Sie sind auf Seite 1von 13

Effect of Surface Tension Anisotropy and Welding

Parameters on Initial Instability Dynamics During


Solidification: A Phase-Field Study
FENGYI YU and YANHONG WEI

The effects of surface tension anisotropy and welding parameters on initial instability dynamics
during gas tungsten arc welding of an Al-alloy are investigated by a quantitative phase-field
model. The results show that the surface tension anisotropy and welding parameters affect the
initial instability dynamics in different ways during welding. The surface tension anisotropy does
not influence the solute diffusion process but does affect the stability of the solid/liquid interface
during solidification. The welding parameters affect the initial instability dynamics by varying
the growth rate and thermal gradient. The incubation time decreases, and the initial wavelength
remains stable as the welding speed increases. When welding power increases, the incubation
time increases and the initial wavelength slightly increases. Experiments were performed for the
same set of welding parameters used in modeling, and the results of the experiments and
simulations were in good agreement.

https://doi.org/10.1007/s11661-018-4663-7
 The Minerals, Metals & Materials Society and ASM International 2018

I. INTRODUCTION solidification structures in the late stages.[7] Therefore, it


is important to investigate the initial instability dynam-
WELDING is an important joining method that is ics of the solidification process in a molten pool.
widely used in many fields of the manufacturing Theoretical modeling and numerical studies have been
industry. The mechanical properties of weld joints carried out to predict the initial instability in the past
depend on the microstructures which result from the few decades. The theoretical analysis was developed
weld solidification, which is a history-sensitive process.[1] from the constitutional undercooling[8] theory, the
Investigating the solidification dynamics can provide a Mullins and Sekerka (MS)[9] theory, and the Warren
theoretical basis for optimizing the welding process. In and Langer (WL)[10] theory. The WL theory provides a
the past few decades, the development of computational method to predict the incubation time and the pertur-
materials science has advanced the understanding of bation frequency of the initial instability. The simulation
solidification behavior during welding.[2,3] However, it is results from the WL model agreed well with the
fundamentally challenging to integrate the whole weld- experimental observations of real-time synchrotron
ing process into one computational model in part due to X-ray radiography,[11] but the detailed solid/liquid
the complex transient conditions in a molten pool.[4] As (S/L) interface morphologies could not be directly
a compromise, most researchers begin microstructural obtained. Wang et al.[12] developed an analytic model,
simulations from the beginning of cellular crystal based on the WL theory, which combines the Fourier
growth, regarding the simulated domain boundary as synthesis method and the time-dependent linear stability
one simple solid layer and artificially seeding the analysis to predict the interface structures. The model
nucleation.[5] These studies have typically ignored the was verified in the steady-state condition of directional
establishment of the solute boundary layer during the solidification.[13] The steady-state model was extended to
initial stage of nucleation and growth, which affects the nonsteady-state by Dong el al.[14] The new model has
final microstructures significantly.[6] Moreover, the pla- been used in the transient conditions found in a gas
nar instability in the early stage could determine the tungsten arc welding (GTAW) process.[15] While there
has been significant research into making the initial
instability predictions, the effect of surface tension
anisotropy and welding parameters on the initial insta-
FENGYI YU and YANHONG WEI are with the College of bility dynamics have not been considered. The surface
Material Science and Technology, Nanjing University of Aeronautics
and Astronautics, Nanjing 210016, China. Contact e-mial:
tension anisotropy and welding parameters have enor-
nuaadw@126.com mous impacts on the S/L interface morphology evolu-
Manuscript submitted November 27, 2017. tion,[16] which could determine the final solidification
structures.[17] Herein, the surface tension anisotropy is

METALLURGICAL AND MATERIALS TRANSACTIONS A


defined by the anisotropy strength and misalignment same set of welding parameters to verify the accuracy of
angle. The misalignment angle is the angle between the the simulation results.
preferred crystalline orientation of the base metal and
the thermal gradient direction at the front of the
interface. The thermal gradient direction changes with II. MODELING AND EXPERIMENT
time during welding because of the heat source move-
ment. Moreover, the growth rate R and thermal A. Phase-Field Model
gradient G are determined by the actual welding The QPF model used in this paper was developed by
process, which both can change according to the Echebarria et al.[27] for the directional solidification of a
welding parameters used.[18] Therefore, it is essential to binary dilute alloy and was modified by Zheng et al.[28]
investigate the effect mechanism of surface tension to be used for the transient conditions in a welding
anisotropy and welding parameters on the initial insta- process. By neglecting the influence of latent heat and
bility dynamics during welding. convection, the so-called ‘‘frozen temperature approxi-
Due to the mesoscale nature of solidification struc- mation’’ is adopted:
tures, the phase-field method is a suitable choice to  Z 
investigate the solidification dynamics.[19–21] Using a
thin-interface asymptotic analysis and the introduction Tðz; tÞ ¼ T0 þ GðtÞ z  z0  Rðt0 Þdt0 ; ½1
of an ‘‘anti-trapping current’’ term,[22] the phase-field
method predicted the dendritic arm spacing,[23] crystal where T0 = T(z0, t) is a reference temperature, and
morphology evolution,[24,25] and equiaxial growth[26] G(t) is along the z-direction. The detailed derivations
quantitatively. The simulation results are consistent with and validations of the QPF model can be found in the
the experimental observations of directional solidifica- study by Echebarria et al.[27] In this study, we present
tion and welding processes. Thus, the quantitative only the equations describing the evolutions of the
phase-field (QPF) model provides a superb tool to phase field and solute field. In the formulation, a sca-
investigate the effect of surface tension anisotropy and lar variable /(r, t) is introduced to determine the phase
welding parameters on the initial instability dynamics. at a fixed point and time. The field / takes on the val-
In this study, we used the QPF model to reproduce ues of / =  1 in the liquid phase and / = 1 in the
the dynamic evolution process of initial instability in a solid phase, and / varies smoothly at the interface.
GTAW molten pool of Al-4 wt pct Cu alloy. Different The solute concentration c(r, t) is characterized by a
misalignment angles are used to reflect the different generalized supersaturation field U(r, t), which is given
surface tension anisotropies as the anisotropy strength is by Eq. [2]:
constant for a specific material. The welding speed is  
investigated by varying the speed while keeping the 1 2kc=c1
U¼ 1 ; ½2
welding power constant. Based on the simulation 1  k 1  / þ kð1 þ /Þ
results, the incubation time and average wavelength of
where k is the solute partition coefficient, and c¥ is the
the initial instability are discussed in detail. The influ-
solute concentration in the far-away field. For cubic
ences of the term G/R on the solidification mode and the
crystals such as in an Al-Cu alloy, the fourfold surface
term G 9 R on the solidification structure size are
energy anisotropy function in a 2D system is approxi-
investigated. Phase-field simulations are performed with
mated by Eq. [3]:
different welding powers at the same welding speed to
 
confirm the effect mechanism of G/R and G 9 R. _
a n  aðh þ h0 Þ ¼ 1 þ c4 cos 4ðh þ h0 Þ; ½3
Finally, experimental observations are made for the

Fig. 1—Diagram of a GTAW molten pool.

METALLURGICAL AND MATERIALS TRANSACTIONS A


where c B. Computational Procedure
4 is the anisotropy strength, h is the angle
between the interface normal and the axis z, and h0 is The growth rate and thermal gradient vary consider-
the misalignment angle. The governing equations of ably across the molten pool due to the movement of the
the phase field and the supersaturation field can be heat source during welding. To investigate the dynamic
expressed by Eqs. [4] and [5], respectively: evolution of the planar interface instability at the initial
" Rt # transient stage, the time-dependent R and G must be
 2 z  0 Rðt0 Þdt0 @/
_
s0 a n 1  ð1  kÞ obtained. The shape of a GTAW molten pool can be
lT @t regarded as the combination of two half ellipsoids. The
    crystals growth begins in the rear ellipsoid, as shown in
! 2!
¼ W2 r a n r / þ /  /3
_
½4 Figure 1.
" Rt # The blue region is the domain used for numerical
z  0 Rðt0 Þdt0 analysis, where al and bl are the depth and rear length of
 kgð/Þ U þ the molten pool, respectively. The value V represents the
lT
welding speed, and a is the angle between the directions
of V and G. The orientation of crystal growth is assumed
 
1þk 1k @U to be always perpendicular to the fusion line in the
 / initial stage. Thus, the relationship between R and V is
2 2 @t
0 1 given by R = VÆcosa. Using the elliptic equation and a
  ! point’s coordinate, the evolutions of R and G are
!B ! _ @/ r / C
¼ r @Dqð/Þr U þ a n W½1 þ ð1  kÞU  A ; expressed in terms of the heat input and welding
@t ! 
r / speed[28]:
1 @/
þ ½1 þ ð1  kÞU al V2 t
2 @t RðtÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2 ½6
½5 V2 t2 al  b2l þ b4l

where the function g(/) = (1  /2)2 is the double-well


function which ensures g( 1) = g(1) = 0, while q(/) = TP  TL
GðtÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ½7
(1  /)/2 dictates that solute diffusivity is 0 in the solid, V t2 þ a2l 1V
2 2 t2
b2
D in the liquid, and varies across the interface. l

lT(t) = |m|c¥ (1  k) k1 G(t)1 is the thermal length,


and m is the liquidus slope. W and s0 are the interface where TP and TL are the central temperatures of molten
width and relaxation time, which are the length scale pool and the liquidus temperature, respectively. The
and time scale, respectively. By neglecting the kinetic values of al, bl, and TP are obtained by the finite element
effect, the phase-field parameters can be linked to the simulation, then the evolutions of R and G with time are
physical quantities according to W = d0 k/a1 and obtained as input for the phase-field simulation. The
s0 = a2 k W2/D, respectively, where a1 = 0.8839 and parameters for simulation under welding conditions by
a2 = 0.6267 are the numerical constants, k is the varying speed and varying power are shown in Tables I
coupling constant, and d0 = C/|m|(1  k) c0l is the and II, respectively.
capillary length. C = c0 TM/L is the Gibbs–Thomson The welded material used is Al-Cu alloy 2A14 and its
coefficient, where c0 is the isotropic part of the surface chemical composition is shown in Table III. This
tension, TM is the melting temperature of the pure material can be considered to be a dilute Al-4.0 wt pct
solvent, and L is the latent heat. Cu binary alloy. The physical properties of the material
and related parameters are summarized in Table IV.[29]

Table I. Parameters for Simulation Under Different Welding Speeds with Welding Power 3600 W

Symbol (Unit) V = 2.0 mm/s V = 2.5 mm/s V = 3.0 mm/s V = 3.5 mm/s
al (mm) 2.95 2.62 2.26 2.03
bl (mm) 4.52 4.37 4.28 4.12
TP  TL (K) 830 680 550 450

Table II. Parameters for Simulation Under Different Welding Powers with Welding Speed 2.5 mm/s

Symbol (Unit) P = 2000 W P = 3000 W P = 4000 W P = 5000 W


al (mm) 1.72 2.45 3.19 3.85
bl (mm) 2.92 4.08 5.67 6.36
TP  TL (K) 380 620 850 1050

METALLURGICAL AND MATERIALS TRANSACTIONS A


The misalignment angles h0 were chosen to be 0, 15, 30 field and the solute field. A time step size is chosen below
and 45 deg due to the four-fold symmetry of the the threshold of numerical instability for the diffusion
system’s cubic crystal structure. equation in 2D system, i.e., Dt < (Dx)2/(4D). The initial
The program coding was written in FORTRAN and phase-field variable is set as /(z,0) = tanh(z/2) along
executed by the Institute for Cyber Science Advanced the normal to the planar interface, and the initial
Cyber Infrastructure (ICS-ACI) at the Pennsylvania supersaturation field U is set to a steady-state diffusion
State University for this study. The most important profile along the normal to the interface.
simulation parameter is the interface width W. With the To consider the infinitesimal fluctuations at the planar
decrease of W, the calculation precision increases but interface in phase-field simulation, a fluctuating current
the computational cost increases dramatically.[30] JU to induce concentration fluctuations is included in
According to Reference 31 the W only needs to be one the governing equation. Its components are random
order of magnitude smaller than the characteristic variables obeying a Gaussian distribution with
length scale of the microstructures. The characteristic variance[32]:
length is about 5 to 10 lm in the initial stage, so W was m
chosen to be 0.11 lm in the simulation. The Eqs. [4] to JU ð~ r0 ; t0 Þ ¼ 2Dqð/ÞFU dmn dð~
r; tÞJnU ð~ r0 Þdðt  t0 Þ ½8
r ~
[5] are solved by an explicit finite difference method with
a fixed grid size of dx = 0.8 W. The Neumann bound- The fluctuating current depends explicitly on the
ary conditions with zero-flux are used for both the phase phase field via the solute diffusivity Dq(/) that is equal
to 0 in the solid and D in the liquid. The magnitude of
FU = F0U[1 + (1  k)U] can be defined by the relation
Table III. Chemical Composition of Al-Cu Alloy 2A14 of Reference 33:
(Weight Percent)
D E
Si Fe Cu Mn Mg Zn Ti Al
D E ðdcÞ2 FU
ðdUÞ2 ¼ 2
 ; ½9
0.6 to 1.2 0.7 4.4 0.4 to 1.0 0.4 to 0.8 0.3 0.15 bal. ðDc0 Þ DV

where the constant noise magnitude


FU = kv0 (1  k)2 N1 A c0
1
is the value of FU for a
reference planar interface at temperature T0(U = 0).
Table IV. Physical Properties of Al–4.0 Wt Pct Cu and
Related Parameters The value Dc0 = c¥(1/k  1) is the concentration jump
across the interface, v0 is the molar volume of the solvent
Symbol Value Unit atoms, and NA is the Avogadro constant.
Liquidus temperature, TL 922.9 K
Alloy composition, c¥ 4.0 wt pct C. Experimental Design
Liquid diffusion coefficient, D 3.0 9 109 m2/s
Liquidus slope, m  2.6 K/wt pct
The experimental observations for the same set of
Anisotropy of surface energy, e0 0.01 welding parameters were carried out to verify the
Equilibrium partition coefficient, k 0.14 phase-field simulations. The base metal was Al-Cu alloy
Gibbs–Thomson coefficient, C 2.4 9 107 Km 2A14, and the dimensions of the plates were

Fig. 2—The evolutions of (a) incubation time and (b) the average wavelength with the misalignment angle h0 for different welding speeds.

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 3—Dynamic evolutions of the planar instability with different misalignment angles: (a) h0 = 0 deg, (b) h0 = 15 deg, (c) h0 = 30 deg, (d)
h0 = 45 deg (the time interval between successive lines is 0.09 s).

200 mm 9 120 mm 9 8 mm. To investigate the effect of observations or by numerical simulations. According to
welding speed, welding speeds of 2.0, 2.5, 3.0, and Reference 14 the incubation time and average wave-
3.5 mm/s were used with a constant welding power of length can reveal the effect of surface tension anisotropy
3600 W. For studying the effect of welding power, the on the initial instability. The evolutions of these two
welding powers of 2000, 3000, 4000, and 5000 were used criteria with misalignment angle for different welding
at the welding speed of 2.5 mm/s. Optical metallography speeds are presented in Figure 2.
was performed on the longitudinal section of the Figure 2 shows that for a given welding speed, the
welding beads to observe the microstructures near the incubation time and average wavelength increase
fusion line. slightly with the increase of h0. These increases result
from the stabilizing effect of the surface tension
anisotropy on the planar interface evolution. Based on
III. RESULTS AND DISCUSSION the local equilibrium, the one-sided model of solidifica-
tion, and with consideration of the surface tension
A. Effect of Surface Tension Anisotropy anisotropy, the free boundary conditions can be
Due to the transient behavior of the interface, it is expressed as[13]:
difficult to distinguish the instantaneous critical inter- @cL
face velocity of the initial instability by experimental Vtip ðcL  cS Þ ¼ D ½10
@z

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 4—The evolutions of (a) interface velocity and (b) solute concentration with misalignment angles h0 (P = 3600 W, V = 2.0 mm/s).

Fig. 5—The evolutions of (a) interface velocity and (b) solute concentration with different welding speeds.

TI ¼ TM þ mcL  jC½1  15c4 cos 4ðh þ h0 Þ; ½11 To demonstrate the different morphologies between
the misalignment angles clearly, the dynamic evolutions
where Vtip is the instantaneous interface velocity, cL and of the interface structures with different h0 values are
cS are the concentrations at the liquid side and solid side shown in Figure 3.
of the interface, respectively. TI is the temperature at the It can be seen that as h0 increases from 0 to 45 deg,
interface, j is the interface curvature, and c4 is the the planar interface remains intact longer and the initial
anisotropy strength. The term [1  15c4cos4(h + h0)] in wavelength becomes larger. The growth orientations of
Eq. [11] can be simplified to [1  15c4cos(4h0)], as the the crystals are along the preferred crystalline orienta-
value of h is infinitesimal in the initial stage. Then the tion but not the thermal gradient direction. This occurs
surface tension stiffness can be expressed as since the existing base metal grains at the fusion line act
c0[1  15c4cos(4h0)], which is positively correlated with as the substrate, and growth occurs by arranging atoms
h0. The stiffness reflects the stabilizing effect of the from the liquid metal upon the substrate grains without
surface tension anisotropy, i.e., the times for the planar altering their preferred crystalline orientations.[34]
interface and the larger initial wavelength increase with To get a more thorough understanding of the effect of
the increasing interface stiffness. This conclusion from surface tension anisotropy on the S/L interface during
the analytic model is consistent with the results in its evolution, the time-dependent characteristic param-
Figure 2. eters need to be investigated. The interface velocity and

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 6—The evolutions of (a) incubation time and (b) initial wavelength with welding speeds.

Fig. 7—The evolutions of (a) growth rate R and (b) thermal gradient G with different welding speeds.

solute concentration in front of interface with different diffusion from the tip position to both sides of the
h0 are presented in Figure 4, where the welding condi- interface. Hence the interface velocity increases sharply
tion was P = 3600 W and V = 2.0 mm/s. (Figure 4(a)) and the solute segregation decreases
Figure 4(a) shows that the evolutions of interface (Figure 4(b)) at the beginning of the cellular crystal
velocity with different h0 in the initial transient stage. The growth. After this critical time, the interface velocity
velocity increases with time and reaches its peak around reaches its peak and then decreases as the cellular crystal
the incubation time and then decreases. In the solidifi- tip and the solute concentration break down. That is to
cation process during welding, the total undercooling say, the inflection points of the curves correspond to the
consists of the thermal undercooling dominated by R critical times for instability. As shown in Figure 4, the
and G, and the constitutional undercooling caused by curves completely overlap with each other before the
solute trapping at the S/L interface. In the initial stage, instability time is reached, revealing that the surface
the interface is planar and advances slowly towards the tension anisotropy does not affect the interface velocity
liquid region. The growth rate and the solute aggregation or solute diffusion during solidification. This conclusion
ahead of the S/L interface increase as time progresses, is consistent with Eqs. [10] to [11], indicating that the
and eventually, the planar interface loses its stability and surface tension anisotropy influences the stability of the
develops into cellular crystals. A small spacing cell has a planar interface by changing the boundary conditions of
large interface curvature which can promote the solute the free boundary problem.

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 8—The evolutions of (a) G/R and (b) G 9 R with different welding speeds.

Nevertheless, the capillary effect in Eq. [11] with a concentration is that the increased interface velocity
small surface tension anisotropy (c4 = 0.01) had only a also increases the degree of solute segregation.
15 pct deviation from the isotropic case, indicating that Figure 6 shows the evolutions of incubation time and
the surface tension anisotropy has only a small influence initial average wavelength of planar instability with time
on the initial instability dynamics. Meanwhile, the for different welding speeds. This illustrates that the
differences between the curves in Figures 2(a) and (b) incubation time decreases, and the wavelength remains
indicate that the effect of h0 is negligibly small compared stable when the welding speed is increased.
to that of the welding speed. In practice, the weld metal The deterministic parameters of solidification are the
is polycrystalline material where the misalignment angle growth rate R and the thermal gradient G for different
is a random variable. The property of the surface welding conditions. By solving Eqs. [6] to [7], the
tension anisotropy in practical applications should be evolutions of R and G with time for different welding
almost constant. Therefore, we can focus on the effects speeds are obtained and shown in Figure 7.
of welding parameters on the initial instability dynamics It can be seen that the growth rate increases with
for the remaining portion of this study, using the increasing welding speed. The thermal gradient
misalignment angle h0 = 0 deg. decreases with increasing welding speed. The values of
R and G will both strongly influence the solidification
dynamics during welding.[34] The ratio of G to R could
B. Effect of Welding Speeds
determine the solidification mode, i.e., whether growth is
In the simulations, four different welding speeds, 2.0, planar crystal, cellular crystal, columnar dendritic crys-
2.5, 3.00, and 3.50 mm/s, were used with the constant tal, or equiaxed dendritic crystal. The cellular crystal
welding power of 3600 W. The evolutions of interface will appear earlier as the ratio G/R, decreases. The term
velocity and solute concentration in the front of G 9 R influences the microstructure size with the
interface with different welding speeds are presented in microstructure size decreasing with the increase of
Figure 5. G 9 R. Based on Eqs. [6] to [7], the evolutions of G/R
As shown in Figure 5(a), the interface velocity varies and G 9 R were calculated for different welding speeds
greatly under different welding speeds. The maximum and are shown in Figure 8.
interface velocity increases as welding speed increases, Figure 8(a) shows that the G/R gradually decreases
and the time before the peak decreases as the welding with the time which reveals that the interface morphology
speed increases. As the heat input decreases when the changes from planar, to cellular, and to dendrite during
welding speed increases, the solidification time is short- welding. The colored dots in the inset of Figure 8(a)
ened which causes this increase of the interface velocity. corresponds to the incubation times of instability, which
Figure 5(b) shows the evolutions of the solute concen- indicate that the incubation time decreases as the welding
tration with time for different welding speeds. The solute speed increases. As the G/R decreases with the increasing
concentrations maximums decrease with increasing welding speed, the cellular crystal growth will appear
welding speed since the increasing welding speed reduces earlier with smaller G/R values. As shown in Figure 8(b),
the molten pool size and can limit the solute diffusion the G 9 R increases with time in the initial transient
process. Another cause of the decreasing solute stage. The colored dots in the inset correspond to the

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 9—The microstructures near the fusion line with different welding speeds: (a) V = 2.0 mm/s, (b) V = 2.5 mm/s, (c) V = 3.0 mm/s, (d)
V = 3.5 mm/s.

incubation time of instability, from which it is seen that yellow dotted line in the inset of Figure 8(b) emphasizes
the values of G 9 R under different welding speeds are this point. Because the G 9 R influences the microstruc-
almost the same regardless of the incubation time. The ture size, the similar values of G 9 R at the incubation

METALLURGICAL AND MATERIALS TRANSACTIONS A


time will result in cellular crystals that have the similar survived dendrites. Figure 10 shows the comparison of
spacing. This conclusion is consistent with the results the initial average wavelength obtained by the experi-
shown in Figure 6(b). mental observations and the phase-field simulations
From the above discussions, it is seen that the with different welding speeds.
increasing welding speed could promote the initial The comparisons between the phase-field simulations
instability while having limited influence on the initial and experimental observations are in general agreement
wavelength. Figure 9 shows the optical metallography as both show that the initial wavelength remains
of the microstructures near the fusion line under stable for the different welding speeds. Furthermore,
different welding speeds. this agrees well with the previous discussion regarding
The micrographs in Figure 9 are located at the the G 9 R term remaining consistent regardless of the
bottom of the molten pool. The red dotted line in these incubation time.
images highlights the fusion line. The wavelength varies
with welding speeds and ranges from approximately 8.0
C. Effects of Welding Powers
to 11.0 lm. The average wavelength was calculated by
counting the number of cellular crystals near the fusion In the simulation experiments, four different welding
line, including the eliminated cellular crystals and the powers of 2000, 3000, 4000, and 5000 W are used at the
constant welding speed of 2.5 mm/s.

Fig. 10—Comparison of the initial wavelengths obtained by


phase-field simulations and experimental observations, respectively,
with different welding speeds. Fig. 12—The evolutions of incubation time with welding power.

Fig. 11—The evolutions of G/R with different welding powers.

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 13—The evolutions of G 9 R with different welding powers.

From Section III–B, the G/R term controls the powers used. The average wavelength was obtained by
solidification mode, and the G 9 R term determines counting the number of cellular crystals near the fusion
the microstructure size. By solving Eqs. [6] to [7], the line. Figure 15 shows the comparison of the initial
evolutions of G/R with time for different welding powers wavelength from the experiments and the phase-field
were obtained and are presented in Figure 11. simulations for different welding powers.
Figure 11 shows the values of G/R increase with The results from the phase-field simulations and
increasing welding power. From the discussion in experimental observations generally agree and show
Section III–B, cellular crystal growth appears earlier as that the initial wavelength increases slightly as welding
the G/R decreases. This means that the planar interface power increases. This result is consistent with the
lasts longer as the welding power increases. To verify previous discussion in this section.
this conclusion, the evolution of the incubation time The wavelength from the experimental results (from
with the welding power from phase-field simulations is 9.0 to 12.0 lm) is a little larger than that from the
shown in Figure 12. phase-field simulations (from 6.0 to 8.0 lm). The
The results from phase-field simulation in Figure 12 difference may result from the following reasons. First,
illustrate that the incubation time indeed becomes the phase-field simulations are done in a 2D system,
longer as the welding power increases. This is consistent while the experimental results are 3D. Second, the flow
with the discussed influence mechanism of G/R. field was not considered in the phase-field model used
To predict the change of the microstructure size with here. Also, the strong nonlinearity of the thermal
the welding power, the evolutions of G 9 R with time gradients during welding may not be fully realized in
were calculated and are presented in Figure 13. the simulation. As the average initial wavelengths from
The colored dots in the inset of Figure 13 correspond the simulation results and experimental observations are
to the incubation time of planar instability. From these within an order of magnitude, the phase-field model can
markers, it is seen that the G 9 R decreases slightly as be regarded as a quantitative model.
welding power increases at the incubation time. Conse-
quently, the initial wavelength should increase slightly as
welding power increases. Based on the discussion in IV. CONCLUSIONS
Section III–B, increasing welding power makes the
planar growth process last longer and makes the initial The effects of surface tension anisotropy and welding
average wavelength to be slightly larger. The optical parameters on the initial instability dynamics during the
metallography of the microstructures near the fusion GTAW of an Al-alloy were investigated by a QPF
lines for different welding powers is shown in Figure 14. model. The simulation results reveal that the surface
The microstructures shown in Figure 14 are located at tension anisotropy and welding parameters affect the
the bottom of the molten pool. Note that the red dotted interface instability evolution process. The following
line shows the fusion line. The wavelength ranges from conclusions are drawn based on the simulation and
approximately 8.0 to 12.0 lm for different welding experimental results:

METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 14—The solidification microstructures near the fusion line with different welding powers: (a) P = 2000 W, (b) P = 3000 W, (c)
P = 4000 W, (d) P = 5000 W.

1. The surface tension anisotropy does not influence interfaces remain intact longer and a larger initial
the solute diffusion process but does influence the wavelength can be obtained.
stability of S/L interface during the solidification of 2. The welding parameters affect the initial instability
welding. Moreover, as the stiffness of the interface dynamics by varying the growth rate and thermal
increases (i.e., larger h0 values), then both the planar gradient. The G/R term determines the solidification

METALLURGICAL AND MATERIALS TRANSACTIONS A


3. L. Wang, Y. Wei, X. Zhan, F. Yu, X. Cao, C. Gu, and W. Ou: J.
Mater. Process. Technol., 2017, vol. 246, pp. 22–29.
4. H. Bhadeshia and L. Svensson: Math. Model. Weld Phenom., 1993,
vol. 1, pp. 109–82.
5. C. Gu, Y. Wei, F. Yu, X. Liu, and L. She: Metall. Mater. Trans. A,
2017, vol. 48A, pp. 4314–23.
6. W.J. Boettinger, S.R. Coriell, A.L. Greer, A. Karma, W. Kurz, M.
Rappaz, and R. Trivedi: Acta Mater., 2000, vol. 48, pp. 43–70.
7. W. Losert, B.Q. Shi, and H.Z. Cummins: Proc. Natl. Acad. Sci.,
1998, vol. 95, pp. 439–42.
8. W.A. Tiller, K.A. Jackson, J.W. Rutter, and B. Chalmers: Acta
Metall., 1953, vol. 1, pp. 428–37.
9. W.W. Mullins and R.F. Sekerka: J. Appl. Phys., 1964, vol. 35,
pp. 444–51.
10. J.A. Warren and J.S. Langer: Phys. Rev. E, 1993, vol. 47,
pp. 2702–12.
11. Y. Chen, A.-A. Bogno, N.M. Xiao, B. Billia, X.H. Kang, H.
Nguyen-Thi, X.H. Luo, and D.Z. Li: Acta Mater., 2012, vol. 60,
pp. 199–207.
12. Z. Wang, J. Wang, and G. Yang: Appl. Phys. Lett., 2009, vol. 94,
p. 061920.
13. Z. Wang, J. Wang, and G. Yang: Phys. Rev. E, 2009, vol. 80,
Fig. 15—Comparison of the initial wavelengths from phase-field p. 052603.
simulations and experimental observations, respectively, with 14. Z. Dong, W. Zheng, Y. Wei, and K. Song: Phys. Rev. E, 2014,
different welding powers. vol. 89, p. 062403.
15. W. Zheng, Z. Dong, Y. Wei, and K. Song: J. Cryst. Growth, 2014,
vol. 402, pp. 203–09.
mode and the G 9 R determines the microstruc- 16. H. Xing, L. Zhang, K. Song, H. Chen, and K. Jin: Int. J. Heat
ture’s size. Specifically, the incubation time Mass Transf., 2017, vol. 104, pp. 607–14.
decreases, and the initial wavelength remains 17. J.J. Hoyt, M. Asta, and A. Karma: Mater. Sci. Eng. R, 2003,
vol. 41, pp. 121–63.
stable as the welding speed increases. As the welding 18. A. Sánchez Osio, S. Liu, and D.L. Olson: Mater. Sci. Eng. A, 1996,
power increases, the incubation time increases, and vol. 221, pp. 122–33.
the initial wavelength increases. 19. L.-Q. Chen: Annu. Rev. Mater. Res., 2002, vol. 32, pp. 113–40.
3. The simulation results are consistent with experi- 20. W.J. Boettinger, J.A. Warren, C. Beckermann, and A. Karma:
mental observations for the same set of welding Annu. Rev. Mater. Res., 2002, vol. 32, pp. 163–94.
21. L. Gránásy, L. Rátkai, A. Szállás, B. Korbuly, G.I. Tóth, L.
parameters. This verified the accuracy of the sim- Környei, and T. Pusztai: Metall. Mater. Trans. A, 2014, vol. 45A,
ulation results. pp. 1694–719.
22. A. Karma: Phys. Rev. Lett., 2001, vol. 87, p. 115701.
23. L. Wang, Y. Wei, and F. Yu: Mater. Sci. Technol., 2017, vol. 33,
pp. 846–53.
24. M. Wei, Y. Tang, L. Zhang, W. Sun, and Y. Du: Metall. Mater.
Trans. A, 2015, vol. 46A, pp. 3182–91.
ACKNOWLEDGMENTS 25. E. Miyoshi, T. Takaki, M. Ohno, Y. Shibuta, S. Sakane, T.
Shimokawabe, and T. Aoki: NPJ Comput. Mater., 2017, vol. 3, p. 25.
The study is supported by the Priority Academic 26. G. Demange, H. Zapolsky, R. Patte, and M. Brunel: NPJ Comput.
Program Development of Jiangsu Higher Education Mater., 2017, vol. 3, p. 15.
27. B. Echebarria, R. Folch, A. Karma, and M. Plapp: Phys. Rev. E,
Institutions (PAPD) and the financial aid for the pro- 2004, vol. 70, p. 061604.
ject from the Fundamental Research Funds for the 28. W.J. Zheng, Z.B. Dong, Y.H. Wei, K.J. Song, J.L. Guo, and Y.
Central Universities NP2016204. The authors acknowl- Wang: Comput. Mater. Sci., 2014, vol. 82, pp. 525–30.
edge Long-Qing Chen and Xiaoxing Cheng at Penn- 29. A.J. Clarke, D. Tourret, Y. Song, S.D. Imhoff, P.J. Gibbs, J.W.
sylvania State University for their help in the Gibbs, K. Fezzaa, and A. Karma: Acta Mater., 2017, vol. 129,
pp. 203–16.
development of the phase-field code. 30. A. Karma and W.-J. Rappel: Phys. Rev. E, 1998, vol. 57,
pp. 4323–49.
31. H. Xing, X. Dong, H. Wu, G. Hao, J. Wang, C. Chen, and K. Jin:
Sci. Rep., 2016, vol. 6, p. 26625.
REFERENCES 32. B. Echebarria, A. Karma, and S. Gurevich: Phys. Rev. E, 2010,
1. F. Yu, Y. Wei, Y. Ji, and L.-Q. Chen: J. Mater. Process. Technol., vol. 81, p. 021608.
2018, vol. 255, pp. 285–93. 33. M. Asta, J.J. Hoyt, and A. Karma: Phys. Rev. B, 2002, vol. 66,
2. H.L. Wei, J.W. Elmer, and T. DebRoy: Acta Mater., 2016, p. 100101.
vol. 115, pp. 123–31. 34. S Kou: Welding Metallurgy, Wiley, New York, 2003, pp. 166–67.

METALLURGICAL AND MATERIALS TRANSACTIONS A

Das könnte Ihnen auch gefallen