Pliazhuk Supplement September 2019.

qxp_Layout 1 8/8/19 4:46 PM Page 251

WELDING RESEARCH
SUPPLEMENT TO THE WELDING JOURNAL, SEPTEMBER 2019
Sponsored by the American Welding Society

In-Situ Monitoring of Transient Strain Formation

in Vertical Welds

Even in a short weld, compressive transversal strains in front of the weld pool
gradually increase as the weld pool moves up the plate

BY M. PLIAZHUK, C. REYES, M. MARTINEZ, J. GOLDAK, H. NIMROUZI, AND D. K. AIDUN

parts are subject to uneven distribution of high thermal gra-

ABSTRACT dients resulting in residual stresses (RS) along the weld
path, thus affecting the mechanical performance of the
This study focuses on the use of optical noncontact ex-
joint. Thus, there is a need to examine and develop physics-
perimental mechanic techniques to capture the transient
strain formation of a single-bead, manually performed verti- based models that capture the complex nature of residual
cal weld. A 300 × 300 × 6.35-mm A36 steel plate was ma- stresses in weldments. This study focuses on analyzing the
chined with a V-groove. Digital image correlation (DIC) cam- transient strain formation in a manual vertical weldment in
eras in combination with a FLIR One Pro camera were posi- the proximity of the fusion zone for an A36 steel plate using
tioned to capture the transient strain development in situ on noncontact experimental mechanics techniques. In addition,
the back of the plate. A manual vertical weld bead from the the experimental results obtained were further used to vali-
bottom to the top of the plate was performed. The results of date a computational weld mechanics (CWM) model, known
this study demonstrate a complex behavior of both tensile as VrWeld, and thus gain insight into the physics of tran-
and compressive regions leading to transversal and longitudi- sient weld dynamics.
nal residual strain distributions. Due to the manual nature of
Zhou et al. (Ref. 1) classified the measurement of residual
the weld, this study focused on the transient strain forma-
tion during the exchange of a filler rod. Experimental results stresses into two main categories: destructive (i.e., hole
show that even in a short 300-mm weld, compressive trans- drilling) (Refs. 2, 3), which are not applicable to a finished
versal strains in front of the weld pool gradually increase as product, and nondestructive (i.e., fiber optic, strain gauges,
the weld pool moves up the plate. Finally, a finite element neutron diffraction), which in many cases tends to be time
model based on a 3D transient Eulerian formulation of the consuming, costly, and in some instances requires the use of
weld pool coupled to a 3D Lagrangian analysis of the com- sophisticated facilities (i.e., neutron diffraction) (Ref. 4).
plete weld that solves the conservation of energy, mass, and The widespread use of noncontact experimental mechan-
momentum with a Gaussian pressure distribution, thermal ic techniques, such as digital image correlation (DIC) and
flux distribution, and mass flux distribution from the arc was thermography, provides an opportunity to measure a 2D
used to verify the experimental results.
field of strain and temperature in time. Zhou et al., in (Ref.
1), made use of DIC during a welding application. In their
study, the authors made use of additional light sources, to
KEYWORDS moderate the electromagnetic radiation, in the form of light
and heat emission emanating from the welding arc, in an at-
• Transient Strain in Welds • Thermography
tempt to reduce the uncertainty of the measurement. Other
• Digital Image Correlation (DIC)
• Computational Weld Mechanics scientists performed welding experiments with DIC and no-
ticed that electromagnetic radiation is one of the biggest
concerns. Eshtayeh, Hijazi, and Hrairi (Ref. 5) tested this
Introduction technique for a similar metal weld and compared results to
thermographic imaging technique, thus concluding that DIC
is a relatively easy method for defect detection in weldments
Starting in the late 19th century, arc welding has been and and, even though it does not allow for identifying the type
continues to be a widespread important method for joining of defect, such as porosity, slag inclusion, incomplete fusion,
metal parts during the manufacturing of metal structures. root concavity, etc., it proved to be quite accurate on assess-
However, due to the nature of this process, welded metal ing the presence of these. Gollnow and Kannengiesser per-

https://doi.org/10.29391/2019.98.022

SEPTEMBER 2019 / WELDING JOURNAL 251-s

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 252

WELDING RESEARCH

Fig. 1 — Experimental setup. Fig. 2 — V-groove and overall dimension of the specimen plate.

Table 1 — Mechanical Properties of the Plate and Filler Metal

Material Tensile Strength (lb/in.2) Yield Strength (lb/in.2) Elongation in 2 in. (%) Reduction of Area (%)
[MPa] [MPa]

A36 Plate 63800 53700 15 40

[439.9] [370.2]
70S-2 wire 83000 71000 27 58
[572] [490]

formed a successful in-situ experiment using DIC with an source, a polarization controller, a beam splitter, an analog-
intent to examine the process of solidification cracking in to-digital converter, a computer, and a fiber-optic sensor
gas tungsten arc welding (GTAW) of similar welded flat pan- that is used to acquire the thermal and total strain measure-
els (Ref. 6). Furthermore, Agarwal et al. (Ref. 7) investigated ments (Ref. 12). One of the advantages of this technique is
the influence that RS had on such cracking in a similar laser the high spatial resolution to measure strain, which provides
weld with the help of DIC. In addition, De Strycker et al. a data point on average every 1.25 mm for a 10-m-long fiber
(Ref. 8) were successful in detecting deformations during at a data acquisition rate of 23.8 Hz. However, at 100-Hz
the welding process with added external loading using DIC data acquisition rate, the same system is able to measure
and mentioned one of the great advantages of this method strain every 5 mm for the same 10-m-long fiber. The fiber-
as being its contactless nature while accounting for thermal optic sensor consists of an anhydrous silica (ASI) single
expansion of the material. Despite the advantages provided mode, polyimide coated with a mode field diameter of 9 mi-
by many authors on the use of DIC, its use is often impeded crometers, a cladding size of 125 micrometers, and a coating
in welding application by intense light emission, where spe- diameter of 145 micrometers, operating at wavelengths of
cial measurements to shield the cameras from the light 1.31 micrometers. This technology has been used to meas-
emitted from the arc are required (Ref. 8). Nevertheless, the ure strain and for structural health monitoring applications
technique still proved to be accurate in many of the studies in composite structures (Refs. 13, 14).
found in the literature. Chen et al. recently published a Matteï et al. (Ref. 15) showed the use of a thermographic
study on the use of 2D and 3D DIC. The authors made use of system for capturing the isothermal contours during hybrid
two cameras in the proximity of the weld pool and obtained laser and gas metal arc welding of an AA2024T3 T-joint. The
results that very closely matched those obtained with strain results were further simulated, and a direct comparison be-
gauges. The study shows the ability of 3D DIC to capture the tween the experimental data and the simulated data demon-
deformation and strain profiles during the welding process strated good agreement between the numerical and experi-
(Ref. 9). mental approximation. In addition, Goldak et al. (Ref. 16)
In addition to DIC, authors have experimented with the demonstrated the use of a FLIR One Pro commercial, off-
use of distributed fiber optics to measure thermal and total the-shelf camera to capture the isotherms produced during
strains. This technology makes use of an optical time do- welding of two 1018 steel plates. The experimental results
main reflectometry (OTDR), an optical frequency domain were also simulated, and a direct comparison of a manual
reflectometer (OFDR), or a polarization OTDR (POTDR), weld was evaluated. The authors found the thermographic
based on the Rayleigh backscattering principle (Refs. 10, system was able to capture temperature values below 400°C
11). Rayleigh backscatter is produced by the interaction of and the heating rate between the two agreed very well for
the light wave with the silica impurities in the fiber. This in- the first 100 s of the welding process. The authors believe
teraction produces a reflection approximately 10–7 m–1, the discrepancies between the isothermal areas in the exper-
which can be detected and analyzed utilizing one of the imental and simulation were attributed to the computation-
above-mentioned systems. The OFDR method, on which the al model not accounting for the convection heat transfer
LUNA ODISI-B system is based, makes use of a tunable laser during the experimental setup, as room temperature air was

252-s WELDING JOURNAL / SEPTEMBER 2019, VOL. 98

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 253

WELDING RESEARCH

A B C

Fig. 3 — A — Full FLIR One Pro thermal image; B — cropped thermographic region of interest; C — converted 2D temperature field
on the region of interest.

blown by a fan toward the back of the plate required for the space-time data from two noncontact sensing systems with
DIC system, which was also being used as part of the au- a 3D parametric computer model of the weld is novel.
thors’ experimental setup.
The development of a physics-based model and the use of Transient 3D Computer Model of Weld
computational techniques, such as finite element methods
(FEM), have also been used to predict the formation of RS The weld heat source model is based on models proposed
(Ref. 16). This methodology was also evaluated by by Ohji et al. (Ref. 18) and Sudnik (Ref. 19). The weld heat
Eisazadeh, Hamedi, and Halvaee (Ref. 17) to predict weld source model is a 3D transient Eulerian formulation that
nugget size and the welding quality in resistance spot weld- solves the conservation of energy, mass, and momentum
ing (RSW), and was experimentally proven to be accurate with a Gaussian pressure distribution, thermal flux distribu-
(Refs. 18–20). Additionally, Chang and Teng (Ref. 21) con- tion, and mass distribution from the electrode. The effects
firmed the results of finite element analysis (FEA) evalua- of surface tension and gravity were included. Patterns for
tion of residual stresses in butt joint welds with x-ray dif- weave welds with weave amplitude and dwell times are sup-
fraction. Eisazadeh et al. (Ref. 22) performed FEA calcula- ported in the model. This heat source model predicts the
tions to predict RS in a dissimilar metal weld; however, weld shape and size as a function of the weld procedure,
these results were not experimentally verified. Zacharia weld joint geometry, and base and filler metal composition.
(Ref. 23) calculated the temperature and strain distribution This is in sharp contrast to the distributed power density
on a metal plate for a horizontal weld and validated the that is often called the double ellipsoid heat source model
analysis with experimental data. that requires the weld pool shape and size be known
Despite the many advancements in understanding the through experimental data (Ref. 20). In our study, the Euler-
physics of weldments, the formation of transient strains ian weld pool model is coupled with a Lagrangian 3D tran-
leading to the development of RS is an ongoing interest for sient model that solves the conservation of mass, energy,
the welding community. Many of the results found in the and momentum.
literature have focused on the estimation and/or measure- The conservation of energy in spatial coordinates is
ment of the RS. As such, the novel nature of the present solved using a semi-Lagrange method:
study has been the use of high resolution spatial and time
data in situ during a manual weldment. Although this is r dh/dt –  ∙ q + Q = 0 (1)
not the first time DIC has been used for analyzing weld be-
havior, the authors believe this is the first time that DIC where r is density, h is specific enthalpy, q is thermal flux,
has been reported on a vertical transient weld from the and Q is power per unit volume. The specific heat and ther-
back side with a larger area of interest. In addition, the mal conductivity are functions of temperature.
combination of this experimental technique and a one to The quasi-static conservation of momentum in spatial
one comparison with the VrWeld model far and close to coordinates was solved using a semi-Lagrange method:
the proximity of the weld pool, provides a unique insight
into the transient strain behavior, which to our knowledge ∙ s + g = 0 (2)
has not been reported until now (Refs. 18, 19). Further- and
more, the authors believe that the correlation of transient

SEPTEMBER 2019 / WELDING JOURNAL 253-s

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 254

WELDING RESEARCH

A B C

Fig. 4 — A — FLIR thermographic image prior to extinguishing the arc for a filler metal rod exchange; B — temperature distribution
of the thermographic image; C — VrWeld temperature distribution.

mental setup and coupon specification are described as

p = g (1/R1 + 1/R2) (3) follows.

where s is the Cauchy stress, g is the acceleration due to Experimental Setup

gravity, p is the pressure drop across the liquid-plasma in-
terface, g is the surface tension, and 1/R1, 1/R2 are the prin- The experimental setup consisted of the following: 1)
cipal curvatures at a point on the liquid surface. The inertial DIC system from Ref. 28, and 2) a commercial, off-the-shelf
force and Marangoni forces are neglected in this model. FLIR ONE Pro thermographic camera (Ref. 29) connected to
The mass flow in the wire is described as pr2wirevwire = an iPhone X (Ref. 30) as shown in Fig. 1. The FLIR One Pro
Afillervweld, where r wire is the radius of the wire, vwire is the wire camera is a fixed emissivity camera with an 8–14 microme-
speed, vweld is the weld speed, and Afiller is the cross-sectional ter spectral wavelength, capable of measuring temperatures
area of the filler metal, to enforce the mass conservation as up to 400°C. The camera has a 160  120 thermal resolution
a constraint in the model. The weld pool model requires all and a visual resolution of 1440  1080, with a pixel size of
parameters of most weld procedures plus additional parame- 12 m and a 8–14 m spectral range (Ref. 29). It has been
ters such as surface tension and arc pressure that are not documented in the literature that the device has an accuracy
usually included in a weld procedure data sheet. The VrWeld of ±5% at a fixed emissivity value of 0.90 (Ref. 31). The au-
model has been developed by Goldak and his team over mul- thors conducted a lab experiment to validate the accuracy of
tiple decades. A detailed description of the mathematical the FLIR One Pro camera. The experiment consisted of cov-
formulations can be found in Refs. 28–30. ering the front of an oven with an A36 steel plate painted
with the same very high temperature paint used in the
Methodology study. The results of these experimental tests are not pre-
sented but confirm that up to 120°C (maximum tempera-
In this study, the authors made use of digital image corre- ture the oven could reach) fall within 5°C when compared to
lation (DIC), a FLIR One Pro thermographic camera con- a K-type thermocouple connected to a National Instrument
nected to an iPhone X, to capture the temperature and data acquisition system. Furthermore, the authors looked at
strain formation during a manual welding process. One A36 several emissivity tables for surfaces painted with white
steel plate machined with a V-groove was positioned in a fix- paint and found the emissivity value ranged from 0.84 (Ref.
ture as to allow for a vertical orientation of the manual weld. 32) to 0.95 (Ref. 33).
A vertical position of the weld was preferred to allow for vi- A single CW-GTA weld bead was performed at the Clark-
sual in-situ monitoring of the weld by the DIC cameras from son University welding laboratory. The arc current (A) was
the backside of the plate. The primary objective of this ex- set to 150 direct current electrode negative (DCEN) and
perimental setup was to instrument the process to capture recorded arc voltage was roughly 14 V. It is important to
as much experimental data as possible to compare the com- note that during a manual weld, the arc current and voltage
putational welding results with the 2D field views of the are constantly changing during the welding process. An av-
welding process. The authors wanted to perform, whenever erage current and voltage, with its corresponding voltage
possible, a one to one comparison between the experimental speed, were used as input into the VrWeld model. The DIC
and numerical findings. In addition, in this study, a gas clock provided an estimate on the location and time for
tungsten arc manual weld was used to assess the variability which the filler rod exchange took place. As such this infor-
typically associated with these techniques in a single-bead mation was used also as input to the VrWeld model.
weld of the steel plate. A detailed explanation of the experi- The DIC cameras were strategically positioned at the back

254-s WELDING JOURNAL / SEPTEMBER 2019, VOL. 98

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 255

WELDING RESEARCH

Table 2 — Thermal Conductivity Properties of the Plate and Filler Metal

Material Temperature Conductivity

F [K] (BTU/h-ft-F) [W/m K]

35.33 [275] (7.5) [13]

71.33 [295] (7.8) [13.5]
116.33 [320] (8.1) [14]
215.33 [375] (8.7) [15]
296.33 [420] (9.2) [16]
395.33 [475] (9.5) [16.5]
470.93 [517] (10.1) [17.5]
560.93 [567] (10.3) [17.8]
658.13 [621] (10.7) [18.5]
749.39 [672] (11.2) [19.4]
838.13 [721] (12.0) [20.7]
928.13 [771] (12.0) [20.7]
1018.13 [821] (12.5) [21.7]
1108.13 [871] (12.9) [22.3]
1198.13 [921] (13.5) [23.3]
1288.13 [971] (14.0) [24.3]
1378.13 [1021] (14.4) [24.9]
1468.13 [1071] (14.4) [25]
1558.13 [1121] (15.0) [26]
A36 Plate 1648.13 [1171] (15.6) [27]
1738.13 [1221] (16.2) [28]
1828.13 [1271] (16.2) [28]
1918.13 [1321] (17.3) [30]
2008.13 [1371] (17.9) [31]
2098.13 [1421] (20.8) [36]
2188.13 [1471] (23.7) [41]
2278.13 [1521] (26.6) [46]
2368.13 [1571] (29.5) [51]
2458.13 [1621] (32.4) [56]
2548.13 [1671] (35.2) [61]
2638.13 [1721] (41.6) [72]
2728.13 [1771] (47.4) [82]
2818.13 [1821] (52.6) [91]
2908.13 [1871] (58.4) [101]
2998.13 [1921] (64.1) [111]
3088.13 [1971] (69.9) [121]
3178.13 [2021] (75.7) [131]
3268.13 [2071] (81.5) [141]
4888.13 [2971] (81.5) [141]

8.3 [260.0] (23.13) [40.0]

368.3 [460.0] (23.13) [40.0]
764.3 [680.0] (21.97) [38.0]
70S-2 wire 1583.3 [1135.0] (14.54) [25.0]
2636.3 [1720.0] (17.34) [30.0]
2834.3 [1830.0] (185.0) [320.0]
 2834.4 [ 1830.0] (185.0) [320.0]

of the plate, while a fan was placed to blow room tempera- Solutions Inc. (Ref. 28) to give an optimal match confidence
ture air toward the back of the specimen. The fan was re- for a given noise level. The VIC-3D software was utilized for
quired to minimize the heat formation in the air path of the the analysis of DIC data. A subset size was set to 29 with a
light to the DIC cameras. This formation of heat of the air step size equal to 7. Engineering strain with a filter size of
interferes with the ability of the DIC software (VIC-3D) to 15 was selected for the strain calculation as an input param-
cross correlate to the reference images (Ref. 34), thus pre- eter for the VIC-3D software.
venting the computation of strain fields during welding. In addition to the use of DIC, a FLIR ONE Pro thermo-
In addition, LED lights were placed on the back of the graphic camera was used in this study. The thermographic
welded plate to create a good contrast between the high- camera was connected as an accessory to an iOS phone (Ref.
temperature base paint and the speckle pattern. The speckle 30) allowing the user to obtain thermographic images using
pattern was applied with a 0.026-in. roller from Correlated the FLIR ONE app (Ref. 35). The primary advantage of this

SEPTEMBER 2019 / WELDING JOURNAL 255-s

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 256

WELDING RESEARCH

A B

Fig. 5 — Strain distribution in the following: A — Transversal; B A B

— longitudinal directions 1 min after the weld started.

setup is that it allowed the authors to convert every pixel of

the image into a temperature point for an area of interest. In
this study, one of the authors positioned himself at the back
of the LED lights of the experimental setup, at approximate-
ly 2 m away from the specimen. Previous attempts had been
made in which the author positioned the iPhone and FLIR
ONE Pro camera at just 0.5 m from the specimen; however,
at this proximity, the radiation from the arc weld would
cause the iPhone to stop working, thus preventing the au-
thor from gathering thermographic resolution images of the
back of the plate during the welding process. It is important C D
to note that the FLIR ONE Pro Camera is only rated to cap-
Fig. 6 — Transversal (x-axis) strain readings during the process
ture temperatures up to 400°C. To analyze the thermogra- of switching the filler rod: A — Right before switching off the
phy data, a separate Matlab™ code was created utilizing the weld torch; B — weld torch switched off completely and liquid
Atlas Software Development Kit (SDK) (Ref. 36). The use of weld pool is solidifying on the back side; C — liquid weld pool
the SDK allowed the authors to convert each pixel of the in- solidified; D — weld torch back on and weld continues.
frared image into a corresponding temperature value by us-
ing the GetPixelsArray function from ImageProcessing set heat treating the paint for 30 min at 128°C followed by a
for each image taken, thus creating a two-dimensional ma- natural convection cooling for another 30 min (Ref. 38). The
trix of temperature values that were later used to represent heat treatment process was followed by a secondary heat
the temperature field as a contour plot. The author contact- treatment at 160°C for 30 min, followed by a secondary
ed FLIR to see the trustworthiness of the data obtained in cooling process for another 30 min.
those regions where the camera was saturated due to higher
temperature values. The customer support team at FLIR in-
dicated that each pixel is independent, and as such values of Experimental and VrWeld Results
temperatures below 400°C regardless of where they take
place in the image can be trusted with an accuracy of ±5%. The following section includes the experimental and Vr-
Weld results. The VrWeld models are presented side by side
Specimen Configuration with the experimental data to do a direct comparison and
thus demonstrate the ability of the models to capture the
One 300  300  6.49 mm A36 steel plate was machined complex physics dynamics of a manual arc welding process.
with a V-groove, as shown in Fig. 2. The plate was welded with A single weld pass of 270 mm in length was performed
a single-pass weld bead using a 11⁄ 6-in. ER 70S-2 filler rod. along the V-groove of the A36 plate. However, the manual
The mechanical and thermal conductivity properties of nature of the weld required the welder to perform two filler
the plate and filler are given in Tables 1 and 2. rod exchanges lasting approximately 8 s each during the
To avoid unknown residual strains during the measure- welding process. The first sub pass covered a distance from
ment, the A36 plate was heat treated for 2 h at 621°C prior 10 mm from the bottom edge of the plate to 130 mm, thus a
to welding, followed by a cooling process to room tempera- total weld length of 120 mm. This pass was computed at an
ture as indicated in Ref. 37. In addition, and due to the need average speed of 0.77 mm/s. The second sub pass covered
to paint the back of the plate for the DIC measurement, a the distance between 130 and 220 mm at an average speed
high-temperature paint commonly used in the automotive of 0.64 mm/s. While the final pass of the weld covered the
industry for painting the exhaust of cars was utilized. The distance from 220 to 270 mm at an average speed of 0.7
very high-temperature (VHT) paint is rated to withstand mm/s. The total time associated with this weld was 368 s.
1093°C (Ref. 38). The cure process for the paint consisted of

256-s WELDING JOURNAL / SEPTEMBER 2019, VOL. 98

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 257

WELDING RESEARCH

A B A B

Fig. 8 — Strain distribution along the weld pool 1.5 min before
the weld was terminated.

metal plate prior to exchanging the filler rod. The gradual

increase in temperature while the weld is in progress can be
observed in the 2D temperature field in Fig. 4B, C. It is in-
teresting to observe the slight temperature variation that
appears on the back of the weld pool as shown in Fig. 4B. In
addition, the VrWeld model captures a similar temperature
distribution — Fig. 4C. The model predicts temperatures
well above 400°C, which in this comparison has not been
C D shown to compare similar isotherms. Another interesting
observation is due to the slow thermal conduction of heat in
Fig. 7 — Longitudinal (y-axis) strain readings during the process the material, the moment the filler exchanges took place is
of switching the filler rod: A — Right before switching off the not visible from the cropped image produced by the FLIR
weld torch; B — weld torch switched off completely and liquid One Pro camera in Fig. 4A, B. However, the moment that a
weld pool is solidifying on the back side; C — liquid weld pool filler rod is exchanged can be observed by viewing the move-
solidified; D — weld torch back on and weld continues. ment of the welder and the position of the welder’s mask in
Fig. 3A. The welder temporarily removes the welding mask,
Thermographic Results while grabbing a new filler rod.

Figure 3A shows the entire thermographic image pro- DIC Results

duced by the FLIR One Pro camera. A subset of the image
was selected by a rectangular plate denoting the area of in- A total of ten reference images were captured prior to
terest. This subset corresponds to 206 by 178 pixels as indi- starting the weld. These reference images were evaluated to
cated at the top left corner of the figure. Figure 3B shows understand the error range of the DIC measurement tech-
the selected region cropped from the original thermographic nique under a nonstrained condition. The results demon-
photo, which recorded the initial temperature of the plate strated that error in the transversal direction were within
(and hence the room temperature) at approximately 24°C. the range of –62 and 54.5 , while in the longitudinal direc-
This image size corresponds to approximately 300  300 tion the range varied between –63 and 72 . Figure 5
mm — the dimensions of the steel plate. Figure 3C is a filled shows the strain distribution in transversal (A) and longitu-
contour plot representing the temperature distribution over dinal (B) directions one minute after the weld started. The
the selected region where every pixel in the area of interest variation in transversal strain values in the beginning of the
was converted into a temperature matrix and plotted in weldment ranges from –9827 to 8405 , as opposed
Matlab™. The position of the selected region in the original to –621 to 3811  in the longitudinal direction. As seen in
image and region’s size is consistent among all cropped ther- Fig. 5A, a compression region in front of the weld pool is
mal camera images. It is worth mentioning that according to formed during the welding process. This compression region
the manufacturer, the maximum temperatures detected is followed by a tensile region immediately above it. Howev-
with the FLIR ONE Pro camera is 400°C, even though the er, the areas to either side of the weld torch show a tensile
readings in some of the images were measured up to 586°C. region, indicative of the prying nature of the transversal
Figure 3C shows temperature contour fields around the up- strain during the weldment as shown in Fig. 5. This tensile
per left and right corner of the image. These contour fields region switches side to side during the manual weld. It is be-
are produced from reflections of LED light used by the DIC lieved that this fluctuation is a result of the weaving process
system being picked up by the FLIR One Pro camera from during a manual welding process.
the steel frame constraining the specimen. In addition to the transversal and longitudinal strain dis-
Figure 4A show the temperature distribution over the tribution at the initiation of the weldment, it was possible

SEPTEMBER 2019 / WELDING JOURNAL 257-s

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 258

WELDING RESEARCH

Fig. 9 — Longitudinal and transversal strain comparison between DIC and VrWeld at the midsection of the plate, prior to the weld
pool passing through (DIC clock 100 s).

Fig. 10 — Longitudinal and transversal strain comparison between DIC and VrWeld at the midsection of the plate, after the weld
pool has passed through (DIC clock 300 s).

to capture the change in both transversal and longitudinal the weld pool while the tensile region along the side of the
strain during the exchange of the filler rod as shown in Figs. weld increases by 1802  from the initiation of the filler
6 and 7. rod exchange process. It is interesting to observe that the
A short description of the succession of events for both extinguishing of the arc causes an increase in both the ten-
Figs. 6 and 7 are described in the following sequence: sile and compression strain fields in the transversal strain.
1. Prior to extinguishing the weld arc. At this mo- However, the longitudinal strain (Fig. 7) shows a reduction
ment in time during the welding process, the transversal of the tensile strain above the weld arc of approximately
strain (Fig. 6) above the liquid weld pool is measured to 5000  (11,780–6715 ), while the tensile region along
be –5698  (negative sign is indicative of the compressive the side of the arc remains relatively the same, with an in-
nature of the strain field) while a tensile strain of 12,424  crease in magnitude of approximately 700  from Fig. 7A.
is measured to the left of the weld pool. The longitudinal 3. The liquid weld pool solidified completely on the
strain readings (Fig. 7) at the same locations with respect to back face of the plate. Transversal (Fig. 6) compressive re-
weld pool are 11,780 and 7096  — both tensile. gion remains constant (very small fluctuation) while longi-
2. The weld arc is extinguished. At this instant of tudinal strain (Fig. 7) at the same location is found to have a
time, the cooling process of the liquid weld pool is initiated tensile strain of 7060 .
causing it to start solidifying on the back side of the plate 4. The weld arc is again turned on to continue the
(thus the oval white region showing light emission decreas- weld. The area near the heat source is fully liquefied once
ing in size). The magnitude of the transversal strain read- again while the weld continues. The strain in the transversal
ings (Fig. 6) increases to 19,634  in compression above direction (Fig. 6) compressive region continues to go into

258-s WELDING JOURNAL / SEPTEMBER 2019, VOL. 98

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 259

WELDING RESEARCH

A B

Fig. 11 — Strain distribution in the plate 20 min after the weld was completed.

compression reaching –19,867  while well above the weld the weld pool. It is also interesting to observe the lack of
pool a tensile region is formed that reaches 506 . Finally, symmetry on the longitudinal DIC data, which was not pro-
the longitudinal strain (Fig. 7) in front of the weld pool now duced nor was expected by the VrWeld results.
measures –2747 . Figure 11 shows the peak values of the residual strains
As the manual weld continues along the vertical weld formed after 20 min once the weld was completed. As seen
path, it was observed that the transversal strain reached its in this figure, the strain can be observed along the weld pool
most compressive value (–26,341 ) in front of the liquid varying from –1819 and 83  in transversal direction
weld pool roughly 1.5 min before the weld was terminated to –1783 and 336  in the longitudinal direction minutes
as shown in Fig. 8. While at the same instant, the longitudi- after weld was completed.
nal strain reached 6810 .
A one-to-one comparison between VrWeld and the DIC Discussion
experimental results was conducted as part of this study.
Two time steps were selected along the path of the weld to The experimental results obtained in this study clearly
perform this comparison. The first time step was selected at show the complex dynamic behavior of both the strain and
100 from the start of the DIC clock. At this time step, the temperature fields during the welding process. However, in
weld pool is below the midsection of the plate at approxi- this study, the authors considered the use of a handheld
mately 20 mm away. As seen in Fig. 9A, the shape of the lon- camera provided unnecessary fluctuations in the FLIR tem-
gitudinal strain predicted by VrWeld and the DIC are similar perature data set. As such, future work will make use of a
in nature. However, the magnitudes of strains captured by tripod to fix the FLIR iPhone in space to reduce the camera
the DIC differs by a factor of 2.5 from the VrWeld model. motion. Although a thermographic camera that can record
The results are even worse when comparing the transversal temperatures above 400°C would provide additional data,
strain in Fig. 9B. The transversal strain at this proximity of the authors were pleasantly surprised by the data this cam-
the weld pool shows actually opposite strain behavior (com- era was able to capture. For example, an area of the 300°C
pressive vs. tensile). isotherm was 10 mm2 while the total area of the plate was 9
However, Fig. 10 shows the same comparison at the same  104 mm2 (300 mm by 300 mm), as such the area not sam-
midsection of the plate, once the weld pool has passed pled by the FLIR camera represents less than 0.01% (10/9 
through. When the DIC weld clock reads 300 s, at that mo- 104) of the total area of the plate. Thus, 99.99% of the plate
ment in time, the weld is still in progress and the weld pool was sampled by the FLIR camera. Because the fields for tem-
is found on the upper section of the plate at approximately peratures below 400°C are very smooth and have low fre-
88 mm away. As seen in this figure, both the longitudinal quency modes, it was not essential to have a high-resolution
and transversal strain profiles from the DIC results and camera. We could fit a polynomial to virtual temperatures
those obtained from VrWeld are within 500 microstrains of below 400°C. We would expect the polynomial would be a
each other. More importantly, the shape of the strain pro- low degree polynomial. A scientific thermographic camera
files matches the change in strain gradients. Finally, as seen would also provide more control over the data acquisition
in this figure, the center section of the DIC results are miss- rate of the images taken. The current model of the FLIR One
ing, as the paint has been burned off through the passing of Pro app is unable to produce thermographic video and, as

SEPTEMBER 2019 / WELDING JOURNAL 259-s

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 260

WELDING RESEARCH

such, the author had to manually press the camera trigger to to compression and back to tensile around the weld pool is
capture the desired images. This process created uncertainty during the filler rod exchange process as seen in Fig. 7. How-
in synching the different thermographic images with their ever, once the welder continues, this longitudinal strain
equivalent VrWeld model. Furthermore, the thermographic fields stabilizes around the weld pool with 3770  tensile
images highlighted certain features on the temperature dis- on the side and 6810  at the front.
tribution shown in Fig. 4B of the tail of the weld pool, which The in-situ monitoring of the weld shows how the
was not captured by the VrWeld model. It is believed these switching on and off of the arc during a filler exchange af-
features could be caused by the difference in thermal con- fects local strain formation as shown in Figs. 6 and 7A–D.
ductivity of the filler rod and the base material. The results clearly indicate an increase in the magnitude of
Furthermore, it was found that while the use of DIC was the transversal strain in front of the weld torch. As shown in
capable of capturing the instant in time where the welder Fig. 6B–D, switching off the weld torch allows the liquid
performed a filler rod exchange, the FLIR One Pro thermo- weld pool to cool and solidify resulting in an increase in the
graphic data was unable to show that instant in time, due to magnitude of the transversal strain to –19,867  through
the slow heat transfer process. The filler rod exchange was the exchange of the filler rod. In Fig. 8A, the magnitude of
captured through analyzing the background field of view, this transversal strain is even higher, reading –26,341 ,
i.e., the movement of the welder. As such, it was possible which confirms the trend of an increase in the compressive
through the FLIR camera images to distinguish the moment region of the transversal strain since the initiation of the
in time that the welder stopped the arc, raised his welding weld. However, despite the many advantages of the DIC
helmet, and exchanged the filler rod. Through the time method, the lack of strain measurement close to the fusion
stamp of each image, it was possible to estimate this action zone is still an ongoing challenge. The loss of cross correla-
by the welder took just over 8 s. However, it is important tion is due to the light emission from the heat source (weld
to note the change in temperature over those 8 s did not arc) and the molten metal penetrating the backside of the
create a significant change in the thermographic/tempera- plate during the welding process.
ture distribution. Finally, Figs. 9 and 10 show a one-to-one comparison be-
Through the DIC results it could be observed that as the tween the experimental and VrWeld transversal and longitu-
arc is turned on, a transversal tensile strain field along the dinal strain data for the same time steps. The authors be-
sides of the weld pool is formed at the initiation of the weld. lieve that part of the challenge in why some of these compu-
This tensile field is not a constant and fluctuates back and tational weld models are not as accurate in capturing the
forward on either side of the weld pool as part of the weav- physics or at least having data that verifies these models is
ing process associated with a manual vertical weld. It is in- the lack of surface strain and temperature data in the prox-
teresting to observe in Fig. 5A that the tensile strain region imity of the weld pool. It has never been the intention of
along the sides of the weld pool is balanced by the formation this study to capture the maximum temperature point at the
of a compressive strain region in front of the weld pool center of the weld pool, as we believe that by capturing
whose magnitude is in the order of –9827 . This combina- 99.9% of the temperature field (below 400ºC within ±5%)
tion of tensile and compressive regions surrounding the provides us a greater insight into the transient dynamic
weld pool is indicative of a prying open action by the arc on weld behavior, thus, providing a wonderful opportunity that
the material surrounding the weld pool. This initial com- may lead to optimization of weld pool models. As such, the
pressive transversal strain continues to increase along the VrWeld model inputs were based on the following: a) geome-
weld interface, as indicated by the DIC images captured at try and material data of the plates being welded; b) informa-
1½ min prior to terminating the weld, where the compres- tion on the filler weld used; c) average speed of the different
sive transversal strain is larger in size and greater in magni- sections of the plates that were welded and location of the
tude (–26,341) as seen in Fig. 8A. filler rod exchanges; and, finally, d) estimates of the power
In contrast, the behavior of the longitudinal strain curves (voltage and current). This power curve estimate was
around the weld interface can be described as predominant- provided as a constant value. It is this information that al-
ly a tensile field at both the sides and in front of the weld lowed the VrWeld team to perform the coupled thermal me-
pool. The magnitude of the longitudinal strain reaches a chanical analysis that leads to the strain predictions shown
range of 3401 to 3811  as seen in Fig. 5B, which is approx- herein.
imately one-third of the compressive strain field associated The combination of experimental datasets with numeri-
with the transversal strain at the same location and time. It cal modelling capabilities, as shown in this study, demon-
is, however, important to note that DIC images are unable strates how the two scientific techniques complement each
to capture the strain field close to the edge of the weld pool other very nicely.
due to the light emission and the burning of the high- The VrWeld plots shown in Figs. 9 and 10 show how addi-
temperature paint. In addition, the longitudinal strain field tional work is still needed by the CWM community to better
surrounding the weld pool gradually decreases to a compres- understand the weld dynamics in the proximity of the weld
sive field in front of the weld pool reaching a magnitude pool. It is clear that as the comparison between the VrWeld
of –621 . As the weld continues, it is possible to observe approaches the weld pool the predicted strain values lose
that the tensile longitudinal strain field on either side of the agreement with the experimental findings. VrWeld, howev-
weld remains fairly constant, while the longitudinal strain in er, is able to nicely capture the strain fields far away (88
front of the weld pool doubles to approximately 6800  mm) from the weld pool. It is hypothesized that the forma-
during the duration of the 300-mm weld interface. The only tion of plastic strain, microstructural phase transformation,
time the longitudinal strain field seems to vary from tensile in addition to complex thermal dynamic effects in the prox-

260-s WELDING JOURNAL / SEPTEMBER 2019, VOL. 98

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 261

WELDING RESEARCH

imity of the weld pool, play a major role in the dynamic be- the metal filler rod, which resulted in a sudden increase in
havior of these strain fields and that further work is still re- the strain magnitude, most likely due to the cooling and so-
quired to produce better correlated results. lidification of the liquid weld pool while the arc is switched
In addition, the change in magnitude of the transverse off during the exchange, was also recorded. Further studies
strain between the start and finish of the vertical weld in and computational weld optimization models will be re-
the compressive region preceding the weld torch, is indica- quired to understand how cooling and plastic strain forma-
tive of the transient nature of the welding process. As such, tion affects this change in strain magnitude. Finally, the use
at least for this study and over a 30-cm plate, the weld never of the VrWeld model was validated and provided insight in
reached a steady-state behavior. It is important to note that regions where the experimental setup was unable to provide
the vertical boundary condition of the weld is believed to reliable data. The monitoring of the back of the weld and the
play a role in the formation of this change in magnitude, as use of digital image correlation in combination with ther-
the welder is constantly battling to maintain the weld pool mographic techniques, although not perfect, open an oppor-
through a series of weaving motions as he/she moves up tunity to examining the complex welding behavior of differ-
through the weld interface. In addition to this dynamic mo- ent alloys.
tion, the plastic strain formation in welds is usually limited
to regions about two to four times the width of the weld
Acknowledgments
pool. This plastic strain formation is dependent on tempera-
ture, boundary condition, and yield stress of the material to
name a few parameters. For example, if a small cylinder nor- The authors would like to acknowledge the support of
mal to a steel plate is heated from 20° to 150°C, it is possible Mark Hebel at the Clarkson University machine shop for
to obtain plastic strain formation depending on the yield helping us perform the experiments. Marharyta and Reyes
stress of the steel. However, if the same plate without the would like to acknowledge the financial support provided by
small cylinder is heated uniformly for the same temperature Ronald E. McNair Post-Baccalaureate Achievement Program
range, there is no plastic strain formation. As such, it is hy- and CIDESI, respectively.
pothesized that it is this formation of plastic strain due to
temperature distribution and material transformation that References
lead to this change in magnitude of the transverse fields,
thus affecting the overall residual stress development in
1. Zhou, X., Chen, H. B., Chen, J., Chen, S. B., and Feng, Z.
time. As such, the authors believe that measuring the tem- 2016. High temperature full-field strain measurement based on
perature field surrounding the weld pool is as important to digital image correlation during arc welding. 2016 IEEE Internation-
an understanding on the formation of residual stresses as is al Workshop on Advanced Robotics and its Social Impacts, pp.
measuring the highest peak temperature inside the weld 203–207. DOI: 10.1109/ARSO.2016.7736282.
pool itself. 2. Schajer, G. 1988. Measurement of non-uniform residual
The residual strain field is well known and common to stresses using the hole-drilling method. Part I — Stress calculation
the welding community, as they are aware that residual procedures. ASME. J. Eng. Mater. Technol. 110(4): 338–343. DOI:
stresses form and exist during the welding process. Howev- 10.1115/1.3226059
er, to the authors’ knowledge, there is a limited understand- 3. Schajer, G. 1988. Measurement of non-uniform residual
ing except through measurements such as neutron diffrac- stresses using the hole-drilling method. Part II — Practical applica-
tion of the integral method. ASME. J. Eng. Mater. Technol. 110(4):
tion, which require access to national laboratories capable of 344—349. DOI: 10.1115/1.3226060
performing such measurements. As such, the use of DIC has 4. Park, M. J., Yang, H. N., Jang, D. Y., Kim, J. S., and Jin, T. E.
been shown to be able to measure 2D residual strain fields. 2004. Residual stress measurement on welded specimen by neu-
This experimental technique is believed to provide confi- tron diffraction. Journal of Materials Processing Technology, Vols.
dence for the reported transient fields in this study. In re- 155–156, pp. 1171–1177. DOI: 10.1016/j.jmatprotec.2004.04.393
cent years, DIC has become much more available in academic 5. Eshtayeh, M., Hijazi, A., and Hrairi, M. 2015. Nondestructive
settings, opening the opportunity for faculty and students evaluation of welded joints using digital image correlation. Journal
to measure residual strain fields in situ. of Nondestructive Evaluation, pp. 34–37. DOI: 10.1007/s10921-015-
0310-z
6. Gollnow, C., and Kannengiesser, T. 2013. Hot cracking analy-
Conclusion sis using in situ digital image correlation technique. Welding in the
World 57(3): 277–284. DOI: 10.1007/s40194-013-0027-5
To summarize, this study has shown the complex tran- 7. Agarwal, G., Gao, H., Amirthalingam, M., and Hermans, M.
sient behavior of both tensile and compressive strain re- 2017. In situ strain investigation during laser welding using digital
gions leading to transversal and longitudinal strains in a image correlation and finite-element-based numerical simulation.
manual arc welding process monitored from the back of the Science and Technology of Welding and Joining 23(2): 134–139. DOI:
plate. This behavior was recorded by two nondestructive 10.1080/13621718.2017.1344373
8. Strycker, M. D., Lava, P., Paepegem, W. V., Schueremans, L.,
strain monitoring techniques: digital image correlation and and Debruyne, D. 2011. Measuring welding deformations with the
thermographic data. We captured an unforeseen significant digital image correlation technique. Welding Journal 90(6):
change in the magnitude of transversal strain (and a lesser 107–112.
change in the longitudinal direction) between the start and 9. Chen, J., and Feng, Z. 2018. Strain and distortion monitoring
finish of the vertical weld in the compressive region preced- during arc welding by 3D digital image correlation. Science and
ing the welding torch. The change in transversal and longi- Technology of Welding and Joining 23(6): 536–542. DOI:
tudinal strain distribution during the process of exchanging 10.1080/13621718.2017.1422634

SEPTEMBER 2019 / WELDING JOURNAL 261-s

Pliazhuk Supplement September 2019.qxp_Layout 1 8/8/19 4:46 PM Page 262

WELDING RESEARCH

10. Mrad, N. 2002. Optical fiber sensor technology. Introduction tations. (Part B), Acta Met. 36(11): 3037–3046. DOI: 10.1016/
and Evaluation and Application. New York, N.Y.: John Wiley & Sons, 0001-6160(88)90186-1
Inc. 27. Ashby, M. F., and Easterling, K. E. 1982. A first report on di-
11. RP Photonics Encyclopedia [Online]. September 6, 2018. agrams for grain growth in the welds. Acta Metall. 82(11):
rp-photonics.com/group_index.html. [Accessed 9/6/2018]. 1969–1978.
12. Kreger, S. T., Rahim, N. A., Garg, N., Klute, S. M., Metrey, D. 28. Goldak, J. A., and Akhlaghi, M. 2005. Computational Welding
R., Beaty, N., Jeans, J. W., and Gamber, R. 2016. Optical frequency Mechanics. Ottawa, Canada: Springer U.S. springer.com/ gp/book/
domain reflectometry: Principles and applications in fiber optic 9780387232874#aboutBook
sensing. Proceedings Volume 9852, Fiber Optic Sensors and Applica- 29. Karlsson, L., and Goldak, J. 2013. Computational Welding
tions XIII, Baltimore, Md. DOI: 10.1117/12.2229057 mechanics. Encyclopedia of Thermal Stresses, Springer. DOI:
13. Martinez Bueno, P., Martinez, M., Rans, C., and Benedictus, 10.1007/978-94-007-2739-7_437
R. 2016. Strain monitoring using a Rayleigh backscattering system 30. Goldak, J., Asadi, M., and Karlsson, L. 2011. Welding Funda-
for a composite UAV wing instrumeted with an embedded optical mental and Processes. ASM Press, pp. 819–829. DOI:
fiber. Advanced Materials Research, vol. 1135, pp. 1–19. DOI: 10.31399/asm.hb.v06a.a0005587
10.4028/www.scientific.net/AMR.1135.1 31. C. S. Inc., Correlated Solutions Inc., June 27, 2018. [On-
14. Ribeiro, F. N., Martinez, M., and Rans, C. 2018. Evaluation line]. Available: correlatedsolutions.com. [Accessed June 27, 2018].
of mode II fatigue disobnding using central cut plies specimen and 32. FLIR, June 27, 2018. [Online]. Available:
distributed strain sensing technology. The Journal of Adhesion. flir.com/products/flir-one-pro/. [Accessed June 27, 2018].
DOI: 10.1080/00218464.2018.1433039 33. Apple Inc., June 27, 2018. [Online]. Available: apple.com.
15. Matteï, S., Grevey, D., Alexandre, M., and Kirchner, L. 2009. [Accessed June 27, 2018].
Using infrared thermography in order to compare laser and hybrid 34. Duranty, E. R., Jacoby, S. A., and Davis Jr., J. H. 2018. Inver-
(laser+MIG) welding processes. Optics and Laser Technology 41(6): stigating process occurring in acoustically levitated ionic liquid
665–670. DOI: 10.1016/j.optlastec.2009.02.005 droplets. ECS Transactions 86(14): 297–308. DOI: 10.1149/
16. Goldak, J., Martinez, M., Techernov, S., Nimrouzi, H., Zhou, 08614.0297ecst
J., Aidun, D., and Eisazadeh, H. Correlating large sets of experi- 35. T. Inc., ThermoWorks, February 15, 2019. [Online]. Avail-
mental data with high resolution computational weld mechanics able: thermoworks.com/emissivity_table. [Accessed February 19,
models. Proceedings of the 12th International Seminar Numerical 2019].
Analysis of Weldability. Graz, Castle Seggau, Austria. 36. T. Inc., Transmetra Inc., February 15, 2019. [Online]. Avail-
17. Eisazadeh, H., Hamedi, M., and Halvaee, A. 2010. New para- able: transmetra.ch/images/transmetra_pdf/publikationen_literatur/
metric study of nugget size in resistance spot welding process us- pyrometrie-thermografie/emissivity_table.pdf. [Accessed February
ing finite element method. Materials and Design 31(1): 149–157. 15, 2019].
DOI: 10.1016/j.matdes.2009.06.042 37. Raabe, D. Dierk Raabe’s site. [Online]. Available: dierk-
18. Ohji, T., Matsutani, F., and Hirata, Y. A mathematical model raabe.com/digital-image-correlation-dic. [Accessed July 7, 2018].
for circumferential GTA welding of pipe. IIW Doc. 212, pp. 38. i. system, FLIR camera app.
892–896. 39. I. FLIR Systems, Atlas SDK. [Online]. Available: flir.cus-
19. Sudnik, W., Radaj, D., and Erofrew, W. Validation of comput- thelp.com/app/answers/detail/a_id/1043/related/1. [Accessed July
erized simulation of welding processes. Mathermatical Modelling of 23, 2018].
Weld Phenomena 4. 40. A. I. Technology. 2015. Manufacturing Coating Specifica-
20. Goldak, J. A., Chakravarti, A., and Bibby, M. J. 1984. A new tion, Advanced Integration Technology.
finite element model for welding heat sources. Trans. AIME, Vol. 41. V. company, VHT, Very high temperature painting. [Online].
15B, pp. 299–305. DOI: 10.1007/BF02667333 Available: vhtpaint.com. [Accessed July 7, 2018].
21. Chang, P.-H., and Teng, T.-L. 2004. Numerical and experi- 42. Bueno, P., Martinez, M. M., Rans, C., and Benedictus, R.
mental investigations on the residual stresses of butt-welded 2016. Strain monitoring using a Rayleigh backscattering system
joints. Computational Materials Science 29(4): 511–522. DOI: for a composite UAV wing instrumented with an embedded optical
10.1016/j.commatsci.2003.12.005 fiber. Advanced Materials Research Vol. 1135, pp. 1–19. DOI:
22. Eisazadeh, H., Achuthan, A., Goldak, J., and Aidun, D. 2015. 10.4028/www.scientific.net/AMR.1135.1
Effect of material properties and mechanical tensioning load on 43. LUNA Inc., June 27, 2018. [Online]. Available:
residual stress formation in GTA 304-A36 dissimilar weld. Journal lunainc.com/product/sensing-solutions/odisi/. [Accessed June 27,
of Materials Processing Technology Vol. 222, pp. 344–355. DOI: 2018].
10.1016/j.jmatprotec.2015.03.021 44. V. brand, M-Bond 200 product. [Online]. Available:
23. Zacharia, T. 1994. Dynamic stresses in weld metal hot crack- vishaypg.com/micro-measurements/list/product-11010/. [Accessed
ing. Welding Journal 73(7): 164-s to 172s. July 9, 2018].
24. Kirkaldy, J., and Venugoplan, D. 1983. Prediction of micro-
structure and hardenability in low alloy steels, phase transforma-
tion in ferrous alloys. Proceedings of International Conference, Oct.
4–6, 1983. M. PLIAZHUK, M. MARTINEZ ([email protected]), and D.
25. Watt, D. F., Coon, L., Bibby, M. J., Goldak, J. A., and Hen- K. AIDUN are with Clarkson University, Department of Mechan-
wood, C. 1988. Modelling microstructural development in weld ical and Aeronautical Engineering, Potsdam, N.Y. C. REYES is
heat-affected zones. (Part A). Acta Met. 36(11): 3029–3035. with the Center for Engineering and Industrial Development
26. Henwood, C., Bibby, M. J., Goldak, J. A., and Watt, D. F. (CIDESI), Querétaro, Qro., México. J. GOLDAK and H. NIMROUZI are
1988. Coupled transient heat transfer-microstructure weld compu- with Goldak Technologies Inc., Ottawa, Ontario, Canada.

262-s WELDING JOURNAL / SEPTEMBER 2019, VOL. 98