das Eisenhüttenwesen (Steel Research) 37 (1966) 5, pp. 391–397.
[91] Oriani RA: ‘The diffusion and trapping of hydrogen in steel’. Acta Metallurgica
18 (1970) 1, pp. 147–157.
[92] Bockris JO'M & Weck W: ‘The effect of stress on the chemical potential of
hydrogen in iron and steel’. Acta Metallurgica 19 (1971), pp. 1209–1219.
[93] Terasaki T, Akiyama T & Ohshita S: Quarterly of the Japan Welding Society
(JWS) 4 (1986) 2, p. 378.
[94] Takahashi E, Iwai K & Horitsuji T: Journal of the Japan Welding Society (JWS)
49 (1980) 2, p. 129.
[95] Terasaki T, Yamashita Y & Satoh K: Journal of the Japan Welding Society (JWS)
48 (1979) 9, p. 678.
[96] Private information. Discussion, R Gee, Leeds University, Leeds, U.K., November
1999.
 210

[97] Private information. Discussion, M Swallow, Leeds University, Leeds, U.K.,

November 1999.
[98] Mota JMF & Apps RL: Welding Journal 61 (1982) 7: July, pp. 222-s – 228-s.
[99] Allen DJ, Chew B & Morris P: Welding Journal 61 (1982) 7: July, pp. 212-s –
221-s.
[100] Takahashi E, Iwai K & Horitsuji T: ‘Relation between occurrence of transverse
cracking’. Journal of the Japan Welding Society (JWS) 48 (1979), pp. 855–872.
[101] Karppi R & Nevasmaa P: ‘Contribution to comparison of methods for determining
welding procedures for the avoidance of hydrogen cracking’. IIW-Doc. IX-1673-
92. VTT Publications 107. Technical Research Centre of Finland, Espoo 1992. 37
p. + Apps. 12 p.
[102] Ornig H, Schutz H & Klug P: ‘Comparison of methods to determine the preheat
temperature for high-strength weld metal’. Welding in the World 41 (1998), pp.
144–148.
[103] Beres L, Beres Z & Irmer W: ‘New equation for calculating the MS temperature’.
Schweissen und Schneiden / Welding and Cutting 46 (1994) 8, pp. E128–E130.
[104] AWS D1.1:1988: ‘Structural welding code – steel’. American Welding Society
(AWS), 1988.
[105] BS 5135:1984: ‘Process of arc welding of carbon and carbon manganese steels’.
British Standards Institution (BSI), 1984.
[106] Takahashi E, Iwai K & Horitsuji T: ‘Prevention of the transverse cracks in heavy
section butt weldments of 2 1/4 Cr-1 Mo steel through low temperature postweld
heat treatment. Correlation between hydrogen concentration and practical welding
conditions’. Journal of the Japan Welding Society (JWS) 48 (1979) 10, pp. 92–99.
[107] Vuik J, van Wortel JC & van Sevenhoven C: ‘Application of very low yield
strength consumables in the root pass of weldments to avoid preheating’. Welding
in the World 33 (1994) 5, pp. 362–369.
[108] Boellinhaus Th, Hoffmeister H & Dangeleit A: ‘A scatter band for hydrogen
diffusion coefficients in micro-alloyed and low carbon steels’. Welding in the
World 35 (1995) 2, pp. 83–96.
[109] Devanathan MA & Stachurski Z: ‘The absorption and diffusion of electrolytic
hydrogen in palladium’. Proceedings of the Royal Society of London No A270.
United Kingdom, 1962. Pp. 90–102.
[110] McBreen J, Nasis L & Beck W: ‘A method for determination of the permeation
rate of hydrogen through metal membrane’. Journal of the Electrochemical
Society 113 (1966) 11, pp. 1218–1222.
 211

[111] Scoppio L: ‘Towards a Unified Approach on a European Scale of Electrochemical

Hydrogen Permeation Test Methodologies on Low Alloyed Steels’. Report No
EUR 16790 EN, Europe, 1997.
[112] Widgery D: Interim Report. ESAB, U.K., 2002.
[113] Yurioka N, Wakabayashi M & Motomatsu M: ‘Data sheet of properties of high
heat input SAW weld metals’. IIW-Doc. IX-1868-97. The International Institute of
Welding. Nippon Steel Corporation, Japan 1997.
[114] Lowe G & Bala SR: Welding Journal 60 (1981) 12: December, pp. 258-s – 268-s.
[115] Thibau R & Bala SR: Welding Journal 62 (1983) 4: April, pp. 97-s – 104-s.
[116] Widgery DJ: ‘Welding of high strength steels – weld metal hydrogen cracking’.
Esab Report TR 907. ESAB -SSAB Meeting, Oxelösund, 23–24 September 1998.
Esab Group (UK), United Kingdom 1998. 10 p.
[117] Håkansson K: ‘Submerged arc welding of Weldox 900E in 50 mm plate thickness
– Hydrogen cracking tests’. IIW-Doc. II-A-117-03. Kockums AB, Laboratory,
Sweden. The International Institute of Welding, 2003. 14 p.
[118] Åström H: ‘Tillsatsmaterial för svetsning av höghållfasta ståhl – utveckling och
problemområden’. Svetsen 60 (2001) 5, pp. 7–14. (in Swedish)
[119] Yurioka N: ‘Prediction of Weld Metal Strength (Report III)’. IIW-Doc. IX-03. The
International Institute of Welding, 2003. 31 p.
[120a] Nevasmaa P, Laukkanen A & Alhainen J: ‘Evaluation of Risk of Hydrogen
Cracking using a Combined Finite-Difference to Finite Element Analysis and
Coupling to Damage Mechanics’. Proceeding of the "9th International Conference
on the Mechanical Behaviour of Materials (ICM-9)". Geneva, Switzerland, 25–29
May 2003.
[120b] Laukkanen A, Nevasmaa P & Alhainen J: ‘Numerical Evaluation of Hydrogen
Cracking using Transient Finite Element Analysis and Concentration Dependent
Damage Mechanics Constitutive Model’. Proceeding of the "7th International
Seminar: Numerical Analysis of Weldability (NAW-7)". Graz–Seggau, 29
September – 1 October 2003. Austria: Institute for Materials Science, Welding
and Forming (IWS) & Graz University of Technology (TUG), 2003. 16 p. (to be
published)
[121] Kim HJ & Kang BY: ‘Effect of Microstructural Variation on Weld Metal Cold
Cracking of HSLA-100 Steel’. ISIJ International 43 (2003) 5, pp. 706–713.
[122] Alcãntara NG & Rogerson JH: ‘A Prediction Diagram for Preventing Hydrogen-
Assisted Cracking in Weld Metal’. Welding Journal 63 (1984) 4, pp. 116-s – 122-
s.
 212

[123] Widgery DJ: ‘Welding Consumable Development – Remaining Challenges’.

Proceedings of the "Welding Engineering Research Centre 2nd International
Conference on Recent Developments and Future Trends in Welding Technology".
(Reprint). Cranfield University, United Kingdom 2003. 10 p.
[124] Lee HW & Kang SW: ‘The Relationship between Residual Stresses and
Transverse Weld Cracks in Thick Steel Plate’. Welding Journal 82 (2003):
August, pp. 225-s – 230-s.
Appendices
 APPENDIX 1/1

Calculation methods predicting the required preheat Tcr

to avoid weld metal hydrogen cracking
1. Okuda-evaluation [2]

Tcr = 1.15 ∗ (5.24 RM + 277 log HD – 482)

where:
RM : tensile strength of weld metal (kgf/mm2)
HD : diffusible hydrogen content in the weld metal (measured by gas
chromatography method) (ml/100g DM)

validity range: 77 < RM < 110 kgf/mm2

1.0 < HD < 5.1 ml/100g
h < 40 mm (for single-bevel groove)
h < 80 mm (for double-bevel groove)

conversion formula: 1 kgf/mm2 = 1 MPa / 9.81

2. Yurioka-Yatake-evaluation [3]

Tcr = 120 + 120 ∗ log (HD/3.5) + 5.0 ∗ (aw – 20) + 8 ∗ (RM – 83)

where:
HD : diffusible hydrogen content of weld metal (measured by JIS glycerine test)
(ml/100 g deposited metal)
aw : weld metal build-up thickness (mm)
RM : tensile strength of weld metal (kgf/mm2)

validity range: 0.1 < HD < 40 ml/100 g DM

15 < aw < 30–40 mm
60 < RM < 90 kgf/mm2

Tcr = 120 + 120 ∗ log (HD/3.5) + 5.0 ∗ (aw – 20) – 0.05 ∗ (aw – 30)2 + 8 ∗ (RM – 83)

validity range: as above, except:

30 < aw < 50 mm
 APPENDIX 1/2
Tcr = 250 + 120 ∗ log (HD/3.5) + 8 ∗ (RM – 83)

validity range: as above, except

aw > 50 mm

conversion formulae: H(JIS) = (1/1.26) ∗ ( HGC – 2.19) ; acc. to Ref.[12]

H(JIS) = 0.67 H(IIW) – 0.68
H(JIS) = 0.67 H(IIW) – 0.80
H(JIS) = (0.77 ± 0.03) ∗ HGC – (0.71 ± 0.19)
H(JIS) = 0.77 H(IIW) – 0.47
 APPENDIX 2/1

Comparison between the preheat estimates T0 given by

EN 1011-2 Method B and the NSC Y-Groove cracking tests
for multipass weld metals

Table 2.1. Weld metals[12] with the tensile strength RM ≈ 800 MPa.
Weld CET aw HD T0 : T0 : ∆ T0
(ml/100g EN 1011-2[1] Y-Groove[12]
(%) (mm) DM IIW) (°C) (°C) (°C)
Y80C 0.32 40 2.0 78 75 ≈0
L80-1 0.31 20 47 75 –28
30 4.3 75 125 –50
40 95 175 –80
L80-2 0.30 20 56 125 –69
30 6.7 85 175 –90
40 105 225 –120
L80-3 0.29 20 69 150 –81
30 10.4 97 200 –103
40 116 > 200 > –100

Table 2.2. Weld metals[12] with the tensile strength RM ≈ 750 MPa.
Weld CET aw HD T0 : T0 : ∆ T0
(ml/100g EN 1011-2[1] Y-Groove[12]
(%) (mm) DM IIW) (°C) (°C) (°C)
L74 0.29 30 4.0 57 75 –18
L74-1 0.25 20 17 50–75 > –33
30 6.7 45 125 –80
40 65 175 –110
L74-2 0.28 20 58 – –
30 9.9 87 150 –63
40 106 200 –94

CET: carbon equivalent according to EN 1011-2 : 2001[1b]

aw: weld build-up thickness in the Y-Groove test
H D: weld diffusible hydrogen content (ml/100g DM) according to
ISO/IIW 3690-1977 E[57]
T0 : EN 1011-2: required preheat temperature according to EN 1011-2 Method B[1b]
T0 : Y-Groove: required preheat temperature according to the NSC Y-Groove
cracking test[12]
 APPENDIX 3/1

Results of the residual stress measurements at Helsinki

University of Technology, Finland

Fig. 3.1. The principles of the Ring-Core test method for determining weld residual stresses in
thick plate multipass welds. Plate thickness 40 mm.
 APPENDIX 3/2

Fig. 3.2. Results of the weld residual stress measurements using the Ring-Core method for the
300 mm long multipass welded specimen (HL40), plotted against the weld depth (mm). Stress
components (MPa) into the weld longitudinal direction (Sa), weld transverse direction (Sc)
and into the direction of 45° between them (Sb). Analysis made using two different damping
calculation functions of HUT Materials Technology Laboratory, Finland (upper figure) and
Siemens KWU, Germany (lower figure).
 APPENDIX 3/3

Fig. 3.3. Results of the weld residual stress measurements using the Ring-Core method for the
800 mm long multipass welded specimen (HL402), plotted against the weld depth (mm).
Stress components (MPa) into the weld longitudinal direction (Sa), weld transverse direction
(Sc) and into the direction of 45° between them (Sb). Analysis made using two different
damping calculation functions of HUT Materials Technology Laboratory, Finland (upper
figure) and Siemens KWU, Germany (lower figure).
 APPENDIX 3/4

Fig. 3.4. The principles of the hole-drilling test method for determining weld residual stresses
in thin plate 2-pass welds. Stress components (MPa) into the weld longitudinal direction (A),
weld transverse direction (C) and into the direction of 45° between them (B). Specimen and
weld length 800 mm, plate thickness 6 mm. Specimens fixed onto the table using bolted
clamps (Appendix 3/5, Fig. 3.5) and welded strong-backs (Appendix 3/6, Fig. 3.6).
 APPENDIX 3/5

Fig. 3.5.
 APPENDIX 3/6

Fig. 3.6.
 APPENDIX 4/1

Description of the procedure applied at the University of

Oulu for deriving DH values from the potentiostatic Ia–t
curve recorded during the electrochemical permeation test
The background of the electrochemical permeation method [5b, 36b]

During the electrochemical permeation test, hydrogen diffusion can be described by the
Fick ´s law:
st
j D gradC (1 law )
H
c
(1)
2 nd
D C ( 2 law )
H
t

If DH is independent on the hydrogen concentration over the distance x, then the Fick´s
equation can be written as:
∂ C
J = − D H
∂ x
(2)
∂ c ∂ 2C
= D H
∂ t ∂ x 2
Fick´s law can only resolved under the given boundary conditions.
Because the hydrogen permeation test can meet the Fick´s law boundary conditions,
the requirements for the test can be written as:

t = 0, C = 0 for 0 < x < L ; t > 0, C = C0 = constant for x = 0, C = CL = 0 for x = L.

Following this, the hydrogen diffusion coefficients can be calculated.

The principles of the determination of DH [5b, 36b]

The potentiostatic curve that displays current on the exit side of the specimen, Ia, as a
function of time t in the permeation test. Whilst Ic is the actual charging current applied
as constant to the entry side of the specimen, Ia is the recorded current from the reverse
(exit) side of the specimen and is hence proportional to the extent and velocity hydrogen
permeates through the specimen during the test. Thus, Ia is influenced by the hydrogen
diffusion coefficient DH of the given specimen. Consequently, it is Ia that shows
correlation with DH, not Ic.
The J/Jmax in the y-axis in Fig. 4.2 is directly proportional to the charging current
Ic/Ic(max) where Ic corresponds to C0, i.e., the mean value of hydrogen concentration on the
 APPENDIX 4/2
specimen entry side, see Fig. 1. As long as Ic is kept constant, the recorded Ia–t curve will
be an exponential curve. Different DH values in different materials then manifest
themselves as different shapes of the corresponding Ia–t curve.
Before deriving DH from the Ia–t curve, the s.c. onset point must be determined. In the
tests of the present study, the time Ic was applied to the specimen entry side was set as the
onset point. After applying Ic, the corresponding Ia–t curve can be recorded. The onset
point in the Ia–t curve is then defined as the original zero: t0 = 0, I0 = 0, from which the
previous part – being treated as an diffusion independent background, is omitted from the
entire transient curve.
There are different ways to calculate hydrogen diffusion coefficients using the Ia–t
curve. The value of DH differs between different calculations, because the Ia–t curve is
not an ideally exponential curve.
• D(t1/2) is defined to calculate the hydrogen diffusion coefficient by using the half of
the whole rising time from onset point (Ia = 0) to the maximum of Ia , and D(t1/2) =
0.138 L2 / (t1/2), where L is the thickness of specimen.
• D(t0.617max) is defined to calculate the hydrogen diffusion coefficient by using the
0.617 times of the total rising time from the onset point (T0 = 0, Ia = 0) to the
maximum of Ia , and D(t0.617max) = L2 / 6t(0,617max) .
• D(t1, t2) is defined to calculate the hydrogen diffusion coefficient by using two points
(t1, t2) in the total rising time from onset point (T0 = 0, Ia = 0) to the maximum of Ia ,
and D(t1, t2) = 0.25 L (t2 – t1) / t2 t1 ln[J2(t2 / t1)0.5/ J2].
• D(t1, t2, max) is defined to calculate the hydrogen diffusion coefficient by using two
points (t1, t2) and the maximum of Ia in the total rising time from the onset point
(T0 = 0, Ia = 0) to the maximum of Ia, and D(t2, t1, max) = L2 ln[(J∞ – J1) / (J∞ – J2)]
/ π2(t2 – t1).
• Dtb(tg) is defined to calculate the hydrogen diffusion coefficient by using the
breakthrough time tb which is the intersection of the time axis with the tangent line to
the permeation curve at its inflection point, and Dtb(tg) = L2 / 19.8 tb(tg).
• Dtb(exp) is defined to calculate the hydrogen diffusion coefficient by using the
breakthrough time tb which is the intersection of the time axis with the exponent line
to the permeation curve, and Dtb(exp) = L2 / 35.2 tb(exp).
Because the differences among the values calculated with different calculation
methods are far too great, a new method to calculate the hydrogen different coefficient
was developed.

Derivation of DH in the present study [5b, 36b]

In the present study, the relaxation time, i.e., the time regarded as being due to finite rate
constant for transfer of hydrogen atoms from the surface into the metal phase,
approximately equal to 10 sec on the basis of the present experiments, was applied in the
analysis. Usually, the point of the first maximum Ia peak detected after the original zero
was defined as the end point, see Fig. 4.2.
 APPENDIX 4/3
The recorded Ia–t curves were processed before they were applied for calculation of
DH according to the requirement of the Fourier´s solution. Values of Dt(0.1), Dt(0.2) and
Dt(0.25) were calculated corresponding to the ratio of J/Jmax to 0.1, 0.2 or 0.3,
respectively[20]. These were found to be more reliable than the commonly determined
values for the time lag, breakthrough time, half delay time, etc.[19, 21].
These values were determined as:
Dt(0.1) = 0.1 L2 / t
Dt(0.2) = 0.2 L2 / t
Dt(0.25) = 0.25 L2 / t
where L = thickness of the specimen; t = the relaxation time

Fig. 4.1. Concentration gradient in a plane sheet during a permeation test (assumption of
constant hydrogen concentration on the entry face) (CNRS).

Fig. 4.2. Characteristic time values for D calculation (CNRS).

Explanation:
J/Jmax in Fig. 4.2 is directly proportional to Ic/Ic(max).
 APPENDIX 4/4
Dt(0.1) = 0.1L2/t
Dt(0.1) = 0.1L2/t
Dt(0.1) = 0.1L2/t
L: the thickness of specimen
t: is the time.
For example, how to calculate hydrogen diffusion coefficient with the record data of Ia
to calculate hydrogen diffusion coefficient:
Ic = 20–24 mA
Then get a set of Ia data:
Table 4.1.
Times (s) Ia (uA)
0 15.27
10 15.27
12 15.27
22 15.27
32 15.28
42 15.31
52 15.39
62 15.50
72 15.60
82 15.73
92 15.85
102 15.96
112 16.06
122 16.16
132 16.25
142 16.32

1) determining the maximum value of Ia: 16.85

2) the onset point value Ia: 15.27
3) the difference of the maximum and the onset value is: 16.85 – 15.27 = 1.57
4) the corresponding value J1(0.2924Jmax): 15.27 + 1.57 x 0.2924 = 15.735
5) determining the time t1:
15.735 is between 80 second (Ia = 15.73) and 90 second (15.85)
t1 = 80 + 2 (15.735 – 15.73) x 10/(15.85 – 15.73) = 82.5
6) determining t2, t3 (same as t1)
t2 = 157
t3 = 194
 APPENDIX 4/5
7) determining the relaxation time: 10 seconds
8) calculate the diffusion coefficient:
D(0.1t) = 0.1 L2 / t1 = 12.45 ⋅ 10–6 (cm2/s)
D(0.2t) = 0.2 L2 / t2 = 12.28 ⋅ 10–6 (cm2/s)
D(0.1t) = 0.25 L2 / t3 = 12.26 ⋅ 10–6 (cm2/s)
where L = thickness of the specimen; 0.95 mm.
Also with the data in Table 4.2 can obtain Ia–t curve:

The hydrogen diffusion coefficient can also be calculated with other ways according to
Fig. 4.2. Notice that onset Ia in the curve is not zero (Ia is 15.27, time has been 0).
 APPENDIX 4/6

Table 4.2.
Times (s) Ia (uA)
152 16.38
162 16.47
172 16.51
182 16.56
192 16.58
202 16.63
212 16.66
222 16.70
232 16.71
242 16.74
252 16.76
262 16.78
272 16.79
282 16.79
292 16.81
302 16.82
312 16.83
322 16.84
332 16.85
342 16.86
352 16.85
362 16.83
372 16.84
382 16.84
392 16.83
402 16.83
the changing onset point: t = 0

If adopting the formula:

D = 0.138L2/t(1/2)
L = 0.95 mm
t(1/2): mean half of the time from the onset point to the maximum (Ia)
the maximum of Ia: 16.86, and the corresponding time: 342 s
the D = 0.138 x 0.0952/(342/)2 = 7.283 x 10–6 (cm2/s).
If adopting other formulas to calculate D, the time (t) must be determined, if the
following formula is adopted,
D = L2/19.8 tb(tg) ,
the time tb(tg) is only determined from the relation curve of Ia with t, because Tb(tg) is
only determined by the intersection of the time axis with the tangent line to the curve Ia–t
at the inflection point. Sometimes because of the unidea permeation curve, it is not easy
accurately to do so.
 APPENDIX 5/1

Y-Groove hydrogen cracking test results of L-80, L-75 and

L-60 weld metals made at Nippon Steel [12]

Filler Weld Weld Weld

material RM HD aw Preheat & Interpass temperature (°C)
ml/100 g
MPa (IIW) mm RT 50 75 100 125 150 175 200
Y-80C 850 2.0 40 z z { {
20 z } {
L-80 800 4.3 30 z } {
40 z z { {
20 z } {
L-80 825 6.7 30 z z z { {
(1) 40 z z z z
20 z z z {
L-80 815 10.4 30 z z z z {
(2) 40 z z z

L-74 735 4.0 30 z { { {

20 z
L-74 745 6.7 30 { {
(1) 40 z z { {
20 z
L-74 745 9.9 30 { {
(2) 40 z z {

L-60 620 5.1 30 { { { {

L-60 645 11.5 30 z { { {
(1)

z : cracked specimens
} : both cracked and crack free specimens
{ : crack-free specimens
 APPENDIX 6/1

Examples of typical smaw weld metal microstructures –

optical microscopy made at the University of Oulu [36a]

Fig. 6.1. All-weld metal microstructure of NSC15 specimen: arc energy 1.5 kJ/mm (50 x).

Fig. 6.2 - All-weld metal microstructure of NSC30 specimen: arc energy 3.0 kJ/mm (50 x).
 APPENDIX 6/2

Fig. 6.3. All-weld metal microstructure of NSC50 specimen: arc energy 5.0 kJ/mm (50 x).

Fig. 6.4. All-weld metal microstructure of VTT1 specimen: arc energy 3.0 kJ/mm (50 x).
 APPENDIX 6/3

Fig. 6.5. All-weld metal microstructure of VTT2 specimen: arc energy 3.0 kJ/mm (50 x).

Fig. 6.6. All-weld metal microstructure of OU1 specimen: arc energy 2.4 kJ/mm (50 x).
 APPENDIX 6/4

Fig. 6.7. All-weld metal microstructure of OU2 specimen: arc energy 2.4 kJ/mm (50 x).

Fig. 6.8. All-weld metal microstructure of OU3 specimen: arc energy 2.4 kJ/mm (50 x).
 APPENDIX 7/1

Chemical composition of the smaw and saw weld metals

associated with the cracking tests
HICC 2/01
Table 7.1. Chemical compositions of the SMAW electrodes used for experimental
multipass weld metals.
APPENDIX 7/2
APPENDIX 7/3
APPENDIX 7/4
APPENDIX 7/5
APPENDIX 7/6
APPENDIX 7/7
 APPENDIX 8/1

Comparison between the Rp0.2/RM ratios calculated

according to the nominal weld metal strength data and
actual measured strength values

Weld metal (WM) Rp0.2/RM ratio Rp0.2/RM ratio

acc. to nominal WM acc. to measured WM
strength data strength properties*)
OK 48.08 0.90 0.86–0.87
OK 74.78 0.92 0.85–0.87
Nittetsu L-80 0.90–0.92 0.75–0.92
OK 75.75 0.92 0.84–0.88
Atom Arc 12018 0.90 0.87
OK 75.78 0.95 0.92
OK Autrod 13.27 0.86 0.85–0.88
Filarc PZ 6138 0.91 0.93–0.96

*) measured data: SMAW welds in the arc energy range: 1.6–5.0 kJ/mm
SAW welds with the arc energy of: 5.0 kJ/mm
FCAW welds with the arc energy of: 0.7 kJ/mm
 APPENDIX 9/1

Comparison of the calculated RM estimates according to the

estimation methods by VTT and ELGA (Åström) [118]
Arbitrary weld metal Pcm range

Applied Pcm RM RM Difference between the

value acc. to acc. to two RM estimates
VTT's estimation Elga's [118] estimation
(%) (MPa) (MPa) (%)
acc. to
Eqs (45a) and (48b)
0.165 695 672 3.3
0.215 836 800 4.3
0.250 935 889 4.9
0.295 1063 1004 5.6

* VTT's estimation method:

RM = 2.85 ∗ HV(ave) ; HV(ave) = 993 ∗ Pcm + 80

RM = 2 830 ∗ Pcm + 228 (MPa)

* ELGA's estimation method (Åström)[118]:

RM = 2 556 ∗ Pcm + 250 (MPa)

 APPENDIX 10/1

Weld t100 cooling time estimation diagram based on the JIS

y-groove Tekken tests acc. to JSSC [3b]

Fig. 10.1. A new cracking parameter for welded steels.

 APPENDIX 11/1

Flow-charts of the procedure for the avoidance of weld

metal hydrogen cracking in multipass weld metal

Step 1
Calculation of the worst case estimate of Hcr
- crack / no crack

Weld metal Weld metal Weld metal

no no
chemical composition hardness strenght

yes Determine: yes

yes no
generate data

Actual weld metal data? no True weld

True weld metal metal data
no maximum hardness no Rp0.2, RM
yes HVmax?

Calculate: Pcm yes no yes

Eq. (43) CET

Calculate: HVave Calculate:

Eq.(45) HRcr = f (HVmax)
Eq. (59)

Calculate: RM Rp0.2 available?

Eq.(48b) Calculate:
Hcr = f(Pcm)
Eq. (56) no yes
Choose: appropriate Rp0.2/R M
-ratio acc. to nominal values,
App. 8 or set value: 0.9

Set: Rp0.2 = resL

Calculate:
Hcr -estimate (s)
Hcr = f ( resL),Eq.(54b)

Take: lowest No WM Compare: WM

Hcr value cracking yes no cracking
Hcr > HD
risk risk!!!

Apply: Reduce: HD
T0/Ti = 100°C
Go to Apply: elevated preheat /
Step 1 interpass temperature
flow2_1.dsf

Go to
Step 2
 APPENDIX 11/2

Step 2
- calculation of a safe estimate of T0/Ti

Weld hydrogen Weld metal Weld Weld build-up

content HD chemical composition hardness HV thickness aw

no no

yes Determine yes no yes yes no

generate
data

Check: HVmax Check: 20 mm <

Check: limits for Calculate: CET or HVave aw < 40 mm
Level 1 (HD, CET) Eq. (43b)

HVave yes
yes
Accept: conservative
no
estimate of T0/Ti
Limits
yes no HVmax
violated
yes
no
no

Apply: Level 1 Apply: Level 2 Apply: Level 2 Apply: Level 2

estimation scheme Option 1 Option 2 Option 3
Eq. (62) Eq. (67b) Eq. (68a)
Eq. (68b)

Prediction as a safe
T0/Ti estimate!
flow2_2.dsf