an appraisal of the tekken test
DESCRIPTION
Ensaio de Trincas por hidrogênioTRANSCRIPT
University of WollongongResearch Online
University of Wollongong Thesis Collection University of Wollongong Thesis Collections
1986
An appraisal of the Tekken testAlan Lionel WingroveUniversity of Wollongong
Research Online is the open access institutional repository for theUniversity of Wollongong. For further information contact the UOWLibrary: [email protected]
Recommended CitationWingrove, Alan Lionel, An appraisal of the Tekken test, Doctor of Philosophy thesis, Department of Metallurgy and MaterialsEngineering, University of Wollongong, 1986. http://ro.uow.edu.au/theses/1606
"AN APPRAISAL OF THE TEKKEN TEST"
A THESIS SUBMITTED IN FULFILMENT OF THE
REQUIREMENTS FOR THE AWARD
OF
THE DEGREE OF DOCTOR OF PHILOSOPHY
FROM
THE UNIVERSITY OF WOLLONGONG
BY
ALAN LIONEL WINGROVEA.S.TC.BSc. MAppSci
DEPARTMENT OF METALLURGY
and
MATERIALS ENGINEERING
MAY 1986
CANDITATE'S CERTIFICATE
This is to certify that the work presented in this thesis was carried out by
the candidate in the laboratories of the Depertment of Metallurgy and
Materials Engineering of the University of Wollongong and has not been
presented to any other university or institution for a higher degree.
Alan L.WingroveA.S.T.C. B.Sc. M.App.Sc.
ACKNOWLEDGEMENTS
The auther would firstly like to thank the Australian
Welding Research Association for the financial support that has been
given to this research. Throughout this work, virtually every member
of the staff of the Department of Metallurgy and Materials Engineering
has been helpful in some way and this has been greatly appreciated.
In particular thanks are due to the Head of Department,
Associate Professor N. F. Kennon for the use of the laboratory
facilaties and for his and Dr. D. P. Dunne's help and encouragement
during their supervision of the research.
Thanks are also due to Mr. F. Groves and the workshop staff
for design and construction of equipment and to Mrs. Ann Webb for
her assistance during the arduous task of preparing this thesis.
The donations of both the steel plate by B.H.P. Steel Slab and
Plate Products Division and the electrodes by Welding Industries of
Australia Pty. Ltd. are also greatfully acknowledged.
SYNOPSIS
Literature relevant to the weldability of steel by manual metal arc welding
has been reviewed. In particular, the process of manual metal arc welding has been
examined and current theoretical models that have been developed to predict the
structure of the weldment are discussed. Weld defects, particularly hydrogen
assisted cold cracking of the heat affected zone (HAZ) are discussed and a review of
weldability tests relevent to H A Z cracking has been made. The Japanese developed
Tekken Test has been examined in detail and the relevance of Cracking Index
measurements has been analysed.
The effects that variations to the extrinsic welding parameters of voltage,
current, and speed on Cracking Index as determined by the Tekken Test have been
examined. Intrinsic welding variables of microstructure and restraint stress have
been examined using optical and electron microscopy and a restraint stress
measuring technique. Cracking Index results have been related to the extrinsic
welding parameters by variations caused to the intrinsic variables of microstructure
and restraint stress.
It was found that increases in welding speed and voltage increased the
value of the Cracking Index and increases in welding current decreased the Cracking
Index value. The interaction of these effects produced a variation of results for
welding at constant heat input.
All three extrinsic parameters were found, independently and collectively
to affect the microstructures of both the H A Z and the weldmetal. It was found that
increases in welding voltage and current and decreases in welding speed increased
the volume fraction of proeutectoid ferrite in the H A Z with an associated decrease in
H A Z cracking. Cracking of the weldmetal was found to be related to the ferrite
morphology; fine lath (acicular) ferrite was found to be the microstructure least
susceptible to cracking. Formation of fine lath ferrite microstrucures was enhanced
by low welding speeds. From the various combinations of weldmetal and H A Z
microstructures that could be achieved a range of Cracking Index values could be
produced.
The influence that root gap of the test piece, the composition ( carbon
equivalent) of the steel plate, and the classification and manufacturing source of the
electrodes had on Cracking Index has also been studied. Increasing root gap and
carbon equivalent increased the value of Cracking Index although in the latter case
the relationship varied with welding conditions. Cracking Index was also found to
vary with electrodes of different classifications and from different manufacturing
sources. It is proposed that these Cracking Index variations are attributable to
strength incompatibilities and differences in electrode flux compositions
respectively.
It is concluded that because of the variation of results, often with
considerable scatter, produced by the interaction of all the variables, the Tekken
Test is not valid as a general method for determining weldability, but rather
provides a useful comparative test of the responses of different steels and electrodes
to welding under strictly specified welding conditions.
TABLE OF CONTENTS
CHAPTER 1 INTRODUCTION 1
CHAPTER 2 THE MANUAL METAL ARC
2.1 Introduction 9
2.2 Covered Electrodes 11
2.3 The Weld 15
2.3.1 The Weldmetal 18
2.3.2 The Heat Affected Zone 22
CHAPTER 3 WELD DEFECTS
3.1 Introduction 31
3.2 Solidification Cracking 32
3.3 Liquation Cracking 33
3.4 Reheat Cracking 33
3.5 Lamellar Tearing 34
3.6 Chevron Cracking 35
3.7 Hydrogen Assisted Cold Cracking 36
3.7.1 The Effect of Hydrogen 36
3.7.2 Microstructure 40
3.7.3 Stress 45
3.7.4 Predicting HAZ Cracking 51
C H A P T E R 4 WELDABILITY TESTS
4.1 Introduction 60
4.2 The Implant Test 62
4.3 The Tensile Restraint Cracking Test 63
4 .4 The Rigid Restraint Cracking Test 64
4.5 The Lehigh Restraint Test 65
4.6 The Controlled Thermal Severity Test 66
4.7 The Tekken Test 68
CHAPTER 5 SCOPE OF THE PRESENT WORK 75
CHAPTER 6 EXPERIMENTAL
6.1 Introduction 80
6.2 Materials 81
6.3 Specimen Preparation 82
6.4 Welding Equipment and Procedures 85
6.5 Determination of Cracking Index 87
6.6 Metallography 90
6.6.1 Optical Metallography 90
6.6.2 Electron Metallography 91
6.7 Restraint Stress 91
C H A P T E R 7 E X P E R I M E N T A L R E S U L T S
7.1 Introduction 94
7.2 The Effect of Welding Variables on Cracking
Index
7.2.1 Introduction 96
7.2.2 The Effect of Voltage 97
7.2.3 The Effect of Current 97
7.2.4 The Effect of Speed 97
7.2.5 The Effect of Constant Heat Input 98
7.2.6 Discussion of Results 99
7.3 The Inter-relationship Between Welding Variables,
Microstructure, Stress and Cracking Index
7.3.1 Introduction 101
7.3.2 The Effect of Voltage 102
7.3.3 The Effect of Current 113
7.3.4 The Effect of Speed 118
7.3.5 The Effect of Constant Heat Input 123
7.4 Other Factors Affecting Cracking Index
7.4.1 Introduction 127
7.4.2 Root Gap 127
7.4.3 Material Composition 129
7.4.4 Electrode Type and Manufacture 133
CHAPTER 8 DTSCTISSTON AND CONrT TT.STnxrs
APPENDIX
REFERENCES
1
CHAPTER O N E
INTRODUCTION
The American Welding Society(1) defined a weld as "a coalescence of
metal wherein coalescence is produced by heating to a suitable temperature, with or
without the application of pressure and, with or without the use of a filler metal.
The filler metal either has a melting point the same as the base metal or has a melting
point below that of the parent metal but above 800°F."
The process of welding dates back to the Middle Ages (2). Forge welding
throughout that time was used in the production of armor, sword blades, and other
military equipment. With forge welding, the two steel parts to be joined were raised
to near white hot in the forge fire, fluxed with sand, and hammered together. The
hammering extruded the liquid oxide, or slag, bringing together clean metallic
surfaces to form a metallic bond - a process of coalescence. Forge welding is
essentially a solid phase process and it was not until the end of the 19th century that
fusion welding emerged.
However, in terms of large structures, timber and stone remained the
major materials for construction of ships, buildings and bridges. Around the middle
of the 19th century ironmaking was developed and for a short period of time cast
iron was used for some structural purposes. However, by the early 1900's steel
production had been developed, and structures of steel began to appear. Alloys of
iron, carbon and other elements emerged as a versatile material and today the scope
and magnitude of steel production and usage is staggering. Gas pipelines of over
2
6000km and supertankers of over 100,000tonne are examples of modern uses. The
fabrication of such structures is in fact a direct result of the development of fusion
welding as a steel fabrication process.
Although gas , electric arc and resistance welding techniques began to
be developed soon after Bessemer and Siemens had developed their steelmaking
processes, riveting and bolting remained the predominant method of joining steel
sections for almost a further 40 years.
During this time significant advances in metal arc welding were made
and from the bare wire electrodes of Slavianoff in 1890 and Kjellberg in 1907(26),
Strohmerger in 1911 developed covered electrodes to be followed in 1917 by Jones
with powdered metal in the flux coating. Electrodes were developed to weld
stainless steel in 1920 and by the 1930's improvements in the flux coatings had led
to the production of weld deposits with mechanical properties commensurate with
those of the steel plate.
Nevertheless, metal arc welding was looked upon with
considerable scepticism and opposition prevailed up to, and even after, the 2nd
World War. The failure of welded structures was thoroughly exploited by the
opponents of welding as evidence of the inferior nature of welding as a joining
process. The loss of the British built ship the Joseph Medill in 1935 and the
collapse of the all welded Hasseft Bridge in Belguim in 1938 were examples to be
cited as proof that welding was unreliable. Similarly, the exaggerated comment that
followed the failures of the World War 2 American built T2 tankers led many people
to believe, for a time, that cracking was to be expected in any welded ship or
structure. The critics, however completely ignored the cracking, or complete failure
3
in riveted structures such as that which occurred with the all riveted U S built ship
the Oakley L. Alexander(27). The investigation into the T2 failures finally led the
United States Admiralty Ship Welding Committee to state that "welding as a process
for building ships has been entirely vindicated. Given sound design, good
workmanship and tough steel, the reliability of welded ships is beyond question".
Yet, even today there are still a few critics.
Defects in welded joints can, however, lead to catastrophic failures. The
weld joint may not fail but the presence of small defects within the weld zone may
lead to failure by other fracture mechanisms that may occur in the steel plate. For
example, in the case of the Alexander Kielland (10), an oil drilling rig in the North
Sea, a 5 m fatigue crack propagated from a small toe crack in a 6 m m fillet weld
which joined a non load-bearing flange plate to one of the main braces; 123 people
died when the platform overturned.
The knowledge that defects in the weldmetal and the heat affected
zone (HAZ) adjacent to the weld can lead to failure of the entire structure has led to
the development of welding procedures and steel compositions and cleanliness that
are directed towards eradicating or minimising the possibility of such defects. In
steel production, effective deoxidation processes, low sulphur and phosphorus
levels, high manganese to sulphur ratios and control of the shape and size of
inclusions, particularly M n S , has led to improved steels quality and reduced welding
defects.
Coupled with the demand for defect free weld zones has been the
demand for welded structures with larger physical dimensions to operate in more
hostile environments and so be constructed of more sophisticated alloys. The
4
developments of super tankers in 1970, commercial nuclear reactors, offshore
drilling rigs and platforms in the North Sea in 1972, and the development of gas
fields in the U S S R in 1975 are examples. The increased demand for improved
mechanical properties is possibly best demonstrated by the example of materials
specifications for U.S. submersibles(ll,12). Prior to the mid 1940's combat
submarines were constructed from low carbon steels having a yield strength (YS) of
220MPa. Between 1943 and 1958 high tensile steels with a Y S of 340MPa were
used. These were replaced in 1958 by quenched and tempered steels with a
minimum Y S of 550MPa. In current use are the H Y 1 0 0 grade steels with a Y S of
690MPa and the next generation of submarines are to be constructed from steels of
the type H Y 1 3 0 having a Y S of 900MPa. In the future the U.S. Navy plans to
build submersibles with a proposed depth capability of 6000m requiring steels with
a Y S of 1200MPa, these are the H Y 1 8 0 type steels (64).
These super strong, ultra clean, fine grained, homogeneous,
fracture tough steels when joined by welding in the fabrication process, are, in the
H A Z of the fusion weld, subjected to a thermal cycle which causes rapid grain
growth, phase transformations, the dissolution and reprecipitation of carbides,
nitrides etc, gas absorption and diffusion coupled with the development and
imposition of a complex stress state. It is to be expected then that adjacent to the
weldmetal, in the heat affected zone (HAZ) some deterioration or variation to parent
plate mechanical properties would occur and could be delaterious to the subsequent
performance of the structure. The extent of this deterioration in the H A Z has led to
the term weldability.
Weldability has been defined (3,4) as " the capacity of a metal to
be welded, under fabrication conditions imposed, into a specific, suitably designed
5
product and to perform satisfactorily in the intended service". The precise meanings
of such terms as good weldability or poor weldability have been the topic of many
committee discussions. Basically there are three extrinsic factors that can be
important in determining the weldability of a steel, namely;
(i) the steel composition,
(ii) the joint,
(iii) the welding process,
The complexity of the interaction between these variables is demonstrated in Fig.
1.1(28) which relates to cold cracking. It is generally agreed that if conditions are
critical and cause an undesirable interaction between the intrinsic parameters of
microstructure, hydrogen content , and tensile stress, cracking will result. The
nature and number of possible combinations of the interactions involved makes
quantification difficult. However, to assess weldability in relation to cold cracking
would require quantitative knowledge of the extrinsic and intrinsic parameters.
The operative clause in the definition is "under fabrication conditions
imposed". Environmental conditions in Saudi Arabia would be different to those in
Alaska so that ambient temperatures and atmospheric moisture could influence the
ultimate weldment properties. Similarly a steel m a y be readily weldable in a
fabrication shop using submerged arc welding, but be difficult to repair by manual
arc welding on location in the North Sea. A s a generalisation then, it could be said
that a steel possesses good weldability if it can be welded reliably and economically.
All steels are weldable provided the necessary welding conditions can be achieved;
the cost of achieving a defect free weld can, however, be prohibitative.
Nevertheless the implication that weldability is a measurable property or can
be derived from a measurable mechanical property, such as m a x i m u m hardness
6
produced by water quenching, or chemical composition, has prompted the
development of many weldability tests(6) and empirical equations (28) and relative
weldability gradings have been proposed(7,13,14). Both weldability tests and
equations are, however, prone to relate to conditions that cause one particular
classification of weld defect, for example carbon equivalent formulae and the C T S
weldability test endeavor to respectively predict and determine conditions for heat
affected zone hydrogen assisted cold cracking. Perhaps it would be more preferable
if steels were assessed for each type of defect, then rated on the most likely type of
defect to be present under the particular welding conditions.
Such a position could, however, impose on the steelmaker an ever
increasing range of specifications with a demand for minimum conditions to be
applied where they are not warranted. Nevertheless the current position is that
designers endeavor to specify numerically the weldability of a steel according to
empirical equations or gradings and there is a growing tendency to relate the results
of weldability tests, often field tests, to the specified grading.
Weldability tests can have three purposes;
(i) to determine the factors that determine good weldability,
(ii) to develop procedures to obviate defects in the welded joint, and
(iii) to compare the relative weldability of various steels.
Many weldability tests were developed to either explain a particular failure
or to prevent a similar occurrence under similar circumstances. In most cases the
tests relate to a complex conjunction of conditions and material properties so that
when test results derived from a particular test are used for specifications, they may
either represent conditions that are too severe or too mild in relation to the actual
7
service condition. This is particularly the situation with self restraining weldability
tests where difficulties also arise with reproducibility. The Lehigh Restraint
Test(16) has been applied as a workshop test for pipe materials (17) and some
correlation between propensity for cracking as indicated by the test and field
experience was claimed (18). Others found difficulties in reproducibility and
interpretation (19).
The Tekken Test(20), developed by the Japanese, has been used
extensively in that country for assessing steel plate susceptibility to cold
cracking(21,22) and to generate data for the development of empirical
equations.(22). Interest in Australia in the Tekken Test led, in 1969, to an
investigation by Hensler et al. (23) w h o found that provided stringent conditions
were met the results could be reproducible. Subsequently an interlaboratory
comparison of the weldability of two steels was carried out by the Australian
Welding Research Association ( A W R A ) . The results of this comparison had a level
of scatter that made comparison of the weldability of the two steels impossible and
cast doubt on the test as a quantitative measure of weldability. Several Japanese (25)
workers have found similar scatter. The A W R A thus concluded that, for the time
being , the Tekken Test not be included in an Australian Specification ASB301 as a
weldability test as had been intended and that a further assessment be carried out
using welding equipment suitable for automatic deposition of the test weld. It was
on this basis that the present program of research was initiated and the results of this
work are covered in the second part of this thesis.
The Tekken Test is a weldability test, used to examine the susceptibility of
a steel to heat affected zone (HAZ) cold cracking, i.e. delayed cracking of the H A Z
after the weld has cooled near to or has been at ambient temperatures for some
THE STEEL THE JOINT THF. WELDING PROCESS (THE EXTRINSIC VARIABLES)
CHEMICAL MECHANICAL THICKNESS SITE TYPE & HEAI FJ^CTROPE ANALYSIS PROPERTIES A, CONATIONS DESIGN INPJII
C O L D C R A C K I N G
FIG. 1.1 Schematic representation of the interaction of variables associated with the manual metal arc welding process (28)
8
period of time. The factors that are known to be associated with such cracking are r
the formation of a detrimental microstructure, usually martensite, the presence of
hydrogen, and the imposition of stress (see Fig. 1.1). The literature review in
Chapters 2 ,3 discusses and relates these factors. Chapter 4 deals with weldability
testing and in particular the Tekken Test. The overall literature of Chapters 2 to 4
inclusive thus serves as a background to Chapter 5 in which the scope and aims of
the present research are presented.
9
CHAPTER TWO
THE MANUAL METAL ARC WELD
2.1 INTRODUCTION
Manual metal arc welding (MMA) is a particularly complex process.
The arc processes which occur when the anode and cathode are refractory metals,
such as tungsten, are relatively well understood. However, when the arc associated
with flux coated electrodes is considered, the various physiochemical, metallurgical,
and electrical reactions that accompany the transfer of metal and slag are complex.
By way of introduction a brief discussion will be provided of the electric arc
associated with refractory metals and inert gas where no metal transfer is involved.
An electric arc has been defined (30) as " a discharge of relatively large
current and relatively low voltage, especially with a low cathode drop". Electrical
conduction takes place in the arc through a gaseous column (see Fg.2.1) called the
plasma, which has high electrical conductivity. The plasma contains a radiating
mixture of free electrons, positive ions, and highly excited electrically neutral atoms.
Because the conductivity of the plasma between the electrodes is maintained by
thermal ionization, the temperatures must be high. It is generally agreed (31) that the
zones of electrical contact between the plasma and the electrodes are quite different to
the general plasma column, so that the arc can be divided into three parts, see Fig.
2.1 (31). Because there exists a higher concentration of charge carriers at the anode
and cathode a non linear potential (or voltage) distribution occurs. Similarly,
temperature gradients occur at the anode and cathode relative to the plasma. High
electrical field strengths exist at the anode and cathode due to the space charge
10
distribution and high currents induce magnetic fields which compress the plasma.
Axial pressure gradients may be produced in the plasma stream which transport the
heat (and material, in the case of consumable electrodes) away from the electrode
(the cathode in Fig.2.1) and to the work (the anode).
It has been found that if the electrode is consumable, the geometry of the
tip of the electrode is unstable as metal droplets are transferred, however, with
tungsten electrodes using adequate cooling it has been possible to reduce these
instabilities (32,33,34) and to study the various current / voltage relationships and
their relationships to the arc. In this respect the tungsten arc inert gas system has
been particularly suitable for study and has been well documented(36).
When the electrode tip is not cooled, heating and melting may occur and
metal may transfer from the electrode to the work and this occurs with consumable
electrodes. The transfer of material may occur by one of two processes. Short
circuiting occurs when a droplet that has formed touches the weld puddle while still
partly attached to the electrode. At this instant metal vaporization occurs in the neck
connecting the droplet to the electrode thereby severing the droplet, and the arc is
re-established. The second process is by droplet transfer i.e.,the molten drop
leaves the electrode and travels in free flight inside the arc plasma to the weld
puddle. The droplet velocity in the arc has been measured and found to be constant
over the arc length and between 380mm/sec and 1300mm/sec.(37), depending on
welding conditions. The velocity of a metal droplet falling under gravity has been
calculated as 32.5mm/sec which indicates that gravity forces alone do not cause
metal transfer.
It has been shown that for M I G metal transfer the electromagnetic
CATHODE
E - TOTAL ARC VOLTAGE
Ei ANODE POTENTIAL
Ep-PLASMA POTENTIAL
Er-CATHODE POTENTIAL
FIG.2.1. Schematic representation of the metal arc indicating the various sections of the arc plasma(35).
T A B L E 2.1.
The ionization potential of a number of elements and compounds(40)
Ca Rb K Na Ca Ti Fa O,;0
3.88 4.16 4.34 5.14 6.11 6.84 783 12.5; 13.5
H;H, H,0 CI, CO; CO, N.N, Ar F He
13.5; 15.4 13.2 13.5 14.1; 14.3 14.5; 15.6 15.7 18.6 24.5
11
force on a 1.5mm droplet m a y be as high as 200xlO"5N compared with the
gravitational force of 25xlO"5N. In the case of covered electrodes used in M M A
welding the forces acting on the droplet may vary depending on the constituents in
the flux. Any one of five forces may dominate; these are surface tension and
viscosity, gravity, electromagnetism (pinch effect), hydrostatic pressure, the action
of the plasma jet, and metallurgical reactions.
2.2 COVERED ELECTRODES
Manual metal arc welding processes are complex because of the nature of
the arc column brought about by the nature of the constituents that comprise the flux
coating. For covered electrodes the voltage difference in the column is small
compared with the difference in the cathode and anode regions^Electrons, because
of their size, are the most mobile particles in the arc column and thus contribute
most to the current flow. Temperatures in the arc and the arc stability are functions
of the ionization of the constituents in the flux coatings. The ionization potentials
(eV) of a number of constituents of welding arcs are shown in Table 2.1 from which
it can be seen that those elements which are normally gaseous at room temperatures
have the higher ionization potentials. Arc temperature depends on ionization
potential according to the relationship(38),
T = 591 V010-087 eqn.2.1
where VQ is the effective ionization potential of the arc atmosphere and I is the arc
current.
12
Those electrodes with fluxes that produce large amounts of constituents
with high ionization potentials, for example H or 1^, as in the case of cellulosic
electrodes, give higher arc temperatures and hence have deeper penetration
characteristics. The easily ionised elements such as Na, K, and Ca are present in
the flux to aid in maintaining arc stability during welding with alternating currents.
The ease of ionization aids in arc reignition when polarity is reversed.
Table 2.2 presents the major constituents of the flux coatings of various
electrodes; all contain compounds capable of producing gases such as carbon
dioxide, carbon monoxide, hydro gen, and hydrocarbons. While the function of
these gases is to protect the molten metal from the effects of atmospheric oxygen and
nitrogen it is not just a simple mechanistic shielding as in the case of inert gas
shielding using argon or helium. Gas metal reaction involving oxidation/deoxidation
reactions at the electrode tip also appear to aid in metal transfer. Wegrzyn(39)
found that even with bare wire electrodes overhead ( vertically upward) welding
with fully killed steel wire was impossible, yet overhead welding with rimmed steel
wire was possible. The reason for the difference was related to the reaction at the
electrode tip (40). With rimmed steels the core usually contains higher carbon and
oxygen levels than the surface layers which are relatively clean and low in carbon
and oxygen. The effervescent reaction at the molten tip of the electrode of
FeO+C = Fe+ CO eqn2.2
caused a vigorous evolution of gas which aids in the projection of the metal droplet
in the arc plasma.
Ingredients
Type of Covering Cellulosic True Rutile Basic
(Mild steel) Low-hydrogen
Metal powders (Fe. Deoxidanis
Cellulose/Rock TiO, CaCOvMgCO, CaF. SiO.Feldspar/Mica Clavs Extrusion aids
Fe Mo)
Weld Metal H.ml/IOOg
0-10 15-19 FeMn
30-45 15-20 18-22 —
s
— —
over 30
0-10 10-15 FeMn
main 0-10 40-60 0-15 — 5-20 4-10 0-5-1 20-30
0-35 8-15 varied
— 0-10 25-50 15-30 0-10
0-4 0-2-1 3-15
T A B L E 2.2. Compositions of the flux coatings of popular and commonly used manual metal arc electrodes (5 8)
be jfiu: .so
h-s,•••./.'. x/ *>
l
FIG. 2.2. Diagram showing the welding configuration in terms of a point source, q, moving at a constant velocity, v .(63).
13
With cellulosic electrodes the flux coatings contain quantities of
combustible organic substances which generate large quantities of hydrogen which
contributes to approximately 6 0 % of the gas in the arc column. The high hydrogen
content in the plasma generates high temperatures due to its high ionization potential.
This results in cellulosic electrodes having high penetration qualities and so, fast
welding characteristics.
The high hydrogen content also produces reducing conditions and iron
oxides may be reduced by hydrogen.
FeO + 2H = Fe + FI^O eqn. 2.3
Hence, only mild deoxidants such as ferromanganese need to be added to the flux.
However large quantities of hydrogen gas in the arc column lead to hydrogen
absorption in the weld metal which in turn can lead to hydrogen embrittlement
cracking in either or both of the H A Z and the weldmetal(59,60). Because of this,
cellulosic electrodes are particularly unsuitable for the rigid requirements set down
for such structures as off shore platforms and pressure vessels.
Rutile electrodes contain considerably less cellulose anclflfe-l^is added
as a deoxidant. Initial developments used !§&$£] deoxidant however it was
found that SiOz caused high Si pickup in the ferritic weld metal and resulted in
weldmetal embrittlement. In addition to the deoxidation characteristics,Ti02 has
low thermionic emission potential which results in smooth arc characteristics
particularly with A C welding. However, rutile electrodes produce weld metal
14
deposits with high oxygen contents (eg. 835ppm) and correspondingly low fracture
toughness (5 8).
The flux coating of basic low hydrogen electrodes contains large
amounts of C a C 0 3 and CaF2. The C a C 0 2 decomposes in the arc to form C 0 2 as a
protective gas and dilutes any present. Carbon dioxide is also an oxidising gas
and in the heat of the arc decomposes to form C O and 0 2 giving rise to a form of
basic oxygen steelmaking in which the basic slag produced aids in the control of
harmful impurities such as S and P. The advantage of low hydrogen electrodes is
the elimination of organic materials in the flux and any hydrogen that does arise is
either produced from moisture associated with atmospheric adsorption or residual
moisture associated with the extrusion aids. Residual moisture may be completely
removed by baking at 700°C, but an electrode in this condition becomes useless for
welding in over head positions presumably because of the reduced gas pressure in
the arc plasma required to transport the metal droplets and the lower level of
reducing reactions(eqn. 2.3) at the electrode tip. Exposure to moisture restores the
capacity for overhead welding. Lancaster(61) cites the conclusion of Wegrzyn and
Beker in that "even low hydrogen electrodes depend for their effective use on the
presence of a certain amount of water in the flux hence hydrogen in the arc plasma".
Electrode Type and Classification.
In Australia electrodes are classified in accordance with AS1552-1973
which designates each class of electrode by the letter E followed by a four digit
number, eg. E4816.
The first two digits indicate the approximate tensile strength of the
15
deposited weldmetal in units of lOMPa. For the above example, E4816 electrodes
have an approximate tensile strength of 480MPa.
The third digit indicates the welding positions the electrodes can be used
in, 1 indicates all positions, 2 indicates flat and horizontal vertical fillet only, in the
example E4816 may be used in all positions.
The final digit indicates the type of electrode flux coating; 0 and 1 indicate
cellulosic electrodes, 2 and 3 indicate rutile electrodes, and 5,6,and 8 mdicatefcfcpw
hydrogen basic electrodes. L o w alloy electrodes have an additional digit which may
be used to denote chemical composition. Morgan(143) has presented a complete
description of the behaviour of welding electrodes to American Welding
Specifications, which are very similar to the Australian Specification and hence serve
as a usefull guide to assess suitability for specific welding applications.
2.3 THE WELD
From the discussion in Section 2.2, fusion welding can be likened in some
respects, to a miniature steelmaking process. However, the short times and high
temperatures associated with the melting and refining stages does not allow
thermodynamic equilibrium to be achieved. The thermal cycle causes very steep
temperature gradients, such that solidification processes in the weldmetal and phase
transformations in the heat affected zone of the parent plate depart significantly from
those that occur nearer to equilibrium conditions.
Numerous theoretical analyses of the weld thermal cycle have been
16
proposed, although most generally tend to be modifications to the analysis
originally proposed by Rosenthal in 1935(62). H e assumed that the energy of a heat
source that moves with a constant velocity v, can be given by
E = nVI eqn. 2.4 AT- n
where V is welding voltage, I is welding current and n is the arc efficiency ( for
manual metal arc welding n=0.8 ). The differential equation for heat flow expressed
in the coordinates identified in Fig2.2 is given by eqn. 2.5 where T is temperature
(K), t is time, and X is the thermal conductivity.
7>XZ dy* dZ2- — ** ^t eqn'25
However, equation 2.5, refers to a system of fixed coordinates, and to relate
eqn.2.5toa system of moving coordinate x may be replaced by e given by
e = x - vt eqn. 2.6
eqn 2.5 becomes
For bead on plate welds where heat flow is three dimensional the solution
of eqn. 2.7 becomes
T -To = J& "*ghHhl -2'8
isotherms
o*
FIG. 2.3. Diagram showing the three dimensional distribution of temperature isotherms produced by an arc weldC62>>
o
soo 1OO0 Temp., C
FIQ.2.4. Diagram showing the variations in thermal diffusivity with temperature for three metals(62).
17
and for thin plates, where heat flow is essentially two dimensional
T-T* -f£«^* (If) — where a is the thermal diffusivity, KQ is a Bessel function of the first kind, zero
order and r is ( e + y2 + z2 ).
The form of the temperature gradients implied by eqn. 2.8 is shown in
Fig. 2.3 from which it can be seen that, as would be expected, extremely steep
profiles exist, particularly at the front of the moving heat source.
The Rosenthal analysis involves four basic assumptions(62)
(1) material properties such as thermal diffusivity are unaffected by temperature; as
can be seen from Fig.2.4 this is not the case; for iron the relationship is complex,
(2) a moving point source represents an arc welding electrode,
(3) latent heat liberated by phase transformations are negligible,
(4) heat losses at the plate surface are insignificant.
To take such factors, particularly thermal diffusivity, into consideration
would obviously complicate the heat flow equations.
Experimentally it is difficult to make temperature measurements in the weld
pool although measurements of the thermal cycle in the H A Z using embedded
thermocouples have been made (76) and in general the measured temperature/time
curves have the form predicted by the Rosenthal equations(62). This correlation
18
of theory and experimental measurement of H A Z thermal cycle has obvious value in
predicting microstructures in the H A Z .
2.3.1 The Weldmetal.
Although electric arc welding is a casting operation, observed microstructures
of weldmetals differ from those produced by more conventional solidification
processes. This can be related to the continuous movement of the arc so that
solidification conditions change as the heat source moves away from the solidifying
metal. Furthermore, the rapid cooling conditions, the higher than normal casting
temperatures, and the high inclusion content of the melt due to metal slag turbulence
all contribute to significant departures from equilibrium both during solidification
and during subsequent solid state transformations. It is not inconceivable then that
weldmetal can have microstructures and mechanical properties that are significantly
different to similar alloys cast under conventional casting conditions .
Mastryukova and Prokhorov (38,41,42) found that as welding conditions
were varied, changes in solidification behavior affected the macrostructure of the
solidified weld metal. This in turn could effect the weld metal properties. They
developed a series of equations for solidification rate and predictions of thermal
gradients(41). The model on which the equations were developed was based on
heat transfer for a two dimensional system and as Waring (42) has pointed out, did
not account for a central equiaxed region, or for the edge of weld epitaxial growth
region.
Epitaxial growth was systematically studied by Savage et al. (44,45) and
has since been studied in more detail by Matsuda(46) and Loper et al. (47). O n the
basis of their results it is n o w well established that because of epitaxial growth
' / / / / / / / / ^
low
FIG- 2.5- T h e influence of welding speed on the isothermal distribution and the subsequent grain structure of the weldmetal(85,15).
low alloy base material
grain growth z o n e - —
transition lin planar—•
cellular cellular-dendritic
dendritic ce'.i'jlar-dendritii
cellular-planar—=)
cram growth" zone
high alloy base ma^e'ia!
FIG-2.6. Schematic illustration of the various types of growth products developed during solidification of the weldmetal as a function of alloy composition,crystal growth rate R, and temperature gradient,TL (80).
19
nucleation in the molten weld pool, does not present a significant energy barrier.
Additionally, it has been found that for both f.c.c. and b.c.c. metals, preferred
columnar growth usually occurs in the <100> direction closest to the direction of
the m a x i m u m thermal gradient. D u e to the moving heat source, the columnar grains
become curved and often do not survive to the centre of the weld pool, before new
grain are nucleated, see Fig. 2.5.
In contrast to ingot solidification, the change in growth pattern from
columnar to equiaxied grain growth is rare in weld pool solidification. Calvo et
al(48), and later in more detail, Savage et al. (44,45), observed columnar to cellular
and cellular to dendritic growth patterns. It was found that it was possible to
generate a range of growth structures in a solidifying weldmetal simply by varying
the welding conditions. Alloy additions also promoted the formation of the dendritic
region which is shown schematically in Fig. 2.5 and Fig. 2.6. Qualitatively, at
least, this is in agreement with the isotherms predicted from the Rosenthal equations.
During cooling from the solidification temperature to room temperature
further modification to the microstructure can occur due to solid state phase
transformations and the precipitation of ferrite from austenite is significant in this
respect.
Depending on the welding conditions, the austenite to ferrite transformation
may produce a range of ferrite morphologies including grain boundary proeutectoid
ferrite, Widmanstatten side plates, fine lath ferrite (often referred to as fine acicular
ferrite ) and bainite. It is generally agreed that fine lath ferrite with reduced
proeutectoid ferrite is desireable in order to obtain optimum mechanical properties,
20
particularly notch toughness and transition temperature (50,51,52). Rassanen and
Tenkula(49) related various ferrite morphologies to cooling rate, whereas
others(50,52,56), have found them to be related to oxygen concentration, see
Fig.2.7. Ricks, Howell and Barrite(53) made a thin foil investigation of "acicular
ferrite" and found that the morphology was in fact lath shaped and not needle shaped
as the term "acicular" implies and that the laths were often associated with oxide
inclusions which led them to conclude that the inclusions were the primary
nucleation sites. This had been proposed previously by Abson and Dolby(54) and
Cochrane and Kirkwood(55), but from an examination of nucleation efficiency(57)
inclusions were energetically less favorable than austenite grain boundaries as
nucleation sites. Ricks et al. (53) proposed that nucleation at inclusions occurred
because of a reduced energy barrier suggesting that the particles acted as "an inert
substrate". After initial precipitation at oxide inclusions the continued ferrite
precipitation was achieved by sympathetic nucleation which led to the interlocking
morphology of fine lath ferrite
The model for nucleation of fine lath ferrite proposed by Ricks et
al (53) is not totally consistent with the available evidence. Kirkwood(55) found
that the number of inclusions smaller than 2 u m increased linearly with oxygen
content so that it would be expected that higher oxygen content would contribute to
nucleation of fine lath ferrite. In fact, from metallographic examinations(51,52), it
has been found that weldmetal with low oxygen content (_200ppm) had a
microstructure which was bainitic, whereas at intermediate levels («300ppm) the
microstructure was fine lath ferrite and was associated with the optimum fracture
properties, (see Fig.2.7). However, weldmetals with higher oxygen contents
(600-800ppm) had a microstructure of grain boundary and Widmanstatten ferrite
CL1 !4U
S 120 C
~S5 N/mm?
98CN/mm:
20'° u ^00 » 500 800 xygen content, ppm
FIG.2.7. Diagram showing the relationship between oxygen content in the weldmetal and weldmetal fracture toughness(50)
hrai ^Mrrtpq ion
FIQ.2,%. Schematic diagram showing the various sub-zones of the H A Z approximately corresponding to the alloy of 0.15%C indicating the temperatures reached during the weld thermal cycle(15).
21
which is not consistent with the proposition that nucleation of fine lath ferrite is
enhanced by the presence of oxide particles. Kirkwood used high welding
voltages, hence high heat inputs and thus reduced solidification and cooling rates,
and it could be argued that heating effects influenced the microstructure produced.
Terashima and Tsuboi(50), examined the influence of flux basicity on
oxygen content during submerged arc welding in which heat input was not a variable
and found an increase in the density of oxide particles with increased oxygen
content which would tend to suggest enhanced nucleation for fine lath ferrite rather
than the transition to coarse ferrite precipitation at the higher oxygen contents. The
dilatometry studies of Ito et al.(52) indicated that precipitation of fine lath ferrite
occurred below 500°C, whereas the formation of grain boundary and Widmanstatten
ferrite occurred at approximately the same cooling rate, in the higher oxygen alloys
between 700°C and 500°C. In all of the studies, although the increase in inclusion
content was related to the increase in oxygen content and ferrite morphology, no
comment was made about the preferred location of oxide particles relative to oxygen
content. It would be expected that the larger of the inclusions would be on grain
boundaries because of, a) the pinning of the grain boundary by the inclusion and/or
b) the increased growth rate brought about by the higher rates of grain boundary
diffusion. It would be expected, therefore, that at the higher oxygen concentrations
the grain boundaries would contain a high proportion of large oxide particles.
Higher diffusion rates at grain boundaries, coupled with a possible lowering of
activation energy for nucleation due to the grain boundary / inclusion surface energy
configurations, would promote the early nucleation and growth of grain boundary
ferrite at a higher temperature before the size of intragranular oxides became critical
with respect to nucleation. Obviously this model is speculative and requires a
detailed examination of inclusion size and distribution.
22
Clearly the formation of fine lath ferrite is important in relation to
weldmetal strength and toughness. However, nucleation and growth reactions may
be competitive and such factors as cooling rate, inclusion size and distribution,
weldmetal flux and chemistry are all important in contributing to a phase
morphology that has beneficial mechanical properties.
2.3.2The Heat Affected Zone
Numerous investigators have studied the H A Z (76,77), measured cooling
rates using embedded thermocouples and endeavoured to simulate microstructures to
measure physical and mechanical properties and hence understand and predict
performance. In order to understand the nature of the H A Z , the relationships
between the initial microstructure and the complete thermal cycle of heating, dwell
time, and cooling is important. It may be convenient to relate the various sub-zones
of the H A Z to the phase equilibrium diagram as in Fig. 2.8 but it is not just the peak
temperature or the cooling cycle that contributes to the final microstructure. The
right hand side of Fig.2.8 is a constitutional equilibrium diagram and indicates that
the heating cycle should involve diffusion controlled processes of
a +Fe3C=oc+ y = y eqn 2.10
Because the heating cycle is rapid it is likely that there is considerable
superheating before diffusion controlled transformations can occur. With rapid
heating, carbides have been found to remain stable well above their dissolution
temperatures(79). It has also been suggested(49) that the ferrite to austenite
transformation, in some cases, could occur by a martensitic type reaction in which
23
case it has been suggested(80) that transformation strains could have an effect on
accelerating primary recrystallization and grain growth.
Because final grain size is so important, the competing factors that
enhance and retard grain boundary migration must be considered in developing any
model to predict H A Z grain size. Easterling(80) has pointed to the importance of
the heating portion of the thermal cycle. Ikawa et al.(63) showed that more than
8 0 % of grain growth occurred during the heating part of the thermal cycle. It is
generally agreed that the driving force for grain growth is a minimization of surface
energy by reducing surface area. Impediments to grain boundary migration include
solute atoms which act as a type of viscous drag on the boundary and particles such
as, carbides, nitrides and carbonitrides which, if of the correct size, "pin" grain
boundary movement. Temperature provides the necessary energy to reduce the
effects of grain boundary migration impediments and aid diffusion processes
necessary for migration. However, in the H A Z of a weld the temperatures are high
and the time at temperature is short and it is contentious whether or not isothermally
derived data can be applied to a theoretical models of weld H A Z .
From the Rosenthal analysis(62) and from experimental measurements(76)
steep temperature gradients exist adjacent to the weld fusion line. A qualitative
description of the behavior of particles in the H A Z would encompass, complete
dissolution adjacent to the fusion boundary, partial dissolution, particle coarsening,
and as the distance from the weldmetal increased, the unaffected precipitate
distribution.
Where precipitate stability is maintained it has been found(65) that grain size
is proportional to the mean distance between particles and can be related empirically
24
by an equation of the form(64)
d = Kr eqn.2.11
I where d is the grain diameter, K is a constant, r is the particle diameter and V f is the
volume fraction of particles.
In the region adjacent to the fusion line, the type of particle will determine the
temperature and rate of dissolution. As a general rule the stability of precipitates
increases with reduced or more negative values of free energy of formation(81).
Furthermore, many particles are complex, particularly when a number of alloying
elements are present so that an analysis of dissolution is difficult. Nevertheless
Ashby and Easterling (66) have been able to develop an analysis for the dissolving
behavior of simpler carbides and nitrides in the H A Z and display them as
"dissolution contours" and functions of peak temperature and time.
Both the analysis of Albury et al. (66) and Ashby and Easterling(67)
assume that grain growth is controlled and requires no nucleation, is driven by
surface energy, and obeys an Arrhenius type relationship. Albury et al. equated the
grain growth equation that was empirically derived by Hannerz and D e Kezinzy(39)
dfn = Kt + d" eqn.2.12
to Arrhenius behaviour, so that for isothermal grain growth,
dj* - df= A[exp-Q/RT] (tf-ti) eqn.2.13
25
where df is the final grain size, d4 is the initial grain size, t} is the initial time, tf
is the final time, T is the temperature, Q is the activation energy for grain growth,
and n is a grain growth exponent determined by Albury et al. to be equal to 2.73.
A weld thermal cycle has no isothermal period but can be represented by a
series of discontinuous steps, the hold time of which is small, constant and equal to
At so that
df = A[exp(-Q/RT) At] + df''5 eqn.2.14
The constants A and Q in eqn. 2.14 were evaluated from isothermal experiments.
Temperature isotherms were calculated from the Rosenthal equations so that at
various locations in the H A Z it was possible to calculate the grain size and compare
the results with those measured over a range of heat inputs.
The Ashby and Easterling(67) analysis takes a similar approach in that
eqn.2.14 is expressed as
A. '•<S<£ - |c?c/^- / ft \/t eqn 2.15
/?C.
and integration yields a similar result to eqn. 2.14 except for the value of n=2 as
opposed to the measured value of 2.73. Ashby and Easterling avoided experimental
determinations of the constants in eqn. 2.15 by measuring grain size for one set of
10
10*
10
10"
PEAK TEMPERATURE C 900 1000 1100 1200 1300 MOO 1500
14
OS Mariensite_
50-Manenstie
Nr MiCROALLOYCD
T0 = 30OK
10"
tf£ 29 "% Nts C D'SSOliil ion
50.*#£l00
nOO 1200 1300 1400 1500 1600 1700 1800 PEAK TEMPERATURE K
FIG.2.9. H A Z diagram for a N b microalloyed steel. The full lines are predicted constant grain size contours, the shaded region shows regime over which carbide dissolution occurs, the broken lines show the volume fraction of martensite,(75)
26
conditions of T and At. Grain size contours for T At relationships were then fitted
to the measured values as shown in Fig.2.9. and the calculated values of T and At
related to heat input by using the Rosenthal equations.
Both analyses have been reasonably successful in predicting grain growth
in the H A Z . The basic problem of course is that of relating an isotherm, which has
no physical width to a grain which has finite size and is often very large. Both
analyses also tend to overestimate the grain size, particularly close to the fusion
boundary. This is understandable from the viewpoint of the physical significance
of the isotherms and the fact that the peak temperatures reached at various points in
the H A Z decrease rapidly with increasing distance from the fusion line. The
grains at or near the fusion line could experience a steep gradation of peak
temperatures and thus a steep gradation of growth rates.
Albury et al.( 66) calculated a value of of 168kJ/mol. for the activation
energy for grain growth compared with 120kJ/mol. from the results of Ashby and
Easterling(67). Both of these values are higher than the activation energy of
109Kj/mol(82) for grain boundary diffusion, yet lower than the activation energy of
260Kj/mol(83) for self diffusion. This suggests that grain boundary migration
involves only partial, or alternatively, intermittent, solute drag. The differences in
the values of Q derived by the the two investigations could be attributed to the
materials studied or alternatively differences in the value of n used.
During the cooling cycle in the HAZ, a variety of phase transformations
may occur because of the different thermal cycles that occur at different locations
away from the fusion boundary. In general, both the peak temperature and the
cooling rate decrease as the distance from the fusion boundary increases.
27
Continuous cooling transformation (CCT) diagrams have been used extensively to
interpret the microstructures produced in the H A Z . Inagaki and Sekiguchi(70)
have developed a wide range of C C T diagrams for structural steels using a high
austentising temperature so as to be specifically relevant to weld H A Z . They chose
a temperature of 1350°C which was later found by Ronningen et al.(71) to coincide
with the peak temperature of an isotherm approximately 50 urn from the fusion line;
a point where dilution effects such as H A Z grain boundary liquation ceased. In
addition to the C C T curves, Inagarki and Sekiguchi produced diagrams relating
microstructure and hardness to the cooling time between 817°C and 500 °C. Such
diagrams are useful to explain the results observed after a particular welding
operation; however, difficulties arise in predicting preheat temperatures that would
be necessary to avoid deleterious microstructures.
Albury and Jones(84) used CCT curves in conjunction with the Rosenthal
equations to develop a diagram incorporating H A Z microstructure as a function of
heat input and joint thickness to predict suitable preheat temperature. Albury and
Jones did not present a range of microstructures that could be produced by the
various welding conditions. They determined the most desirable microstructure and
the diagram that was developed described the conditions appropriate for obtaining
the required microstructure.
Although CCT diagrams can be useful in assessing qualitatively the likely
phases in the H A Z , quantitative techniques are better. The concept of carbon
equivalent, (Ceq), was introduced in 1940 by Dearden and O'Neill (72). A carbon
equivalent is derived from the composition of the steel and is based on the principle
that other alloying elements in addition to carbon contribute to the hardness of a
28
quenched steel. The proposal was that above a certain critical value of C which
was related to the hardness of the H A Z , the microstructure would be susceptible to
cold cracking. This will be discussed in Chapter 3.
The cooling time between 800°C and 500°C (denoted as^Tg/5) has also
been used to describe the role of the thermal cycle on H A Z microstructure.
Rose(193) proposed that the cooling rate in this temperature range controlled the
transformation behavior of austenite during welding of C-Mn steels. Rose showed
that^Tg/5 could be correlated with an energy equivalent, L where,
L = E , eqn. 2.16
/rTTT
and E is the arc energy (see eqn. 2.4), h is the plate thickness, n is a heat transfer
factor( n=l for butt joints, 2 for bead on plate, 3 for fillet welds). Determinations
of ATg/5 have also been attempted (194) using the Rosenthal analysis(62). Inagaki
and Sekiguchi(70) have combined ATg/5 with C C T diagrams that were adapted for
welding.
Predictions of the volume fraction of phases present in the HAZ, and hence
the hardness of the H A Z has been attempted by Pavosker and Kirkaldy(73). They
used the Rosenthal equations to predict the peak temperatures and cooling rates in
the H A Z , then, modifying the hardenability formulae of Maynier et al.(74) the
volume fraction of martensite, bainite, ferrite and pearlite could be calculated from
the alloy composition. Similarly, the hardness of each phase could be calculated
from the composition using the linear regression formulae of Meynier et al. (74).
"
E -» 3 10 •-
z > o LU OS
z Ui
0 (
1 1 rr-i j — 1 /• '8
tncrvtting r L9 ' -1* "*~- * h- ! » ^*^/e
*J^ ' FUSION ' /••sj ,5. ZON€ / / ** 27>
/ / ' / / ' '
' Si** t *
^''v'jXv A> 'MO 2oj" 2§v^
*^^ '*} ' ' ^eZ" \**^^ < i
3 1 2 3 DEPTH mm
/ / / / i
•_/ • ?*? ZOt / /
yS- .
1
4
-
/ * > / V-N
PARENT PLATE
' 5 (
FIG.2.10. A diagram relating hardness to energy input for the H A Z of a N b microalloyed steel. Full lines are isotherms, broken lines show hardness contours calculated from the model, data points show Vickers hardness values for two test welds. (75).
15
2 10
3 c z > O c
0 5
Increasing
_ , — • • , r~.
' It J> £> &' -S* re ' ' ' _
/ ' / Dissolution
NbC
' ioo-- j>o-<
FUSION
ZONE
3 4 DEPTH mrr. "
FIG. 2.11. A diagram relating grain size and heat input for the H A Z produced in a N b microalloyed steel. Full lines are isotherms, broken lines are grain size contours, and the shaded region corresponds to the zone for carbide dissolution.(75).
29
Thus the microstructure of the H A Z was predicted from the chemical composition
of the plate and the heat input of the welding process.
Ions, Easterling and Ashby(75) recently used a similar approach to
calculate the volume fraction of phases. Beginning with the data of Inagaki and
Sekiguchi(70) they used the carbon equivalent formulae adopted by the International
Institute of Welding ( H W ) to relate composition to critical cooling then using a form
of the lohnson- Mehl equation (77) they related cooling rate to volume fraction of
phases present. The hardness of each phase and thus of the H A Z was calculated
using an approach similar to that used by Pavosker and Kirkaldy(73).
The analysis of Ions et al.(75) was comprehensive in that it
incorporated the analysis of Ashby and Easterling(67) to predict precipitate
dissolution and austenitic grain growth during the heating part of the weld cycle.
The influence of grain size was incorporated in the determinations of volume
fractions of the phases present and presented diagramatically, the analysis allows
the prediction of grain size and hardness contours as a function of heat input and
distance from the fusion boundary. Examples of this analysis are shown in
Fig.2.10andFig.2.11.
Both the analyses of Pavosker and Kirkaldy(73) and of Ions et al.(75)
offer tolerably good descriptions of the thermal processes to predict the
microstructure in the H A Z of weldments. Both analyses however, rely on empirical
equations which, by their very nature are generally limited to the conditions from
which they were derived. Use of C is a particular example in this respect.
Derivations of C are generally based on one of three factors:
(i) cracking behavior, as determined by a particular weldability cracking test,
30
(ii) the hardness of the H A Z , or
(iii) the transformation characteristics of the steel.
For the last two factors it is basically a differentiation between hardness and
hardenability. The IIW version of carbon equivalent used by Ions et al. (67) was
derived from cracking test data on a wide range of carbon and low alloy steels while
the Mayner equations (74) used by Pavosker and Kirkaldy relate to hardenability.
Pavosker and Kirkaldy did not take into account the effects that grain size had on
transformation kinetics in their analysis and neither of the analyses take into account
the effects of autotempering. In Fig 2.10 it can be seen that differences between
measured hardness and the calculated contours exist particularly at the higher
hardness which suggests that autotempering had occurred.
Numerous other criticisms could be leveled at the analyses used to
predict the weldment H A Z , but there is good correlation between experimental
observations and calculated models. Undoubtedly as they are applied to a wider
range of steels further modifications will occur. Concepts used to develop the
mathematical treatment are also based on recognised mechanisms and
transformations rather than simply fitting an equation to a set of experimental
observations.
31
C H A P T E R T H R E E
WELD DEFECTS
3.1 INTRODUCTION
The purpose of this Chapter is to review briefly several of the more common
types of weld cracks, and in particular, to discuss in detail the formation of
hydrogen assisted cold cracks in the H A Z of welds.
As pointed out in Chapter 1 the presence of a crack in, or immediately adjacent
to, a weld need not necessarily lead to failure however, small cracks may act as
nuclei for catastrophic failure by becoming a stress concentrator for brittle fracture
or the commencement of a fatigue crack.
Cracks may be produced in the weld and the adjacent weld zone during
service by a deterioration of physical properties, or by subjecting the welded joint to
service conditions beyond the design capacity. The types of cracks to be discussed
in this Chapter are those produced as a direct consequence of the welding process
and usually appear prior to the component being commissioned into service. These
cracks may occur in the lvspSftie»tl, such as solidification cracking (Sect.3.2),
liquation cracking (Sect.3.3), reheat cracking(Sect3.4), lamellar tearing,(Sect3.5)
chevron cracking(Sect.3.6), and as previously mentioned, H A Z cold cracking
(Sect.3.7). In all cases, there are two basic prerequisites for the formation of
cracks; namely, the prersence of a susceptible microstructure, and the imposition of
a stress, usually, but not always, brought about by the joint configuration and the
FIG.3.1. Photomacrograph showing interdendritic solidification cracking in a weldmetal.X 5.(224)
FIG. 3.2. Photomicrograph showing liquation cracking produced at ghost boundaries in a boron treated AISI304 stainless steel X400, (224).
32
effects of thermal contraction. The development and distribution of stress will be
discussed in Section 3.7.3.
3.2 SOLIDIFICATION CRACKING
A solidification crack is shown in Fig.3.1. The cause of solidification
cracking is generally well understood (86,87,88) and is primarily the result of the
partitioning and rejection of alloying elements at columnar grain boundaries and
ahead of the advancing solid/liquid interface. Cracking occurs in the terminal stages
of the weld pool solidification. Because of alloy segregation in the liquid the final
liquid to solidfy does so at a much lower temperature. The solid dendrites have
contracted during cooling hence as the final liquid film solidifies a fracture surface is
formed and has a morphology of a dendritic form. In the case of steels, high
sulphur levels may give rise to low melting point phases such as FeS(mp 1190°C)
or FeS-FeO eutectic (mp 940°C) which invariably partitions into the liquid and due
to low surface energy spreads along grain boundaries to form an almost continuous
three dimensional network of material with a lower strength than the metal.
The phenomenon of solidification cracking is often referred to as " hot
tearing" (89,90) or "super-solidus" cracking(88) which tends to indicate that
cracking might be expected to occur when the grain boundary film is liquid. This
is not always the case. There is evidence(91) to suggest that the segregation of
intentionally added alloying elements can form grain boundary films which can
fracture at temperatures well below the solidus. A n example of this is the formation
of N b C films in low C, 6%Cr-Mn-Mo-V-Nb weld metals (239,240). It would
appear therefore that the temperature at which fracture occurs is not the controlling
FIG. 3.3. Photomicrograph showing ductility dip cracking in the H A Z of an AISI304 stainless steel, X150 etchant 1 0 % HC1 electrolytic (224).
S-sod dus
R-- r'ecf ystollisoiion
U S U H G U M P [ t U l U n [
xS
FIG. 3.4. Schematic representation of two temperature ranges in which the H A Z microstructure is conducive to cracking and poor ductility.
33
factor but rather it is both the strength of the phase produced due to segregation and
the extent of thermal contraction necessary to develop the fracture stress.
3.3 LIQUATION CRACKING
Liquation cracking may occur in the HAZ, or in the weldmetal during
multipass welding processes. Cracking may occur when low melting point grain
boundary films become liquid or, solid state films, due to precipitation, may fracture
under the stresses due to thermal contraction.
Examples of the deleterious precipitates or phases that can cause this type of
cracking are,
a) sulphide(lOO) andphosphide(lOl) inclusions,
b) carbides such as NbC(102), TiC(103), and Zr(CN) (104),
c) boron-carbides M ^ C B ^ (105),
d) borides M 3 B 2 (103), Ni4B 3 (119), and
e) intermetallic phases as occur in some Al alloys (120)
Cracking need not necessarily be associated with existing grain boundaries, but can
occur at preexisting grain boundaries which, have been decorated by films or
precipitates (as shown in Fig.3.2). N e w grains may form by recrystallization but
the network remains and can lead to cracking when the thermal contractional stresses
are appropriately high.
3.4 REHEAT CRACKING
Reheat cracking, which is often referred to as "ductility dip" cracking,
occurs in the H A Z of weldments as shown in Fig.3.3. The cracks occur some
*£•*.*»
?.*>» >-V
^ •
:#.'"* j?
•
' . ' • - •
• f
FIG. 3-5. Photomicrograph showing an example of lamellar tearing adjacent to the martensitic region of the H A Z in a butt welded structural steel( 18).
Ca Ti
FIG- 3.6. Scanning electron photomicrograph showing decohesion at a sulphide inclusion and a fractured silicate in EH-33 steel (95).
34
distance from the weldmetal interface and closely resemble the triple junction "wedge
type" cracking associated with creep rupture.
The case of ductility dip cracking shown in Fig. 3.3 was observed in the
same material as the liquation cracking shown in Fig.3.2. Diagramatically the
ductility of the H A Z related to the temperature reached at various locations away
from the weld fusion line could have the form shown in Fig.3.4. Close to the
fusion boundary, the reduced ductility is associated with liquation cracking,
previously discussed (see Sect.3.3). Ductility cracking usually occurs under
conditions in which a section of the H A Z reaches a temperature approximately one
half of the absolute melting point. The stress developed and imposed on the H A Z is
accommodated by grain boundary sliding and often results in cracks being formed at
grain boundary triple junctions. At positions closer to the fusion boundary where
temperatures are higher grain boundaries migrate away from any cracks that may
form and thus limit crack growth.
3.5 LAMELLAR TEARING.
An example of lamellar tearing is shown in Fig.3.5 . The distinguishing
features of lamellar tearing are the horizontal and vertical paths of the crack in the
base metal that are just outside the hardened martensitic/bainitic region of the H A Z .
Lamellar tearing usually occurs when a tee or corner joint is welded in such a
configuration that the fusion boundary of the weld is parallel to the rolling £0Pgfc
of the steel plate.
Lamellar tearing became a problem of significance in the 1960's and mid
1970's in the construction of pressure vessels (96), ships(97), offshore
t / »
FIG. 3.7. Photomacrograph showing chevron cracking in a longitudinal section of the weldmetal of a submerged arc weld(163).
35
platforms(98), and nuclear power plants(99). Investigators generally agree (94,95)
that crack nucleation occurs by decohesion at sulphide particles or by fracturing of
silicates as shown in Fig.3.6. Crack propagation occurs to link cracks on the same
plane, and vertical ductile tears link^c^^t^gara3iel^ii||^s^ *i
Lamellar tearing can also be time dependent and it is thought(95) that both
the possibility of hydrogen diffusion to the crack, and reduced ductility due to strain
aging, could contribute to the crack propagation mechanism.
Since inclusions provide nuclei for lamellar tearing, any increase in steel
cleanliness, or modification to inclusion shape and distribution, would be beneficial
in minimising lamellar tearing. Continuously cast steels have increased resistance
to lamellar tearing presumably because of the smaller inclusion size and uniform
distribution produced as a direct consequence of the melt turbulence and rapid
solidification during casting. This effect, and small additions of calcium to
spheroidise sulphide inclusions, has reduced the occurrence of lamellar tearing.
3.6 CHEVRON CRACKING.
Chevron cracking of weld metal is illustrated in Fig.3.7, and is reported to
occur mainly in submerged arc welds(178,179) although a number of cases have
been reported in M M A welds(180). Circumstantial evidence suggests that it is a
form of hydrogen cracking(181) because increasing the baking temperature of the
fluxes of electrodes from 500°C to 800°C reduced, but did not eliminate, the
problem. The morphology of the cracking was not found to be characteristic of
hydrogen induced cracking in that intergranular cracks were joined by fine
transgranular cracks. The surfaces of the intergranular cracks were thermally
^-'^
'¥?-*r.
FIG. 3.8. Photomicrograph showing examples of hydrogen cracking in the H A Z of steel welds
a) Toe crack, X5, nital etch. b) Root crack, X5, nital etch. c) Underbead Cracking, X250, nital etch.
36
faceted indicating that the crack surfaces had been exposed to high temperatures.
Hydrogen assisted cracking is normally associated with temperatures between
200°C and ambient at which thermal faceting is not known to occur. However
hydrogen is known to aid in the formation of thermal facets in iron , at temperatures
of about 700°C to 800°C (60).
3.7 HYDROGEN ASSISTED COLD CRACKING
Hydrogen assisted cold cracking is probably the most serious, least
understood, and most widely encountered weld cracking problem. Examples of
HAZ cold cracking are shown in Fig.3.8 a,b, and c.
In Chapter 1 it was pointed out that the factors that contributed to HAZ cold
cracking were the presence of hydrogen, a susceptible microstructure, and the
application of a stress.
3.7.1 The Effect of Hydrogen.
Determinations of the basic mechanisms of hydrogen embrittlement is the
subject of continuing research and has been covered in numerous reviews a 21,122).
These reviews serve to emphasize the complexity involved in understanding the
behavior of hydrogen in the HAZ of welded steel joints. Owing to this complexity,
improvements in the control of hydrogen cracking of the HAZ have depended on the
accumulation of experience with new grades of steels and an understanding of
hydrogen generation in consumables rather than in developing a fundermental
understanding of the mechanisms by which hydrogen effects the properties and
fracture mechanisms of the HAZ.
FIG. 3.9. Schematic representation of the diffusion of hydrogen relative to the movement of the arc. A and B refer to temperature fronts in the weldmetal and the base metal respectively (182).
FIG. 3.10. Schematic representation of microscopic fracture modes observed as a function of decreased stress intensity factor and concomitant decreased cracking rate, a) high K, microvoid coalescence, b) intermediate K, cleavage , c) low K, intergranular, d) intergranular with assistance from H 2 pressure (107).
37
The influence that hydrogen has on the mechanical properties of steels is a
topic of controversy. Tetelman (149), Gibala(150), and more recently Hirth (151)
found no significant effect on the yield stress of steel. O n the other hand
Rogers(152) and Bastien et al. (153) found that the upper yield point of mild steel,
to be either reduced or eliminated, and the work hardening exponent to be
reduced(153). Grant and Lundsford(154) observed that the stress required for a
particular increment of plastic strain was lowered by the presence of hydrogen.
Hydrogen is known to diffuse more readily than other elements through
body centred cubic iron and it has been estimated that at 20 °C, hydrogen atoms are
capable of diffusing 10 1 2 times faster than carbon or nitrogen atoms(106).
Furthermore the activation energy for diffusion of hydrogen in ferrite decreases
from 32.7kJ/mol below 200°C to 13.4kJ/mol above 200°C. The reason for the
change in activation energy is not clear but there is good evidence to suggest that at
low temperatures hydrogen is trapped at defects such as dislocations, grain
boundaries, inclusions, and microcracks. Hydrogen precipitation occurs as
hydrogen gas and diffusion below 200°C is controlled in some way by the
re-solution of diatomic hydrogen and diffusion of hydrogen atoms from one defect
site or "sink" to another. Furthermore, while the diffusivity of hydrogen atoms in
ferrite is high, the solubility is low by comparison with austenite (106), in which
the diffusivity is relatively low. The significance of this diffusivity/solubility
relationship is important in the formation of a zone of hydrogen enrichment
adjacent to the weld fusion line. As pointed out in Sect. 2.2, Chapter 2, hydrogen
in the weldmetal can arise from the dissociation of water vapor in the welding arc.
As the weld metal cools, the mismatch in transformation temperatures, due to
composition differences between the weldmetal and the H A Z , results in diffusion
from the ferritic weldmetal to the austenitic H A Z . This is shown schematically in
rtK3 *QC ^ W CCNC£.NT~^'ON cc HYOPCOE.N DIS3CLVED
IN •"?*£? T P MA7£plA:-
FIG.3.11 Diagram showing a suggested relationship between stress intensity, dissolved hydrogen concentration and fracture mode in a macroscopically small volume of crack tip material ( M C V - microvoid coalescence, Q C - quasi cleavage, IG inter granular)(107).
.oft
' 1 - ' - " I I 1 L 1 1 1,1111 1 i l i u m i i i « • .tt ' ' ' .o
F*. i 'j r*• T.m* Mr
FIG. 3.12. Diagram showing the experimentally determined failure time as a function of hydrogen content (reduced by increasing baking time of electrodes), and applied stress intensity.(109).
38
Fig. 3.9. The resulting high hydrogen content is trapped when this austenite
subsequently transforms at a lower temperature to the body centered phases.
Although there is disagreement regarding the effects that dissolved hydrogen
have on the yield behavior of steels, it is generally agreed that hydrogen is
detrimental and promotes premature fracture. O n the basis of his own research, and
an extensive literature review, Beacham(107) proposed that hydrogen diffuses into
the structure just ahead of a crack and aids whatever deformation the structure will
allow. This proposal is similar to the propositions of Johnson(182) and
Troiano(108), however, Beacham introduced the concept of stress intensity factor
(K), in conjunction with the hydrogen concentration at the crack tip to determine the
fracture mechanism. The Beacham model is shown schematically in Fig.3.10. It
can be seen that for high values of K microvoid coalescence is the predominant
fracture mechanism, while for lower values of K the fracture mode varies from quasi
cleavage to intergranular. It is further suggested in the model, that hydrogen,
instead of "locking" dislocations in place, "unlocks" them and allows them to move
and/or multiply at reduced stresses. As such, it would be expected that hydrogen
concentration would also be significant in determining values of K and the fracture
mode. Beacham explains this relationship qualitatively as shown in Fig.3.11. In
the diagram the dashed lines represent critical combinations of K and hydrogen
concentration needed to cause crack growth by the three fracture mechanisms. The
numbered solid lines represent loading conditions, curves 8, 9, 10, represent
situations where precracked or stress-concentrating notches exist; for example curve
8 represents conditions under which a pre-existing crack or notch is held at constant
load conditions while hydrogen diffuses to the crack tip causing a shift in conditions
until the critical conditions of hydrogen concentration and K are reached for crack
10 12
FIG. 3,13- Diagram showing the hydrogen distribution in a single bead on plate weld as a function of distance at various times after welding(113). 60,
50-
<
05 10 15
HTDROCEN, cm1 /I00q
FIG- 3.14. Diagram showing ductility of smooth tensile specimens as a function of hydrogen contentat five strength levels( note lton/sq in= 15.44MPa). (115)
X
39
propagation by the particular fracture mechanism. O n the other hand for high
hydrogen concentrations the conditions described by curves 2,3,and 4 are proposed.
In terms of hydrogen content, stress concentration factor (K), and time to
fracture ( time for hydrogen diffusion), the model agrees with experimental
observations, at least qualitatively. The work of Gerberich and Hartbower(109),
shown in Fig.3.12. is consistent with this general trend. From a quantitative
standpoint, there are difficulties in expanding the model to test existing data and
applying it to practical situations. Accurate determinations of hydrogen content are
difficult. Furthermore in H A Z cold cracking hydrogen may be "residual", i.e.
trapped in sinks or "diffusible" that is, situated in the structure as an interstitial and
able to move by interstitial jumps. Diffusible hydrogen is generally considered to
be detrimental in cold crack propagation( 110,111). It is claimed that methods are
available to determine diffusible hydrogen content, however considerable conflict
exists in results obtained in different laboratories with welds produced under
identical conditions(l 12). Theoretical relationships between hydrogen concentration
in the overall structure and the crack tip may be derived but experimental verification
could be difficult. Furthermore, there are indications that diffusible hydrogen
concentrations are variable with time and temperature. Christenson et al.(113)
determined the hydrogen content of thin slices that were taken from weldments
during cooling and found that initially the highest concentration of hydrogen was in
the centre of the weldmetal, see Fig. 3.13. As the weldmetal cooled the peak value
decreased in magnitude and shifted in location to the weldmetal/parent metal interface
This example is for one particular set of welding conditions and it would be
expected that other variations to that shown in Fig. 3.13 could occur. The
distribution of hydrogen contents within the weldment would be expected to depend
in complex way on hydrogen solubility and diffusivity, temperature and
40
microstructure, as well as the nature of stress induced diffusion, and mechanisms
and efficiencies of trapping of hydrogen at "sinks". Therefore to develop analytical
solutions, the assumptions and simplifications necessary, would tend to limit the
practical significance of predictions of absolute values of hydrogen in the
HAZ(114).
3.7.2 Microstructure.
It is evident, from Section 3.71, that, as a general observation, hydrogen
impairs the properties of steels. The work of Farell and Quarrel( 115), indicated that
the degree of impairment of mechanical properties increased with increasing strength
of the steel, see Fig.3.14. From this, and other work, two deductions are possible.
First, strength and ductility have a direct relationship with microstructure so that it
could be concluded that some microstructures are more susceptible to hydrogen
embrittlement than others, and secondly, because strength and hardness are often
empirically related, measurements of hardness of the H A Z would be a convenient
means of categorising the relative susceptibilities of H A Z s to hydrogen assisted cold
cracking(116,117). Generally it has been found that cracking does not occur in
structures with a hardness below about 350HV. However this is only a general
empirical rule, for cracking has been observed to occur in a H A Z having a hardness
as low as 250HV, and other steels may be crack free in a H A Z with a hardness as
high as 450HV. It would appear then that hardness and susceptibility to cold
cracking are not always directly related. Depending on the alloy composition and
cooling rate, different volume fractions of microsructural constituents may be
produced that can have the same overall value of hardness. BoniszevAski and
Watkinson (118) investigated the relationship between microstructure, hardness, and
susceptibility to hydrogen assisted cracking(HAC). and from their work the British
41
Welding Research Association adopted the concept of the Embrittling Index (EI)
where
EI = NTS^- LCS eqn 3.1
MT5Hr
where N T S H F is the notch tensile strength in hydrogen free conditions, and LCS is
the lower critical stress applied in constant load rupture tests in the presence of
hydrogen ( Sect. 3.12). Clearly, the index can vary from 0 for steels with a low
susceptibility to hydrogen assisted cracking(HAC) to 1 for steels that are very
susceptible to H A C . Boniszeviski and Wilkinson used both their own results
(118,189), and those of previous workers( 119,120,121,123), to relate EI to
hardness for a number of C-Mn, Ni-Cr-Mo, Ni-Mo-V, and Cr-Mo steels. These
results are shown in Fig.3.15 from which it can be seen that for a particular
hardness value the steel composition can have a significant influence on the
susceptibility to cold cracking as indicated by the value of EL
BoniszeWski and Watkinson (118) also related the results shown in
Fig.315. to the microstructures of the steels. They found that for hardnesses, of
500HV and above, for which EI approached unity, plate martensite was the
predominant microstructural constituent present. At lower hardness, the spread of
EI values from 0.2 to 0.7 was found to be associated with a variety of constituents
in the microstructure. They concluded that microstructures containing upper bainite
were more susceptible to H A C than those containing lath martensite, which in turn
had about the same susceptibility as "granular bainite". Granular bainite is formed
when carbon enriched austenite ahead of the advancing ferrite interface transforms
to plate martensite(146) producing a microstructure of carbide free bainite and
martensite . The observations of Boniszewski and Watkinson (118) covered a
Wela mptol Sinair un Reneated
A C.Mn • i
b Ni:G:Mo
C Ni.MoV
~ D Ci:Mo
High 0-6%) Mn lte.1
1 °-A~ *B rff^l
. fir L)"^ j^^ Conventionol Heels M n ,
Mo
200 300 400 500
Horanejs o.f constant loori ruoture ir>*cim*rn H V
FIG, 3.15. Diagram showing the effect of H A Z hardness on hydrogen Embrittlement Index for a number of steels of different alloy composition(118).
FIQ- 3.16. Photomicrograph showing an underbead crack associated with a non metallic inclusion, X1000 etchant 2.5% nital (184).
42
range of conventional C-Mn, and Ni-Cr-Mo steels together with low carbon (0.1%),
Cr (up to 5%)-Mo(up to 1%) steels and a number of low carbon high M n and high
Ni steels.
The analysis of microstructures by Boniszevifski and Watkinson (118) was
not quantitative but was derived from observations of carbon extraction replicas
examined by electron microscopy at magnifications up to 10,000X. The main
feature used to identify plate martensite was the parallel arrays of carbides that had
precipitated during autotempering and were decorating the twin boundaries.
Martensites formed in the types of alloy steels examined would be expected to be
complex and difficult to interpret. Boniszevisiki and Watkinson attributed the high
values of EI for high M n ( 7 % ) low C(0.016%) to the formation of plate martensite
(189). It is known that(190) for Fe-Mn alloys two forms of martensite can
coexist,the second martensite having formed by the austenite transforming to a close
packed hexagonal phase. In such a system, differentiation of phases, using carbon
extraction replicas would be difficult. The work of Boniszeviski and Watkinson
nevertheless does point to the important role that microstructure plays in determining
susceptibility to H A C .
From the work of Boniszev^jki and Watkinson, it would appear that the
nucleation of cold cracking could be analysed from two approaches, namely, crack
nucleation in the presence of plate martensite, and crack nucleation in the abscence of
plate martensite. Normally, plate and lath martensites are associated with high and
low carbon content respectively. Martensites formed in commercial steels are
known to be complex, so that a complete description of the formation and properties
of martensites is yet unavailable(133). Nevertheless, it is recognised that plate
martensite is susceptible to the formation of microcracks. Davies and Magee(144)
43
working with high carbon steels observed microcracks, both along the edges and
across the martensite plates. The formation of microcracks in plate martensite is
thought to be associated with the transformation strains, but the exact mechanism
has not been clearly defined. Davies and Magee found that both the density and size
of cracks increased with the martensite plate size. The plate size is a function of the
austenitic grain size which in the weld H A Z can be very large and consequently a
high density of microcracks could occur in the H A Z . Furthermore, for plate
martensites the M s temperature is usually below 200°C, so that the transformation
from austenite (with a high solubility for hydrogen) to martensite (with reduced
diffusivity) can result in entrapment of hydrogen gas in these microcracks. A
considerable partial pressure of hydrogen (PH) would be expected so that the
Griffith relationship for a microcrack of length 2c
iv? <ti=7xi ein-3-2
becomes
C£r2 + PH = [ £ E ) 2 e(ln3-3
where a is the average applied stress for crack growth without hydrogen, V is
the surface energy andOT? is the average applied stress for crack growth with
hydrogen present. Fast(161) estimated that at 300K the partial pressure of
hydrogen gas that would be generated by 5mg/100ml of atomic hydrogen
converting to molecular hydrogen would be 2X10 1 1 atmospheres. Although this
44
pressure appears to be very high the evolution of hydrogen gas from cracks has been
observed by a number of investigators (145,146) so that pressure assisted growth
would, in part, account for the lower stress levels required to propagate cracking or,
when hydrogen content is high, crack growth could be entirely due to hydrogen gas
pressure. The presence of hydrogen gas is also known to reduce values of surface
energy and so further reduce the value of applied stress.
In the absence of plate martensite, bainite was found to be the
microstructural constituent most susceptible to hydrogen assisted cracking.
Considerable controversy surrounds m a n y aspects of the formation of
bainite(133,146), but it is the generally agreed that upper bainite is comprised of
heavily dislocated laths of ferrite with cementite particles precipitated between the
laths and aligned with them. Lower bainite forms at lower temperatures and
consists of heavily dislocated plates of ferrite containing parallel arrays of fine
carbides at approximately 55° to the longitudinal axis of the plate. Lower bainite is
considered to have a higher fracture toughness than upper bainite(146).
Boniszfia&ki and Watkinson(l 18) concluded that both upper and lower bainite were
equally susceptible to hydrogen cracking and that "granular" bainite was the most
desireable bainitic structure. Boniszeviski and Watkinson also concluded that it was
the geometric irregularity of the granular bainite that reduced its "capacity to nucleate
transgranular cleavage cracks along and across the bainite colonies". Because of the
irregular nature of granular bainite it would be expected that crack propagation
would be the controlling process and the presence of plate martensite would be
expected to aid the nucleation process.
The nucleation of HAZ cold cracking in the absence of plate martensite has
been attributed to non-metallic inclusions such as oxides and sulphides. A number
45
of workers(183,184) have found that underbead cracks in the H A Z are often
associated with non-metallic inclusions which suggests that they are sites for
nucleation. Hart(147) carried out an extensive review of literature relating to the
effects of sulphur on H A Z cracking and concluded that reduced sulphur levels
increase the risk of cracking. This effect was found to be related to increased
hardenability and the consequential formation of more susceptible microstructures.
The addition of rare earth metals ( R E M ) and calcium have also been investigated as a
means of modifying the shape of inclusions. It has been found that R E M caused
problems with welding arc stability(148) and there has been considerable conflict
regarding the benefit of such additions. O n the other hand calcium additions have
been found to not influence arc stability but to cause modifications to inclusion
shape by spheroiodizing them. In a limited study, Wilson(151) claimed to have
increased the resistance to H A Z cracking by the use of calcium inoculation. The
results however were very scattered and unconvincing and Wilson conceded that
further work is needed to demonstrate the effects conclusively. Changes to
inclusion shape and distribution have already found to be benificial in reducing
lamellar tearing, see Section 3.5.
3.7.3 Stress
Satoh et al.(155) pointed out that during fabrication of welded structures
two types of stresses are produced;
(1) residual welding stresses, and
(2) reaction stresses caused by external restraint.
Residual stresses occur in welded structures for a variety of reasons and
when superimposed can lead to a complex stress distribution. Residual stresses
that arise from forming operations such as rolling, bending or shearing, can be
x-niriN<nunmii—A^aiiiiimnmiiin:
a) Butt weld
b) Distribution of a,along YY
-1 Reaction stress
I ?" _-- Curve 2
Tension Curve 1 v
IllUillll l!i!ll r
Compress i on
,,:,
c) Distribution o< o along XX
FIG. 3-17. Diagram showing typical distribution of residual stresses in a butt weldment, curve 2 represents the imposition of restaint stress(164)
46
complicated by the residual stresses arising from welding and the accompanying
uneven temperature distribution so that in severe cases distortion of the component
can occur. For a simple butt joint, free from external restraint, the distributions of
residual stresses caused by welding are shown in Fig.3.17. It is generally agreed,
that in the longitudinal direction of the weld, tensile stresses exist at the centre of the
joint and compressive stresses at the ends. The transverse distribution of stresses is
considered to be caused by the interaction of shrinkage, quenching and phase
transformation strains; however they need not always be additive and so lead to
high values of residual stress. O n the contrary, residual stresses have been found to
be approximately 50MPa, which is relatively low in magnitude(185,186), and can
be considered to contribute to hydrogen assisted cracking when hydrogen
concentrations are high.
Reaction stresses can be large (approximately 200MPa) and it is these
that, in general, induce cold cracking. Reaction stresses arise from the clamping
effects of the other components in a structure and prevent free thermal contraction.
Reaction stresses are long range, compared with residual stresses, and are not
necessarily reduced or relaxed by crack formation. Consequently stresses of this
type are dangerous in the practical sense. Beginning in the 1930's Naka(156)
developed a theory of reaction stress produced in butt welds with external restraint.
H e introduced a parameter called the " restraint intensity", and denoted it K,
obviously derived from the lapanese word kosoki, meaning restraint. H e defined
K as, "the force per unit of weld length caused by the elastic change of a unit in root
gap, uniform along a butt welded joint". More recently the symbol R F has been
used to symbolize restraint intensity to avoid confusion with stress intensity factor
K, used in fracture mechanics.
1& th y ^ n w ; j '
Contraction S
Reaction +~ p H X b\ mm> fore. P
to) Simple butt weid under restraint
Xb
FIG. 3.18. Diagram depicting the shrinkage of a butt weld under
restraint(157). h/l
0 0.02 0.04 0.06 0.08 0.10
h/l
0 0.02 0.04 0.06 0.08 0.10 900
800
700
600
500
400
300
200
ICC
i i i i I _ 0= 16 kJ/cm
A h =15mm — o h = 20mm —
vh = 30mm - oh= 40mm _<*
- r4#-tr -
- z °\
v ,
1 = 20mm '6 12 kJ/cm_ o 16 kJ/cm -o- 18 kJ/cm_ -<>-20 Kj/cm k £ 32 H J / C T -
i
3 1000c zee::
a) INTENSITY OF REETRAi',: = , (:,/-
FJC '• 3.19. Diagram showing rel
900
800
700
600
1 500
~ 400 * to 300
200
100
I _ Q =
-
—
n
-n:; z) :NTENS;
ationship b
I I I I 16 kJ/cm 0 -
0—-
So -o/
a' ° qg/fc.
°^ A h = 15mm ^ o h=20mm
/ a h=25nm _
1 , i
10000 2c:c:
TY OF RECTRA'.f.T -. f '.N/rr.r
etween final reaction stress and restraint intensity for a mild steel and H T 80 steel (157).
47
Satoh and Matsui(157) considered a butt joint, as shown in Fig 3.18 where
the plate material was fixed at both ends. The hindered contraction on cooling
developed a reaction force, P, which caused elongation of the base metal X BM and
the weldmetal X y^^ so that the total elongation, S, is given by,
S = X vm + \ M eqn.3.4.
In Fig.3.18b the line O Y M represents the relationship between P and X y^. The
line ON represents the relationship between P and X BM and the slope of the line
O N is the restraint intensity R F
R F tan 0 = Eh_ eqn. 3.5. L
where E is Youngs modulus, h is the plate thickness, and L is the restraining
length.
The relationship between final reaction stress CJNET and the restraint
intensity was deduced by Satoh et al. (54) to be
ONET = RFa *T~t< • \\ • V«~ fa 1.* eqn.3.6
•V/-V, '.'
* S -
<. < <
s £ 5>
21Zi
ULLLLU ttt t t M
cr0
£>>>*
£ i/2S
f S
FIG. 3.20. Diagram showing free joint (left) and constrained joint (right). The degree of constraint is the spring constant(158).
T A B L E 3.1. Stress concentration factors at the root of welds for various joint geometries and root angles(179).
Varuble Groove
:ypo
1/2 V
Gnnive type
y Rj.it AT.ZIC
Y X
'•12 1 grrcve
^ - =>
l • y r^
I i~M t • — . .
Geometric
Groove
male
J(°)
60
60
60 60
/ 3 re
i
^
configuration of
Root
ingle
9(°)
40 60
90
100 114 120
80
90 120
3v»
/ !
Th
hi
i^uove
itkness
mm
30
30
30 30
,
)
we
E
1 groove
2
r» V —
. — r f /
Id
:centxicity
n
0
0
0 0
t
t
Throat
thick-
new & w(mm)
5
' 5
5 5
X ;r~
y'
i_\ »•.' i • •}
1 /<
Ki
6.5 5.8
4.3
4.7
4.0 3.5
4.7"
4.2 3.7
•.*.
/ t
v.^ »
48
provided that
and ° NET <^ ° YSWM
where a is the coefficient of thermal expansion, T w s is the solidus temperature of
the weldmetal, H is the specific heat of the weldmetal, p is the groove angle, C is
the specific heat of the plate metal, Q is the net heat input, p is the density of the
metal, and OYSWM is the yield strength of the weld metal. From equation 3.6 it is
clear that reaction stress is independent of plate thickness and heat input, as found
by Satoh et al. and shown in Fig.3.19.
The analysis of Satoh and Matsui(157) was reliant on the plate being rigidly
fixed at points A and A' (see Fig.3.18). Masubuchi(158) considered the
situation where A and A1 were not rigidly fixed but could be subject to a degree of
movement. H e defined this capacity for movement as the degree of restraint (Ks)
and pointed out that welded joints may be classified as free joints, as shown in
Fig.3.20a, or constrained joints, as shown in Fig.3.20b, in which the constraining
constant is represented as sets of springs. During cooling, transverse shrinkage of
magnitude 8 occurs, and a uniform stress o per unit length of weld is developed
so that the total load, P = o L. Masubuchi(158) proposed that aQwas
49
proportional to 6 and the proportionality constant or "spring constant" K be
defined as
Ks = _Sa = P eqn. 3.7
The degree of constraint has practical significance in that a structural joint
is rarely rigidly fixed and some degree of elastic flexibility is always available.
Analyses based on the degree of constraint concept have been carried out for
various weld geometries(187). Weldability tests have also been developed to
provide a range of degrees of constraint 18 8).
As early as 1940 the concept of degree of constraint, ( Ks), was
considered of importance by Stout et al.(188) during the developement of the Lehigh
weldability test. It was realised that weldment cracking could be induced by
increasing restraint; however, Ks was expressed in terms of specimen width, see
Chapter 2 Sect.4.4.
In the mid 1970's Matsui et al. (159) and later Kalev(178) pointed out
that the risk of hydrogen cracking was also dependent on the groove geometry and
other geometric factors not taken into account by RF. Attention was thus directed
towards the nature of localized stresses in the crack initiation zone which is normally
the root of the weld. Previously, Stout etal.( 188) pointed to stress concentration
as a significant factor in any quantitative analysis of weldability. Nevertheless it
was not until 1978 that Satoh et al. (179) and Togoda (180) proposed the "apparent
r r
=.r_ ' L
RS-type
1 'D
1 L 0 30
a)
60
9, (°)
90
0
eL(°)|9L(°) Ol 40 101120
A< 66 IV|13Q DI90
HNr 1 0 60 90 120 150 180
b) e o
1 -
0.5
0 LN^ ^ — l
0 90 150
c)
110 130
9 L(°)
FIG. 3.21. Diagrams showing the variations in Fj with groove
geometries and root angles, 9J162).
50
elastic stress concentration factor", Kt, as a parameter for stress concentration at the
root of the weld. Using photo elastic techniques and a finite element method (FEM)
analysis of the local stress was determined under plain strain conditions and Kt was
defined as the ratio of the local stress to the average net stress. Table 3.1 shows
various calculated values of Kt for a number of groove geometries.
Karppi(162) extended and refined the stress concentration factor of Satoh
et al(179) to define the stress field parameter, Fj. Karppi used FEM to determine the
stress field at the root notch of welds and so relate the stress component in the
loading direction s to the net stress, a by the relationship
0"Y = I^S-SEX- eqn.3.8
where r is the distance from the notch tip to the point in the stress field relative to
a . The parameter Fr was also examined as a function of groove geometry, and
like Kt was found to vary as can be seen in Figs 3.21,a,b,c. Variability in both Kt
and Fj with root angle can have significant effects on the fracture stress and has
particular importance in analysis of results from various quantitative weldability
tests. This point will be considered in greater detail in Chapter 4.
51
3.7.4 Predicting H A Z Cracking
Accurate predictions of the susceptibility of a steel to H A Z cracking would
require a quantitative knowledge of the manner in which the extrinsic variables of the
steel, the joint, and the welding procedure effect the intrinsic factors of
microstructure, hydrogen concentration, and stress.(see Fig. 1.1). The critical
combination(s) of these intrinsic factors would then determine the predictability of
H A Z cold cracking. The problem of predictability can be devided into two areas,
namely:
a) the interaction of the extrinsic factors that determine the importance of the
intrinsic factors, and
b) the range of interactions between the intrinsic factors that can cause cracking.
An exact solution of a) obviously would be difficult. Control of one of the
intrinsic factors in b) is more manageable. In this respect microstructure has
recieved a considerable amount of attention. As mentioned in Chapter 2, Sect.2.32
the concept of carbon equivalent (C ), was one of the earliest attempts to predict
conditions conducive to H A Z cold cracking . The value of carbon equivalent was
determined as an assessment of the combined effects of various alloying elements in
a steel on the transformation characteristics, including the M s temperature and was
expected to indicate the hardness of the H A Z and whether or not the H A Z would be
susceptible to cold cracking. A n empirical value of 0.35 was generally considered to
be the critical value of the carbon equivalent, and corresponds to a H A Z hardness of
350HV. It was generally considered that welding could produce a H A Z susceptible
to cracking for steels with compositions that resulted in a carbon equivalent value
higher than 0.35.
52
The proposition that steels with a carbon equivalent higher than 0.35 are
difficult to weld without H A Z cracks is a generalization. As can be seen from
Fig.3.15, susceptibility to cracking, as indicated in this case by EI, can vary for a
constant H A Z hardness depending greatly on the microstructural constituents that
contribute to the hardness value. T o cover wider ranges of steel compositions
numerous carbon equivalent values and formulae have been developed from the
basis that H A Z cracking may be detected by one of the many weldability tests
available. It is generally implied that the cracking test results are related to the carbon
equivalent by some feature of the microstructure.
Winterton(165) reviewed published formulae derived before 1961 and
related cracking in controlled tests(166) to the M f temperature. Pickering(146)
disagreed and suggested that the M s temperature would be more appropriate. In
1968 a formulae for carbon equivalent was developed by the International Institute
of Welding ( U W ) and incorporated into a British Standard(167).
CM = C + Mn + Cr+Mo+V + Ni+Cu eqn.3.9 q 6 S /5
which was derived from data using carbon and low alloy steels, and is generally
applied to such steels.
Numerous other carbon equivalent formulae have been developed based
on the alloying elements present and the consequent phase transformation
characteristics. These have been reviewed recently by Yurioka(148) who
categorised the more widely used formulae into three broad groups. Group A, for
which the H W fomula (eqn 3.9) is generally used and can be generally applied to
53
carbon and low alloy steels, Group B, for low carbon(0.05-0.3%C) low alloy steels
for which the Ito-Bessyo carbon equivalent (Pcm) formula(168) is used:
P c m = C + Si + Mn+Cu+Cr + V + M o + Ni +5B eqn.3.10 so 20 /o 60
Group C, where it is considered that there is an interaction between carbon and the
other alloying elements, an example is the formula for the carbon equivalent(CEN)
developed by Yurioka et al.(169).
C E N = C +A(C){ Si + _Mn +Ni +£u+ Cr+Mo+V+Nb + 5B} eqn.3.11 & h ZO /$ ,5
whereA(C) = 0.25 [3+ I " e*P t'40 Q' °''*)\ ]
All of the carbon equivalent formulae ( eg C (HW), P c m C E N ) that
have been developed tend to encompass the same alloying elements that contribute to
hardness and hardenability. The differences in the formulae generally concern the
relative contribution that each element makes. A point of interest is that none of the
formulae include the effects of sulphur, phosphorus, and cobalt. For example
cobalt is known to reduce hardenability (170) whilst low S levels have been found to
increase the susceptibility of the H A Z to cold cracking(171). and phosphorus, even
though an impurity, has been found by Grange(172) to have about twice the effect
of carbon on hardenability up to a saturation value of about 0.05%.
54
Carbon equivalent formulae are empirical, and are limited in range of
application to the alloy steels from which they were derived. The limitations can be
appreciated from the fact that steels used to develop P c m (eqn.3.10) were low alloy
and produced a fully hardened H A Z , whereas the IIW C (eqn.3.9) was based on
C - M n steels which have lower hardenabilities.
Krieg(189) and Brisson et al.(190) pointed out that carbon equavilant
formulae are not identically related to microstructure because they neglect the effects
of cooling rate which can vary depending on welding heat input and plate section
size. Correlations between carbon equivalant and implant test results have also been
found to be poor, particularly for H S L A steels(191). Yurioka et al.(169) found that
the IIW C was also unsuitable for 0.02-0.05% carbon bainitic pipeline steels.
Furthermore, carbon equivalent does not take into consideration the other variables
involved in cold cracking, namely hydrogen content and stress, so that in some
situations it may be too conservative, so that unwarranted and costly welding
proceedures might be implemented, or alternatively, the risk of catastrophic failure
may be underestimated.
Ito and Bessyo(173) derived predictive equations which include, with
the carbon equivalent P c m , terms to account for joint restraint and hydrogen
content. They related these terms to a cracking parameter, P w and by using
sawcuts in the Tekken Test piece (see Sect. 4.7 Chapter 4), similar to the Lehigh
Constraint Test (Sect. 4.5), they were able to vary the intensity of restraint, RF. By
combining hydrogen concentration,[H], and P (eqn. 3.10), they defined a new
E E E E
1/1
*
c
6 VI
0)
b
5 c c
3000
2000
1000
20°C Initial temp
a CO u
_ y \
o No crocking •» t - 10% • II- 5 0 % * 51-100%
Weld crocking percentage in section
\ OO ci*3 9 CK>« • 1
\
\ O C H •
\ Y o o cno •
c>rr"> rri 11 n i 111 • n • iti^^^wri i * .
"Jo c\ o o
~ \ * 4
O \ * * O o \
o \a»MO • • A \ * *
1 1 1 02
_,. _ Si Mn
0.3
Cu Ni Cr Mo
0.1
30 20 20 60 20 15 10 56* 60 ' %
FIG. 3.22. Diagram showing the relationship between the intensity of restraint,Rf and the cracking parameter, Pc.
55
term Pc, where
Pc =Pcm +IH1 eqn. 3.12
60
The relationship between Pc and RF was determined experimentally and is shown in
Fig. 3.22. From this relationship the cracking parameter, Pw, was determined:
Pw = Pcm+ IH1 +JS£_ eqn.3.13
60 400
where [H] is the hydrogen concentration determined by the IIW method, and Rf is
the intensity of restraint.
Following the Ito-Bessyo equations, there have been a series of predictive
formulae for a cracking parameter, usually denoted PH that have been developed
by the Japanese. These include
Ito et al.(173)
PH = Pcm +0.0931ogHD + RF/4320 eqn.3.14
Inagaki et al. (174)
PH = Pcm +0.1621ogHD + RF/4160 eqn.3.15
56
Satoh et al. (175)
PH = Pcm + 0.2141ogHD + RF/4280 eqn3.16
PH = Pcm +0.0591ogHD +RF/4280 eqn3.17
and Suzuki(176)
PHA = Pcm +0.1431ogHD+RFY/2400 eqn3.18
where HDis the effective value of diffusible hydrogen Rpy is equivalent intensity of
restraint for the Tekken Test geometry and Pj^is the versatile cracking parameter.
Suzuki (176) recently modified the PHA analysis (192), however it still contains
Pcm thus limiting the range of applications to low carbon alloy steels.
The analysis by Suzuki(176,192) contains in the determination of HD and
RpY procedures and multiplication constants to interrelate cooling rate, diffusion
and restraint characteristics to cracking tests other than the Tekken test. The
proposition is that the more specific cracking parameters (such as PH) are derived
from the more general and versatile cracking parameter Pj^.
Pavoskar and Kirkaldy(73) developed an analysis to determine the critical
stress for HAZ cold cracking. The analysis is based on a HAZ Index which
57
incorporates both the hardness of the H A Z and the percentage of martensite present
i.e.
HAZ index = 1565 -10(%martensite)-max HAZ hardeness.
The H A Z index they found to correlate well with the critical stress o. for H A Z vl
cracking when a term for hydrogen content was added.
G =( H A Z Index)172 . (Hydrogen Index) cr
ac r =[ 1565-10(%M)-HV]1/2 . [31-15.51og[H]] eqn3.19
They proposed that when the Welding Index, given by
W I = a fcalc. , eqn3.20 cr 6oO
was less than unity, the steel could be regarded as susceptible to cracking. Pavoskar
and Kirkaldy (73) used the experimantal data of Evans and Christensen(177), and of
Ito and Bessyo(173) to verify the validity of the acr calculation which had been
derived from implant and restraint cracking tests . There are however difficulties in
determining stresses associated with joints in fabricated structures, so that the
significance of the critical stress derived from calculations is questionable.
58
Cracking parameter formulae have similar limitations to the carbon
equivalent formulae. Their predictive capacity relates to one or more particular
cracking tests and the type of steel used in the determination of experimental data.
The relationships that the cracking tests have with welded structures is dubious.
As the diversity of low alloy steel strengthening mechanisms increases it would be
anticipated that the diversity of both carbon equivalent and cracking parameter
formulae would increase thus adding to what is an already confusing situation.
Furthermore, being able to predict in detail the magnitude all of the intrinsic variables
that contribute to critical conditions for HAZ cold cracking does not offer alternatives
for avoiding such conditions. Attempts have been made to develop predictive
equations for necessary preheat temperatures, however these are, again empirical,
and thus only apply to the range of alloys and welding conditions from which they
were derived.
From an examination of Fig.3.22 it can be deduced that an increase in Rf
of 4X103 MPa is equivalent to a decrease of 0.01% in Pc. thus it would appear that
chemical composition , Pcm of the parent plate is the most influential factor in the
predictive equations. The importance of Pcm indicates that chemical composition is
a controlling factor by influencing the microstructure of the HAZ. The approach of
Ashby and Easterling(67), Albury et al.(66) and Ions et al.(75), using fundamental
equations for heat flow and diffusion do not at present provide predictive equations
that encompass the three intrinsic parameters of stress, hydrogen content, and
microstructure, but instead predict, in some detail, welding microstructure diagrams
for the HAZ. (see Section 2.32). There are voids in these analyses also. For
example, the possible effects of precipitation hardening and the effects of
autotempering have not been considered. However, in principle, the approach can
59
be used to predict the microstructure in any region of the HAZ(75) that was
produced by any combination of extrinsic variables. Hence, accepting the basis on
which the analyses are based, there does not appear to be any reason why they
cannot be developed to encompass all possible microstructures in the H A Z . The
ultimate aim is thus to select welding conditions to avoid the formation of crack
sensitive microstructures, or alternatively to asses the possibility of achieving
desirable structures.
60
CHAPTER FOUR
WELDABILITY TESTS
4.1 INTRODUCTION
In the early 1950's, Granjon(123), at the request of the HW, collated and
classified the weldability tests in use at that time. Sixty tests which were used to
determine various aspects of weldability were mentioned. Numerous refinements
and additions to the list have been made in the past 30 years and further surveys
have been published by Granjon(124,125) and Linden(126). In general, there has
been increased emphasis on weldability tests to determine cold cracking of the H A Z
welded steels.
Weldability tests to assess cold cracking may be used for three purposes,
namely:
(i) to determine the predominant variables that govern cold cracking,
(ii) to compare the relative susceptibilities of steels to cold cracking, and
(hi) to develop proceedures to prevent cold cracking.
A large variety of tests have been developed to attain one or more of the
above uses and , in general, may be classified into one of the three following
categories:
(a) self restraining tests, such as the Controlled Thermal Severity (CTS) test, the
Lehigh Test, and the Tekken Test,
(b) externally loaded tests, such as the Implant Test, the Tensile Restraint (TRC)
Test, and the Rigid Restraint (RRC) Test, and
61
(c) simulated tests such as the Constant Load Rupture (CLR) test.
In Chapter 3 it was pointed out that cold cracking was caused by three
interacting intrinsic variables, microstructure, stress and hydrogen concentration.
The abovementioned categories of test attempt to incorporate these variables and the
results obtained from the tests should be interpreted strictly in terms of the variables.
With the self restraint tests the results are generally expressed in terms of the
welding conditions, heat input, and preheat, and the extent of cracking is often
expressed as a percentage of the weldment longitudinal cross-sectional area. The
external loading tests provide a means of changing and measuring the load
independent of the other variables. Hence, at least one intrinsic variable, namely
stress can be expressed quantitatively as a test result. The simulation tests are more
of scientific value. For example, using resistance heating, a simulated thermal cycle
may be used to produce a particular H A Z microstructure in a specimen of sufficient
size that can be hydrogen charged to a known concentration and the fracture stress
measured. The relationship between data generated by these tests has significance in
generating criteria for mathematical modelling rather than relating directly to real
welds in service.
There have also been numerous attempts to correlate the results of the
various tests with each other and with real weld situations, but the results of these
comparisons have often been contradictory and confusing. For example,
Evans(127) suggested that a good correlation existed between the C T S test and the
Implant Test, whereas the opposite conclusion was drawn by Hart( 128). Satoh et
al. (129) indicated that a good correlation existed between Implant Test results
and those from R R C and T R C tests; Fikkers(130) found no correlation at all.
WILD-- ^SPECIMEN PVATE
*LlO»vENT. •ASMCR
^ \ SPECIMEN PLATE SHOW*
[_ff" 1 SECTONCD AT IM*LA*T
BAU. AMO SOCKET JOWT
TO PNEUMATIC LOAOIMC CTLWOCR
FIQ. 4.1.Diagram showing the details of an Implant Testing machine(221).
F IG- 4-2- Photomacrograph of a section of an implant specimen showing the the helical notch in the H A Z . (132)
62
Additionally, test results from the same weldability test often differ because of
different testing proceedures (131,132,134).
Weldability tests used to determine HAZ cold cracking susceptibility are
usually of the self restraining, or externally loaded type. The purpose of this
Chapter is to describe a number of the more widely used weldability tests and to
discuss their applicability to weldability and the development of weldability
equations. The external loading type tests, such as the Implant Test is discussed in
Section 4.2, the T R C Test in Section 4.3, and the R R C Test in Section 4.4. The
more common self restraining tests such as the Lehigh Test is presented in Section
4.5, the C T S Test in Section 4.6 and the Tekken Test in Section 4.7.
4.2 THE IMPLANT TEST.
The Implant Test operates on the principle of applying a known stress to a
real H A Z . The implant type of test was originally used by Granjon(206) who
subsequently modified the technique (207) with the object of quantifying the
welding behavior of high strength steels.
In this test, a helical notched cylindrical implant specimen is machined
from the steel to be investigated, and is inserted in a close fitting smooth hole in a
steel plate. A test weld bead is deposited over the end of the specimen, and when
the weld cools to a predetermined temperature, a load is applied to the free end of the
specimen for a long period of time, or until fracture occurs. Fig.4.1 shows a
schematic form of the implant test machine. The notch produces a stress
concentrator, and because of the helical form, the location of the stress
concentration occurs at some point in the H A Z , see Fig.4.2. By carrying out tests
FIG.4,3. Diagram showing the results of Implant Tests. The diagram also shows the reproducibility of test results derived from a separate series of cracking tests(142).
S M ) 86 25mm tWdt, E3«ctTO€%» BH-KT 4mm cfta. 170 Amp., 150 mm/mn, no pr*>«o1
m
e E v. M.
m m
s M M
t t c
GO
50
*0
30
20
10
0
t
\ \ \ \ \ Oock \ \inttiatlon •
rv1 &
« t i . . 1 i
3 3 10 20
•
K)0% \cradsHj \ \ .
\ \ \
4 \
! !•!..! 50 ICO 1 ' 1 2
Loading tun*
Crock initiation
o Restroinad oft«r wtldlng Rwlroin«d
during wtkling
\
, - \ A *
O
1 , . ' • 1 . ,| 200 ,r™M° «»
! ' i t 3 IO 20
hri
00% crock •
*
IP" *fc
i 1 2000
_1 30
F1Q-4.4. Diagram showing the effect of tensile load and rupture time in T R C tests of steel weldments(137).
63
at various stresses, a diagram of the type shown in Fig.4.3 can be produced.
The Implant Test has been found to be useful in grading steels,
particularly H S L A steels. Gordine(208) compared the weldability of several Arctic
grade line-pipe steels using the Implant Test, and the C T S test, and compared the
results with measured hardnesses of the H A Z , and cold cracking predicted using the
IIW and Ito-Bessyo carbon equivalents. The Implant Test was found to be more
sensitive in grading the susceptibility of the steels to cold cracking than the C T S test
and carbon equivalents, particularly, the IIW version, which correlated poorly.
Unfortunately standardization of the Implant Test procedure has not been
achieved, although, attempts have been made by the IL Committee of Japan (209)
and the H W Subcommission 1 X B (210). Therefore there are many Implant Test
results available from various sources that are not readily comparable.
4.3 THE TENSILE RESTRAINT CRACKING TEST.
The Tensile Restraint Cracking (TRC) test, developed by Suzuki et al.
(137) involves the butt welding of a pair of steel plate samples, and, immediately
after cooling the application of a constant tensile load for a long period of time until
root cracking occurs in a manner similar to the Implant Test technique. Figure 4.4
shows an example of test results. W h e n a specimen is subjected to a certain stress
a crack initiates at a certain time and growth is time dependent. As the stress level
decreases, the time for crack nucleation and propagation increases. Below a certain
level of stress, termed the critical stress, cracking does not occur. The higher the
preheat temperature and the smaller the diffusible hydrogen content of the weldmetal
the greater the critical stress becomes until it finally reaches the yield stress of the
weldmetal(137,138).
3£
lovable r.hurk r-L
"* 1=
V ^ Gage length of
restraint l —
Fixed chuck
FIG.4.5. Diagram showing the developement of restraint stress during cooling and cracking in the R R C test(164).
100
•o -
so
P 40
20
RvJiuinf length I
L-300 mm
StHCHB-A 2O0O ion RRC Ml MOOTTt Mffpvohn
Spoomtn vitfffc. 1000 inn
l«500mm CRACKING ZOHC
» No cracking
Crocking
.' " >'" v.*i?^1-120OJ5*. " V ' , - - ..— - _i-- I-1300m*
J L_i_L i 10 20 40 SO
TM
1 t I too 200 400 600 COO 2000 4000
FIG.4.6. Diagram showing the restraining stress developed during cooling and cracking of RRC test specimens(164).
64
4.4 T H E RIGID R E S T R A I N T C R A C K I N G TEST.
Introduced by Satoh et al. (157) the Rigid Restraint Cracking (RRC) test is
based directly on the concept of restraint intensity (see Section 3.73). A certain
restraining length is kept constant during the cooling cycle of the weld by means of
an automatic control mechanism.
Diagramatically the experimental arrangement is shown in Fig. 4.5. By
repeating the test for various constraint lengths, the critical restraint intensity and the
critical net stress can be determined for a particular material, groove geometry, and
welding conditions. The type of results obtained from the R R C Test is shown in
Fig.4.6. Unlike the Implant Test, and the T R C Test, the R R C Test is based on the
natural transient stress produced on the joint itself during cooling. It incorporates
not only the cumulative effects of the weldmetal and parent plate contraction, but
also the reducing effects that transformation expansion has on the reaction stress.
The T R C and Implant Tests do not incorporate the role that transformations play in
producing the stress state(129).
Specimen size for the RRC Test depends primarily on the size of the
hydraulic test machine. The most powerful hydraulic R R C Test machine currently
in use is at Kawasaki Heavy Industries Ltd., and has a capacity of 2 0 M N so that
specimens 1000mm wide and 3 0 m m thick can be tested(135). Generally specimens
with a width of 100mm are used(129,136).
"I
L-.f-1-f-l
PtatB
Groove for w*idng
4" 900
Omit n pct»< 3 rr«3t
FIG. 4.7. Diagram showing the shape and dimensions of the Lehigh Restraint Test specimen(211).
FIG. 4.8. Photomacrograph showing typical crack formation in the Lehigh Restraint weldability test(213).
65
4.5 T H E L E H I G H R E S T R A I N T TEST.
Developed by Stout et al.(211), the Lehigh Restraint test ( Fig. 4.6)
was used to quantitatively determine weldability, by measuring the degree of
constraint required to produce weldment cracking. The degree of constraint can be
varied by changing the length of the slots along the edges of the plate. The width of
specimen that would cause cracking was expressed as a measure of weldability in
terms of restraint. The purpose of the test was then to rate steels and electrodes
quantitatively and to express their weldability in a numerical form.
In terms of a quantitative determination of restraint stress. Stout et al.(211)
pointed out that there was little point in the exercise because of the difficulty " in
evaluating the nature of the stress in the weld itself, particularly in the critical notch
area of the weld root" and " it is almost impossible to predict quantitatively the
restraint which a given joint in a structure will possess".
Kihara et al. (212) carried out an examination of the weldability of high
strength steels and electrodes using both the Lehigh Restraint Test and the Tekken
Test. They concluded from the location of cracks observed in transverse
metallographic sections that the Lehigh Test was preferable for examining weldmetal
properties rather than the steel plate properties. This conclusion was based on the
observation that for the Lehigh Tests cracking occurred almost exclusively in the
weldmetal. (see Fig. 4.8).
Bithtrmd test weld
Bottom plote
^ffh ' 7 &
plote
KJ L L J "i'd.otw*.
FIG. 4.9. Diagram showing the configuration and dimensions of the Controlled Thermal Severity (CTS ) test specimen.
66
4.6 T H E C O N T R O L L E D T H E R M A L S E V E R I T Y TEST.
The Controlled Thermal Severity (CTS) test was adopted by the
British Welding Research Association as a means of assessing the weldability of low
alloy steels with yield strengths in the range 300MPa to 600Mpa. The composition
range covered steels with a total alloy content of less than 4 % and carbon in the
range 0.1% to 0.3%. The testing technique was based on the the recognition that
the mass of the joint determined the cooling rate for a particular heat input and the
composition of the steel determined the phase transformation characteristics.
The CTS test was developed by Cottrell et al. (201,202,203) and consisted
of a fillet weld on two sides of a 75 m m square plate on a larger restraining plate,
Fig.4.9. The restraining welds were allowed to cool before the test fillet welds
were made on the remaining sides of the test plate which could be positioned so that
one test weld could be made near the end of the base plate, giving cooling by heat
conduction along two thicknesses of plate, a condition known as a bi-thermal weld.
The other test weld was placed near the center, so that heat was conducted along
three thicknesses ie., a tri-thermal weld. In this manner the cooling conditions of the
H A Z could be varied. After welding, and cooling for an extended period to allow
cracking to occur, the test pieces were sectioned, and assessed for cracking by
metallographic examination.
Cottrell(204) developed empirical equations for the cooling rate
introduced by the joint mass and which was expressed as a Thermal Severity
Number (TSN). The T S N was simply 4 times the total thickness of the plate, in
inches, through which heat could flow away from the weld, ie. a butt weld in 1/4"
plate has a T S N of 2. Cottrell developed an empirical equation for the rate of
67
cooling, R, at 300°C (the M s temperature) for a weldment produced with an arc
energy, E, and a TSN of N, expressed as
R = 1 eqn. 4.1
By correlating the TSN, and a carbon equivalent denoted CE), of the form
CE = C + Mn_ + Ni + ( Cr + Mo + V> eqn4.2 ZO /S /O
it was possible to grade a steel with a weldability index, and using tabulated data for
electrode size (heat input), the weldability of the complete configuration could be
determined, and, if necessary, preheat prescribed.
The objective of the total analysis was to predict the welding conditions
necessary for all joint configurations in a welded structure. Correlation of thermal
severity, composition and heat input with weld H A Z cracking was achieved from
C T S Test results; the C T S Test was also used as a means of checking the calculated
conditions for welding. Unfortunately the degree of constraint of the C T S test is
low and invariable. In welded structures the restraining stress is often high so that
the C T S test often underestimated the weldability of a steel in a practical joint
configuration.
A number of workers have endeavoured to increase the degree of
constraint by welding "strongbacks" to the base plate(183), or by incorporating a
1. 6 m m gap between the plates at the weld root(218,219). However, relating these
joints to joints in welded structures is difficult because it is difficult to determine the
Unit: mra
)))))))>
u-
2M
• T»»» viliH.** •
_ •*
— X
so
Cootfrmin*
X«((((«(
-60-
Section AA'
T"
I ^ _! !_
"V //
^
B
FIG. 4.10. Diagram showing the dimensions and configuration of the Tekken Test specimen(212).
Unit: nun
Colt. nun
1.A ?••
:ii((!(((-^EiMgMgiii
FIG. 4.11. Diagram showing the weld configurations available for the performence of the Tekken test(141).
68
magnitude of stresses in the test weld configuration and to compare these stresses to
those in the welds of a fabricated structure.
The suitability of the CTS test for weldability came into question after the
failure of the Kings Street Bridge in Melbourne in 1962. It was found that brittle
failure had originated from preexisting toe cracks at fillet welds across the ends of
cover plates on the tension flanges of girders. The C T S tests carried out did not
produce such toe cracks and it was only with considerable modifications to increase
the degree of constraint in the C T S test that such cracks could be formed(183).
This result cast considerable doubt on the C T S for assessing susceptibility to cold
cracking.
4.7 THE TEKKEN TEST.
The Tekken Test was devised by the Technical Research Institute of the
Japanese National Railways(139). Initially, the test piece consisted of two parts,
each 3 3 0 m m long and 75 m m wide although the original version was subsequently
modified by Kihara et al.(140) in 1962, and incorporated as a Japanese Standard in
1966(141).
The test piece is shown in Fig. 4.10, and consists of two parts, each
2 0 0 m m X 7 5 m m which are joined by constraint welds to form a central V section
containing a 2 m m gap at the root for the deposition of a test weld. T w o methods of
depositing the test weld are available and these are shown in Fig.4.11. and the
specification(141) requires that at least 48 hours elapse after welding before the test
weld is examined for cracks. Three procedures shown in Fig 4.12, are described
for crack measurement and Cracking Index is expressed as the average of these
restraint welds rcrr.ovei)
Bool rr««>"0 Cr = tOORrt^l.
3e*d crjcvrq Cb; 100 BlA \'.
l-^rrxy- 'tiring
."Of« ACi = lC t.C r. Col/3 "7.
„<jr\| 3»»S r>».qrl
Section near gnd o< crack
FIG. 4.12. Illustration of a partially cracked weld, showing the intersection of the crack with the root and surfaces of the weld bead (23).
0 4 8 12 Cooling Rate at 300 C
FIG. 4.13. Diagram showing the relationship between CI and cooling rate at 300°C (in °C/sec) derived from Tekken Tests (212).
1UU • •
80 X to
n £60
I SAO
i i f y ^
• \ \
i
ion
-is-
STEEL
O 1022
• 1045A
A '067
1 •
•
200 400 600 800 900 Maximum Vickers hardness in heat
affected zone . Vhn
FIG; 4-14- Diagram showing the relationship between CI and maximum hardness in the H A Z for the Tekken test.(23).
69
three measurements. Hensler et al. (23) found that by heat tinting the crack surfaces
at 400°C, fracturing the specimen, then measuring the area of crack and expressing
this as a percentage of the weldmetal longitudinal cross sectional area, a value of CI
could be obtained equally as reliable as that obtained using the more laborious
techniques described in the Specification(141).
Cracking Index has been related to various aspects of the welding
process, and characteristics of the weldment. Kihara et al.(212) related CI of a
number of weldments to the cooling rate at 300°C, hardness in the H A Z , and
preheat temperature( eg see Fig 4.13). Hensler et al. (23) developed similar
relationships(Fig. 4.14). From the results of these workers it is interesting to note
that most values of CI are either the maximum(100%) or minimum(0%) with very
few values between 3 0 % and 70%. Inagaki et al.(25) found that it was necessary
to examine four test welds to ensure consistency for a " go or no go test and that at
least eleven specimens were necessary in order that cracking percentage might be
estimated in detail"(25). Hensler et al.(23 ) quotes reproducibility of the order of
+ /- 1 0 % as does Kihara et al. (212), however to achieve such results Hensler et
al. point out that heat energy input, E, (eqn. 2.4) must be maintained at +/- 5%.
Such close tolerances on E require that welding voltage be maintained within the
required range which in turn would require the arc gap to be maintained at +/-
0.15mm. This limit would be difficult to achieve consistently by a manual welder.
Values of CI measured using the Tekken Test are a consequence of crack
nucleation and subsequent propagation. From Chapter 2, it is evident that
microstructure, hydrogen concentration, and stress are the three important intrinsic
parameters which contribute to the formation of a cold crack. However, the work
of Satoh et al.(179) and Karppi et al.(162) indicated that stress concentration, rather
S3i -C '•'I '-I.
LICLMO FLJX
/ /' / ' / 7 -' SOLID "E'AL
CASE n
FIG. 4.15. Schematic representation of equilibrium interfacial energies associated with a weld bead during welding.
FIG. 4.16. Schematic representation of the effects of interfacial energy on the Tekken test weldmetal geometry.
70
than stress, is important particularly in regard to crack nucleation in the region of the
weld root of the Tekken testpiece. The work of Satoh et al.(179) and Karppi et
al.(162) also indicate that the stress concentration can vary considerably with root
angle, (see Table 3.1 and Fig.3.21c). Such variations could contribute significantly
to variability in crack nucleation conditions. The shape of the weld bead which has
been found to be determined by the interfacial energy configuration^ 14),
determines the root angle. An analysis of the interfacial energy forces reveals that
the bead shape is not a function of one value of surface tension, but is determined by
the balance of three such forces. This situation is illustrated for two cases in Fig.
4.15 in which the three interfacial energies are: y the interfacial energy between
the molten flux and the solid metal, yr_- the interfacial energy between molten flux
and molten steel and, y the interfacial energy between molten steel and solid steel,
For case l(see Fig.4.12)
Y. = Y cos a + Y cos B eqn4.4 '5\ *(! V\
and for case 2,
Y = Y cos B - Y cos 9 eqn4. 5z n f
Thus y must be greater than y5. The expected effects that differences in
liquid flux/ solid metal interfacial energy would have on the weld beads formed in
the Tekken Test welds are shown schematically in Fig. 4.16a from which it can be
seen that high values of y5 cause small root angles and consequently high stress
(b)
FIG.4.17. Photomacrographs showing cracks and bead shape produced in the Tekken test using (a) low hydrogen electrodes, and (b) high titania electrodes. Note the the change in weld bead geometry and root angle, X5, (212).
71
concentrations. Conversely, low values of y s would tend to contain the weld
bead, resulting in large root angles, and low stress concentrations, Fig.4.16b.
Amongst the various factors which can effect surface energy, the
composition of the weld slag is probably the most influential. Electrode flux
composition not only effects the nature of the gas shielding of the welding arc and
the metal/slag refining reactions that were discussed in Chapter 2, but also the
composition of the resulting slag and interfacial energy relationships. In this latter
respect it has been found that the flux composition can effect weldmetal penetration
into the parent plate(214) and weld bead shape(215) by variations to the interfacial
energy configuration. From Table 2.2 it can be seen that the type of electrode is
based on the composition of the flux. Kihara et al.(212) found that bead shape and
cracking behavior changed with electrode type as can be seen from Fig.4.17a and b
but they proposed nb reason for the change in location of the crack when the high
titania electrodes were used. It can be seen however, that there is a shift in the
location of the stress concentrator. It is generally accepted that cracking of the type
shown in Fig.4.17a is typical of Tekken Test cracks for low hydrogen electrode
welding. With modern electrodes of a particular type and conforming to a particular
specification, e.g. basic low hydrogen type, E4816 classification, differences in flux
coating composition can occur, particularly from different manufacturing sources.
Specifications for electrodes refer only to mechanical properties of the weldmetal,
not to penetration characteristics and weld bead geometry. Hence variations in flux
composition would be expected to contribute to variability in general weld bead
shape hence root angle and thereby could contribute to the determination of
weldability.
72
The Japanese Specification for the Tekken Test(141) does not
specify welding conditions. A number of workers have chosen an arbitrarily heat
input of 1.7kJ/mm(173,175,176), however this value is not used universally
(23,24). Although Ito and Bessyo(173) considered that test results were reliable
over the range 1.7kJ/mm to 3.0kJ/mm, Hensler et al. (23) found that close control
of heat input was necessary. Changes in welding conditions can also influence
weld bead geometry(234,235). Such changes are not a function of heat energy
input, but optimum weld geometries are achieved by correct combinations of
welding current, voltage and speed. The variations to weld bead geometry
presumably arise due to surface tension effects caused by differences in the arc
temperature causing higher temperatures in the molten weld pool. In Tekken Tests
that are performed by manual operators, variations to heat input energy(23) and weld
bead geometry(24) are likely to occur.
In the experimental work carried out by Karppi et al.(162,216) to
determine Fj (see sect.2.73), welds were deposited using an automatic weld
deposition technique. From a perusal of the results of Karppi et al. (Ref.162, Table
3), the standard deviation of ¥x was extraordinarily high for the Tekken Test weld
geometry (0.16) if compared with other joint geometries examined (0.08). The high
standard deviation of F : values suggested possible reason for scatter in Tekken Test
results.
The variability of root angle and the consequent effect on the magnitude of
stress concentration would appear to have a significant effect on crack nucleation.
However these effects are complex and at present unpredictable not just in relation
to the Tekken Test but also in relation to welded structures.
73
Regarding crack propagation, it can be seen from Fig. 4.17a that the
crack follows a path, in general, normal to the tensile stress direction. For this to
occur the crack path necessarily passes through a portion of the H A Z in the plate
material, and then continues in the weldmetal. Therefore it would appear that
measured values of CI relate to the fracture characteristics of both the H A Z of the
plate, and the weldmetal. It is conceivable therefore, that a steel highly susceptible
to H A C , could register low values of CI, if welded with suitable electrodes, because
of the difficulty of crack propagation in the weldm£t<3 /. The low value of CI in
such a situation would not indicate a true assessment of the weldability of that steel.
The argument put forward by Masubuchi(164) questions the value
of CI data for assessing how a steel will resist cracking. H e pointed out that for a
small crack of length 2C, in a weld length L, the CI would be
CI = 2C. eqn.4.5
As the crack grows in length the stress intensity factor K, at the crack tip, which is
given by
K = G J K C eqn4.6
also increases. However as the crack increases in length, the length of the weld
supporting the load decreases and as a consequence a increases, ultimately leading
to unstable crack growth. This suggests that self restraining weld tests will have no
cracks, or only short cracks, or 1 0 0 % cracking which appears to be in agreement
74
with the results shown in Figs. 4.13 and 4.14 indicating that very few CI results are
within the range 3 0 % to 7 0 % .
There appear to be several important points regarding the reliability and
pertinence of the Tekken Test results for weldability. First, the results of the
Tekken Test apparently incorporate a high level of scatter which could be related to
either or both crack nucleation and crack propagation. The extrinsic factors that can
effect these two phenomena in the Tekken Test have, as yet, not been fully
investigated. Furthermore, in the Tekken Test the crack follows a path involving
the H A Z and the weldmetal. It would therefore seem more appropriate to relate
weldability test results to the total weldment which incorporates both the weldmetal
and the H A Z .
With the above points in mind, it should be pointed out that workers have
tended to be conservative (e.g. see Fig. 3.22) in applying Tekken Test results to the
development of weldability formulae.
75
SCOPE OF THE PRESENT WORK
In this Chapter, the reasons for the choice of materials, the techniques used,
and in particular, w h y the Tekken Test was chosen as a topic for study, are
discussed. Although several weldability tests have been reviewed in Chapter 4, a
brief mention of some of the shortcomings of weldability tests will be considered to
serve as a background for a discussion of the proposed program of work.
Possibly, one of the greatest criticisms of weldability tests relates to their
origin. Granjion(123), after a review and classification of weldability tests,
commented that," the greatest part of these tests at present in use have been devised,
after a failure had occurred during welding or in actual service, in order to explain
these failures or prevent their recurrence. These failures are in most part the result
of a complex conjunction of circumstances and material properties; and the tests
themselves are difficult to interpret". If weldability tests were applied only to those
areas from which they originated, then problems could be minimized. However,
there is a tendency that the tests begin to be applied in areas for which they are not
suited. For example as pointed out in Chapter 4, Sect. 4.6 the C T S test was
envisaged to apply to steels with tensile strengths in the range 300 to 600MPa. The
construction of the Kings Street Bridge in Melbourne utilized material outside this
range (650MPa), but more to the point the C T S test did not reproduce restraint
conditions commensurate with those of the structural components of the
bridge(183).
76
The C T S test and the Kings Street Bridge failure indicate two points of
concern with weldability tests; first the relationship, if any between the test weld
configuration and conditions and the welds in the actual structure, and secondly, the
range of conditions over which test results are reliable.
To complicate matters further, weldability tests have been used to develop
empirical equations and formulae to predict H A Z cold cracking. The Tekken test has
been used extensively by Ito and Bessyo( 168,173) to develop such equations. The
results of weldability tests have been obtained using low alloy steels and the
equations would only be applicable to those steels. However the criterion for
cracking relies on quantitative data (of the type shown in Fig. 3.22) and as pointed
out in Chapter 4, Sect. 4.7 the significance of relating a Cracking Index value of
5 1 % to the parent plate where cracking of the weldmetal is also involved.
In relating a restraint cracking test to an actual welded structure two
considerations must be made. First, it must be assumed that the restraint intensity
of the actual joints in the structure is less than that generated in the restraint test
piece. Secondly, that the cooling rate generated in the test piece is more severe than
that of the actual structure so that the microstructures formed in the H A Z of the test
piece are more susceptible to H A C than the service microstructures.
Masubuchi(164) considered that the upper limit of the restraint intensity for practical
joints to be 40h, (where h is the plate thickness) whereas the restraint intensity
generated by the Tekken Test has been measured to be 70h(173). Restraint
intensities have been measured experimentally for a number of structures, for
example welded ships(78) and pressure vessels(164), and, in general, values of the
order of 40h and lower were most common. However, at several locations in a
pressure vessel, butt welded joints had restraint intensities in excess of 70h. The
77
relevance of intensity of restraint is also a point of contention because stress
concentration would seem to be more relevant (from chapter4) but more difficult to
predict or control. This factor is apparent from the results of Karppi et al.(162) in
relation to the Tekken Test and would be expected to apply to actual joints in welded
structures.
Nevertheless, the extensive use of the Tekken Test in Japan, and in
particular, its continued use in the development of empirical equations aroused
considerable interest in Australia. This interest increased, as various laboratories in
Australia endeavored to apply the Tekken test to grade the weldability of steels only
to be confronted with the scatter of results discussed in Chapter 4(217). The
Australian Welding Research Association ( A W R A ) , requested the Institute of
Materials Research in 1967 to investigate the Tekken test, with the view to
standardizing the test procedure, and conditions, and if found reliable,
recommending the test and associated procedures as a field or workshop test for
weldability and to have it covered by an Australian specification.
Hensler et al. (23) carried out the investigation for the Institute of Materials
Research and examined a range of specimen dimensions and test procedures, and
made specific recommendations in those respects. They also pointed out that, for
consistent, reproducible results, close tolerances of welding conditions were
necessary. Subsequent to the publication of the work of Hensler et al. (23) the
A W R A initiated an interlaboratory Tekken Test program which incorporated the
Tekken Test results from eight laboratories comparing two steels. Welding current
and speed were specified, and electrodes were supplied from the same
manufacturing source. The scatter of results that was obtained is indicated in Table
5.1., where it can be seen that between laboratories, the results ranged from 0 to
TABLE 5.1.
The test results reported by the participating laboratories for Tekken test welds carried out with a preheat of 50°C.(24).
Preheat
50°C
Labi 7 3 4 5 , g 7
8
C
PLATE
71 0 38 0 25 43 0 1.5
racking
A
86 0 23 0 55 0 0 0
Index 100 *
PLATE B
88 0
. 54 0 79 0 75 100
86 0
100 0 94 0 25 100
FIG.5.1. Photo macrographs of transverse sections of a number of the Tekken test welds carried out in the interlaboratory comparison, preheat 50°C.(24).
78
1 0 0 % cracking, and for independent laboratories, differences of 4 0 % were not
uncommon. These results cast considerable doubt on the Tekken Test as a general
test for weldability.
From an examination of macrosections of the weldments produced by the
various laboratories in the comparison, it was apparent that considerable differences
existed in the weld bead size and geometry, Fig.5.1.
An apparent causal relationship between weld bead differences and test
results scatter was assumed and prompted the A W R A to consider a further
appraisal of the Tekken Test using a mechanical technique of weld deposition. It
was anticipated that the mechanical welding approach would eliminate variations
due to manual welders. The current program of work was thus embarked upon.
From Fig. 5.1, it is apparent that not only were weld bead size and
geometries often different, but that cracking often occured in different locations
(compare A l and A 6 in Fig.5.1). If variations in welding conditions do influence
test results, then the influence is related by the intrinsic parameters of microstructure
and stress. This conclusion is based on the assumption that all electrodes have been
dried in an identical manner so that weld metal hydrogen concentrations were
constant. It was therefore proposed, that the aim of the research to be carried out
would be to examine, in detail, the effects that variations of welding current,
voltage, and speed had on measured values of CI. To maintain consistency, a
mechanical deposition technique was required, to provide independent control over
each of the welding parameters. The aim was to interpret the CI extrinsic
parameter relationships in terms of observed variations to the intrinsic variables of
microstructure and stress. O n this basis, a research plan was developed, Fig. 5.2,
Welding Variables Current Speed Voltage
JMtc rostmcture Cracking Index • • — g ) — Stress
Significance of
The Tekken Test
FIG- 5.2. Diagram showing the planned approach with which the research was to be carried out The numbers indicate the Stages and the sequence of the work.
79
in which can be seen that the work was divided into a number of Stages to ensure
programmed development. The Stages are identified numerically in Fig.5.2 and
were the basis of the sequence of the experimental work.
In an endeavour to maintain a degree of consistency with the previous
work, the steel chosen for the study was of the type and grade of one of those used
in the interlaboratory comparison, namely A S 1204-350. Added advantages in the
use of this steel include that:
(i) it has a carbon equivalent of 0.35 which according to the A W R A weldability
guide (7) can be welded successfully without preheat, but does require a degree of
care to avoid H A Z cracking, and
(ii) it is a grade of steel commonly used in welded structures, so that any new
information generated regarding the weldability of the steel, would have obvious
advantages.
The electrodes chosen were also the same grade and type as those used in the
interlaboratory comparison, namely 4 m m E 4 8 1 6 basic low hydrogen, and from the
same manufacturing source. This choice also offered the advantages of use of
electrodes that are in general use and because they were of the type and grade used in
the Interlaboratory Comparison(24) results could be beneficial in developing an
interpretation of the results of the Interlaboratory Comparison.
By carrying out the research according to the plan shown in Fig. 5.2, it
was intended to relate CI to variations in the extrinsic parameters of current, voltage
and speed and interpret these in terms of the intrinsic parameters of microstructure
and stress. Another objective was to examine the scatter of results and to interpret
these in relation to the intrinsic parameters.
80
C H A P T E R SIX
EXPERIMENTAL
6.1 INTRODUCTION.
The general and specific aims of the research have been described in
Chapter 5. To facilitate such a programme of work; materials, equipment and, in
particular, techniques had to be developed in order to achieve reliable results. Given
the inhomogeneous nature of the MMA welding process and the reported, and often
conflicting, scatter in results from other investigators, reproducibility was
considered to be most important. This Chapter is thus devoted to the experimental
phase of the work, and describes not only the techniques and equipment, but also
the difficulties that were encountered and how these were overcome or minimised.
In a number of cases, investigations diverged from the main stream of the
work. For example, the majority of test welds were carried out using one grade of
electrode, from one manufacturing source. Similarly, the general research involved
only one grade of steel, however, two other steels, one with a higher and one with a
lower carbon equivalent were also examined. Nevertheless, the main thrust of the
research was directed along the lines set out in Fig. 5.2.
As pointed out in Chapter 5 two intrinsic parameters were examined,
namely microstructure and stress. Although both were interrelated by the welding
variables of current, voltage and speed, they were examined separately in two sets of
experiments. The overlap that occurred in the two sets of experiments was found to
be useful in examining reproducibility.
(a)
(b)
n. *
*~si r -».<<«* *«£ ,' 4, J-.». ~. 4
§ssi iii^^^^^s£^>^ r *
x<>r* - -^yf • , j> Z2~r-*^-^ ^ —4*
^&^f^-^ > •
;
(c)
% y " ^"l.'% '"•^;f,«*:«r*B*K—•
FIG. 6.1. Photo micrographs showing the microstructures of the three steels used a) 0.3 Ceq, 140HV(20), b) 0.36 Ceq, 153 HV(20),
c) O.Ceq, 157HV(20). X150; etchant, 2.5% nital.
T A B L E 6.1 The three grades of steel used in the current investigation, showing the chemical analysis, carbon equivalenuirW), and the mechanical properties.
Mechanical Props.
Y.S(MPa) T.S. Elong
250 410 22%
350 480 18%
520 16%
Steel Grade
AS 1204-250
AS 1204-350
AS 1204-400
c« .3 0
.35
.41
Chemical Composition
C Mn
.14 .96
.16 1.16
.16 1.43
Si
.13
.23
.40
P
.025
.025
.026
S
.01
.01
.005
T A B L E 6.2. The chemical composition and the mechanical properties of the weld deposits produced from the three grades of electrodes used in the present work.
Grade
E4816
E6218M
E7618M
Chemical Composition
C Mn Si Ni
.12 .9 .60
.07 1.0 .44 1.6
07 1.5 .50 2.1
Cr Mo
-
.3
.2 .4
Mechanical Props.
Y.S. T.S.
480 550
600 680
760 830
Elong
31%
26%
23%
81
This Chapter provides a description of the materials used in the
investigation, (Sect 6.2), the methods of Tekken Test piece preparation, (Sect. 6.3),
and the welding equipment and procedures, (Sect. 6.4). The experimental
techniques involved in analysing test welds, including measurements of CI, are
presented in Sect. 6.5, the metallography associated with the weldment in Sect. 6.6
and the techniques for stress measurement, Sect. 6.7.
6.2 MATERIALS.
The steel plate used for the research was a commercial grade C-Mn steel
covered by the Australian Specification, A S 1204-350 and the composition,
mechanical properties and carbon equivalent are shown in Table 6.1. A limited
number of test welds were also carried out using two other steels, the compositions,
mechanical properties and carbon equivalents of which are also shown in Table 6.1.
All plate was 2 0 m m thick in the rolled condition with microstructures comprising
ferrite and pearlite as shown in Fig. 6.1. Test specimens were cut from the plate in
such a way that the test welds were deposited transverse to the rolling direction.
The electrodes used were commercially produced 4mm, E4816,basic low
hydrogen type and the composition and mechanical properties of the weld deposits
are shown in Table 6.2. A number of test welds were also carried out using two
other grades of electrode, namely E6218M and E7618M and the compositions and
mechanical properties of these are also shown in Table 6.2. In addition, a limited
comparative examination was carried out using electrodes of the same specifications
but from a second manufacturing source. These were denoted as being from
"source B ", but because they conform to the specifications previously mentioned,
compositions and mechanical properties are as shown in Table 6.2. Prior to welding
FIG. 6.2 Photograph showingthe Tekken Test piece clamped in the jig holding the two sections in place for the deposition of restraint welds.
TEST WELD-
RESTKAINT BOLTS
(with strain gauges attached
to
measure restraint loads)
'HIKIIM'ii-iilll'l.K
'[constrained piece)
•r7 . "V TEKKEN TEST PIECE (fixrd piece J
FIG. 6.3. Schematic representation of the experimental arrangement proposed to measure restraint loads.
82
all electrodes were dried in accordance with the manufacturers recommended
procedure by baking for 2hours at 250°C followed by storage at 110°C.
6.3 SPECIMEN PREPARATION.
Two specimen configurations were used for Tekken Test pieces in the
current work. The standard Tekken Test piece ,( see Fig. 4.10) was used in Stages
1,2, and 3 (refer to Chapter 5) and a modified version of the Tekken Test piece was
used in the experimental work associated with the restrdning load measurements.
For the research in Stages 1,2 and 3 the two parts of the test piece shown
in Fig. 4.10 were machine sawn from the plate material so that the bevel edge was
transverse to the rolling direction. Prior to assembly, the two parts were deburred
and the bevel edge examined for accuracy using a set of profile gauges. Section
dimensions were also examined for conformity to the tolerances set down by
Hensler et al. (23). Specimens that did not conform were rejected and, if possible,
subsequently machine milled to the correct dimensions. The technique described by
Hensler et al. (23) was used initially to construct the Tekken Test piece. This
involved placing the two parts of the test piece in a jig, Fig. 6.2, and setting the gap
using a 2 m m thick, flat strip of metal as a feeler gauge. The test piece parts were
clamped in position by tightening the the retaining nuts, see Fig.6.2, the jig and the
specimen were then heated in a furnace for lhour at 200°C. Upon removal from the
furnace the restraining welds were deposited using 4 m m E6218M electrodes. It
was found that using this technique the specified 2 m m gap could be maintained
reliably. Cambell(218) had similar difficulties but, by machining the length of the
specimen where the restraining weld was to be deposited so that abutment of the two
83
parts occurred, closure of the 2 m m gap was eliminated when the restraining welds
were deposited. Investigators at A.I.S. Port Kembla(217) inserted 2 m m thick and
5 m m wide pieces of steel in the gap at the two points where the gap between the
restraining welds and the test weld existed. This latter technique was found to be
satisfactory, mainly because it was convenient and required no extra costly
machining. Furthermore, it was possible to vary the thickness of the strip and so
investigate the influence of the root gap on cracking behaviour. Because of the
location of the steel strips at the beginning and end of the test weld, it was possible
to start the arc on a steel strip and finish the test weld on a steel strip so that defects
at the weld ends, such as crater cracking could be avoided, or at least minimised.
To measure restraining loads several possible experimental techniques
were examined. First it was possible to fill the V-joint at the restraining weld
completely with weldmetal, machine both surfaces, then attach strain gauges. For a
large number of test welds this can become expensive and difficulties exist
regarding load calibration. Nevertheless, a number of test welds were performed
using this approach to assess the load regime expected to be encountered.
A second experimental arrangement is shown schematically in Fig. 6.3.
This arrangement is similar in principle to the Rigid Restraint Test developed by
Satoh(157) in that it not only allows self developed restraint to be measured but can
also be used to impose higher or lower restraint loads by tightening, or loosening the
nuts on the restraint bolts which also function as load cells. This experimental
configuration could be used as a R R C or T R C test device. The concept embodied
in this device was that cracking could be monitored, and that by rapid unloading at a
particular level of load during the crack propagation period, the results could be used
to develop crack length/ stress relationships. A prototype was produced and it was
FIG 6.4. Schematic representation of the second alternative experimental arangement proposed to measure restraint loads.
Midi Volts
FIG. 6.5. Diagram showing the load/milli-volt calibration relationship for the two load cells.
84
found that the magnitude of the loads recorded were approximately 1 0 % of the load
measured when the direct measuring technique was used. Furthermore, cracking
could not be induced under welding conditions that were known to produce
substantial cracking in the standard Tekken Test piece. It was therefore concluded
that the equipment was "soft" and the thermal contraction developed by the cooling
weld was being accommodated as flexure in the equipment and not as restraint on
the weld and the H A Z . It was considered that while the general concept was valid
there would need to be considerable time spent and cost involved in developing the
equipment to such a stage that reliable results could be achieved. Therefore the
third possibility was considered.
The third experimental arrangement investigated is shown schematically in
Fig.6.4. Load cells consisted of steel blocks with strain gauges glued one to each
side of the block in accordance with the gauge manufacturers recommendation.
Using a Wheatstone Bridge circuit coupled to a pen recorder, the load cells could be
calibrated in compression in a calibrated testing machine and the load (in kN) vs
voltage (mV) calibration is shown in Fig. 6.5 for both cells. To accommodate the
load cells the two parts of the Tekken Test piece were accurately machined so that
when the test piece was assembled the required 2 m m root gap was produced. The
experimental configuration was clamped together using two 1 0 m m high-tensile bolts
(see Fig.6.4). In order that loads due to thermal expansion could also be measured
the bolts were tightened to a preload of 20kN prior to deposition of the test weld.
Thermocouples were inserted in the steel load cells adjacent to the strain gauges so
that temperature rises that might occur due to the test weld could be measured. The
temperature rises detected were found to be less than 50°C. Temperature was
monitored in the load cell for each test weld and corrections were made to the
recorded voltages for temperature rises. The correction factor quoted by the strain
FIG. 6.6. Photograph of the equipment used for the mechanical deposition of the test welds showing A- work table B- electrode feed C- electronic controller D- welding transformer E- pen recorder
85
gauge manufacturer was l^m/m/K. Between test welds the load cells were checked
for conformity to the kN/ m V calibration.
A chromel/alumel thermocouple was also located at the centre of the test
weld length in part B of the test piece see Fig4.10, and the thermocouple bead was
positioned approximately 1 m m below the surface of the bevel face. This was
achieved by drilling a 1.5mm diameter hole from the back surface of the specimen.
After the test was performed the exact location of the thermocouple bead was
determined by preparing a metallographic section.
6.4 WELDING EQUIPMENT AND PROCEDURE.
For the deposition of test welds an automatic welder was used which was
developed by Steelmains Pty. Ltd. and is shown in Fig. 6.6. This equipment
allowed welding current, voltage and sp <$ed to be preset to desired values, and so
remove the possible variations that arise during the manual operation of covered
electrode welding.
Essentially, the equipment consisted of a work table (A), capable of travel
speeds from 130mm/min to 700mm/min. through a variable speed drive from a 3
phase electric motor. In the present work welds were performed in the downhand
position with the table horizontal. However, the table could be set at a range of
angles so that welding could be carried out at positions of downhand, vertical to
horizontal overhead. Similarly, the electrode feeder mechanism (B) could be set at a
range of angles with respect to the work. All welds were carried out an angle of 12°
to the work piece. The electrode was clamped in an insulated electrode holder and
200 -
20 25
VOLTS
30
FIG. 6.7. Welding current/voltage relationship for the welding transformer used in the present work (droop characteristics).
86
the feed was accomplished through a screw drive from a D C servo motor controlled
by an electronic controller.
The electronic controller (C), provided overall control and coordination of
the electrode feeder, the work carriage and the welding generator as well as
displaying the welding current and voltage during the welding operation.
Throughout the welding process the controller not only sensed and displayed the
welding current and voltage but also regulated the motion of the electrode feed to
maintain the electrode voltage at a constant set value. The preset voltage was
achieved through a lockable 10 turn dial and preset push button which caused the
preset voltage to be displayed on the digital meter.
During operation, control of the electrode drive was achieved by the
difference between the actual welding voltage and the set voltage. For example, if
the welding voltage was above the set voltage the controller would accelerate the
electrode downward to shorten the arc and reduce the welding voltage. If the
correction process was carried too far the reverse process occured. By suitable
adjustment to the gain of the amplifier in the controller the welding voltage could be
held at the set level by a constant downward drive speed.
The welding current was controlled by means of the current setting
displayed on the front panel of the arc welding transformer unit which was a
Welding Industries of Australia Pty. Ltd. unit type Weldarc 230. The droop
characteristics of this unit are shown in Fig. 6.7. It was often found that current
settings did not correspond with measured values displayed on the controller and
several trial welding runs were often necessary to adjust the desired preset current
and voltage conditions.
87
O n the electronic controller a recorder outlet enabled the voltage and current
signals to be displayed as a function of time on a recorder (E). Voltage and current
measured were true R M S values accurate to less than 0.5volt and 5amp respectively.
Both the table drive and the electrode feed could be operated independently
allowing the work and the electrode tip to be moved to any desired location for the
commencement of welding.
To carry out a test weld the following procedure was followed: the table
speed, welding current, and voltage were set to the required values after which the
specimen was placed in position such that the location of the electrode tip
corresponded with the edge of one of the metal strip used to set the specimen gap.
A small ball of steel wool (approx. 8 m m dia.) was placed immediately below the
electrode tip and the electrode was slowly moved down to partially compress the
steel wool. By pressing the "start" button the current flowed, the steel wool melted
and the arc started. A short ( and adjustable ) delay enabled the arc to stabilize
before the work table was activated, after which the automatic deposition of the test
weld commenced. W h e n the welding arc reached the second metal strip in the
specimen the "arc stop" push button caused the work table to stop and
simultaneously interrupted the welding current. Actual weld length was measured
using a vernier gauge on the test weld. From the chart record produced, voltage,
current and welding time could be derived and welding speed calculated.
6.5 DETERMINATION OF CRACKING INDEX
After deposition of the test weld, specimens were left on the table of the
88
automatic welding rig, undisturbed, for 24 hours. This period differs from that
specified in the Japanese Standard, JIS 3158, but is consistent with the work of
Hensler et al.(23) and that specified and used in the A W R A interlaboratory
comparison(24). A number of specimens that cooled in turbulent air were found to
have values of CI higher than those cooled on the table of the welding machine in
conditions of still air. The standardised condition given above was thereafter
adhered to. Following cooling the restraining welds were machine sawn from the
specimen, taking care not to add to or extend existing cracks which was achieved by
placing the specimen in the machine saw so that the clamping action of the vice
maintained a transverse compressive force on the weld during sawing. The saw cuts
were made at right angles to the weld line and passed through the metal strips that
had been inserted to set the root gap.
As pointed out in Chapter 4 the Japanese standard for the Tekken Test
offers several methods for determining and expressing quantitatively the degree of
cracking. These are shown in Fig.4.12. and include the surface crack ratio ( C s )
which involves measurement of cracking with the unaided eye, or another suitable
method, direct from the surface of the weld bead. Often, when cracking is not
extensive, the crack will not propoagate to the upper surface of the weld bead. Bead
cracking, C B , can be measured from 5 cross sections of the weldment. After
suitable metallographic preparation the average height of the bead root cracks can be
expressed as a percentage of the bead height. The third and fourth methods involve
dye penetration of the crack so that when the remainder of the weldment was
fractured, either by bending or tensile loading, the extent of cracking can be
deduced. This can be expressed as root cracking C R or the section area cracking C A.
The average of C s , C B and C R , is usually defined as the cracking index, CI.
™ - 6-3- Photograph showing the technique used to fracture Tekken test piece specimens, after heat tinting, in order to measure cracking index
FIG. 6.9.Photograph of the fracture surfaces produced after heat tinting a Tekken Test weld. The dark area, A, represents the cold crack, and the lighter area, B, was produced by fracturing the weld by the technique shown in Fig 6.8, X1.25.
89
Hensler et al. (23) examined the probability of errors from each technique
and concluded that heat tinting the weld crack surfaces and measuring C A did not
differ from the more lengthy determinations of CI.
The technique chosen was that of heat tinting the fracture surfaces in a
manner similar to that described in A W R A Report N o P4-71-79 and then
measuring C A after complete fracture of the weld. The advantages of using this
approach were that it was simple, and involved very little specimen preparation.
The process of determining CI involved:
(i) removing the restraining welds
(ii) heating in a furnace one hour at 400°C,
(iii) fracturing the cooled test piece along along the weld line in the hydraulic press
shown in Fig. 6.8.
This procedure gave fracture surfaces of the type shown in Fig.6.9 consisting of a
darkened heat tinted area (A) which corresponded to the cold crack produced by
welding and a lighter fracture surface (B) caused by the overloading in the hydraulic
press.
Cracking index was measured by photographing the fracture surfaces, as
in Fig. 6.9, then using a digitising tablet and Apple H E microcomputer to determine
and calculate the fraction of total cross sectional area that was cold cracked.
90
6.6 M E T A L L O G R A P H Y .
Both optical and electron metallographic techniques were used to examine
various features of the weldments. Bright field optical metallography was used to
examine polished and etched sections of the weldmetal and the H A Z . Both scanning
electron microscopy and transmission electron microscopy were used to examine
the surface of cracks in the weldment by direct examination and the use of replicas
respectively.
6.6.1 Optical Metallography.
Transverse sections were cut from the test welds then abraded and
polished according to the technique described by Samuels(219). In specimens that
had been completely fractured or had been broken open to measure CI the two parts
were placed together and strips of steel welded top and bottom to retain them in
their approximate positions prior to fracture as shown in Fig. 6.10. Specimens were
macroscopically examined which enabled an assessment of weld profile and crack
location to be made then photomacrographs were prepared for a permanent record
using a bellows type camera with vertical illumination. For microscopic examination
specimens were examined Bausch and L o m b Research metallograph before and after
etching in 2.5% nital.
To examine the characteristics of the electrode tip produced by different
welding conditions 1 5 m m of the end of the used electrodes was cut off using a
jewellers saw and mounted in cold setting resin under vacuum to enable the resin to
penetrate the flux coating, so that during metallographic preparation the friable flux
coating remained intact.
FIG. 6.10. Photomacrograph of a broken Tekken Test specimen showing the method used to hold weld sections together in their relative positions for metallographic preparation and examination. X1.6; etchant, 2.5% nital.
FIG. 6.11. A typical load/time and temperature/time trace recorded using the experimental arrangement shown in Fig.6.4. The trace covers only the early stages of the cooling process. Reduced X2/3.
91
6.6.2 Electon Metallography-
Scanning Microscopy.
In order to examine characteristics of cold cracking the weldment section
was cut from the Tekken Test piece using an abrasive cutting wheel with adequate
coolant to avoid heating. The specimens were ultrasonically cleaned in acetone and
mounted on discs for examination in a Hitachi S450 scanning electron microscope.
Transmission Electron Microscopy
To resolve the finer detail of the fracture surfaces two stage replicas
were prepared using the technique described by Scott and Turkalo(220) which
involved pressing a piece of cellulose acetate sheet that had been moistened with
acetone onto the fracture surfaces. After the cellulose acetate sheet had dried it was
stripped off the fracture surface and shadowed with platinum and carbon in a
vacuum evaporation unit. The cellulose acetate sheet was then dissolved in acetone,
leaving a positive platinum-carbon replica of the fracture surface. Replicas were
exarnined in a J O E L J E M 100U electron microscope operating at 120 K V .
6.7 RESTRAINT STRESS.
The procedure to measure restraining stress used the experimental
configuration shown in Fig.6.4. and involved the following steps:
(i) prior to tightening the through bolts the Wheatstone Bridge was balanced,
(ii) the bolts were then tightened to a preload of 20kN (20mV)
(iii) thermocouples were placed in the various locations and taped in position
(iv) the assembled test piece was placed in position on the welding rig in the correct
location for welding and asbestos fabric was used to cover all strain gauges and
wires to avoid electric short circuits that may be caused by weld spatter.
(v) load/time from the two load cells and temperature/time from the central
Time
FIG. 6.12 Diagram showing displacement/time relationships for the two ends of a Tekken Test piece. Trace A, displacement transducer at the beginning of the test weld, trace B displacement transducer at the end of the test weld.
92
thermocouple were recorded prior to, during , and after the test weld on a three pen
recorder,
(vi) weld voltage and current /time and the temperature of the two load cells were
recorded on two separate two pen recorders.
A typical recording of the three pen recorder is shown in Fig.6.11,
in which it can be seen that the start and finish of the test weld are indicated by
pulses in the trace caused by the start and stop of the table motor drive. Under
normal test conditions the chart was allowed to record for 24 hours or until
fracture occurred. In Fig.6.11 the trace displays the first 100 seconds of the cycle.
Load cell A was located at the beginning of the test weld and load cell B was at the
finish end of the the test weld. From Fig.6.11 it can be seen that load cell A initially
recorded a compressive load followed by tensile restraint, however, more
importantly the load at the start end of the weld was always smaller than that
recorded at the finish end of the test weld (load cell B). Load / time relationships
recorded by strain gauges attatched directly to the Tekken Test piece were similar in
form and magnitude to those shown in Fig. 6.11.
To investigate this difference in restraint between the two ends of the test
weld a further series of welds were carried out without the imposition of restraint
using the experimental equipment shown in Fig. 6.3. The "constrained" section of
the test piece was set at 2 m m and any movement of this section was recorded at each
end by two displacement transducers. A typical result is shown in Fig.6.12 from
which it can be seen that greater displacement occurred at the finish end of the test
weld suggesting that previous load/time measurements were not a function of
instrumentation but rather that rotational stress were introduced during the welding
cycle and these ultimately led to a higher stress at the finish end of the test weld.
93
Kihara and Masubuchi(219,229,221) studied rotational distortion in butt
welds and found that rotation was affected by both heat input and welding speed.
From this work it was anticipated that relative loads recorded at each end of the test
weld would vary with welding conditions and this was found to be the case during
the course of the investigation. Although the influence of rotational stresses on the
CI results is difficult to predict, it would be expected that because of the higher load
generated at the finish end of the test weld, the finish end of the test weld would be
the preferred location for crack nucleation. Observations throughout this research
indicated that this was the case. Furthermore, the work of Hensler et al. (23) (see
Fig.6 ref 23), suggests that rotation is a general condition of the Tekken Test.
The distrubution of stresses along the length of the test weld was also
considered. Karppi et al. (238) found that the stress at the centre was higher than at
the ends in a R R C test weld. Ito and Bessyo (236,237) found that for Tekken Test
specimens the intensity of restraint was higher at the ends of test welds. In the
present work it was assumed that the load at the centre of the test weld was the
average of the two loads recorded by the the two load cells A and B. Transverse
sections were cut at the centre of the test weld length for metallographic preparation,
to locate the exact position of the thermocouple and to measure the weldment cross
sectional area from which the restraint stress at the centre of the weld was calculated.
94
C H A P T E R 7
EXPERIMENTAL RESULTS
7.1 INTRODUCTION
The purpose of this chapter is to report and discuss the results of the
experimental work which was carried out according to the plan described in Chapter
5, Fig 5.2. For the sake of convenience Fig. 5.2 has been reproduced and is
included in this chapter for easy reference. Stage 1 of the work was aimed
primarily at examining the hypothesis that variations of the welding parameters could
influence the values of Cracking Index (CI) as measured by the Tekken Test. The
subsequent stages were then undertaken to analyse systematically the manner and
mechanisms by which the the extrinsic parameters of welding voltage, current and
speed affected the intrinsic parameters of microstructure and stress which, in turn,
resulted in the measured values of Cracking Index (CI).
Although the experimental work proceeded according to the planned
approach shown in Fig 5.2, the results and discussion are not presented in the
same sequence. The results from Stage 1 of the work are presented in section 7.2
in which CI is related to variations in the welding parameters and some of the
implications of both the welding parameter/ CI relationships and the scatter of results
are discussed. However, the results of Stages 2, 3, 4 and 5 have been combined in
a presentation of the independent effects that each of the parameters of voltage,
current and speed have on microstructure and stress and their relationship with CI.
Welding Variables Current Speed Voltage
Microstructure TJ3]—»• Cracking Index •*—O2}— Stress
Significance of
The Tekken Test
F I G- 5-2- Diagram showing the planned approach with which the research was to be carried out. The numbers indicate the Stages and the sequence of the work.
95
This has been achieved by treating each parameter separately so that the effects of
welding voltage in Section 7.3.2, current in Section 7.3.3, and speed in Section
7.3.4, relate the respective single extrinsic parameter to the intrinsic factors of
microstructure and stress and to measured values of CI. The purpose of this
approach was to present an overview which would be particularly useful in
considering the interaction of the parameters in relation to constant heat input,
(Section 7.3.5) and in the discussing the significance of the Tekken Test as a
measure of weldability (Stage 6, Fig 5.2) which is presented in Chapter 8.
As with all research, a number of interesting secondary features emerged
during the course of the work. While these were found to be significant in their
influence on measured values of CI and thus important as possible areas of future
research, they were considered to be peripheral to the main aim of the present work
and were not pursued in depth. The results of these excursions are presented and
discussed both in several of the abovementioned sections and in Section 7.4.
Caution should be exercised here regarding the extrapolation of results
obtained in the present work to other combinations of welding conditions and
materials. The observations reported and discussed in this chapter should be
considered as relating only to the steel plate and consumables used in the current
research program. In the small number of experiments where different materials,
particularly electrodes, were used the results were not entirely consistent with those
results obtained using the electrodes chosen for the main experimental work. It
may be that the results reported and the mechanisms discussed cannot be considered
as necessarily universal to the Tekken Test when applied to other plate materials and
electrodes.
96
7.2 T H E E F F E C T O F W E L D I N G V A R I A B L E S O N C R A C K I N G I N D E X
7.2.1 Introduction.
To examine the individual effect that each of the extrinsic welding
parameters had on measured values of CI it was necessary to maintain two of the
parameters constant whilst the third was varied. To achieve this it was necessary to
set reference values for the three parameters. By carrying out a trial set of test
welds it was found that welding conditions of 20.5 volts, 180 amps, and 160
mm/min. produced a weld bead without excessive undercut, with adequate
penetration and containing a small root crack. With the exception of welding
voltage, the welding parameters could be increased above, or decreased below these
reference values to produce acceptable welds with respect to weld bead profile. It
was found that on welding at voltages below 20.5volts the electrode flux coating
often dragged on the sides of the vee joint of the test specimen and caused
inconsistent travel speeds. By holding any two of the reference values constant
and by varying the third, three series of test welds were performed with currents
ranging from 130 to 200amps, voltages of 20 to 30 volts, and speeds of 130 to
220mm/min. Whilst these conditions produce heat energy conditions ranging from
1 to 2kJ/mm, nevertheless, they represent conditions over which welds with
different, but acceptable, profiles could be produced by manual welders in practical
situations. These conditions also extended beyond the ± 0.08kJ/mm limits
suggested by Hensler et al. (23) (see Chapter 5).
100
• 0
^° P60
e>30
U •40
3 M
20
to 20
Voltage 25 3'
FIG. 7.1. Diagram showing the effect of welding voltage on cracking index for a welding speed of 160rnm/min. at 180amps
100
90
80
^.70
| 60
-60
u 40
30
20
10
m
1 0 0 120 140 160
Current amps 180
FIG- ,7-2- Diagram showing the effects of welding current on cracking index for a constant voltage of 20.5 volts and a speed of 160mm/min. Xote_ below 130amps welds failed by lack of fusion
2 indicates specimens which, , after welding, were cooled by forced air flow.
97
7.2.2 The Effect of Voltage
To examine the influence of welding voltage on measured CI, test welds
were performed at a constant welding current of 180amps and a constant welding
speed of 160mm/min. The manner in which CI varied by increasing the voltage
from 20 to 30volts is shown in Fig. 7.1. It can be seen that over this very narrow
range weld cracking increased from approximately 30% at 20.5 volts to 100% at 27
volts. It was found that below 19 volts the arc length was reduced to the extent that
the electrode dragged on the specimen groove and rendered the results unreliable.
Above 30volts, because of the increased arc length, unstable arc conditions occurred
with an associated intermittent weld deposition.
7.2.3 The Effect of Current
The welding speed was set at 160mm/min and the welding voltage at 20.5
volts and a series of test welds was carried out at currents ranging from 140amps to
190amps. The measured CI as a function welding current is shown in Fig.7.2. It
can be seen that as welding current was increased from 150amps to 190amps CI
progressively decreased from 100% to 30%. This general decline is not without
considerable scatter of results.
7.2.4 The Effect of Speed
By maintaining the welding voltage and the welding current constant the
effect of welding speed on CI was determined. Two series of welding speed/CI
experiments were carried out, one at 180amps and a second at 170amps both at
20.5volts and with welding speeds ranging from 130mm/min to 220mm/min. The
100
90
60
v. 70 \.
60
a -a £ 50
" i 0
o 2 30 U
20
10
—
-
/
" •/
- /
s
/'
y
* •
v. /
/
• /
/
/ ^ *
f .
N * 0 ^
i I i
y ^
L
^
.. 1
•
•
•
. .,L, ., 120 140 160 180 200
Weldinq Speed m m / „ ^ /m in
FIG. 7.3. Diagram showing the effect of welding speed on cracking index for two constant welding currents (170amps and 180amps) at a constant voltage of 20.5volts.
100
Welding Speed mm/n;n
20 0
F K f 74- D i a S r a m showing the effect of welding current-voltage ^nTo,s?eed o n c r a c k i ng index (CI) for a constant heat energy of 1.3 ok J /mm
results of these two series are shown in Fig 7.3 in which it can be seen that
increasing welding speed, for both welding currents, was associated with an
increase in CI.
7.2.5 The Effect of Constant Heat Energy
From the results of Sections 7.2.2, 7.2.3 and 7.2.4 it would appear that
the three extrinsic parameters have an effect on CI. However, as each of these
parameters changes with the other two constant it can be seen from equation 7.1 that
heat energy input (E) varies:
E = 60 V I eqn. 7.1 103S
where V is the voltage, I is the current (amps), and S is the welding speed
(mm/min).
A series of test welds was therefore carried out in which heat energy
input was maintained constant at 1.38kJ/mm ( which corresponded to the reference
conditions) over the range of voltages, currents and speeds previously investigated.
The value of each of the parameters was appropriately varied to maintain heat energy
input constant. The results of this series are shown in Fig.7.4 in which the CI has
been superimposed on the current-voltage/ speed relationship of 1.38kJ/mm.
Sound welds were produced at low welding speeds, currents and voltages.
However, at higher values of welding current and speed susceptibility to cracking
increased, ultimately to a CI of 100%.
99
7.2.6 Discussion of Results.
From equation 7.1 it can be seen that by increasing values of welding
current and voltage and by decreasing welding speed the value of input heat energy,
E, may be increased. It has been found (6) that the incidence of cold cracking can
be reduced by increasing the heat energy input. The results shown in Figs. 7.2 and
7.3 also indicate that cold cracking, as measured by the CI values, decreased with
increasing current and decreasing speed. However, the results obtained for voltage
variations, Fig 7.1, appear to be contrary to that predicted. Cracking increased with
increased heat input ( brought about by the increase in voltage). A possible
explanation could be derived from the relationship between arc length (1) and
welding voltage(V), provided in equation 7.2 (6)
V = K +1 d P eqn 7.2 10
where K is a constant dependent on the electrode metal, and is 12 for steel, d is the
electrode diameter, and P is the current density. From eqn.7.2, it can be deduced
that welding voltage is proportional to the arc length. Therefore welding at high
voltages involves long arc lengths which create possible problems of contamination
of the arc plasma with atmospheric gases. These gases could ultimately be retained
in the weldmetal and H A Z . Several gases are known to be deleterious to weldment
mechanical properties^).
100
Assuming the anomalous effect of voltage on CI occurred due to gaseous
contamination, and if the welding voltage was held at low values and only the
current and speed were varied to maintain a constant heat energy input (ref. eqn.
7.1), then it would be expected that CI would remain constant. However, for the
reference conditions of ISOamps, 20.5volts, and 160mm/min. (1.38kJ/mm) CI was
measured to be 3 0 % ; yet for the same heat energy input but conditions of 150amps,
20.5volts, and 133mm/min, CI was found to be zero, see Fig.7.4. These results
suggest that welding speed and/or current have an effect on weldment cracking
other than by the direct influence on heat energy.
The effect of welding speed(S) on CI is shown in Fig 7.3 and suggests the
existence of a linear relationship of the form
CI = kS eqn 7.3
Allowing for the scatter of results observed, the value of the proportionality
constant (k) increases with decreasing welding current. Thus, if welding speeds are
reduced towards a minimum, CI will also tend toward a minimum. However, for
low currents (150amps) CI would be very sensitive to increases in welding speed. It
would appear that the influence of welding speed on CI is beyond the scope of the
effects of heat energy input alone. Nevertheless, as a general rule the beneficial
effects of increasing welding current to to reduce CI (Fig 7.2) can be be offset by
increasing welding speed and voltage (at constant E, Fig 7.4).
The scatter of results recorded (see, for example, Fig 7.2 ) raises
doubts regarding the reliability of quantitative data derived from the Tekken Test.
Masubuchi (164) has shown that with restraint type weldability tests, once a crack
101
has reached a critical size, unstable crack growth follows; so that measurements of
cracking in the 40 to 7 5 % region would be unreliable. In the present work the
scatter of results occurred in this region.
As a general conclusion the CI value is decreased by increases in current
and decreases in speed both of which contribute to increases in heat energy input.
However, increasing welding voltage, although increasing heat input, has a
detrimental effect and increases the measured values of CI. Furthermore, welding
current and speed appear to have an effect on CI that is additional to the input heat
energy factor.
7.3 THE INTER-RELATIONSHIP BETWEEN WELDING VARIABLES
M I C R O S T R U C T U R E . STRESS A N D C R A C K I N G INDEX.
7.3.1 Introduction
The results reported and discussed in this section relate the extrinsic
parameters of welding voltage (Section 7.3.2), current (Section 7.3.3), speed
(Section 7.7.4) and heat input (Section 7.3.5), to the intrinsic parameters of
weldment microstructures and contractional stress.
Specimens used in the metallographic examinations were those used to
develop the welding parameter/CI relationships reported in Section 7.2, in addition
to a second set of specimens for determining the relationships between welding
voltage and the contractional stress developed (see Section 6.3). Several duplicate
specimens were often used.
102
By combining the results of Stages 2,3,4,and 5 relating weldment
microstructures, stress conditions and the mode of weldment cracking to the welding
variables, more detailed discussion of the mechanisms contributing to the measured
CI values was possible. This discussion is presented in the following sections.
7.3.2 Effects of Voltage.
Macroscopic examination of transverse sections of the weldments revealed
that, depending on the welding voltage, cracking occurred in two distinctly different
locations. For weldments produced at low voltages (20.5 volts) cracking occurred
first in the H A Z then propagated into the weldmetal, as can be seen in Fig.7.5. At
higher voltages (23 to 29 volts) cracking nucleated on the part B face of the test
piece(see Fig.4.10) and propagated entirely in the weldmetal (Fig 7.6). High
values of CI were associated with this type of cracking.
From Figs 7.5 and 7.6 it can be seen also that the geometry of the
weldmetal changed with increases in welding voltage. For the higher voltage
(Fig.7.6), considerably greater penetration of the parent plate occurred and the
weldment had an increased width to height ratio compared with the weldment
produced at lower voltage (Fig. 7.5). Similarly the H A Z extended further into the
parent plate for welds deposited at higher voltages. The root angle 0 L which
determines the stress concentration on the H A Z , as defined by the stress field
parameter F1? (see Fig 3.20c and ref. 216), also increased from 110° at 20.5volts to
130° at 25.7volts.
Fracture Modes
To determine the mechanism by which cracking of the weldments occurred,
the fracture surfaces were examined directly using a scanning electron microscope
(SEM) and replicas of the fracture surfaces were examinned in a transmission
FIG. 7.7. Transmission electron fractograph, of a carbon replica of fracture surface m a weld produced at 25.7volts 180amps, and 164mm/min. The "dimple" structure is indicative of microvoid coalescense and ductile fracture. X3800.
103
electron microscope(TEM). From these examinations it was established that
cracking of welds produced at high voltages was entirely ductile, Fig 7.7, with the
classical "dimple" structure indicating a microvoid coalescence mechanism. At 20.5
volts fracture occurred in the HAZ by intra- and trans-granular cleavage, Figs 7.8a
and 7.8b respectively, whilst crack propagation in the weldmetal occurred by quasi
cleavage and microvoid coalescence, Figs 7.9a and 7.9b respectively.
HAZ Microstructure
From a metallographic examination of the weldments it was determined that
the microstructure of the HAZ adjacent to the fusion boundary was different in
weldments produced at low voltages compared with that produced at high welding
voltages. At low voltages the microstructure consisted of lath martensite with a
grain boundary network of ferrite and some Widmanstatten ferrite, Fig 7.10.
Higher welding voltages resulted in a HAZ with a larger grain size, compared to the
lower voltage welds, see Fig 7.11, and with an increased amount of grain boundary k
and Widmanstatten ferrite.
Weldmetal Microstrucftire
Metallographic examination also established that the microstructure of the
weldmetals were different in several respects and that these differences were also
related to the welding voltage. First, weldmetal deposited at high voltages
contained a larger number of oxide particles than welds deposited at the lower
welding voltages, compare Fig.7.12 and Fig.7.13. Secondly, the microstructure of
weldmetal deposited at low voltage (20.5volts) consisted of grain boundary
proeutectoid ferrite and fine lath ferrite (often referred to as acicular ferrite), Fig
7.14. Weldmetal deposited at high welding voltage (28volts) contained a
microstructure of coarse grain boundary ferrite with large Widmanstatten ferrite side
plates Fig 7.15. In these latter weldments the fracture path generally followed the
grain boundary ferrite allotiomorphs. By repolishing and etching the weldment
FIG. 7.12. Photomicrograph showing the inclusion distribution observed in the weldmetal deposited at 28.7volts, 180amps and 160mm/min. X150
FIG- 7.13. Photomicrograph showing the inclusion distribution observed in the weldmetal deposited at 20.5 volts, 180amps, and 160mm/min. X150
FIG. 7.14. Photomicrograph showing the microstructure of weldmetal deposited at 20.5volts, 180amps and 160mm/min. X500; etchant 2.5% nital. '
FIG. 7.15.Photomicrograph of the weldmetal microstructure produced at 25.7volts,180ampsand 160mm/min.X500; etchant 2.5%nital.
104
samples produced at high voltages in a caustic/ chromic acid solution(229) it was
found from metallographic examination that, in addition to the high oxide content,
substantial oxygen enrichment of the weldmetal had occurred, Fig.7.16. From
chemical analyses of the weldmetals it was determined that the oxygen contents of
welds deposited at high welding voltages were of the order of 800ppm compared
with approximately 300ppm for welds deposited at lower welding voltages.
Welding Stress.
To determine the reaction stress developed and imposed on the test weld
during cooling the experimental procedures and equipment described in Sect. 6.5
were used. The stress-time relationships during cooling are shown in Fig.7.17 for
welds deposited at 20.5volts and 28volts and it can be seen that there is a slightly
higher reaction stress generated for the higher voltage welding conditions.
Thermocouples suitably embedded in these specimens (see Sect 6.6) also measured
the thermal cycle of the H A Z during the course of the welding and cooling cycle.
The measured thermal cycles for welds deposited at 20.5 volts and 28volts are
shown in Fig. 7.18. It can be seen that the time to cool from 800°C to 500°C (the
eutectoid temperature range) and from 500°C to 200°C (the martensitic and bainitic
transformation temperature range) are both longer for welds deposited at 28volts,
indicating lower cooling rates for weldments produced at 28volts than for weldments
produced at 20.5volts.
Electrode Characteristics.
It was observed that the geometries of the tips of the electrodes were
different after welding at low and at high voltage, Figs.7.19a and 7.19b. It is
evident that at low welding voltages, Fig.7.19a, a cone formed from the flux and the
frozen metallic tip of the electrode was contained within the cone. The electrode tip
formed by welding at higher voltages is shown in Fig.7.17b, from which it can be
seen that a large frozen metal globule extends beyond the flux coating.
FIG- 716- Photomicrograph showing oxygen enrichment (light areas ) in the weldmetal produced by welding at 28volts, 180amps, and 160mm/min. X250; Etchant, boiling caustic/ chromic solution (229).
FIG. 7.17. Diagram showing the stress-time relationships for two Tekken test welds performed at (a) 20.5volts, 180 amps ,and 160mm/min. and (b) 28volts, 180amps, and 160mm/min.
IX>C-
ccx-
u aoc-
60C-
arjO-
jpo
FIG. "-18. Diagram showing the temperature -time relationships developed m the H A Z of two welds during the welding cycle for conditions of (a) 20.5volts, 180amps, and 160mm/min. and (b) 28volts, 180 amps and 160mm/mi n.
(a) X 3 (b)X3 FIG. 7.19.Photomacrographs of the tips of electrodes formed during welding at 180amps and 160mm/min. and at (a) 20.5volts and (b) 28volts.
ico-
80-
60
40
20
0 15 20
Volis
FIO-7.20. Diagram showing the effect of welding voltage on CI for nominal constant conditions of 180amps and 160mm/min. for electrodes trom a second manufacturing source (source B).
105
It is known that different electrode manufacturers vary the relative
proportions of flux constituents to develop particular arc characteristics of the
electrode, mainly to appeal to individual manual welders. Consequently, it is
conceivable that the observed high voltage effects were peculiar to electrodes having
a particular flux composition. Therefore, to investigate the voltage effect further,
electrodes of the same classification, E4816, but from a second manufacturing
source (denoted B ) were used in a second series of voltage/CI experiments. The
results of this series of tests on the source B electrodes are shown in Fig 7.20. and
are similar to those observed for source A electrodes (see Fig. 7.1). However, test
welds performed at low voltages (20.5volts) had a CI of 0 % compared with 3 0 %
measured for source A electrodes operating under the same conditions.
Photomacrographs of transverse sections of welds produced at low and
high voltages using source B electrodes are shown in Fig 7.21 and Fig. 7.22
respectively. At high welding voltages fracture occurred entirely in the weldmetal,
in a similar manner to that observed for source A electrodes(see Fig. 7.6).
Fracture surfaces of cracks produced in weldments obtained using source
B electrodes at high voltages were examined using a S E M . It was deduced that
crack propagation had occurred by ductile mechanisms as indicated by the
characteristic "dimple" structure observed.
By comparing Fig. 7.21 with Fig. 7.5 and Fig7.22 with Fig. 7.6 it can
be seen that a distinct weldmetal geometry is associated with electrodes from each
manufacturing source. Source B electrodes produced a weldmetal deposit with a
distinct protruberance in the centre of the top surface and deeper penetration into the
plate material. Furthermore, the root angle measured on a number of welds was
(a)X3 (b)X3 FIG.7.23. Photomacrographs of the tips of electrodes from manufacturing source B after welding at (a) 21 volts, 180amps, and 160mm/min (b) 28volts, 180amps and 160mm/min.
_______ TABLE 7.1. Analysis of flux coatings of E48T6 type electrode from two different
manufacturing sources, denoted A'and B.
ELEMENT Mg Si K Ca Mn Fe Ti
3l 24l) 20.0 44.4 L2 L6 ~
B - 10.1 14.1 59.7 - 1.8 7.7
106
approximately 130° which was larger than that measured for welds produced using
source A electrodes, (105°).
From a metallographic examination of transverse sections of
weldments formed at high welding voltages and using source B electrodes it was
found that microstructures similar to those produced using source A electrodes were
produced for the same welding conditions(see Fig.7.11 and 7.15). The weldmetal
contained a larger number of oxide particles and had a microstructure of course grain
boundary and Widmanstatten ferrite; the H A Z was martensitic with grain boundary
and Widmanstatten ferrite.
The geometry of the tips of the electrodes was also examined and
photomacrographs of sections of these are shown in Fig.7.23a and 7.23b for low
and high voltages respectively. Although the general tip configuration was the
same as for the source A electrodes(Fig.7.17a and 7.19b), the cone of source B
electrodes used at low voltages was found to be consistently longer (compare
Fig.7.23a and Fig.7.19a).
A metallographic examination of weldments produced at 20.5 volts
revealed that the microstructure of the weldmetals deposited using both source A and
B electrodes were similar, Fig.7.14 and consisted of grain boundary ferrite and fine
lath ferrite. It was found however, that the microstructure of the H A Z produced
using source B electrodes, Fig 7.24, was different from that produced using source
A electrodes( compare with Fig.7.10) in that it contained a mixture of martensite,
bainite and ferrite and the nature of etching indicated a degree of autotempering.
FIG.7.24. Photomacrograph of the H A Z from a weldment produced at 21 volts, 180 amps, and 160mm/min. using source B electrodes. X250; etchant, 2.5%nital.
400-
300-
I •5 w
a I 200
FI(^ 7-Ir D i agram showing the results of microhardness test traverses weldments produced at 21 volts, 180amps, and 160mm/min. using source A ( + —- * # ) a n d source B (__— _ ____^ eiectrodes.
in
107
Microhardness traverses were carried out on weldments produced at
low voltages using both electrode. The results of the traverses are shown in
Fig.7.25 from which it can be seen that in the HAZ immediately adjacent to the
fusion line the hardness values measured for source B electrodes are approximately
100 hardness points below the measured hardness values for source A electrodes.
Using energy dispersive x-ray analysis the relative concentrations of
the major elements present in the electrode fluxes were determined. These results
are shown in Table 7.1.
Discussion.
From the photomacrographs shown in Figs.7.5 and 7.6 it is apparent
that cracking in weldments produced at high and low welding voltages occurred in
two distinctly different locations. For welds produced at low voltages cracks that
had propagated in the HAZ were arrested when they entered the weldmetal, Fig 7.5.
Cracks that formed in the weldmetal after welding at high voltages propagated
entirely in the weldmetal, Fig 7.6, and subsequently led to high values of CI,
Fig.7.1.
The microstructure of both the HAZ and the weldmetal was found to
be different in weldments produced by welding at high and low welding voltages.
Microstructures in the HAZ (Figs 7.10 and 7.11) were found to differ in that high
voltage welding produced a HAZ microstructure with both increasing grain size and
content of proeutectoid ferrite. The increased heat energy input at high voltages
(equation 7.1) was found to be associated with reduced cooling rates (see Fig 7.18),
and would have two effects: first, the increased time above 200°C would enhance
the diffusion of hydrogen from the weldment and secondly, the increased time
108
between 800 °C and 500°C would result in microstructures containing constituents
less susceptible to hydrogen assisted cold cracking(l 18). These factors are reflected
in the absence of HAZ cracking in weldments produced at high welding voltages.
Weldments produced at the lower welding voltages cooled at a higher rate, contained
less proeutectoid ferrite and were cracked in the HAZ.
Although the reaction stress varied little with increased welding voltage
(Fig 7.17 ) the root angle was observed to be different for welds produced at high
and low voltages. The measured root angle increased from 110° at 20.5 volt to
130° at 25.7 volts( see Fgs.7.5 and 7.6). From the analysis of Karppi (216) the
stress field parameter F1 would decrease from 0.75 to 0.4mm1/2 (see Fig. 3.20c),
corresponding to a 46% reduction in the stress concentration on the HAZ when
welding voltage was increased from 20.5 volts to 25.7 volts. Therefore, in
weldments deposited at higher voltages,the HAZ has a microstructure with a
reduced susceptibility to crackinabecause of both the microstructural constituents
present and the reduced hydrogen content (both being attributable to lower cooling
rates), and a geometry which reduced stress concentrators for possible crack
nucleation.
From Fig.7.14 and Fig 7.15 it is evident that welding voltage also
affected the microstructure of the weldmetal. The microstructure of coarse grain M
boundary and Widmanstatten ferrite produced by high voltage welding(Fig.7.15)
also coincided with the high values of CI measured. The association of a high
oxygen content in the weldmetal, the increased inclusion content and the coarse grain
boundary and Widmanstatten ferrite with an increased susceptibility to fracture is in
agreement with the results of Kirkwood (51) and others (52,56). Kirkwood(51)
and Ito et al. (52) also found that weldmetal with lower oxygen content (200ppm)
109
had a microstructures of fine lath ferrite and that weldmetal with this microstructure
was more resistant to fracture than weldmetals containing higher oxygen content
(800ppm) and coarse ferritic microstructures. Results of the present work which
relate qualitatively the number and size of oxide particles, oxygen content,
microstructure and weldmetal fracture are in general agreement with the results of
the previous work of Kirkwood(51) and others (52,54).
It seems likely that welding voltage can affect the value of CI by
affecting the nature of the microstructure of both the H A Z and the weldmetal and by
influencing the stress concentration at the root of the weld ( as determined by the
stress field parameter). L o w welding voltages, although producing conditions
conducive to H A Z cracking (susceptible microstructure and high value of Fx), also
produced a weldmetal microstructure that apparently had an increased resistance to
fracture so that crack propagation from the H A Z into the weldmetal was terminated
resulting in a low value of CI. By welding at higher voltages (hence higher heat
input) the beneficial effects of producing a H A Z with a microstructure with
increased resistance to cold cracking are offset by the detrimental effects of a
microstructure in the weldmetal which has a low resistance to fracture. As a result
crack propagation occurred entirely in the weldmetal, resulting in high values of CI.
The evidence for ductile fracture occurring in the weldmetal produced by
high voltage welding suggested that it is not the microstructure alone that
contributes to the weldmetal cracking. Ductile fracture mechanisms are generally
considered to require higher levels of strain energy than other mechanisms for crack
propagation. Non-metallic inclusions are known to promote ductile fracture at low
macroscopic strains(225). The high non-metallic inclusion content would aid the
110
crack propagation mechanism by enhancing the nucleation of microvoids which
could be significant in contributing to the increased value of CI.
In Section 7.22 it was proposed that lengthening of the welding arc due to
increased welding voltage caused atmospheric gases to be absorbed in the arc plasma
and incorporated finally, in the solid weldmetal. Because of the high weldmetal
oxygen content and the large numbers of oxide particles in the weldmetals,
contamination of the arc plasma with atmospheric oxygen appears to have occurred.
Considering the geometry of the electrode tips, additional mechanisms to simply
increasing the arc length would seem to be involved. For low welding voltages
(20volts) the tip of the electrode was found to be cone shaped (Fig.7.19a) while at
higher welding voltages the flux cone was absent and a frozen metallic tip protruded
beyond the flux coating ( Fig 7.19b). Essers et al (226) in a study of flux
composition and electrode deposition characteristics, found that the formation of a
cone at the electrode tip coincided with optimum welding conditions. This occurred
for two reasons: first, high melting rates were achieved by a process of early "
pinch" effects producing small droplets, and secondly, the cone shape aided in
directing and containing the arc plasma in a dense column. In the absence of this
cone (as with electrodes operating at high welding voltages) a diffuse arc plasma of
increased length would allow atmospheric gases to penetratethe arc column.
Formation of large metal droplets was also found by Essers et al to reduce melting
rate by acting as a barrier to heat transfer to the unmelted metal core of the electrode
which resulted in overheating of the droplet. Transfer of large overheated metal
droplets could account for the increased penetration of the parent plate and the
changes in geometry of the weldment. It could also be argued that it was the
higher temperatures that caused the lower rates of cooling that contributed to the
111
formation of the type of microstructures observed for both the H A Z and the
weldmetal.
The effect of high welding voltage on the microstructures of both the HAZ
and the weldmetal was found to be independent of electrode flux composition.
Consequently, the high voltage deposition of weldmetal with a microstructure
susceptible to fracture occurred using both source A and source B electrodes and
produced welds with a CI of 100%. However, at low welding voltages the use of
source B electrodes was found to be associated with welds having a CI of 0%.
The reduction in CI can be related to the nature of the H A Z . Source B electrodes
formed a H A Z with a reduced hardness. In Table 7.1 it can be seen that differences
exist in the flux composition, particularly in relation to Si, K, Ca and Ti.
Although the compounds which contained these elements were not determined, it is
known that variations to the content of various elements in the flux not only
influence the arc characteristics, but can also effect the recovery rate of the metal
deposited(214,215) and the weld bead morphology (214) . It can be seen in
Fig.7.5 and Fig.7.21 that, for source B electrodes, the recovery rate (mass of metal
deposited) is higher and a more convex weldmetal profile is produced.
Hazlett(214), Patchett(227), Jansen(228), and more recently
Schwemmer et al.(213) have shown that flux composition can influence the
weldmetal geometry by variations to slag viscosity and the relative values of
interfacial tension. Schwemmer et al. found that for the FeO- Si0 2-MnO system
expansion of the weld pool was restricted as the interfacial tension between the flux
and the parent metal decreased. Similarly, as the interfacial tension between the
flux and the liquid metal increased, the liquid tended to minimise surface area to
112
form a more rounded bead. The result of this type of behavior is to confine the
weld bead and so to contain the heat input and, in general, to increase penetration.
To determine whether these general principles apply to the present
work, bead on plate test welds were deposited using electrodes from sources A and
B and photomacrographs of transverse sections of these are shown in Fig7.26 and
Fig.7.27. It can be seen that depth of penetration (determined as the width to
depth ratio) is greater for type B electrodes which suggests that, although the flux
composition is different, the situation is similar to that for the previously described
work of Schwemmer et al. (213). The containment of the heat input to increase
penetration would also be expected to affect the cooling cycle in a manner similar to
that brought about by high heat energy input welds. This would cause the
formation of microstructures with a reduced susceptibility to cold cracking. The
microstructure shown in Fig.7.24 and the hardness measurements shown in
Fig.7.25 are in agreement with this hypothesis.
In addition to the microstructural differences observed in the HAZ of
weldments produced by the two electrode types, the root angle was also found to be
larger in weldments produced from source B electrodes. It can also be argued that
this arose due to the differences in the flux/metal interfacial energy relationships that
were developed 6y the different electrode fluxes. This was most probably brought
about by the restrictions imposed by the flux upon liquid metal movement on the
straight face of part B of the Tekken Test piece joint
The recorded differences in CI with welding electrodes of the same
classification (E4816), but manufactured with fluxes having different composition,
would appear to be associated with the effect of weld bead geometry caused by
113
interfacial energies between the slag and both the liquid metal and the parent plate. If
the weld bead is contained, high localized heat input causes the formation of a H A Z
microstructure less susceptible to cracking as was found for source B electrodes.
The shape of the resultant weld bead can also reduce the stress field parameter and
the combined effect of both of these factors can lead to a reduction in cracking as
was found for weldments produced using source B electrodes.
7.3.3 The Effect of Current.
Figure 7.2 shows that as welding current increased a general decrease in
CI occurred. Furthermore, over the range of currents investigated (150amp to
195amp) considerable scatter of CI results occurred.
Weldment Macrostructure
Macroscopic examination of transverse sections of the weldments indicated
that cracking in all cases had a similar form to that shown in Fig.7.5 and Fig.7.28.
Cracks initiated at the root of the weldment in the pointed section of the test piece,
propagated into the H A Z , and changed direction. Continued propagation occurred
in the weldmetal in a direction approximately normal to the stress direction, and
towards the upper surface of the weldmetal, Fig.7.28.
Fracture Mode.
From an examination of the fracture surfaces of weldment cracks using a
S E M it was apparent that, for the range of currents investigated, fracture behavior
was similar to that depicted in Figs.7.8 and 7.9, i.e.trans -and intercrystalline
fracture in the H A Z , and microvoid coalescence and transgranular cleavage of the
weldmetal. A s a general observation it was also noted that the fracture surfaces of
the weldmetal produced at low welding currents contained higher proportions of
areas of transgranular fracture. Quantitative measurements were not made to
determine the precise proportions of transgranular and ductile fracture.
FIG.7.29. Photomicrograph of the H A Z microstructure of a weldment produced at low welding currents (150amps, 20.5volts, 160mm/min.). X250; etchant, 2.5% nital.
F!0- 7-30- Photomicrograph of the H A Z produced in a weldmetal deposited at high welding currents (195amps, 20.5volts, 160mm/min.) X25; etchant, 2.5% nital.
•4&H9MI
FIG.7.31. Photomicrograph of the microstructure of weldmetal produced by welding at low currents (150amps, 20.5volts,160mm/min.) X50; etchant, 2.5% nital.
FIG.7.32. Photomicrograph of the microstructure of weldmetal produced by welding at high currents, ( 195amps, 20.5volts, 160mm/min.) X500; etchant, 2.5% nital.
114
H A Z Microstructure
Metallographic examination of the weldments indicated that the microstructure
of the HAZ adjacent to the fusion boundary was different in weldments produced at
low welding currents compared with that produced at high welding currents. At
low currents the microstructure consisted of lath martensite with a very fine grain
boundary network of ferrite, Fig 7.29. The grain size produced in the HAZ of
these weldments was also small compared with that produced at higher currents,
Fig. 7.30. It can also be seen in Fig.7.30 that the grain boundary ferrite network
was thicker and Widmanstatten ferrite was present.
Weldmetal Microstructure.
The microstructures of the weldmetals were found to be different when
produced using different welding currents. At low welding current the
microstructure contained very fine grain boundary ferrite (Fig. 7.31) and a
distribution of fine ferrite laths within grains of bainite. High welding currents
produced weldmetals with a microstructure containing grain boundary ferrite,
Widmanstatten side plates and fine lath ferrite, Fig.7.32. It was also found that
weldmetals produced at the lower welding currents(150amps) were harder
(approx.295HV20) than those weldmetals deposited under other welding conditions
(approx. 220HV20).
From the measured thermal cycles the cooling time over two temperature
ranges, namely 800°C to 500°C and 500°C to 200°C were determined for
various welding currents. The results of these measurements are recorded in
Fig.7.33 which shows that reduced welding current resulted in shorter cooling
times in both temperature ranges (i.e. higher cooling rates).
5 u5'-
Speed
uo 150 170
Current
180
amps
FIG. 7.33. Diagram showing the relationships that welding current and the inverse of welding speed have with cooling times in the temperature ranges 800°C to 500 °C and 500 °C to 200 °C.
a—§-
0
500 - 200 °C
800-500 °C
115
Welding Stress
Measured values of restraint stresses that were developed and imposed on
the test weld during cooling were subject to considerable scatter. The average stress
over the range of welding currents investigated (130 to 195amps) was calculated to
be 220 +/_ 20 M P a . It was found however, that restraint stress measurements from
repeat specimens welded at higher welding currents (180amps) were more consistent
and the scatter was only +/_ lOMPa.
The results from measurements of root angle were also subject to
considerable scatter over the range of welding currents investigated. A correlation
could not be established between the scatter of CI results( Fig. 7.2) and the scatter of
root angle measurements.
Discussion.
From equation 7.1, the general effects of increased welding current are to
increase heat energy input into the weldment and to increase the mass of metal
deposited according to a relationship of the form(6),
m = a_I eqn7.4
where m is the mass of weldmetal/ unit length of weldmetal,v is the velocity of
welding, I is the welding current and a is the fusion constant. The value of a varies
with the type of electrode and wire diameter, but is generally of the order of
0.17gm/amp.min.
116
The combination of equations 7.1 and 7.4 indicate that as welding
current is increased both the thermal energy input to the weldment and the mass of
the weldment increased proportionally. The extent of thermal contraction during
cooling of the weldment would therefore also be expected to increase as the welding
current was increased. Associated with this increase in thermal contraction would be
an increase in restraint load. However, the increase in weldmetal mass and hence
weldmetal cross sectional area would result in a constant restraint stress as welding
current was increased. Equation 3.6 predicts that reaction (restraint) stress is
independent of heat input and the results of Satoh et al.(157) verify this prediction.
The results obtained in the present work also suggest that reaction stress is
independent of heat input although the scatter of results render a definite conclusion
difficult.
Increases in welding current (heat input) were found to cause a decrease
in the cooling rates of weldments, as expressed as the elapsed time in the eutectoid
transformation temperature range (800°C to 500°C) and the bainitic/ martensitic
transformation temperature range (500°C to 200°C ) see Fig. 7.33. It was also
found that the microstructures of both the H A Z s and the weldmetals changed as the
welding currents increased. Increased welding current from 150amps to 195amps
was found to be associated with a progressive increase in the content of proeutectoid
ferrite in the substantially martensitic H A Z , (see Figs 7.29(150amp),
Fig7.10(180amp) and Fig7.30(195amp)). As a consequence the fraction of
martensite would be reduced, resulting in a microstructure with reduced
susceptibility to H A C of the H A Z .
The lower cooling rates developed by welding at higher currents also
allow increased time for the diffusion of hydrogen away from the weldment. By
117
welding at high currents, the overall effect on the H A Z of the weldment is to
produce a microstructure containing a reduced proportion of phases such as
martensite, that are susceptible to H A C and to reduce the concentration of hydrogen
present in the structure. This procedure is considered to be the optimum method for
increasing the resistance of the H A Z to H A C .
It was observed also that the microstructure of the weldmetal was
different when produced at different welding currents. The microstructure was
predominantly bainitic when produced at low welding currents (150amps),
Fig.7.31, fine lath ferrite and grain boundary ferrite when produced at intermediate
currents (180amps), Fig.7.14, and fine lath ferrite with coarse grain boundary ferrite
at high welding currents(195amps), Fig.7.32. These microstructures can be related
to the cooling rate of the weldment. Rasanen and Tenkula(49) have shown that at
very high cooling rates, as with weldmetals deposited at 150amps, bainite and lath
martensite can be produced. Bainite and martensite are known to be more
susceptible to H A C than fine lath ferrite(l 18).
Weldments produced at 150amps contained a HAZ microstructure that
was predominantly martensitic and a weldmetal microstructure that was
predominantly bainitic. Both martensite and bainite are susceptible to H A C (118)
and contribute to the high CI values shown in Fig. 7.2. In comparison, weldments
produced at high welding currents, contained H A Z s with microstructures that were
more resistant to H A C because of the reduced content of martensite. Weldmetal
microstructure formed by high current welding, had a large volume fraction of fine
lath ferrite which is known to increase the resistance to H A C (118) and to be
fracture tough (51). The combined effects of the desirable microstructures
118
developed in both the H A Z and the weldmetal with the reduced hydrogen content in
the weldment resulted in the lower measured values of CI.
7.3.4 The Effect of Welding Speed.
From Fig.7.3 it can be seen that as welding speed was increased, CI
increased. Scatter of results was also observed to increase as the welding speed was
increased.
Weldment Macrostructure.
Macroscopic examination of transverse sections of cracked weldments
indicated that cracks followed a path similar to that shown in Fig.7.28. Cracks
initiated at the root of the weld, propagated normal to the stress direction in the
H A Z , changed direction into the weldmetal, continued propagation in the weldmetal
approximately normal to the stress direction and then propagated towards the top
surface of the weldmetal.
Fracture Mode.
Examination of the surfaces of weldment cracks was carried out by
S E M . Over the range of welding speeds investigated (130mm/min to 220mm/min.)
observations indicated that the fracture behavior was similar to that evident in the
electron photomicrographs shown in Figs. 7.8 and 7.9. Cracking had occurred in
the H A Z by both trans- and intercrystalline fracture and in the weldmetal by
microvoid coalescence and trans-granular cleavage.
H A Z Microstructure
Metallographic examination of transverse sections of weldments
established that the microstructures of both the H A Z and the weldmetal were
different and the differences were related to welding speed. At low welding speeds
119
(130mm/min.) the microstructure of the H A Z comprised lath martensite and
bainite, with grain boundary and Widmanstatten ferrite, Fig.7.34. At higher
welding speeds the HAZ microstructure contained less proeutectoid grain boundary
and Widmanstatten ferrite and had a smaller prior austenitic grain size, Fig.7.35.
Weldmetal Microstructure
Typical microstructures of the weldmetal produced over the range of the
welding speeds investigated (130 to 220mm/min.) are shown in Figs.7.36, 7.14
and 7.37 . It can be seen that for low welding speed, Fig.7.36, the microstructure
is essentially fine lath ferrite with a grain boundary network of ferrite and some
Widmanstatten side plates. At the highest welding speed Fig.7.37, the weldmetal
had a microstructure consisting of grain boundary and Widmanstatten ferrite with
large intergranular ferritic laths in a bainitic matrix. Cracking in the weldmetal was
often observed to be subsidiary to the main crack and tended to be either aligned
with the large ferritic laths, Fig. 7.37 or, occurred along the ferritic grain
boundaries, Fig.7.38.
Using the thermal cycles recorded from the embedded thermocouples,
cooling times over the temperature ranges 800°C to 500°C and 500°C to 200°C
were determined. Because heat input is related to the reciprocal of welding speed
the cooling times have been shown as a function of the inverse of welding speed in
Fig.7.33. It can be seen that the relationship between reduced welding speed and
increased cooling time is similar to that for increased welding current and cooling
time.
Welding Stress.
The manner in which restraint stress varied with welding speed is shown
in Fig.7.39 and although considerable scatter exists at the higher welding speeds
' ^ «
.wt
FIG.7.38. Photomicrograph showing weldmetal cracking associated with grain boundary ferrite. X300; etchant, 2.5% nital.
240
5 220
(A
200
IPO 14o 160 180 20C, 220
Weldinq Speed mm/ . 3 /min
FIG-7,39- Diagram showing restraint stress as a function of welding speed.
120
there appears to be a general increase in the restraint stress generated with increase in
welding speed.
Measured values of root angle could not be correlated with welding
speed but were found to be 105° +/- 15° over the range of welding speeds
investigated.
Discussion
From Fig.7.3 it can be seen that CI increased as welding speed was
increased. As with welding current, welding speed also affects both heat input to
the weldment and the mass of weldmetal deposited. Equations 7.1 and 7.4 indicate
that inverse relationships exist between welding speed and both heat input and
weldmetal mass. Increased welding speed, therefore, reduced both the heat input
to the weldment and the mass of weldmetal deposited. Using a similar approach to
that presented for welding current (see Section 7.41) it can be argued that reaction
stress should remain constant and independent of welding speed. Equation 3.6
predicts such a result. However, it can be seen from Fig. 7.39 that as welding
speed was increased there was a general increase in reaction stress. The increase in
stress with welding speed shown in Fig. 7.39 is not in agreement with the predicted
behavior. It is known that the arc fusion constant, a, (see equation 7.4) can vary
with welding conditions (6) so that as welding speed was increased, any decrease in
a would result in a disproportionate reduction of the mass of metal deposited relative
to the weld heat input. The necessary reduction in weldmetal cross sectional area
could lead to an increase in the restraint stress.
It can be seen in Fig 7.33 that as the welding speeds was increased from
130mm/min to 220mm/min (1/S=7.7xl0"3 to 4.5X10-3min/mm) there was a
121
decrease in the cooling rate of the weldment, expressed as time in the temperature
ranges 800°C to 500°C and 500°C to 200°C. The influence that the increased
cooling rate had on the microstructures of the H A Z can be seen in the comparative
photomicrographs of Fig.7.34 (130mm/min), Fig.7.10 (160mm/min) and Fig.7.35
(220mm/min): there was a decrease in proeutectoid ferrite content of H A Z
microstructure as the welding speed increased. The consequential increase in
martensite content resulted in a H A Z microstructure susceptible to H A C .
Increased cooling rate of the weldment brought about by increased
welding speed can also have an effect on the hydrogen content of the weldment,
particularly the H A Z . Although it was assumed that the initial concentration of
hydrogen in the weldment was constant, the rate at which hydrogen diffuses from
the weldment depends on time and temperature. The effect of increasing welding
speed would cause a reduction in diffusion time above 200°C and a higher
concentration of hydrogen retained in the structure of the H A Z .
Microstructures of the weldmetal were also found to be different when
deposited by welding at different speeds. However, the comparative
photomicrographs shown in Fig.7.36 (130mm/min), Fig.7.14 (160mm/min), and
Fig.7.37 (220mm/min) do not appear to follow the trend expected on the basis of
cooling rates. The coarse nature of the ferrite in the microstructure produced at
high welding speeds, Fig.7.37, and the fine ferrite present in the microstructure
produced at low welding speeds, Fig. 7.36 do not appear to conform with the
corresponding high and low cooling rates. It would be expected that at low
speeds, high heat input and low cooling rates would enhance the formation of a
coarse ferritic microstructure. The reverse is in fact observed, as can be seen by
comparing the fine lath ferrite in Fig.7.14 (160mm/min) and Fig.7.36 (130mm/min)
122
in which the lath size is smaller after formation at lower speeds. L o w welding
speeds are known to enhance weldmetal turbulence(80), which would be expected to
reduce segregation of non metallic inclusions and increase the number of nucleation
sites for fine lath ferrite.
At the higher welding speeds(220mm/min) the effects of weldmetal
stirring would be reduced and the cooling times were found to be short. Although
grain boundary precipitation of ferrite would be expected to occur as a direct
consequence of reduced intragranular nucleation sites, higher cooling rates would be
expected to lead to the formation of bainite. Rasenan and Tenkula(49) proposed
that depression of the austenite to ferrite transformation temperature can lead to
massive transformations to ferrite, or, the formation of upper bainite which may
occur in a morphology resembling finely spaced Widmanstatten plates. Weldmetal
formed by welding at high speeds also contained numerous cracks in addition to the
main crack. Examples of these are shown in Fig. 7.37 and Fig. 7.38 and suggest
that fracture by either inter- or transgranular mechanisms can occur easily in the
microstructure produced.
The genesis of the weldmetal microstructure depends on both the
solidification process and on the cooling rate thereafter. Weldmetal containing large
ferritic laths and bainite are more susceptible to fracture than microstructures of fine
lath ferrite that are produced during low speed welding. Weldments produced by
welding at the higher speeds have H A Z and weldmetal microstructures susceptible
to H A C . In addition, rapid cooling would be expected to retain a high hydrogen
content in the weldment. Together these two conditions are consistent with the high
values measured of CI.
123
7.3.5 Effect of Constant Heat Input
By increasing welding current, voltage and speed, a constant value of
heat energy input can be maintained. The effect on CI of manipulating the welding
parameters to maintain a constant heat input of 1.38kJ/mm is shown in Fig.7.4 in
which it can be seen that CI does not remain constant. Weldments produced in this
series of experiments were therefore examined to relate variations in microstructure
of the weldment to the variability of the CI results obtained.
Weldment Macrostructure
Macroscopic examination of transverse sections of weldments from low
current, low voltage and low speed welds revealed that when cracking occurred it
was similar in form to that shown in Fig. 7.5, i.e. crack initiation in the H A Z and
propagation into the weldmetal. At higher values of current, voltage and speed,
cracking was entirely in the weldmetal as shown in Fig.7.40. Although the welds
shown in Figs.7.5 and 7.40 were both deposited with a heat input of 1.38kJ/mm the
geometry of the weldmetal bead and the path followed by the weldment crack were
different.
Fracture M o d e
Examination of crack surfaces revealed that for low currents, voltages and
speeds, similar characteristics to those shown in Figs.7.8 and 7.9 were observed,
indicating intergranular and transgranular cleavage in the H A Z and microvoid
coalescence and transgranular cleavage in the weldmetal. Cracking in weldments
deposited at high currents, voltages, and speeds, showed evidence of both ductile
fracture and trans-crystalline cleavage.
Weldmetal Microstructure.
Weldmetal deposited at low current, voltage and speed was found to have a
microstructure similar to that shown in Figs.7.14 and 7.36, which consisted of
400-
300-
200 -
FIG 7-44. Diagram showing the results of microhardness traverses on weldments produced at constant heat input, 1.38kJ/mm, which were achieved by different currents, voltages and speeds.
O
150amps, 20.5volts, 133mm/min
180amps, 20.5volts, 160mm/min
180amps, 25.5volts,200mm/min.
124
grain boundary and fine lath ferrite. At higher welding current, voltage and speed,
and the same heat input, the weldmetal contained a microstructure similar to that
shown in Fig. 7.41: grain boundary ferrite, Widmanstatten sideplates and coarse
intragranular lath ferrite. At the highest current, voltage, and speed the weldmetals
had a microstructure similar to that shown in Fig.7.15.
H A Z Microstructure.
Microstructures in the H A Z of weldments produced at constant heat input
were also found to differ with different values of the welding parameters. At low
current, voltage and speed the H A Z microstructure contained a mixture of
martensite, bainite and ferrite, Fig.7.42. As the welding parameters were
increased, at constant heat input, the variations in the microstructure of the H A Z are
typified by Figs.7.10 and 7.43. which show that the grain boundary ferrite
network has thickened, together with an increase in the Widmanstatten ferrite
content.
Microhardness Measurements
Microhardness traverses were carried out on the H A Z of three weldments
produced at constant heat input of 1.38kJ/mm, but different values of welding
current and speed. These three weldments had CI's of 0 % ( 150amps, 20.5volt,
133mm/min.), 3 0 % ( 180amps, 20.5volts, 160mm/min) and 9 5 % ( 180amps,
25.5volts, 200mm/min). The results of the microhardness traverses are shown in
Fig 7.44 from which it can be seen that the hardness of the H A Z is highest when
the weldment is produced at 180 amps 20.5 volts and 160mm/min.
Discusion
The cracking behaviours of the weldments produced by varying the
extrinsic welding parameters can be classified into three groups. The first group is
that for which cracking did not occur or for which low values of CI were
125
recorded and is typified by weldments produced at 150amps, 20.5volts and
133mm/min. The second group involved those weldments for which 3 0 % to
6 0 % cracking occurred and can be typified by weldments produced at
180amps, 20.5volts and 160mm/min. In the third group, high values of CI were
recorded, and typical of this group were weldments produced at 180amps,
25.5volts and 200mm/min with a CI of 95%.
In the first group, which showed no cracking, the microstructure of
the H A Z was found to have a hardness below that measured for the other two
groups, Fig.7.44, yet the microstructure shown in Fig. 7.42 does not contain an
appreciable amount of ferrite. This observation together with the globular shaped
weld bead suggests that autotempering of the microstructure had occurred in a
manner previously described for the source B electrodes (Section 7.3.2, page 111).
The formation of a soft H A Z does not facilitate crack nucleation and propagation
and leads to the measured CI of 0%.
Weldment cracking in the second group is typified by Fig. 7.5 and it can
be seen to involve cracking of the H A Z in a direction normal to the stress direction,
followed by propagation into the weldmetal. Hardness of the H A Z for this group
was measured to be higher than for the H A Z in either of the other two groups. In
Fig.7.5 it can be seen that the section of H A Z normal to the stress direction
constituted approximately 2 5 % of the weldment cross sectional area so that cracking
measurements above 2 5 % was weldmetal cracking. The fine lath ferritic
microstructure of the weldmetal resisted or rninimised further crack growth.
The third group, which had high values of CI, was associated
predominantly with cracking in the weldmetal (Fig.7.40). These welds were
126
performed under conditions of high current, voltage and speed so that it would be
expected that the combined effects of both high voltage and speed would be to the
detriment of the weldmetal microstructure ( see Sections 3.7.2 and 3.7.4). The
combination of high voltage and speed produced weldmetal mictrostructures which
contained a high volume fraction of coarse ferrite with inferior fracture properties.
The geometry of the weldbead deposits also changed as the welding
variables were increased. Weldments typical of group 1 were observed to be similar
to that shown in Fig.7.21, group 2, Fig.7.5, and group 3, Fig.7.40. In addition
to the effects that weldbead geometry may have had on weldmetal and H A Z
microstructure there was a decrease in weldmetal cross sectional (compare Figs. 7.5
and 7.40) so that it is not inconceivable that increased contractional stresses were
generated which contributed to the increase in CI values and the change in location
of the weldment crack.
It therefore appears that for welding conditions of constant heat input a
causal relationship exists between welding variables, weldbead geometry, the
microstructures of both the weldmetal and the H A Z , and the subsequent weldment
cracking as determined by the CI.
^ c
X LU
Q -O 2 <S e D
2
c u 4 0
30
SO
lO
-
• .
IXZT \-. .'i.e. -V
/ m
/ •
1
,
/ /
% l
2 d
GAP fmmj
FTG.7.45. Diagram showing the effect of root gap on cracking index for welding conditions of 180amps, 20.5 volts and 180mm/min.
FIQ.7.-16. Photomicrograph of a test specimen containing a 3 m m root gap, welded at 180 amps. 20.5 volts and HCmm/min. X5: etchant, 2.5Tc niral.
127
7.4 O T H E R F A C T O R S A F F E C T I N G C R A C K I N G I N D E X .
7.4.1 Introduction
In addition to the general program of work relating to the welding
variables of current, voltage and speed on CI, several other aspects of the Tekken
Test were examined. As mentioned earlier, these were not directly related to the
main aim of the research and were therefore were not examined in depth. The
results obtained, however, are pertinent to the Tekken Test as a weldability test in
general but detailed explanations of the results have not been developed.
The areas examined include the influence of root gap on CI, the results of
which are described and discussed in Section 7.4.2, the effect of plate composition
on CI, Section 7.4.3 and the effect of electrode classification on CI in Section 7.4.4.
7.4.2 ROOT GAP.
To examine the effect that root gap has on CI, a series of test welds were
carried out at 180amp, 20.5 volt and 160mm/min. The root gap of the Tekken Test
pieces was varied using metal tabs of thicknesses different to the 2 m m tabs which
were used in the previous studies (see Section 6.3 ). As shown in Fig.7.45, as
root gap increased, CI increased.
From an examination of transverse sections it was apparent that the fracture
path changed as the root gap increased. For a root gap up to 2 m m , fracture occurred
in the manner depicted in Fig. 7.5. However, for a root gap of 2.5mm to 3 m m
fracture occurred entirely in the weldmetal, Fig.7.46. Cracking had nucleated at
128
the root of the weld, but instead of propagating in such a way as to maintain a path
close to 90° to the tensile direction as shown in Fig.7.40, crack propagation was
closer to 45° to the tensile direction.
Examination of the weldment crack surfaces revealed that for test specimens
with a root gap less than 2 m m fracture of the weldments occurred in a manner
similar to that shown in Fig.7.8 and 7.9. However with 3 m m root gaps the
fracture surfaces contained a dimple structure characteristic of ductile failure.
Metallographic examination indicated that the structure of the HAZ
contained an increased volume fraction of proeutectoid ferrite at the larger root gaps,
compare Fig.7.47 ( 3 m m ) with FIg.7.10 ( 2 m m gap.). Furthermore, the
microstructure of the weldmetal contained a coarser distribution of ferrite in
weldments produced with larger root gaps, compare Fig. 7.14(2mm) with Fig. 7.48
(3mm).
Discussion.
The relationship between root gap and CI found in the present work was
not in agreement with that found by Hensler et al. (23). Kihara and Masubuchi
(222) have shown experimentally that transverse shrinkage of weldments increased
with increased root gap. This result suggests that a qualitative model based on
constant volumetric contraction of a constant weldmetal volume can be developed as
follows. A s the root gap increases the the cross sectional area of the weldmetal
would decrease proportionally. Although the total volumetric contraction would
remain constant and independent of root gap, the proportion of thermal contraction
in the transverse direction would be expected to increase as root gap increased. The
combined effect of reduced cross sectional area and increased transverse contraction
would be expected to increase transverse stress on the weldment
129
Obviously the above model for thermal contraction and generated stresses
ignores the influence that the observed metallurgical factors have on the Tekken
Test results. The decrease in cross sectional area must have an influence on cooling
rate of the weldment by reducing the area of the heat path. This is reflected in the
microstructures of both the weldmetal and the H A Z . The effect of decreased cooling
rate on the H A Z has been to increase the amount of proeutectoid ferrite, Fig.7.47.
This result has been beneficial in reducing H A Z cold cracking. However, the
effect of reduced cooling rate, as a consequence of root gap, on the weldmetal
microstructure has been to increase the proportion of coarse ferrite, Fig.4.8, thereby
increasing the susceptibility of the weldmetal to ductile fracture.
A precise model for describing the relationship between root gap and CI is
not yet available. However, the limited results obtained in this investigation do
point to a relationship between changes in stress and mechanical properties of the
weldment components. Mechanical property changes are brought about by
microstructural modifications, resulting from changes in cooling rate.
7.4.3 MATERIAL COMPOSITION
As pointed out in Chapter 1, one of the purposes of weldability tests is to
assess the relative weldability of different steels. In this brief series of experiments,
three C - M n steels were used to examine possible C e q (IIW version) / CI
relationships. The use of the Tekken test was also examined as a means of
determining conditions for safe welding.
The Tekken Test specimens were prepared from steel plate having the
compositions and C as shown in Table 6.1 in Chapter 6. The reference welding
conditions chosen for the previous work, (180amps, 20.5volts, and 160mm/min)
Carbon
FIG. 7.49. Diagram showing a relationship between Cc q (HW) and CI
for welding conditions of 180amps, 20.5 volts, and 160irrni/rmn.
FIG.7.50. Photomacrograph of specimen, plate composition C 0.41,
welded at 180amps, 20.5volts and 160mm/mrn. X5; etchant, 2.5% nital.
FIG. 7.51. Scanning electron micrograph of the fracture surface of the H A Z crack produced in 0.41 C steel welded at 180amps, 20.5volts, and 160mm/min. X200.
FIQ, 7.52. Scanning electron micrograph of the fracture surface at the root of a H A Z produced in 0.41Ceq steel welded at 180amps, 20.5 volts and 160mm/min. X20.
130
were used because of the acceptable weld profiles produced and the tolerable scatter
in CI results obtained previously.
The values of CI measured are shown as a function of C in Fig.7.49,
where it can be seen that CI increases linearly with C for the welding conditions
chosen.
Weldment Fracture
Macroscopic examination of the weldments confirmed the CI measurements;
however, it was observed that for the steel with the highest C (0.41) cracking
occurred entirely in the H A Z , Fig.7.50, and not in the manner previously observed
when test specimens had a CI of 1 0 0 % (see Fig.7.28). Cracking of the 0.35Ceq
steel is shown in Fig. 7.5 and characteristics of the weldment fracture are shown in
Figs. 7.8 and 7.9. The fracture surfaces of cracks produced in the H A Z of the
0.41Ceq steel were predominantly intergranular, with some transgranular cleavage,
Fig.7.51. At the root of the weld, a band of crack surface was often observed
which contained a "dimple" structure characteristic of ductile fracture and extended
for the length of the test weld, Fig.7.52.
Weldment Microstructure
Metallographic examination of transverse sections of weldments
indicated that the microstructures of the weldmetals were similar to that shown in
Fig.7.14. Microstructures of the H A Z were found to contain reduced contents of
proeutectoid phases as the C e q of the plate material became higher (compare
Fig.7.53 with Fig.7.10 and Fig.7.54). The hardness of the H A Z was measured for
each of the three steels and found to be 264HV(10) for 0.3Ceq, 300HV(10) for
0.36Ceqand 330HV(10) for 0.41C It was observed that the crack path in the
131
0.4C steel weldment had, at several locations, begun to propagate into the
weldmetal,Fig.7.55, and at the root of the weld the crack had occurred in the
weldmetal to a depth which corresponded to that of the ductile fracture band shown
in Fig. 7.52.
A range of test welds were carried out using Tekken pieces prepared from
the three steels at a constant voltage of 20.5volts, but varying welding current and
speed. It was found that for conditions of 165amps and 220mm/min 1 0 0 % CI was
recorded for test welds using all three steels. At 195amps and 140mm/min all
welds recorded a CI of 0%.
Discussion.
It can be seen in Fig.7.49 that CI appears to be linearly related to the
C (IIW), of the three steels. Metallographic examination of the H A Z also indicated
that the higher the C the smaller was the proportion of ferrite in the H A Z
microstructure (compare Figs. 7.53, 7.10, and 7.54). H A Z hardness was also
found to increase as the C„„ increased. These qualitative metallographic eq
observations are consistent with the observed relationship between H A Z cracking
and C However, welds deposited under different welding conditions were eq.
found to yield different CI/C relationships.lt therefore does not seem reasonable to
propose that as a general rule, there is a direct relationship between C e q (HW) and CI
value.
Several other factors which emerged in this series of experiments are of
interest. The first of these relates to crack nucleation. From Fig.7.52 it is inferred
132
that crack nucleation could have occurred at two possible sites. First, crack
nucleation could have occurred at the surface in the weldmetal at a suitable stress
concentration site, with subsequent propagation involving microvoid
coalescence(ductile fracture) in the weldmetal followed by inter- and transgranular
cleavage in the H A Z . A second possibility could be nucleation internally in the
H A Z and subsequent propagation in both directions, producing ductile fracture in
the weldmetal because growth of the crack in the H A Z had developed a tip stress
concentration of such magnitude that ductile fracture could occur. This internal
nucleation site would appear to be a more favoured nucleation site than the surface
because the large root angle( see Fig. 7.50) would reduce the stress concentration at
the surface. It is therefore conceivable that crack nucleation occurred in the H A Z at
some pre-existing nucleation site in a highly susceptible microstructure, followed by
further brittle fracture along the H A Z and ductile fracture through the weldmetal to
the root
The direction of the crack path in the HAZ of the 0.41 C steel is also of
interest, (see Fig.7.55). Termination of the crack path in the weldmetal and the
continuation in the H A Z indicates the relative fracture strengths of the H A Z and the
weldmetal. From this it could be concluded that the fracture strength of the H A Z o>, HflI
was below cos300Cy#, where <xis the tensile stress developed by thermal contraction.
Using a similar argument it is proposed that for the 0.36Ceq steel, because crack
propagation was terminated the fracture strength of the HAZ, a , would be in the
range,
°H ) <w> Vos3°
133
7.4.4 E L E C T R O D E T Y P E A N D M A N U F A C T U R E R .
The purpose of these investigations was to examine the effects that
electrodes of different classifications and from different manufacturing sources had
on the value of CI. Therefore, in addition to the E4816 grade electrodes generally
used, two grades of higher strength electrode, namely E6218M and E7618M, were
investigated to determine the effects of weldmetal strength on CI. Previous work
(see section 7.3) had shown that manufacturing source can also effect Tekken Test
results. It was therefore decided to investigate the behavior of these electrodes from
two manufacturers. Electrodes from the two different manufacturers have been
denoted source A and source B as in Section7.3 for the E4816 electrodes.
The chemical composition and mechanical properties of the weldmetals
are shown in Table 6. Test welds were carried out on the Tekken Test pieces at
180amps, 20.5volts and 160mm/min ie. the reference conditions. The measured
values of CI are shown as a function of the quoted values (241) of ultimate tensile
strength for both source A and source B electrodes and are shown in Fig.7.56. It
can be seen that although electrodes conform to a particular specification their
performance in relation to weldment cracking differ considerably for each electrode
grade.
Although a detailed investigation was not carried out for the E6218M and
E 7 6 1 8 M electrodes, energy dispersive x-ray analysis of the flux coatings revealed
that the compositions were different in electrodes obtained from the two sources,
Table7.2. It was also noted that cracking of the weldment produced using the
E7618M electrodes, source B, was entirely in the weldmetal.
100 -
c-o
BO
70
„60 • -^50
c 2 •'O u 0 (J 30
20
»C
_
-
-
"
• - A
• -B
1 * 500 600 700
W»|dm«tal U T S M P a ftoo
FIG. 7.56. Diagram showing the relationships between CI and the U T S of electrodes of different classifications and from different manufacturing sources.
electrodes from source A
electrodes from source B
T A B L E 7.2
Analysis of flux coatings from electrodes, grades E6218M and E7618M derived from two manufacturing sources denoted type A and B
Electrode Analysis
A E5218M
B
A E7618M
B
Na
2.2
4.7
2.2
Si
20.1
12.8
15.1
11.9
K
14.0
21.0
9.8
20.0
1
Ca
54.0
50.0
51.0
=6.0
Ti
2.1
3.3
5.9
' i
...
Mn
1.9
1.3
3.1
1 c ...
Ni
_
5.7
1.5
1
134
Discussion
The relationships between weldmetal ultimate tensile strength (UTS) and
CI are not the same for type A and B electrodes and neither relationship is in
agreement with that determined by Hensler et al.(23) who proposed, without
supporting evidence, that the relationships they observed were based on weldmetal/
steel plate strength compatibility. In the case of the E4816 electrodes (Section7.3)
evidence indicates that weld bead geometry resulting from flux coating and
consequent slag surface energy relationships were involved. From the differences
in flux analysis shown in Table 7.2 a similar situation could exist for the other
electrode types. However, at this stage and without further evidence, any
proposition would be speculative.
135
C H A P T E R 8
DISCUSSTON
In Chapter 7 the results arising from variations to the values of the
extrinsic welding parameters were presented and discussed as separate entities. The
purpose of this chapter is to present an overview of the combined results of
variations to all the parameters and discuss the effects relative to weldability.
For the steel plate and electrodes used in the greater part of the
investigation (AS-1204-350 plate and E4816 electrodes) the crack in the weldment
had one of two possible forms. Depending on the welding conditions (current,
voltage and speed), cracking occurred entirely in the weldmetal, Fig 7.4, or had a
form similar to that shown in Fig. 7.28, in which nucleation occurred at the root of
the weld, followed by crack propagation, first in the H A Z , and then in the
weldmetal. If crack propagation terminated in the weldmetal, as shown in Fig. 7.5,
then the value of CI was low. Whether or not weldmetal cracking commenced in
the weldmetal or, propagated into the weldmetal from the H A Z , depends on three
main factors: stress level, hydrogen content and microstructure.
Measured values of restraint stress were found to have considerable
scatter, generally about 10%. Furthermore, when a general relationship between a
welding parameter and restraint stress was observed ( as with welding speed, Sect
7.3.4, Fig.7.39) there was only about a 1 0 % increase in restraint stress over the
range of values for which the welding parameter was varied. It would therefore
seem reasonable to conclude that the differences in restraint stress generated by
variations to the welding parameters were not controlling in the overall sense.
136
The second possibility is that hydrogen content might be the controlling
process. Hydrogen content of the weldmetal pool was assumed to be constant
initially. This assumption was based on the proposition that since all welding rods
were baked in an identical manner, moisture content of the flux would be constant
and hence the initial hydrogen content of the weldpool would be identical.
Measurements of hydrogen content of the final weldment were not made to
determine if variations in the final hydrogen content at room temperature occurred as
a consequence of welding parameter variations. Furthermore, because of the
inconsistencies in hydrogen diffusivity in ferrite, see Fig. 8.1, calculations of
hydrogen content after weldment cooling can be difficult. Teraski (230) developed
empirical equations of the form;
HR1Q0 = exp-A(ZD.AT) eqn. 8.1.
where A is a numerical factor, equal to 95 for first pass butt welds, HR100 is the
remaining diffusible hydrogen in the weld at 100°C and H0 is the initial hydrogen
content of the weldpool. The term (X D .AT) is the thermal factor and is derived
from an empirical relationship used to overcome changes in diffusivity with change
in temperature and is of the form;
(LD.AT) = [4.2AT15_2 +2.73 AT15.L5 -13] X 10"5 eqn8.2.
200 500
T C O
FIG. 8.1. Diagram showing the variation of hydrogen diffusivity with temperature for steels (59)
Increasing alloying elements
Decrecsmq oxvaen content
FIG. 8.2. Diagram showing the proposed form of C C T diagrams for weldmetals as suggested by Ito et al. (231).
137
where A T 1 5 _ 2 is the cooling time from peak temperature to 200°C and A T ^ . j 5
is the cooling time from peak temperature to 150°C.
The significance of the Teraski equations (230) is that they relate cooling
rate to the ratio of final and initial hydrogen concentrations, so that although
absolute values of hydrogen content are not known, cooling rate data does yield an
indication of the relative hydrogen content present. Equation 8.2 was derived
directly from experimental data and points to the importance of the low temperature
of the thermal cycle. From Fig. 7.18 it can be seen that, although large differences
occur for ATg_5, differences in time lapse between 200°C and 100°C are small.
Hence significant differences in hydrogen content as a consequence of variations to
the welding parameters would not be expected. It should also be mentioned that
while the hydrogen content may not vary with welding conditions, the manner in
which a constant hydrogen content affects different microstructures can be related to
mechanical properties, see Fig.3.15. Therefore hydrogen cannot be ignored and can
contribute to a greater or lesser extent depending on microstructure.
It is not unreasonable to conclude that in the present work the
microstructure in the weldments is the controlling factor in determining the
occurrence and the extent of cracking in Tekken Test pieces. As previously pointed
out cracking may take one of two possible forms; entirely in the weldmetal or, in the
H A Z and the weldmetal. The location and extent of cracking will depend on the
fracture strengths that the H A Z and the weldmetal develop, as a consequence of
their respective microstructures, relative to the magnitude of the restraint stress. If
the fracture strength of the weldmetal is above that of the H A Z , cracking of the H A Z
138
will occur and vice versa. From Fig. 7.5 and Fig. 7.28 it can be seen that the
m a x i m u m contribution to the total CI made by the H A Z was approximately 25%.
The additional 7 5 % of crack growth needed to produce the 1 0 0 % CI figure shown
in Fig. 7.28 was attributable to weldmetal cracking. For propagation of the crack
from the H A Z into the weldmetal the fracture strength of the weldmetal relative to the
stress concentration at the crack tip would be the determining factor. From the
geometry of the weldbead and the test piece configuration it is obvious that the
length of the crack as it enters the weldmetal would be approximately constant so
that values of CI above approximately 2 5 % can be considered to be a measure of
weldmetal susceptibility to cracking.
Weldmetal microstructure can be considered to have a major effect on the
value of CI. In Sections 7.3.2, 7.3.3, and 7.3.4 it is shown that all extinsic welding
parameters influenced weldmetal microstructure. It was also observed that the most
desirable microstructure was fine lath ferrite with a minimum of grain boundary
ferrite in agreement with the results of other workers(51,52). The microstructures
of ferritic weldmetals have been studied and it is generally agreed that an optimum
oxygen content of approximately 200ppm is required for the formation of fine lath
ferrite(51). Others found that cooling rate can influence the formation of fine lath
ferrite (49). In the present work welding voltage was found to increase the oxygen
content above the value considered to be the optimum, and to result in a lower
cooling rate, see Section 7.3.2. Ito et al.(231) suggested that weldmetal oxygen
content effects both the position and shape of the C C T curve shown diagramatically
in Fig.8.2 in which it can be seen that the lower the oxygen content the further to
the right the C C T curvef s located and the larger is the region in which fine lath
ferrite (acicular ferrite) forms. In the present work weldmetal deposited at
increased welding voltages contained the range of microstructures represented by
speed (rnm/f
220
160
133
m 150amp 180amp 195amp
mG- 8-3- Photomicrographs of the microstructures of the weld metal produced by welding at 20.5volts and at different currents and speeds. X500; etchant, 2.5%nital.
139
the diagram shown in Fig.8.2. High welding voltage would be represented by a
weldmetal cooling curve to the right in Fig.8.2 ( higher heat input) and the C C T
curve to the left ( higher oxygen content) and the diagram then predicts grain
boundary ferrite and Widmanstatten side plates, which were observed, Fig7.15.
Figure. 8.2 does not appear to be totally applicable to the results for
variations of welding current and speed. In Fig. 8.3 the various microstructures
produced by increasing the welding speed (vertical) and by increasing the welding
current (horizontal) are shown. All welds were made using 20.5volts and are the
accumulated results of Sections 7.3.3 and 7.3.4. The heat input (in kJ/mm) to each
weldment is shown on the photomicrograph.
At low welding current (150amps) the effect of increasing welding speed
(decreased heat input) followed a trend predicted in Fig.8.2.; namely that at low
speeds (133mm/min) the microstructure comprised grain boundary and fine lath
ferrite, but at higher welding speeds (reduced heat input) the weldmetal
microstructure was found to be bainitic (160mm/min) or bainitic and
martensitic(220mm/min). However, at higher welding currents (particularly
195 amps) the effect of reduced heat input, caused by increased welding speed, did
not affect the microstructure in the manner predicted by Fig. 8.2. Fine lath ferrite
was produced at low welding speeds (133mm/min) and high heat input yet grain
boundary ferrite and Widmanstatten side plates were formed at the high speeds
(220mm/min), with a consequent lower heat input. There are several possible
explanations of the apparent departure from the microstructures predicted using the
C C T curves of Fig. 8.2. However, it must be remembered that Fig. 8.2 presents a
diagramatic representation of phases present for a particular steel relative to the
temperature and cooling time. It does not describe the preferred sites for, or the
140
mechanism of, phase nucleation and precipitation or how variations to weldmetal
deposition conditions can change the mechanisms and kinetics of nucleation and
precipitation. Higher welding currents can contribute to higher arc temperatures
(equation 2.1) although overall heat input may be lowered. It could be argued that
the higher arc temperatures, caused by higher welding currents can influence
iron/oxygen reactions which could lead to a higher weldmetal oxide content.
Obviously oxygen analysis and an examination of the oxide distribution would be
necessary to validate this hypothesis.
Alternatively, it has already been proposed that low welding speeds
(133mm/min) cause a degree of weld pool turbulence during the solidification
process so that nucleation sites for fine lath ferrite are more uniformly dispersed,
reducing the dependence on grain boudary nucleation. This proposition also would
require a quantitative metallographic examination of the oxide particle size and
distribution.
The resistance to fracture of fine lath ferrite has been investigated by Ohkita
et al. (232) w h o presented evidence to indicate that the interlocking laths and the
large angles between the laths caused resistance to the propagation of cleavage
cracks. The numerous changes in direction of the crack propagating in fine lath
ferrite would cause cleavage cracking to be more difficult than the propagation of
cleavage cracks along large ferrite plates or grain boundaries. For the latter
structure, reduced fracture strength and low fracture toughness was observed (51).
In the present work it was found that under conditions of high voltage welding,
microstructures with large ferrite precipitates fractured in a ductile manner, (see
Section 7.2, Fig. 7.7). Such fracture is normally considered to involve plastic
strain energy; however, as proposed in Section 7.3.2 the high oxide content could
141
be responsible for the greater ease and larger numbers of microvoids nucleated and
hence the lower fracture strength.
In the interpretations of the HAZ microstructures emphasis has been placed
on the volume fraction of ferrite in the H A Z . It is obvious that quantitative
measurements of ferrite content are more desirable than qualitative metallographic
observations. However, it should be noted that with normal etching techniques
difficulties are often experienced in clearly distinguishing between some
morphologies of ferrite and similar martensitic and bainitic lath structures.
Reproduction of the etch can also be a problem because of variations in the etching
conditions. To overcome these problems Navaro and Easterling(233) used
interference layer metallography to distinguish clearly between the various
constituents of the H A Z microstructure and carry out quantitative metallographic
analyses. Nevertheless, in the present investigation it was possible, from qualitative
metallographic observations to relate ferrite content of the H A Z to variations in the
welding parameters.
If two of the extrinsic welding parameters were maintained constant and
the third varied, ferrite content of the H A Z increased with increased heat input for
each series. However, cross comparisons are difficult and are best exemplified by
the results of the constant heat input series of test welds. Test welds performed at
150amps, 20.5volts, and 133mm/min were found to have a CI of 0 % , yet the
microstructure of the H A Z appeared to contain less proeutectoid ferrite than test
welds carried out at at the same heat input(1.38kJ/mm), but at a higher welding
current (180amps) and speed (160mm/min), see Section 7.5. Microhardness
measurements indicated that the H A Z was softer in weldments produced at the
lower welding speed , Fig. 7.44, and the microstructure of the H A Z contained less
142
ferrite, Fig 7.42. It was suggested that tempering of the martensitic H A Z had
occurred, Section 7.5. Weld bead morphology was also found to change with
welding current: rounded and more globular weld beads being related to low
welding currents. Shoda and Doherty (234) pointed to the relationship between
welding parameters and bead geometry, and McGlove and Chadwick(235) have
been able to to develop predictive equations for weld bead geometry based on
welding current, voltage and speed. These equations are empirical and do not relate
to the H A Z . Electrode flux coating was also found to influence weld bead geometry
and a similar H A Z microstructure, hardness and reduced CI value was observed, see
Section 7.3. Relative interfacial energies appear to be the important common factor
in both the cases of different electrode fluxes and the low current welds. In the
latter case it could be proposed that the reduced arc temperature (see equation 2.1)
could influence the surface energy configuration and by restricting the weld bead to a
more globular shape cause tempering of the H A Z . Obviously this proposal is
speculative, but it is clear that a causal relationship exists between weld bead
geometry and autotempering of the H A Z , even though the exact relationship is not
clear at this stage.
The variability of root angle, stress intensity factor, Fx, and unstable
crack growth have been proposed as possible causes of the scatter in the results.
The hypothesis of Karppi(162) for the stress intensity factor would suggest that
scatter of results was related to the nucleation of the crack, whilst the analysis of
Masubuchi (164) suggests unstable crack growth. During the course of the present
work it was found that root angle varied considerably, but could not be related to the
scatter of results. For welding conditions leading to low or zero values of CI little
scatter of results was observed, see Fig. 7.2. However, for CI values above
approximately 3 5 % it can be seen from Fig. 7.2 that scatter was large. This
143
suggests that the crack propagation hypothesis of Masubuchi(164) is the controlling
factor. Furthermore, because the critical value of CI appears to be in the range 25 to
3 5 % it is possible that weldmetal microstructure also contributes to the scatter in CI
results.
Although it was possible to distinguish between three steels of different C
in relation to their weldability, it was equally possible to choose conditions that
would not allow such a distinction to be made. In all cases, welding conditions
were within the range of conditions normally associated with the electrodes used. It
is significant that welds performed using the reference welding conditions caused
weldment cracking entirely in the H A Z for both the 0.36 C _ steel and the 0.41 C
steel. The CI values measured did not incorporate a contribution from weldmetal
fracture. The results for the three steels, under the reference welding conditions,
could thus be considered comparable. However, under other welding conditions
which produce a CI of 1 0 0 % for 0.41 C both H A Z and weldmetal cracking
occurred, similar to that shown in Fig. 7.28.
The results of the present work have shown that not only can welding
conditions change the results obtained from the Tekken Test but that specifying
constant heat input cannot guarantee reproducible results. Although each of the
welding parameters contributes to the heat input, they also have independent effects
on weldability test results. It appears that the concept of welding being a process of
heat and mass transfer oversimplifies the situation when the final mechanical
properties of the weldment are considered. Welding voltage variations can cause
changes to the weldmetal chemistry and can increase the oxygen content of the
weldmetal, current variations influence the weld bead geometry by affecting the arc
144
temperature, and welding speed contributes to weld pool stirring. The added
complications brought about by such factors as variations to electrode flux
composition developed by different commercial manufacturing sources restrict the
Tekken Test results to the exact welding conditions and materials used in the test
welds.
145
C O N C L U S I O N S
From the results of this investigation it is concluded that the Tekken
Test is complicated by the interaction of many variables. Interpretation of CI results
would be expected to be difficult and from the experience of the present work quoted
values of CI could only reliably relate to the materials, electrodes and welding
conditions used to generate the test results. From the CI results obtained from the
Tekken Test welds that were performed under different welding conditions, the
metallographic observations obtained from the weldments, and measurements of
thermal cycle and restraint stress can be summarised as follows.
[1] For the conditions and materials used in the present investigation,
quantitative measurements of weld cracking, as defined by the Cracking Index (CI),
using the Tekken Test can vary with welding conditions and can scatter
considerably.
[2] In the present work CI was found to increase with increased welding voltage
and speed and decrease with increased welding current.
[3] The dependency of CI on welding parameters is a consequence of fracture
susceptible microstructures in the H A Z and in the weldmetal. The CI is reduced by
conditions which promote the formation of fine lath ferrite in the weldmetal and
coarse ferrite particles in a martensitic H A Z .
[4] Because of the relative fracture susceptibilities of both the HAZ and the
weldmetal the value of CI measured may relate to weldmetal cracking only, H A Z
146
cracking only, or incorporate contributions from both the H A Z and the weldmetal.
[5] Because the CI measured in the Tekken Test can involve cracking of both
the H A Z and the weldmetal, comparisons of the susceptibilities of steels to H A C of
the H A Z is difficult.
[6] Additional factors such as root gap of the test piece, electrode flux
composition and electrode classification were also found to have an effect on the
Tekken Test results and so add further doubt to the reliability of test results.
The purpose for which weldability tests are intended as outlined in Chapters 1
and 4 are;
(i) to determine the factors that determine good weldability,
(ii) to develop procedures to obviate defects in the welded joint, and
(iii) to compare the relative weldability of various steels
However, the ultimate objective of a test is prediction of the reliability of a
welded structure.
From the results of the present work weldment cracking in the Tekken Test
appears to be sensitive to small variations in welding conditions and parameters with
considerable scatter of results. Thus there is considerable doubt whether the Tekken
Test is a reliable test for H A Z cracking and for determining weldability.
Although these conclusions should be related only to carbon-manganese
steels described in this thesis and not to other materials such as the low alloy steels
used by Itoand Bessyo(236) it is however probable that the conclusions for this
work are generally applicable.
147
APPENDIX 1
Publications by the candidate presented in support of this thesis.
(1) The Morphology of Creep Cavities in a-Iron , D.M.R. Taplin and A.L. Wingrove, Proceedings of the 2nd International Fracture Conference, Brighton, 1969.
(2) The Morphology and Growth of Cavities in a-Iron , A.L. Wingrove and D.M.R. Taplin, Jnl. of Materials Science, 4, 789, (1969).
(3) Grain Boundary Sliding and Cavitation in a-Iron, A.L. Wingrove and D.M.R. Taplin, Scripta Metallurgica, 3, 649, (1969).
(4) The Energy Required for High Speed Shearing of Steel, T.A.C. Stock and A.L. Wingrove Jnl. Mech. Eng. Sci., 13, 10, (1971).
(5) A Note on th Structure of Adiabatic Shear Bands in Steel, A.L. Wingrove, Jnl. Aust. Inst, of Metals, 16, 67,(1971).
(6) A Device for Measuring Strain- Time Relationships in Compression from Quasi Static to Dynamic Strain Rates. , A.L. Wingrove, Jnl. of Physics E, Scientific Instruments,4, 873,(1972).
(7) The Forces for Projectile Penetration of Aluminium, A.L. Wingrove, Jnl. Physics D, Appl. Phys., 5, 1294(1972).
(8) A Note on the Formation of Chip Fragments due to Adiabatic Shear, S.A. Manion and A.L. Wingrove, Jnl. Aust. Inst of Metals, 17,158, (1972).
(9) Some Aspects of Relating Structure to Properties of Heavily Deformed Copper, A.L. Wingrove, Jnl. Inst. Metals, 100, 313, (1972).
(10) The Influence of Projectile Geometry on Adiabatic Shear and Target Failure, A.L. Wingrove, Met. Trans. 4, 1829, (1973).
(11) The Use of Mild Steel in the Controlled Fragmentation of Experimental Warheads, A.J. Beadford and A.L. Wingrove, Defence Standards Tech. Report No. 506 (1973).
(12) Some Aspects of Target and Projectile Properties on Penetration Behavior, A.L. Wingrove and G.L. Wulf, Jnl. Aust. Inst, of Metals, 19, 167, (1973).
(13) The Phenomenon of Adiabatic Shear Deformation, A.J. Bedford A.L. Wingrove and K.R.L. Thompson, Jnl of Aust. Inst. Metals 19, 61, (1974).
(14) The Penetration of Metals by Projectiles, A.L. Wingrove, Metals Australia,5 251, (1973).
(15) The Tekken Test- The Influence of Welding Variables , A.L. Wingrove, D.P. Dunne and N.F. Kennon, Aust. Weld. Res., 14, 8, (1985).
148
REFERENCES
(1) Welding Handbook, 7th ed., American Welding Society, (!976).
(2) R.F.Tylecote, Metallurgy in Archeology , The Metals Society, London(1962).
(3)Glossary of Metallurical Terms and Definitions, Aust, Standard A.S. 25-1968.
(4) Definitions of Metallurgical Terms , Aust. Inst, of Metals, 3rd ed., Melbourne
1971.
(5) H. Herbiet and KV.Salkin, Zvaranie , 16, 203, (1967).
(6) D. Seferian, The metallurgy of Welding, Chapman & Hall, London, 1962 .
{l)The Weldability of Steels, A W R A Technical Note No.l, May 1982.
(8) Arc Welding Low Alloy Steels, B.W.R.A. publication, August 1956.
(9) J. MacCutcheon, Weld. Jnl, 27, 52, (1948).
(10) Alexander L. Kielland - ulykken (in Norweigan), Norges offenthige
utredning N O U 1981:11 University Press Oslo 1981.
(11) K. Masubuchi, and K. Terai, Future Trends Materials and Welding
Technology for Marine Structures, presented to the 1976 Spring meating of the
Society of Navel Architects and Marine Engineers, Philadelphia P. A. June 1976.
(12) K. Masubuchi, Materials for Marine Engineering, M.I.T. Press, 1970.
(13) C.L.M. Cotterell and B.J. Bradstreet, Brit. Welding Jnl., 34,305, (1955).
(14) B.J. Bradstreet, The Arc Welding of Low Alloy Steels, B W R A publication,
1956.
(15) K. Easterling, Introduction to The Physical Metallurgy of Welding
Butterworth, London, 224,(1983)
(16) R.D Stout, et al., Weld. Jnl. Res. Supl., 25,522s,(1946).
(17) R.D. Stout, R. Vasudevar and A . W Pense,. ibid, 55, 89s, (1976.
(18) R.D. Stout, Welding of Line Pipe Steel, Welding Res. Council publ., New
York, 1977.
(19) A.K. Church, E. Blackwood and K.W. Gunn, Aust. Welding Jnl., 24, 25,
(1980).
(20) M.J. Onati, Jnl. Jap. Welding Soc, 25,277, (1956).
(21) K. Hiroski et al, Weld. Jnl. ResSuppl., X ,36s, (1962).
(22) H. Kihara, H. Suzuki and H. Tamura, Soc. ofNav. Arch, of Jap.,, 60th
Anniv Ser., 1, 1, (1957).
(23) J.H. Hensler, J.W. Graham, and G.V Cullen, Aust. Welding Res., 1,
1.1970.
(24) Report by Project Pannel P4, A W R A Report No. P4-71-79 Aust Welding
Research, 8, 53, ( 1980).
(25) M. Inagaki,. H.Suzuki,, and H.Nakamura, Jnl. Jap. WeldRess , 33,41,
(1964).
(26) Annales de L'Institute Technique du batment et des travaux public A 3, 39,
(1938).
(27) E.M. MacCutcheon, Weld Jnl, 27,52, (1948).
(28) S. Matsui and M. Inagaki, IIW Doc. 1X-970-76.
(29) N. Yurioka, International Congress of Welding Research, May 1984.
(30) W. Finkclnerburg and H. Maecher Electronic Arcs and Thermal Plasma ,
Handbook der Physk, 22, 254, (1957).
(31) W. Mantel, Schweissen und Schneider,, 8,278, (1956).
(32) H.N. Olsen, Paper presented to the 10th Annual Gaseous Electronics
Conference N.Y., October 1957.
(33) J.F. Lancaster, Brit. Weld. Jnl, 1,412, (1954).
(34) H.N Olsen, Paper presented to The 11th Annual Gaseous Electronics
Conference N.Y. 1958.
(35) C.E Jackson, Weld. Jnl. Res. Suppl.,39, 177s, (1960).
(36)_C.E. Jackson, Weld. Jnl. Res. Suppl., 39,129s, (1960).
(37) R.H. Gillette and R.T. Breymeier, Weld. Jnl. Res. Suppl, 30,146s, (1951).
(38) N.N. Prokhorov and N.N. Mastryukova, Weld. Prod., 2,7, (1961).
(39) J. Wegrzyn, UW Doc. 212-292-73.
(40) J. Wegrzyn, Met.Const., 12, 326, (1980).
(41) A.S Mastryukova, and N.N. Prokhov, Automatic Welding, 7,6,(1963).
(42) N.N. Prokhorov,and A.S. Mastryukova, ibid, 18, 17, (1965).
(43) J. Waring, Aust Welding Jnl, 11, 15,(1967).
(44) W.F. Savage et al., Weld. Jnl. Res. Suppl, 44, 175s, (1965).
(45) W.F. Savage and A.H. Arinson, ibid, 45, 85s, (1966).
(46) F. Matsuda et al., Trans. Nat. Inst. Metals. Japan., 11, 43, (1969).
(47) C.R. Loper et al., Weld. Jnl.Res. Suppl, 46, 94s, (1969).
(48) J. Calvo et al., Acta. Met., 8 , 898, (1960).
(49) E. Rassanen and J. Tenkula, Scand. Jnl. Met.,l, 75, (1972).
(50) H. Terahima and J. Tsuboi, Met. Const.,14, 648, (1982).
(51) P.R. Kirkwood, Metal Const. ,8 260, (1978).
(52) Y. Ito, M. Nakanishi and Y. Komizo, Metal. Const.,14,472, (1982).
(53) R.A. Ricks, P.R. Howell, and G.S. Barrite, Jnl Mat. Sci.,17,732,(1982).
(54) D.J. Abson. and R.E Dolby,Welding Inst., Report No. 19/1978. London
1978.
(55) R.C. Cochrane and P.R. Kirkwood, " Trends in steel consumables for
welding "Welding Inst. Conf. Abingdon ,1978,103.
(56) H. Ohkita, S. Tsushima and N. Mori, Aust. Weld. Jnl, 29, 29, (1984).
(57) H.I. Aaronson and C. Wells, Trans. AJ.M.E.,210, 1216, (1956).
(58) T. Bomszwski,The Metallurgist.,11, 640 , (1979).
1
(59) F.R. Coe, "Welding without hydrogen cracking " , Weld. Inst. publ.
Cambridge,1973.
(60) H. Suzuki, Trans. Jap. Weld. 5oc.,(1978), also IIW Doc. 1X-1074-78.
(61) J.F. Lancaster, Brit. Weld. Jnl, 3, 370, (1971).
(62) D. Rosenthal, Congress National des Sciences, Comptes Redix Bruxelles , 2,
1277, (1935) and Trans. A.S.M.E., 68,49, (1946).
(63) H. Ikawa, H. Oshige and S. Noi, Jnl. of Welding (Japan) ,4,317 (1977).
(64) S. Matsuda and N. Okumera, Trans. I.S.U., 18, 198 , (1978).
(65) C.S. Zener, Personal communication to C.S. Smith, reported in A.M.M.E.,
175, 15, (1949).
(66) P.J. Albury, B Chew and W.K.C. Jones, Met Technology, 4, 317, (1977).
(67) M.F Ashby and K.E. Easterling Acta Met., 3041969, (1982).
(68) Results summerised by J.E. Dorn, Creep & Recovery, Cleveland, ASM, ,255,
(1956).
(69) N.E. Hannerz, and R. De Kazinezy, J.I.S.I., 208,475, (1979).
(70), M. Inagaki and H. Sekiguchi, Trans Nat. Res. Inst. Met. Japan, 2,102,
(1960).
(71) J.A. Ronningen, T. Simonsen and N. Christensen cara/. Jnl. of. Met ,3, 87,
(1973).
(72) J. Dearden, and ,H.A. O'Neill, Trans. Inst. Weld. (UK),3, 204 (1940).
(73) V. Pavosker and J.S. Kirkaldy, Scand. Jnl of Met., 11,256, (1982).
(74) P. Maynier et al., Proc. AIME Conf.Symp. Oct. 1977,518.
(75) J.C. Ions, K.E. Easterling and M.F Ashby, Acta Mef.,32, 1949, (1984).
(76) E.FNippes, Weld. Jnl. Res. Suppl. ,38,1s, (1959).
(77) W.A. Johnson and R.F. Mehl, Trans AIME, 35,416, (1939).
(78) M. Wantanbe et al., Jnl. Soc. Nav. Arch of Japan, 104,153, (1958).
(79) R.F. Mehl, Trans. AIME, 54, 318, (1943).
(80) K.E. Easterling, Introduction to the Physical Metallurgy of Welding,
Butterworth, London, Boston 1983.
(81) H.L. Schich, Thermodynamics of certain Refractory Compounds, Academic
Press 1966.
(82) K. Huntz et al., Mem Sci. Rev. Metall., 66 , 85 (1969).
(83) P.J. Albury and C.W Howarth, Met. Sci. Jnl. ,8, 407, (1974).
(84) P.J. Albury and W.K.C. Jones, Metals Technology , July 1977,360.
(85) J.F Lancaster, Metallurgy of Welding, Third Ed. George Allen and Unwin
Ltd 1980.
(86) A.R.E Singer and P.H Jennings, J.I.M., 73,273, (1947).
(87) H.F.Bishop et al, Trans AFS, 68,518, (1960).
(88) J.C. Borland, Brit.Weld. Jnl, 7, 508, (1960).
(89) E.C. Rollason and D.F.T. Roberts, Trans .Inst Weld., 13 ,129 (1950).
(90) J.A. Williams and A.R.E. Singer, Jnl Aust. Inst. Met. ,11, 2, (1966).
(91) J. VanEeghem and A. DeSy, Modern Casting, plOO, July 1965.
(92) T. Boniszevoski et al., Brit. Weld. Jnl. ,13,558, (1966).
(93) T. Boniszevoski and S.E. Le Diew, Brit. Weld. Jnl, 14,132, (1967)..
(94) W. Roberts, B. Lehtinen and K.E. Easterling, Acta Met. ,24, 745, (1976).
(95) S. Ganesh and R.D. Stout, Weld. Jnl. Res. Suppl, 55,341s, (1976).
(96) A.H Goodger, BSINews, 10, Sept. 1966.
(97) M. Watanabe, Symp., Welding and Shipbuilding, Inst, of Welding, 219,
(1961).
(98) Y. Takeski,Welding and Metal Fab., 43,740, (1975).
(99) Nucleonics Week, World News, April 17,1969.
(100) T. Boniszewski and R.G. Baker, J.I.S.I. ,202, 921, (1964).
(101) D.J. Widgery and T. Boniszewski, J.I.SJ., 204,53, (1966).
153
(102) R.N. Younger, J.C. Borland and R.G. Baker, Brit. Weld. Jnl, 8, 575,
(1961).
(103) D.S. Duvall and W.A. Owezarski, Weld Jnl. Res. Supp., 32, 423s (1967).
(104) W.F. Savage and B.M. Krantz, ibid ,31, 13s (1966).
(105) D.I. Roberts and D.J. Pucknell, Inst. Weld. Conf. Oct., Paper No 7,
London, 1967.
(106) J.D. Fast, Gases in Metals, Phillips Technical Library, 1976.
(107) CD. Beacham, Met. Trans. ,3,437 (1972).
(108) A.R. Troieno, Trans. A.S.M. ,52, 54, (1960).
(109) W.W. Gerberich and C.E. Hartbower, Proc. Conf. Fundermentals of Stress
Corrosion Cracking, NACE,1967.
(110) F.R. Coe and Z. Cheno, Hydrogen distribution and removal for a single bead
weld during welding, Part 3, Weld Res Inst., (1975).
(Ill) W. Schwarz andH. Zitter, IIW Doc 11A-258-70, (1970).
(112) T. Boniszewski and J.A. Moreton, B W R A Rep P/19/66.
(113) N. Christensen et al, Brit. Weld Jnl, 6, 272, (1958).
(114) Japan Weld Eng. Soc. and Soc. of Nav. Arch of Japan, Conf on Weldability
of High Tensile Steels, Osaka Japan, 1980, Discussion.
(115) K. Farell and A.G. Quarrel, JISI, ,202 , 1002, (1964).
(116) L. Reeve, Trans. Inst. Weld.,3, 111, (1940).
(117) J.C. Ball and C.L.M. Cottrell, JISI, 169,321, (1951).
(118) T. Boniszewski, and F. Watkinson, Metals & Materials, 7, 90, (1973).
(119) K.C.Wuand WR.E. Herfert, Weld Jnl Res. Suppl. ,32, 32s, (1967).
(120) G.E. Metzger, Weld. Jnl. Res. Suppl. ,32, 457s, (1967),
(121) J.P. Hirth, Met. Trans. ,11A, 861, (1980).
(122) M. Smialowski, Hydrogen in Steel, Addison-Wesly,Reeding, Mass. 1962
(123) H. Granjon.BnV. Weld Jnl, 1,105, (1954).
(124) H. Granjon, Welding World, 1,105, (1963).
(125) H. Granjon, ibid, ,17, 81, (1979).
(126) H. Linden et al., IIW Doc. 1X-779-72.
(127) G.M. Evans and N. Christensen, SINTEF Report, Project 340424 (1971).
(128) P.M. Hart, Weldlnst. Rep., Rep. No. 12/1976/M. (1976).
(129) K. Satoh etal.., U W Doc. 1X-874-74.
(130) A.T. Ficker and T. Muller, IIW Doc. 1X-977-76.
(131) N. Christensen, U W Doc. 1X-824-73.
(132) J.M. Sawhill, Weld. Jnl. Res. Supp. ,53, 554s(1974).
(133) N.F. Kennon, /. Aust. Inst. Met. ,19, 3, (1974).
(134) P. Hart and F. Watkinson, IIW Doc. 1X-889-74.
(135) K. Satoh et al. H W Doc. 1X-737-71.
(136) R. Karppi et al. Trans. Jap. Weld. Soc, 12,14,(1981).
(137) H. Suzuki et al. Jnl. Jap. Weld Soc, 32,44,(1963).
(138) M. Inagarki et al. Jnl. Jap. Weld. Soc.,34, 801, (1965).
(139) M. Otani, ibid ,25, 277,(1956).
(140) H. Kihara et al., Weld Jnl. Res. Supp.,41, 36s, (1962).
(141) Japanes Industrial Standard J.I.S. Z23158-1966.
(142) P. Hart and N.F. Watkinson, Weld Jnl Res. Suppl., 51, 349s, (1972).
(143) D.W. Morgan, Welding Jnl, 55, 1035, (1976).
(144) R.G. Davies, and C.L. Magee. Met. Trans. ,3,307,(1972).
(145) A.P. Granjon, Weld. Jnl. Res. Suppl. A.P.,41, pis, (1962).
(146) F.B. Pickering, Physical Metallurgy and Design of Steel ,App Sci. Pub.
1978.
(147) P.H.M. Hart, IIW Doc. 1X-1308-84.
(148) N. Yurioka, Int. Congress on Welding Research, May 1984.
(149) A.S. Tetelman, Int. Fracture Conf., Interscience Pub. NY,1963.
155
(150) R. Gibala, Trans AIME ,239,1574, (1967).
(151) R.K. Wilson, Weld. Jnl Res. Suppl, 61,182s, (1982).
(152) H.C. Rogers, Acta Met., 4,114,(1956).
(153) P. Bastien and P. Azou, Proc First World Met. Conf, A S M Cleveland, 535,
(1952).
(154) N.J. Grant and J.L Lumsden, Iron Age , 175,92, (1955).
(155) K. Satoh, Y. Ueda andH. Kihara Weld World, 11, 133, (1973).
(156) T. Naka, Shrinkage & Cracking of Welded Joints, Komie Kogyo Co.,Tokyo,
1950,142.
(157) K Satoh and S. Matsui, H W Doc. 1X-574-68.
(158) K. Masubuchi and N.T. Ich, Weld Jnl Res Suppl, 49, 166s, (1970).
(159) S. Matsui et al., IIW Doc. 11-512-69.
(160) H.S. Rawdon and T. Burglund, U.SA. Stands. Bureau Scientific Papers ,
22,649, (1928).
(161) J.D. Fast, Phillips Res. Report, 5, 37, (1950).
(162) R.A.J. karppi. et al, Scand. Jnl Met. ,13, 66, (1984).
(163) J.M.F. Mota and R.L. Apps, Weld Jnl Res Suppl., 61, 222s, (1982).
(164) K. Masubuchi, Analysis of Welded Structures, Pergamon Press, 1980,
Oxford, NY, Sydney, Paris.
(165) K. Winterton, Weld Jnl Res Suppl. ,40, 253s, (1961).
(166) C.L.M. Cottrell, Brit. Weld. Jnl ,1,409, (1954).
(167) L. Reeve, Brit. Weld. Jnl. & Met. Const. ,1,106, (1969).
(168) Y. Ito and K. Bessyo, Jnl. Jap. Weld. Res. ,37, 983, (1968).
(169) N.Yurioka et al., A W R A Conf., Pipeline Welding in the 80's, Melbourne,
1981.
(170) R.F. Mehl and W.C. Hagel, Prog, in Met. Phys., 6, 74, (1956).
(171) N. Smith, and B.I. Bagnall, Brit. Weld. Jnl, 15, 63, (1968).
156
(172) R.A. Grange, Met. Trans. ,4, 2231, (1973).
(173) Y. Ito and K. Bessyo, Weldability Formulae of High Strength Steel Related
to HAZ Cracking, U W Doc. .1X-576-68.
(174) Inagaki et al., Jnl. Jap. Weld Soc, 34, 801, (1965).
(175) K. Satoh et al, Jnl. Jap. Weld. Soc , 48,504, (1979).
(176) H. Suzuki et al., IIW Doc 1X-1195-81.
(177) G.M. Evans and N. Christensen, Scand. Jnl. Met. ,2, 91, (1973).
(178) L. Kalev et al., U W Doc. 1X-1013-76.
(179) K. Satoh and T. Terasaki, IIW Doc. IX-1075-78.
(180) M. Toyoda et al., Trans. Jap weld. Soc, 9,36, (1978).
(181) J.M.F Mota andRX. Apps, Weld Jnl. Res Suppl., 61, 222s, (1982).
(182) H.H. Johnson, Fundamentals of Aspects of Stress Corrosion Cracking
NACE, Houston 1969.
(183) J.G. Bailie, Brit. Weld. Jnl, 14, 51, (1967).
(184) A.L. Wingrove, Univ. of Wollongong Consulting Report, May 1985.
(185) R. Gunnert, Residual Welding Stresses, Uppsala, Almqvist & Wilksells,
Bokryckerri, AB 1955, p71.
(186) N. G. Leide, Residual Stresses and their Effect.Jnt Conf. London, 1977,
Weld Inst, Cambridge pl07.
(187) Japanese Studies on Structural Restraint Severity in Relation to Weld
Cracking Welding in the World,, 15, 155, (1977).
(188) R.D. Stout et al., Weld Jnl. Res. Suppl., 25,522s, (1946).
(189) J. Krieg, Daskohlenstoff aguivalent und seine praktische Bedeutung. Prakt.,
23, 206, (1971).
(190) J. Brisson, P.H. Mayneir, J. Dollet and P. Bastein, Weld Jnl. Res. Suppl,
51, 208s, (1982).
(191) A.J. Bryhan, Weld. Jnl Res. Suppl. ,59, 169s, (1980).
1
(192) H. Suzuki, Trans. Jap. Weld. Soc. ,16, 40, (1985).
(193) A. Rose, Stahl. Eison., 86, 663, (1966).
(194) J. Kas and T.T. ADrichan, Schweissen Schneiden , 21, 199, (1969).
(195) H. Sekiguchi and M. Inagarki, IIW Doc. 1X-262-60.
(196) P.L. Threadgill, Metals and Mat. , 20, 422, (1985).
(197) H. Granjon, Int. Symp. on Fracture in Welds, 1971 Conf. Proc. Jap weld
Soc Tokyo.
(198) S.S. Taliani, Wed Res Int. ,6, 19, (1976).
(199) R.A.Farrar, ibid. ,7, 85, (1977).
(200) J.M.F. Mota, R.L. Apps and J.E. Jubb, Proc. Conf. on Trends in Steels
and Consumables for Welding", Welding Int Abingdon (1978).
(201) IIW Guide to Welding and Weldability ofC-Mn and C-Mn microalloyed
steels., IIW Doc. 382-71.
(202) C.L.M. Cottrell and P.H.R. Lane, B W R A Welding Research, 6, 2,(1952).
(203) C.L.M. Cottrell and M.D.Jackson, ibid, 6, 31, (1952).
(204) C.L.M. Cottrell, Weld Jnl. Res Suppl. ,32, 257s, (1953).
(205) C.L.M. Cottrell and B.J. Bradstreet, Brit. Weld Jnl., 2, 305, (1955).
(206) H. Granjon and R. Gaillard, Soudage et Techniques Connexes, 18,343,
(1964).
(207) H. Granjon, Metals Const, 1,509, (1969).
(208) J. Gordine, Weld Jnl. Res Suppl, 56, 201, (1977).
(209) V Inagarki, IIW Doc. 1X-1151-80.
(210) U W Subcommission 1XB, Draft standard on implant cracking method
suggested by the French delegation to the IIW Comm. IX Oporto, Portugal,1981.
(211) R.D. Stout, S.S Toss, L.J. McGready and G.E. Doan, Weld Jnl. Res.
Suppl, 25,522s, (1946).
1
(212) H. Kihara, H. Suzuki and H. Nakamura, Weld Jnl. Res. Suppl.,4, 36s,
(1962).
(213) H. Suzuki and N. Yurioki, Aust. Weld. Jnl.,9, Autumn Issue (1982).
(214) D.D.Schweimmer, D.L. Olsen and D.L.Williamson, Weld Jnl. Res.
Suppl, 58, 153s, (1979).
(215) T.H. Hazlett, ibid, 36, 18s, (1957).
(216) R.A. J. Karppi, M.Togoda and K.Satoh, U W Doc. 1X-1191-81.
(217) Private discussions, welding research staff Aust. Iron & Steel P/L Port
Kembla.
(218) W.P. Cambell, Welding Jnl Res. Suppl, 55, 135s, (1976).
(219) L.E. Samuels, Jnl. Inst. Met., 81,471, (1952).
(220) R.L. Smith and A.M. Inkalo, Proc ASTM, 57, 536, (1957).
(221) K. Masubuchi, Welding Res Council Bull, 149 (April 1970).
(222) H. Kihara and K. Masubuchi, Rep. Trans and Tech Res. Inst., No. 20
(1956).
(223)H. Kihara, K. Masubuchi and Y. Ogara, Jnl. Soc. ofNav. Arch of
Japan.,100, 163, (1956).
(224) B. Hansworth et al., Met. Const. ,1,5,(1969).
(225) R.E. Reed-Hill, Physical Metallurgy Principles, D Van Nostrand and Co.
Inc., Princ. N.J. 1976.
(226) W.G. Essers, G. Jelmorini and G.W. Tiechelaar, Brit. Weld Jnl, 3(4), 151,
(1971).
(227) B.M. Patchett, Weld Jnl Res Suppl, 53, 203s, (1974).
(228) H.E. Janzen, MSc. Thesis, Colarado School of Mines, (1970), Cited by
Schweimmer Ref. 214.
(229) L. Fine, Metals Progress, 49, 108, (1946).
(230) T.Teraski, R. Karppi and K. Satoh, Trans. Jap Weld Soc, 10, 53 (1979).
(231) Y. Ito, N. Nakanisiki and Y. Komozo, Met Const, and Brit. Weld Jnl, 14,
472, (1982).
(232) S. Ohkita et al., Aust. Weld Jnl, 28, 29, (1984).
(233) E. Navaro and K.E. Easterling, Scand. Jnl. Met., 11, 169, (1982).
(234) T. Shinoda and J. Doherty, Weld Inst. Report, No. 74/1978/PE.
(235) J.C. McGlove and D.B. Chadwich, Weld Inst. Repoprt, No. 80/1978/PE.
(236) Y. Ito, K. Bessyo, Jnl. Jap. Weld. Soc., 37, 983, (1968).
(237 ibid., 38, 1134, (1969).
(238) R. Karppi, D Tech Thesis, Helsinki University of Technology, Espoo
Finland, 1982.
(239) T. Boniszewski et al., Brit. Weld. Jnl., 13,558, (1966).
(240) T. Boniszewski and S.E. Dieu, Brit. Weld. Jnl, 14,132,(1967).
(241) Australian Specification, AS 1552-1973.