You are on page 1of 9

Computational Materials Science 63 (2012) 182–190

Contents lists available at SciVerse ScienceDirect

Computational Materials Science


journal homepage: www.elsevier.com/locate/commatsci

Hot cracking in stainless steel 310s, numerical study and experimental verification
A.R. Safari a,⇑, M.R. Forouzan a, M. Shamanian b
a
Department of Mechanical Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran
b
Department of Materials Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Hot cracking is a serious problem in welding of many alloys such as high strength steels, austenitic stain-
Received 29 February 2012 less steels and aluminum alloys. It takes places during the last stage of solidification where mechanical
Received in revised form 8 June 2012 deformation develops in the mushy zone while the material has low ductility. In this study, hot cracking
Accepted 11 June 2012
initiation and propagation of austenitic stainless steel 310s was studied. A viscoplastic constitutive model
Available online 7 July 2012
was proposed and implemented in finite element simulation. Solidification shrinkage, viscosity, anneal-
ing at high temperature and melting effect of fusion zone and a criterion for hot cracking initiation and
Keywords:
propagation are the main features of the solution. Numerical results were compared with some experi-
Hot cracking
Austenitic stainless steel
ments accomplished in this study in order to verify the proposed method. The results showed that max-
Viscosity imum transverse mechanical strain criterion could predict both initiation and propagation of the hot
Constitutive model cracking. Annealing and melting had the most effect in predicting crack length. Omission of these param-
Annealing eters leaded to underestimated results. Elimination of viscosity effect leaded to overestimation of the
crack length while elimination of solidification shrinkage leaded to underestimated crack length.
Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction [1]. Weld metal solidification cracking has the same root of crack-
ing in casting. In fact, many ideas for weld metal solidification
In the welding process, a concentrated moving heat is applied to cracking come from studies about cracking in casting which is
the joint. This results in a non-uniform transient temperature field usually referred to as ‘‘hot tearing’’. More details about hot crack
in the structure. Because of rapid cooling rate, the solidification of formation can be found in [2,3]. A summary of hot cracking mech-
fusion zone is a non-equilibrium process with dendritic micro- anisms can be found in [4–6].
structure in the solidified weld metal. During the last stage of the Various factors have effect on materials susceptibility to hot
solidification, low melting-point constituents segregate between cracking. Low melting point constitutes increase hot cracking sus-
solidifying dendrites and form liquid films over the interface of ceptibility. During solidification, this constitutes are rebutted by
dendrites. This stage of solidification in which both liquid and solid solidifying dendrites and concentrates at the weld centerline
coexist is called brittle temperature range (BTR). The BTR starts increasing crack susceptibility. Solidification temperature range,
from a temperature below the liquidus temperature called coher- temperature gradient, solute content at the last stage of solidifica-
ency temperature (TC). Below the TC, dendrite arms coalesce; solid tion, surface tension of interdendritic liquid, impurities, ductility of
skeleton form and material can transmit stress. In the BTR, the duc- weld metal in mushy zone, thermal contraction, solidification
tility of material is low. At the same time, material contracts from a shrinkage, extent of restrain, grain shape, structure and size, are
state of zero stress and zero strain and tensile deformation devel- among the most important parameters affecting materials suscep-
ops due to external restrains. Hot cracking occurs when tensile tibility [1,5].
deformation exceeds from material strength and reflow of melt The metallurgical aspects of hot cracking have been studied in
cannot heal them. Weld metal contracts due to both thermal con- much more details than the thermo mechanical aspects. A success-
traction and solidification shrinkage. High thermal contraction of ful description of hot cracking should take into account both met-
austenitic stainless steels makes them prone to solidification allurgical and mechanical aspects. Metallurgical aspects consider
cracking. In aluminum alloys, both thermal contraction and solid- microscopic features and weldability tests consider these effects.
ification shrinkage are high. Therefore, some aluminum alloys Mechanical aspects consider macroscopic features and numerical
especially those with wide BTR are prone to solidification cracking methods such as finite element can be used to estimate them.
Finite element simulation of weld solidification cracking re-
turns to early years of 1990s. Feng made a considerable contribu-
⇑ Corresponding author. Tel.: +98 311 3915235; fax: +98 311 3912628.
tion to the precise modeling of hot cracking [7]. He studied a
E-mail addresses: ar.safari@me.iut.ac.ir (A.R. Safari), forouzan@cc.iut.ac.ir
(M.R. Forouzan), shamanian@cc.iut.ac.ir (M. Shamanian). bead-on-plate weld of an aluminum 2024 plate and evaluated

0927-0256/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.commatsci.2012.06.015
A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190 183

the mechanical strains near the weld pool. Yang et al. simulated a
trailing heat sink as a hot-cracking mitigation technique [8]. They
used finite element method and ductility curve in the BTR for sim-
ulation. Wei et al. studied hot cracking of stainless steel 310s [9].
They used finite element method to model mechanical strain be-
hind the weld pool as driving force and ductility curve as material
resistance. Ductility curve was extracted by method of in situ
observation (MISO). Xu et al. proposed trailing impactive electro-
magnetic force to control hot cracking [10]. Olivier et al. evaluated
the capability of various kinds of hot tearing criteria to predict the
initiation of hot tears in casting [11]. According to their results,
Prokhorov and Won et al. criteria agreed well with the experimen-
tal observations.
Most studies for simulation of hot cracking are based on the
prediction of crack occurrence but not on its propagation. Dike
et al. used finite element method with interface element and dam-
age model of Bammann to study weld solidification cracking of
6061-T6 aluminum [12]. Good agreement was obtained between
simulation and experiment for location of crack initiation and
extent of cracking. Shibahara et al. proposed a temperature-depen-
dent interface element for simulating occurrence and propagation
of hot cracking [13]. In this method, crack propagates when required Fig. 1. Schematic of ductility curve in the BTR.
energy for formation of new surface is provided. Hilbinger et al. used
liquid element for accounting local tensile stress in middle of the
weld seam [14]. These elements had very low yield strengths in Cracking occurs when HCIprokhorov P 0. Prokhorov criterion is purely
the BTR range. They used a criterion based on the maximum defor- mechanical and do not consider metallurgical aspects.
mation that the liquid elements in the solid–liquid region can sus-
tain. This critical deformation was determined experimentally. 2.2. Clyne and davies criterion
In this study, hot cracking of austenitic stainless steel 310s was
studied. The main contribution was to include not only crack initia- Clyne and Davies divided the mushy zone into liquid feeding
tion but also crack propagation. Maximum transverse mechanical zone (0.4 < fs < 0.9) and brittle temperature zone (0.9 < fs < 0.99)
strain was successfully used for the first time as the criterion for where fs is solid fraction [16]. Cracks formed in the liquid feeding
modeling crack propagation. Another important contribution was zone are healed by surrounding melt, whereas cracks formed in
to study the combined effect of annealing at high temperature and the BTR zone cannot be healed, because the dendrite arms are close
melting of fusion zone on the crack propagation. The effect of enough to resist feeding of the surrounding liquid. They introduced
welding parameters, solidification shrinkage and viscosity on hot a criterion equal to tv/tr. tv is the time spent by the mushy zone in
cracking was also studied. Finite element method was used and a the feeding region and tr is the time of solidification in the BTR in
viscoplastic constitutive model with internal state variables and which solid content increases from 0.9 to 0.99. Therefore, the
temperature dependent material properties was developed. Elastic Clyne’s hot cracking index is defined as:
and viscoplastic strains were also treated as state variable (a total
t0:99  t0:9
number of 26 solution dependent state variable). Viscoplastic HCIclyne ¼ P0 ð2Þ
constitutive behavior of material was extracted by static and
t0:9  t 0:4
dynamic hot tensile tests. The material behavior of fusion zone Indeed, this criterion is based on thermal criterion only and no
was also included in the developed constitutive behavior. Numeri- mechanical aspect is considered.
cal result was verified by some experiments accomplished in this
study. 2.3. Yamanaka criterion

2. Hot cracking criteria According to this criterion, Cracking occurs when accumulated
inelastic strain of mushy zone exceeds from a critical value [17].
2.1. Prokhorov criterion By means of tensile hot cracking test and MISO technique, Matsuda
et al. obtained the critical strain for austenitic stainless steel 310s
Prokhorov proposed the first quantitative criterion for hot tear- [18]. The results depended on the strain rate and were in the range
ing [15]. Prokhorov criterion is based on ductility curve as material of 0.01–0.019. Won et al. suggested a simple empirical relation for
strength and deformation of the liquid–solid region as driving the critical strain of low-alloyed carbon steels [19]. It takes into ac-
force. Using a rapid tensile type test, he extracted ductility curves count the strain rate and brittle temperature range.
of aluminum alloys in the BTR. Fig. 1 shows the ductility curve typ-
u
ically. Weldability tests with controlled strain such as Trans-Vare- ecr ¼ _ m n ð3Þ
straint test (TVT) can also be used to extract ductility curve.
e BTR
According to this criterion, cracking occurs when strain rate of u, m and n are material constant and can be found experimen-
material in the BTR exceeds the critical strain rate for temperature tally. According to Eq. (3), with increasing strain rate or increasing
drop (CST). The CST is obtained from ductility curve. the BTR, the critical strain decreases and so the crack susceptibility
Prokhorov hot cracking index (HCI) is defined according to the increases which is in agreement with experimental observations.
following equation: Again vulnerable region (0.9 < fs < 0.99) is considered as the hot
 cracking susceptible region and accumulated inelastic strain is ac-
emin _  counted for this region. Hot cracking index is according to the fol-
HCIprokhorov ¼ e_  jTj ð1Þ
BTR lowing equation:
184 A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190

fs X
¼0:99 in processes such as welding and casting that the behavior of mate-
HCIwon ¼ Dei:e:  ecr P 0 ð4Þ rial at high temperature is intended, a viscoplastic constitutive
fs ¼0:9
model should be used.
Based on the excess stress theory, Perzyna proposed a visco-
2.4. RDG criterion plastic constitutive equation in the form [24]:
1 @F
Rappaz et al. proposed a criterion based on the pressure drop at e_ ijv p ¼ hWðFÞi ð9Þ
l @ rij
the dendrite root [20]. Pressure drop is due to insufficient liquid
feeding for compensating deformation and solidification shrinkage where W(F) is a function of the static yield function F. Varies func-
and results to the formation of a void. The void then develops into a tions for W(F) has been proposed. In a special case, W(F) = F and
crack. According to this criterion hot cracks initiate when the local F ¼ f ðrij ; T; eijv p Þ  RðT; HÞ where R is yield strength and H is harden-
pressure between the dendrite tips and roots drops below a given ing parameter [25]. In this special case, viscoplastic strain rate is
cavitation pressure. It accounts the deformation of the coalesced according to following equation:
solid skeleton (mechanical deformation) and shrinkage perpendic-
rffiffiffi * 12 +
ular to dendritic arms. The maximum pressure drop can be ob- 31 3 Sij  aij
_ vp
eij ¼ ðSij  aij ÞðSij  aij Þ  RðT; HÞ ð10Þ
tained by the following equation: 2l 2 kSij  aij k
Dpmax ¼ Dpme þ Dpsh
Z Z
aij are back stress tensor components. Wang and Inoue showed that
180 ð1 þ bÞl TL
EðTÞfS ðTÞ2 180 v T bl TL
in this situation, constitutive relationship of Perzyna could describe
¼ 2 dT þ
k G TS ð1  fS ðTÞÞ3 k2 G TS the materials behavior over a wide range from inelastic solid to vis-
cos fluid [25]. More general representation of Perzyna viscoplastic
fS ðTÞ2 model is as follow:
 dT ð5Þ
ð1  fS ðTÞÞ2
R @F
where EðTÞ ¼ G1 fS ðTÞe_ P ðTÞdT, b is shrinkage factor, l is liquid vis- e_ ijv p ¼ KhWðFÞim ð11Þ
cosity, k is secondary dendrite arm spacing, G is thermal gradient, @ rij
vT is isotherms velocity, fS(T) is solid fraction at any temperature For fluids, the exponent m tends to unity and K = (3l)1.
T, TL is liquidus temperature and TS is coherency temperature. Hot Other constitutive equations have also been proposed for steels
cracking susceptibility is defined as the inverse of the maximum at high temperature. Based on curve fitting of experimental results,
strain rate sustainable by the mushy zone. The RDG hot cracking in- Kozlowski et al. proposed four one-dimensional constitutive equa-
dex is defined according to the following equation: tions to model mechanical behavior of plain carbon steel in the
HCIRDG ¼ ðDpmax  Dpcr Þ P 0 ð6Þ austenite phase [26]. According to their results, model III had the
best compatibility with experiment. Eq. (12) presents three-
dimensional extension of model III.
2.5. Ploshikhin criterion rffiffiffi
3 S
e_ ijv p ¼   aan im ij
Kh r ð12Þ
Ploshikhin et al. presented an integrated mechanical–metallur- 2 kSij k
gical approach to model solidification cracking in welds [21]. It
where
undertakes the effects of strain accumulation at the final stage of
solidification, microstructure of the mushy zone and thermo-

Q
dynamical properties of the welded material. According to KðTÞ ¼ CðcÞ exp 
T
Ploshikhin criterion, hot tearing takes place when deformation of
the interdendritic liquid film exceeds from some critical value. Max-
imum deformation is a function of welded joint geometry, welding a_ ¼ e_ v p
parameters, thermo physical properties of base and parent metal
and parameters of the microstructure. Ploshikhin hot cracking index CðcÞ ¼ 46550 þ 71400c þ 12000c2
is defined according to the following equation:
  Q ¼ 44650
HCIPLOSHIKHIN ¼ dmax cr
acc  dacc P 0 ð7Þ
Other hot cracking criteria can be found in [4,6,22,23]. a ¼ 130:5  5:128  103 T

3. Constitutive behavior
n ¼ 0:6289 þ 1:114  103 T

Considering additive decomposition of strains, total strain rate m ¼ 8:132  1:54  103 T
could be decomposed into its components. Q is activation energy constant and c is carbon content.
e_ Tot _ e _ v p _ th _ Trp _ Trv Comparison of varies constitutive laws showed that most of
ij ¼ eij þ eij þ eij þ eij þ eij ð8Þ
them have forms identical to the following equation:
The terms on the right hand side of the above equation are elas- 1
tic, viscoplastic, thermal, transformation plasticity and volumetric r ¼ RðT; ev p Þ þ LðTÞe_ v p =mðTÞ ð13Þ
strain rate due to phase transformations respectively. n
where R ¼ ry ðTÞ þ KðTÞev p ðTÞ is static yield surface. For example,
At relatively low temperatures, an elastic–plastic constitutive
compare to Kozlowski model III:
relationship can describe the material behavior well. Power rela-
tionships are usually used for this purpose. However, with increas-
1=mðTÞ
nðTÞ 1 Q
ing the temperature, viscosity effect increases so that beyond the RðT; ev p Þ ¼ aðTÞev p and LðTÞ ¼ expð Þ
CðcÞ T
melting point, the material behaves like a viscous liquid. Therefore,
A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190 185

At low temperature or low strain rate, the second term of Eq. 5. Annealing and melting effects of the heat source
(13) vanishes and the model reduces to a simple power law.
For treatment of mushy zone, two rheological behaviors was In thermo mechanical processes, when the material tempera-
proposed by Koric and Thomas [27]. In the first approach, an elas- ture exceeds from 0.6 to 0.7 of the melting temperature (Tm),
tic-perfectly plastic rate-independent constitutive equation with annealing starts [28]. In annealing, previous history of work hard-
small yield strength was assigned to the mushy zone. The assigned ening clears. This means that annealing resets the equivalent plas-
yield strength was in the order of 0.01 Mpa to enforce negligible tic strain and any hardening [29]. On the other hand, during the
stresses in the mushy zone. In the second approach, a viscoplastic movement of the welding arc, boundary of the fusion zone contin-
constitutive model according to Eq. (14) was proposed. uously changes and so the elements along the fusion line enter and
c exit from the weld pool in a dynamic nature. When elements enter
e_ v p ¼ s hjcs r
 j  ry i ð14Þ the weld pool, their previous strain and stress is relaxed. Behind
lv
the fusion zone, material starts to solidify and stresses and strains
Where accumulate from a state of zero. In this study, for assessing fusion
( emax zone, strain relaxation method is implemented. In this method, be-
j
emin emax P emin
cs ¼ : fore mechanical analysis, temperature history of all elements from
emin
jemax
emax < emin thermal analysis is modified and temperatures above Tm are trun-
8 cated to Tm. Melting temperature is also defined as the thermal
l1
v is a large enough coefficient in the order of 10 . In this study, expansion reference temperature. Therefore, elements will be
the first approach was adopted for mushy zone.
expansion-free until the temperature of nodes is cooled down be-
low Tm. This method has the same effect of element birth and death
4. Integration algorithm method for modeling addition of weld metal. Modification of tem-
perature history was accomplished by a specially developed sub-
In this study, a temperature and strain rate dependent kine- routine. For structural analysis, in the developed constitutive
matic hardening model incorporating viscosity effects was devel- subroutine, annealing was treated as a linear function of tempera-
oped and incorporated as the constitutive behavior of material. ture, starts from 0.6 Tm and completes at Tm. For incorporating the
Stress–strain relaxation of fusion zone was also included. To carry melting, when the temperature reaches to Tm, elastic–plastic
out relaxation, elastic and plastic strains were set as state variable. strains and stresses are set to zero.
A total number of 26 solution dependent state variables were used.
The radial return method was employed to calculate the stress
6. Model geometry and material properties
components. According to this method, for each increment n + 1:
nþ1
sij ¼ nþ1 strij  3nþ1 Gn gij Dep ð15Þ Many weldability tests have been developed to quantify hot
cracking susceptibility. Some tests such as varestraint use an exter-
nally applied load (predefined stress or strain) as the driving force
where and some others such as houldcroft test are self-restraint. Maxi-
nþ1 tr mum crack length or summation of all crack lengths is commonly
sij  nþ1 bn aij
n
gij ¼ used as a measure of crack susceptibility.
r tr In this study, three nominally identical specimens were manu-
nþ1
factured from a 3 mm thick plate of AISI Type 310s steel. Test spec-
b ¼ nþ1 H=n H imens were cut similar to the houldcroft test model. Dimensions of
G is the shear modulus. For bilinear kinematic hardening model, the the test specimens are illustrated in Fig. 2. To eliminate fabrication
evolution of back stress is according to Eq. (16) from which back residual stresses, specimens were solution heat treated at 1050 °C
stress at the end of each increment can be find. for 45 min and then cooled in water. For each specimen a single
autogenously fusion line was deposited along the centerline. The
d aij  2 p arc started from tab plate. When fusion reaches to the bridge, tab
¼ e_ ij ð16Þ
dt H 3 plate falls and a steady fusion continues on the test specimen.
For each specimen two K-type thermocouples were used to re-
nþ1
aij ¼ nþ1 bn aij þ nþ1 HDepn gij ð17Þ cord the temperature. To minimize the effect of shielding gas on
the records, thermocouples were installed on the back of plate
From Eqs. (15) and (17):
with 10 mm distance from weld centerline. Thermocouples were
nþ1  tr
r  nþ1 r  ð3nþ1 G þ nþ1 HÞDep ¼ 0 ð18Þ inserted in holes with 1 mm diameter and 1 mm depth. Pure nitro-
gen gas was used as the shielding gas. For protecting the fusion
In addition, from Eq. (13) for kinematic hardening:
nþ1
mðnþ1 TÞ
r  ry ðnþ1 TÞ
Dev p ¼ Dt ð19Þ
Lðnþ1 TÞ
Combination of Eqs. (18) and (19) leads to the following equa-
tion that should be solved to find nþ1 r
.
nþ1
mðnþ1 TÞ
nþ1  r  ry ðnþ1 TÞ
r  nþ1 r  ð3nþ1 G þ nþ1 HÞ Dt ¼ 0 ð20Þ
Lðnþ1 TÞ
 ; Dev p can be find from Eq. (19).
Obtaining nþ1 r
In this study, the Von Misses yield surface and its associated
flow rule with bilinear kinematic hardening model was used. Large
deformation was accounted in all analyses. Various consistent tan-
gent modules were examined to increase the rate of convergence. Fig. 2. Dimensions of the test specimen (mm).
186 A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190

temperature history of nodes. Following the thermal analysis, a


sequentially coupled thermal-stress analysis was performed

8. Thermal analysis

The Goldak double ellipsoid volumetric heat flux and Gaussian


surface heat flux was examined. Both models had approximately
identical results but surface heat flux was more compatible with
the experiments. It seems to be due to omission of consumable
electrode and small thickness of specimens. The model uses fol-
lowing equation to describe the heat source:
( )
3Q 3½ðx  ðx0 þ v ðt  t 0 ÞÞÞ2 þ y2 
qðx; y; tÞ ¼ exp ð21Þ
pr20 r20

where Q = gVI is the power input. g is the arc efficiency, V is the


voltage and I is the current. r0 is a characteristic length and is such
Fig. 3. Test setup. that the heat flux at the boundaries falls to 5% of its maximum va-
lue. v is the arc speed, t is the time and t0 is a lag factor denoting the
dwell of torch at the start position. The user subroutine DFLUX in
zone from oxidation, a box were constructed and installed on the ABAQUS was used to introduce the heat flux. g and r0 were selected
back of the plate to blow nitrogen. Welding was performed at such that temperature prediction at the thermocouples position and
the ambient temperature of 10 °C. Type of welding was automatic predicted fusion dimension comply with experimental results.
tungsten inert gas welding (TIG). Heat input was 585 J/mm, weld- Dwell time was selected such that a stable fusion width was
ing speed was 1 mm/s and arc time was 80 s. Total weld length was obtained.
80 mm. Fig. 3 shows the test setup. Thermal conductivity was increased by a factor of two in liqui-
Parameters of Eq. (1) were obtained by static and dynamic ten- dus–solidus temp range to account stirring effect and fluid flow in
sile tests at low and high temperatures. A serious of tension tests the weld pool. Thermal boundary conditions include both convec-
were accomplished at a velocity of 1.5 mm/min (strain rate tion and radiation heat lost through all sides of the plate exclude
0.001 s1) and temperatures of 22 °C, 250 °C, 500 °C, 700 °C, symmetry plane. Near the fusion zone, the effect of shielding gas
900 °C and 1000 °C. Other tension tests were accomplished at a increases the convection but dominant effect of emissivity com-
velocity of 225 mm/min (strain rate 0.15 s1) and temperatures pensates this. Convection coefficient and emissivity coefficient
of 700 °C, 900 °C and 1000 °C. are assumed 15 (W/m2/°C) and 0.9 respectively. Initial tempera-
ture of specimen and ambient temperature were assumed 10 °C.
7. Finite element analysis The effect of phase change in the thermal analysis was taken
into account by means of the latent heat of fusion. Release rate
In this study, a 3D finite element model was used. Because of of the latent heat was assumed to be in direct proportion to the so-
the symmetry about the weld centerline, only half of the specimen lid fraction. Fig. 5 shows the solidification fraction of sus310 as a
was modeled. Fixed boundary condition was used for the grip area. function of temperature.
The specimen was meshed so that at the position of each thermo- According to Eq. (22), Latent heat can be incorporated in the
couple one node was located. Fig. 4 shows the finite element mesh. specific heat as an equivalent specific heat.
Finer mesh was used near the weld centerline (WCL). Elements
@T
away from the WCL had larger size to decrease solution time. Q ¼ qC p;eq ð22Þ
@t
Thermal stress analysis was performed in two distinct steps
using the commercial software ABAQUS [29]. Since the tempera- where
ture field is independent of stress or displacement solution,
transient heat transfer was performed first to determine the

Fig. 4. Mesh density for thermal and stress analysis. Fig. 5. Variation of solid fraction with respect to temperature [18].
A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190 187

as an additional thermal expansion and was modified in expansion


coefficient. The solidification shrinkage was neglected above the
coherent temperature where the material is treated as a liquid
with relaxed stress/strain. Fig. 7 shows modified thermal expan-
sion coefficient.
For elements that enter the fusion zone, the yield strength was
assumed 0.01 Mpa and elastic/plastic strains were set to zero. In
this study for elements that satisfy hot cracking index, yield
strength of 0.01 Mpa was assigned. This causes that this elements
treat as a very soft material after cracking. Won et al. criterion was
used for prediction of cracking [19]. For initiation and propagation
of hot cracking, critical strain was assumed 0.01. This is in the
range that was obtained by Matsuda et al. by tensile hot cracking
test [18].
For mechanical analysis, continuum three dimensional 8-node
brick elements (C3D8) were used.

Fig. 6. Equivalent specific heat.


10. Results and discussion

L @fs Fig. 8 shows test specimen 1 and its fusion shape. Experimental
C p;eq ¼ C p 
q @T fusion width was about 6.05 mm on the top of the specimen and
5.24 mm on the back of that (after subtracting crack width).
L is latent heat per unit volume. This leads to very nonlinear specific
According to Fig. 9, Numerical FZ dimensions were 6 mm on the
heat increasing solution time and convergence problems. Fig. 6
top of specimen and about 5 mm on the back of that. The results
shows the equivalent specific heat. As a more convenient way, la-
are in close agreement with experimental results.
tent heat can be discretized in temperature steps proportional to
Fig. 10 shows measured temperature at thermocouple positions
the content of the liquid fraction.
and simulated ones. It can be observed that the temperature values
8-node linear heat transfer brick elements (DC3D8) were used
obtained with the numerical model coincide well with those mea-
in thermal analysis. Fixed time stepping was used for the solution.
sured experimentally. It seems that more accurate modeling of

9. Mechanical analysis

The FE mesh used in the mechanical analysis was the same as


that used in the thermal analysis. Mesh compatibility was ac-
counted to import nodal temperatures from thermal result file.
The temperature history from thermal analysis was corrected by
an especially developed subroutine to be used by successive struc-
tural analysis. In the structural analysis, the developed constitutive
subroutine was used to represent material constitutive behavior.
During solidification, due to liquid–solid phase transformation
and density difference between solid and liquid phase, the material
contracts in excess of thermal contraction which is called solidifi-
cation shrinkage. For austenitic stainless steel with FCC structure,
the volumetric solidification shrinkage is about 4%, which is equiv-
alent to a linear contraction of 1.3% [30]. In this study, solidification
shrinkage was defined as a function of temperature in solidification
temperature range and proportional to solid fraction. It was treated

Fig. 7. Equivalent thermal expansion coefficient. Fig. 8. Fusion zone of test specimen (dimensions in mm).
188 A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190

Fig. 9. Dimensions of fusion zone from numerical result (dimensions in mm). Fig. 11. Experimental crack length of test specimen 1.

Table 1
Comparison of experimental and numerical crack length.

Specimen Crack length (mm)


Experiment 1 33
2 35
3 32
Simulation – 32

Fig. 12. Simulated crack area.

distance equal to 32 mm, TEmax is very low therefore, this curve


can be used as an index to distinguish crack length. Large strain
at the beginning of the diagram is due to large deformation of
Fig. 10. Comparison of simulated temperature with experimental results.
the cracked elements.
Elimination of latent heat effect leads to a crack length of
heat loss from surfaces especially convection effect of torch shield- 31 mm (only 3% error) but welding parameters that influence on
ing gas and back plate shielding gas could improve the results. In the weld pool geometry has significant effect on the crack length.
simulation of the welding, fusion width was more affected by g An increase of 30% in heat input leads to a crack length of
and r0 while temperature at thermocouples was affected by g, r0 52 mm (over 60% difference) while doubling the welding velocity
and heat loss from free surfaces. from 1 mm/s to 2 mm/s leads to a crack length of 12.6 mm. These
Fig. 11 shows crack length of test specimen 1. To insure repeat- results show that hot cracking is very sensitive to heat input. The
ability of results, three specimens were constructed and tested and Prokhorov criterion can describe this behavior. With increasing
Table 1 shows the crack length of them. Good agreement of results the heat input, temperature gradient behind the weld pool de-
denotes that tests have been done under similar and controlled creases and lower strain rate is required to satisfy this criterion
condition. while increasing weld velocity has reverse effect.
Fig. 12 shows the specimen after cracking. Dark area shows ele- Fig. 14 shows the effect of various modeling parameters on
ments that their center satisfies hot cracking index. The predicted predicted crack length. According to this figure, elimination of
crack length was about 32 mm that had good compatibility with annealing, melting and solidification shrinkage, leads to lower
experimental results. Fig. 13 shows maximum transverse mechan- crack length prediction. Annealing and melting has similar effect
ical strain (TEmax) along the WCL. According to this figure, at a and elimination of them leads to the most underestimating crack
A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190 189

description of hot cracking should take into account both metallur-


gical and mechanical aspects. For indexing hot cracking various
theories have been proposed most of them predict the occurrence
of hot cracking but cannot predict crack propagation. In this study,
theory of a maximum transverse mechanical strain was examined
as a successful criterion for predicting occurrence and propagation
of hot cracking of austenitic stainless steel 310s. Finite element
method was used for analysis and extracting driving force. A ther-
mo viscoplastic constitutive model with 26 solution dependent
state variables and temperature dependent material properties
was developed. Parameters of constitutive model were extracted
by static and dynamic hot tensile tests. For thermal analysis, latent
heat of fusion was introduced as a function of solid fraction that
again reveals metallurgical aspects. In structural analysis, thermal
expansion coefficient was modified taking into account liquid–so-
Fig. 13. Maximum transverse mechanical strain along the WCL.
lid transformation. Simulation results were verified by experimen-
tal tests. Three specimens was constructed and the method of
welding was autogenous TIG welding, a method which is common
in hot cracking susceptibility tests. According to this study solidifi-
cation shrinkage, melting effect of fusion zone, annealing at high
temperature and viscosity had significant effect for prediction of
hot cracking. Omission of the first three parameters leaded to
underestimating crack length while omission of the viscosity
leaded to overestimated results.

References

[1] S. Kou, Welding Metallurgy, second ed., John Wiley & Sons, Hoboken, New
Jersey, 2003.
[2] H. Fredriksson, M. Haddad-Sabzevar, K. Hansson, J. Kron, Materials Science and
Technology 21 (5) (2005) 521–529.
[3] S. Kou, Journal of Minerals, Metals and Materials 55 (6) (2003) 37–42.
[4] D.G. Eskin, Physical Metallurgy of Direct Chill Casting of Aluminum Alloys
(Advances in Metallic Alloys), first ed., CRC Press, Taylor & Francis Group, 2008.
[5] C.E. Cross, On the origin of weld solidification cracking, Hot Cracking
Fig. 14. Effect of various modeling parameters on predicted crack length. Phenomena in Welds I, Springer-Verlag, Berlin Heidelberg, 2005.
[6] Y.M. Wong, H.N. Han, T. Yeo, ISIJ International 40 (2) (2000) 129–136.
[7] Z. Feng, A Methodology for Quantifying the Thermal and Mechanical
Conditions for Weld Solidification Cracking, Ph.D. Dissertation, Ohio State
length. Elimination of viscosity overestimates crack length. The University, 1993.
rheological behavior of classical viscoplastic solids can explain [8] Y.P. Yang, P. Dong, J. Zhang, X. Tian, Welding Journal 79 (1) (2000) 9–17.
the above behaviors. Consideration of solidification shrinkage leads [9] Y.H. Wei, Z.B. Dong, R.P. Liu, Z.J. Dong, Modeling and Simulation in Materials
Science and Engineering. 13 (2005) 437–454.
to more thermal strain behind the weld pool. Because of external
[10] W. Xu, H.Y. Fang, J.G. Yang, X.S. Liu, W.L. Xu, Materials Science and Engineering
restrain, this leads to more driving force and larger mechanical A 488 (2008) 39–44.
strain will develop that leads to longer crack length. At constant [11] C. Olivier, C. Yvan, B. Michel, Journal of Engineering Materials and Technology
130 (2008) 021018-1.
thermal strain and thus driving force, consideration of viscosity
[12] J.J. Dike, J.A. Brooks, D.J. Bammann, M. Li, Thermal–mechanical modeling and
leads to less mechanical strain because the rate dependent part experimental validation of weld solidification cracking in 6061-T6 aluminum,
of viscose behavior specify some of driving force to itself. With in: ASM International European Conference on Welding and Joining Science
consideration of annealing and melting, the material behind the and Technology, Madrid, Spain, March 10–12, 1997.
[13] M. Shibahara, H. Serizawa, H. Murakawa, Transactions of JWRI 28 (1) (1999)
weld pool contract from a state of zero strain and tensile deforma- 47–53.
tion will develop in material because While elimination of these [14] R.M. Hilbinger, H.W. Bergmann, W.Köhler, F. Palm, Consideration of Dynamic
parameters lead to compressive strains behind the weld pool. Mechanical Boundary Conditions in the Characterization of a Hot Cracking Test
by Means of Numerical Simulation. Mathematical Modeling of Weld
In the next step, the effect of critical strain on crack length was Phenomena 5, IOM Communications Ltd., London, pp. 847–862.
studied. According to the results, the more the critical strain, the [15] N.N. Prokhorov, Welding Production 4 (1962) 1–8.
smaller crack length will develop. Specification of a fictitious large [16] T.W. Clyne, G.J. Davies, British Foundation 74 (1981) 65. See also: Brit. Found.
68 (1975) 238.
critical strain, demands from crack initiation and propagation. [17] A. Yamanaka, K. Nakajima, K. Yasumoto, H. Kawashima, K. Nakai,
According to Sections 2.1 and 2.2, the susceptible solid fraction Measurement of critical strain for solidification cracking, in: M. Rappaz, M.R.
range for hot cracking is from fS = 0.9 to fS = 0.99. To account this Ozgu, K.W. Mahin (Eds.), Modelling of Casting, Welding and Advanced
Solidification Processes, vol. V, Davos, Switzerland, TMS, Warrendale, PA,
region in the analysis, the equivalent temperature range was ex-
1990, pp. 279–284.
tracted from the solid fraction curve (Fig. 5) and was implemented [18] F. Matsuda, H. Nakagawa, S. Tomita, Transactions of JWRI 15 (2) (1986) 125–
in the constitutive behavior. Elimination of this behavior and con- 133.
[19] Y.M. Won, T.J. Yeo, D.J. Seol, K.H. Oh, Metallurgical and Materials Transactions
sideration of the full solid–liquid temperature range as the hot
B 31B (2000) 779–794.
cracking susceptible temperature range (fS = 0 to fS = 1) resulted [20] M. Rappaz, J.M. Drezet, M. Gremaud, Metallurgical and Materials Transactions
to a crack length of 34 mm. A 30A (1999) 449–455.
[21] V. Ploshikhin, A. Prikhodovsky, M. Makhutin, A. Ilin, H.-W. Zoch, Integrated
Mechanical–Metallurgical Approach to Modelling of Solidification Cracking in
11. Conclusion Welds, Hot Cracking Phenomena in Welds I, Springer-Verlag, Berlin
Heidelberg, 2005.
[22] D.G. Eskin, Suyitno, L. Katgerman, Mechanical properties in the semi-solid
The metallurgical aspects of hot cracking have been studied in state and hot tearing of aluminium alloys, Progress in Materials Science 49
much more details than thermo mechanical aspects. A successful (2004) 629–711.
190 A.R. Safari et al. / Computational Materials Science 63 (2012) 182–190

[23] L. Katgerman, D.G. Eskin, In Search of the Prediction of Hot Cracking in [27] S. Koric, B.G. Thomas, International Journal for Numerical Methods in
Aluminium Alloys, Hot Cracking Phenomena in Welds II, Springer-Verlag, Engineering 66 (12) (2006) 1955–1989.
Berlin Heidelberg, 2008. [28] M.C. Smith, A.C. Smith, International Journal of Pressure Vessels and Piping 86
[24] P. Perzyna, Thermodynamic Theory of Viscoplasticity, Advance of Applied (1) (2009) 79–95.
Mechanics, vol. 11, Academic Press, New York, 1971. [29] ABAQUS/standard user’s manual. Version 6.9.1. Providence. RI: ABAQUS Inc.,
[25] Z.G. Wang, T. Inoue, Materials Science and Technology 1 (1985) 899–903. 2009.
[26] P.F. Kozlowski, B.G. Thomas, J.A. Azzi, H. Wang, Metallurgical and Materials [30] Doru Michael Stefanescu, Science and Engineering of Casting Solidification,
Transactions A 23 (1992) 903–918. second ed., Springer, New York, 2009.

You might also like