Modeling of Carbon Steel Duplex Stainless Steel GTA … of Carbon Steel Duplex Stainless Steel GTA Weld Pool A study investigates the spot gas tungsten arc (GTA) welding of dissimilar

WELDING JOURNAL / JULY 2014, VOL. 93262-s

WELDING RESEARCH

Introduction

Joining dissimilar alloys could sig-nificantly reduce the cost and weightof production without compromisingthe safety and structural requirements(Ref. 1). The difference in the thermo-physical, mechanical, and metallurgi-cal properties of the base metals, alongwith the solution and phaserestrictions, are the challenges in dis-similar welding. The fusion and heat-affected zones (HAZ) are asymmetricdue to differences in the melting tem-peratures, thermal diffusivities, andcompositions of the base metals, andaccordingly, the deepest penetration isusually located off the weld joint (or

interface) (Ref. 2); in other words, thecenterline shift (CLS) is not zero (Ref. 3). Experimental study of heat transferand fluid flow inside the arc and theweld pool (WP) during the weldingprocess is difficult due to the hightemperature of the arc and the moltenmetal, the small size of the WP, andshort welding time. Therefore, applica-tion of mathematical models and nu-merical simulations provide powerfultools for investigation of arc weldingprocesses. Fundamental study of thetransport phenomena involved in thearc welding process of dissimilar met-als/alloys is essential for prediction of

the geometry of the fusion and HAZ,dilution and mixing, cooling rate, andthermal stress residuals. Electromag-netic force (Lorentz force) and temper-ature-dependent surface tension anddensity (Marangoni and buoyancy ef-fects) are the effects that circulate thefluid in the WP (Ref. 4). These effectshave been numerically modeled in thewelding process of similar base metals.A two-dimensional axisymmetricmodel is most often used in numericalsimulation of spot welding of similarmetals (Ref. 5). Some of the modelssimulate the transport phenomena inthe workpiece separately from the arc(Refs. 6–8). In these models, the heatflux and current density from the arcis applied as external inputs at the sur-face of the workpiece. On the otherhand, some researchers modeled thearc and workpiece as a unified systemto better understand the interactionsbetween the arc and the workpiece(Refs. 5, 9–12). Gas tungsten arc welding 304Lstainless steel plates with differentsulfur concentrations was numericallyand experimentally studied by Mishraet al. (Ref. 3). The presence of sulfur inthe stainless steel was shown to affectthe heat transfer and fluid flow in thewelding process. The results indicated,when 304L stainless steels with differ-ent sulfur contents are welded to eachother, maximum penetration occursinside the low-sulfur stainless steel,not at the joint (CLS ≠ 0). The resultsshow flaring of the arc toward low-sulfur steel leads to shifting of thedeepest penetration from the joint. Itwas experimentally and numericallyfound that the sulfur mixes rapidly,and the gradient of the sulfur concen-

Modeling of Carbon SteelDuplex Stainless Steel GTA Weld Pool

A study investigates the spot gas tungsten arc (GTA) welding of dissimilar magnetic ferrous alloys 1018 carbon steel to 2205 duplex stainless steel

BY A. BAHRAMI AND D. K. AIDUN

ABSTRACT The fluid flow and heat transfer in dissimilar spot gas tungsten arc weld-ing (GTAW) of carbon steel to duplex stainless steel is numerically simulatedin this research study. A three-dimensional (3D) model, which considers thedifferences in the material properties of the base metals, and includes theeffects of the electromagnetic, buoyancy, and Marangoni convections, wasprepared by using the finite element software COMSOL Multiphysics® 4.3b.The history of the fluid flow in the weld pool (WP) is investigated by track-ing the travel path of individual fluid particles. An asymmetric weld profileis obtained in which the deepest penetration is shifted toward the stainlesssteel. The numerically calculated weld geometry reasonably matches the ex-perimentally obtained weld profile. To achieve symmetrical weld geometry,the arc location is shifted toward the steel side. It was observed that relocat-ing the arc moves the deepest penetration toward the weld joint, but it doesnot significantly increase the depth of penetration of the weld.

KEYWORDS• Dissimilar Welding • Gas Tungsten Arc Welding (GTAW) • Weld Pool• Fluid Flow • Heat Transfer • Weld Penetration • Weld Profile

A. BAHRAMI ([email protected]) is a PhD candidate and D. K. AIDUN is professor and chair of the Mechanical and Aeronautical Engineering (MAE) Dept.at Clarkson University, Potsdam, N.Y.

Final Bahrami Layout_Layout 1 6/13/14 10:36 AM Page 262

WELDING RESEARCH

tration at the top surface of the weld isinsignificant. However, it should bepointed out, in their case, the materialproperties of the base metals are thesame, and the only asymmetricvariable was the sulfur content in the304L base metals (BM). The transport phenomena in laserwelding dissimilar couples of Cu to Ni(Refs. 2, 13–15) were numericallystudied with the emphasis on mixing.All these studies show a fairly uniformdistribution of Ni in the WP. Mukherjee et al. (Ref. 16)investigated transport phenomena inthe laser welding of Fe-Al dissimilarcoupled with a layer of Ta as a diffusionbarrier. The results showed a greaterdegree of melting happened in the Feside. Consequently, an asymmetric WPwith asymmetric velocity and tempera-ture fields was created. The heat and mass transfer in laserwelding of 304 stainless steel to Ni, adissimilar couple, was studied by Hu etal. (Ref. 17). It was observed that thedistribution of the element Fe in theWP became uniform after 90 ms(0.090 s). In the absence of magneto-hydrodynamics (MHD) in the laserwelding of dissimilar alloys, the fluidis basically driven by the Marangoniand buoyancy in the WP. As a result,the velocity at the top surface of theweld is significantly higher than thatinside the WP. However, in dissimilararc welding, the Lorentz forceincreases the velocity of the meltinside the WP. A higher velocity insidethe WP leads to a more uniform mix-ing (especially along the depth), result-ing in a more homogenous

composition, which is vital to achievegood weld quality. In a previous study,the authors developed a numericalmodel to investigate transportphenomena in spot GTAW of low-carbon steel to duplex stainless steel(Ref. 18). The distribution of alloyingelements across the weld wasdetermined using EDS analysis.Accordingly, the degree of dilution ofeach element in the weld zone was cal-culated. The results showed themixing was quite uniform across theweld zone. The weld geometry was nu-merically simulated, and a favorableagreement between the numerical andexperimental results was obtained.Both numerical and experimental re-sults show the deepest penetrationdoes not happen at the joint but insidethe stainless steel side (CLS ≠ 0). In this paper, a more detailed 3Dsimulation of the fluid andtemperature fields in spot GTAW steelto duplex stainless steel is presented.The effects of Marangoni convectionand Lorentz force on the fluid motionin the WP are presented. The historyof the time-dependent fluid motion inthe weld is demonstrated by trackingseveral fluid particles in the WP usingthe “pathlines” (Ref. 19). To study theeffect of the arc location on the asym-metric geometry of the WP, the simu-lations are repeated for different arclocations, and geometrical features ofthe WP were investigated.

Mathematical Model of the WP The difference in the properties ofthe weldments leads to asymmetricelectromagnetic-induced flow and

temperature fields; therefore, a 3Dmodel is required to simulate thetransport phenomena in dissimilarwelding processes. To mathematicallymodel the spot GTAW process in thisstudy, the following assumptions havebeen made: • The fluid flow in the WP islaminar. • Gaussian distributions of the heatand current density inputs are appliedat the top surface. • The Boussinesq approximation isused to take into account thebuoyancy convection. • The molten materials mix rapidly;therefore, within the bulk of the WP,all of the alloying elements including Sare uniformly distributed, and thusconcentration gradient will be zero. • The effects of surface deformationare neglected. Based on these assumptions, acomputational model was developed,which includes electromagnetics,solid-liquid phase change, fluid flow,and heat transfer. The classicMaxwell’s equations are used toresolve the electromagnetic field andthe Lorentz force, then the flow andtemperature fields are analyzed by ap-plying the conservation equations ofmass, momentum, and energy. Themathematical formulations of thesephysical phenomena and the boundaryconditions are described in detail inthe literature (Ref. 18). One of the complexities of the nu-merical simulation of the dissimilarwelding process is handling the mate-rial properties of the solution in theWP and mixing of diluted species. Toaccurately calculate the material prop-

Fig. 1 — Configuration of the weldment and coordinate system.

Fig. 2 — Lorentz force density field in the weld pool. Carbon steel isto the left and stainless steel is to the right.

JULY 2014 / WELDING JOURNAL 263-s



WELDING RESEARCH

erties in dissimilar welding, the trans-port equations for all diluted speciesshould be solved fully coupled witheach other and with other governingequations that were mentioned before.In the case of welding carbon steel toduplex stainless steel, as featured inthe present study, there are a total of12 alloying elements in the weldment(see Table 1); therefore, 12 transportequations are added to the model. As mentioned in the introduction,Hu et al. (Ref. 17) studied laser weld-ing Ni to 304SS and showed that thealloying elements uniformly mix after0.090 s. In the current study the arctime is 15 s, which is large enough toneglect such an initiation time. Also,Mishra et al. (Ref. 3), in the study ofwelding stainless steels with differentsulfur contents, showed that a recircu-lation in the WP takes about 0.035 s,which is about 0.08% of the 4 s totaltime that the metal is in the liquidphase. Based on these observations, inthis study, it is assumed that the meltmixes rapidly to produce a uniformmaterial throughout the WP. This as-sumption eliminates 12 partial differ-ential equations and significantly sim-plifies the model. The material properties of the solu-

tion are functions of the properties ofthe base metals and the mass fractionof each metal in the mixture (Ref. 20).To determine the material propertiesof the solution in the WP based on theassumption of uniform mixing, thefollowing equation is used:

Aw = DBM1ABM1 + DBM2ABM2 (1)

In this equation, subscripts w, BM1,and BM2 denote the weld, base metal1, and base metal 2, respectively; A isthe material property, which could bereplaced by the density ρ; thermal ex-pansion factor β; thermal conductivityk; heat capacity Cp; viscosity; and elec-trical conductivity σ and D is thedegree of dilution. The dilution, D, canbe calculated from metallographicmethods as well as numerical methods(Ref. 18).

Numerical Method

The finite element software COM-SOL Multiphysics (v.4.3b) is used tonumerically simulate the transportphenomena in the welding process ofdissimilar alloys. This software is capa-ble of resolving the models, which in-clude several physical phenomena, andtakes into account the interactions

among different physics. To calculate the Lorentz force in theWP, electric currents and magneticfield modules are applied. A Gaussiandistribution of the current density atthe top surface of the weldment simu-lates the presence of the arc, and thebottom surface is grounded. The mod-ules heat transfer in fluids with phasechange and laminar flow are employedto resolve the flow and temperaturefields. The Lorentz force, which waspreviously calculated, along with thebuoyancy effects, is applied to the lam-inar flow module as a volume force.The Marangoni effect is imposed tothe model as a surface stress boundarycondition. A smoothed Heaviside func-tion is applied to smooth sharpchanges in the material properties atthe joint. Applying this function con-siderably improves the convergenceand quality of the results. The degree of polynomials in the fi-nite element discretization of all of thepartial differential equations is set toquadratic. An unstructured tetrahedralmeshing is applied to all of thedomains — Fig. 1. The mesh size isvery fine with maximum element sizeof 0.08 mm in the part of the domainthat includes the WP. The mesh is rela-

Table 1 — Composition of 1018 Steel and 2205 Duplex Stainless Steel (wt%)

Fe Cr Ni Mo C Mn Si S P Co Cu N

1018 Bal. — — — 0.18 0.75 0.22 0.03 0.03 — — —2205 Bal. 22.51 5.38 3.12 0.021 1.76 0.50 0.002 0.028 0.20 0.43 0.170

Fig. 3 — Marangoniinduced shear stresses and temperaturecontours (K) on the weld pool surface. Fig. 4 — Velocity field in the weld pool at 15 s of arc time.


WELDING RESEARCH

tively coarser at the remaining parts ofthe model with maximum element sizeof 0.4 mm. The total number ofelements is 198821. A fully coupled,time-dependent solver with maximumtime step of 0.1 s is employed toresolve the system of partial differen-tial equations. An HP® Z220 worksta-tion with a CPU frequency of 3.4 GHzand a memory of 32 GB is used to re-solve the models. Each simulationtook approximately 115 h to perform.

Results and Discussion

The magnetic ferrous Alloys 1018steel and 2205 duplex stainless steelare chosen as base metals for thisstudy (see Table 1). Coupons of steeland stainless steel with a size of 26 ×13 × 6 mm are to be joined to form 26× 26 × 6 mm welded samples, asshown in Fig. 1. The origin of the coor-dinate system is located at the centerof the top surface of the weldment.The spot GTA weld is simulated for anarc current of 150 A, voltage of 14 V,and an arc time of 15 s. An arcefficiency of 75% is used in thenumerical simulation.

Solution Properties

The determination of the degree ofdilution is required to calculate the ma-terial properties of the solution. AnEDS analysis was carried out at differ-ent locations across the WZ (Ref. 18)to measure the concentration of alloy-ing elements. The results showed a

fairly uniform distribution of theelements in the WZ, which confirmsthe corresponding assumption in thisstudy. The following equation was usedto determine the degree of dilution foreach alloying element (Ref. 20):

where Cifz, C

ibm,1, and Ci

bm,2 are theconcentration of ith element in the fu-sion zone, BM 1 and BM 2,respectively. The dilutions of the ele-ments, from the steel side to the WP,vary from 0.27 to 0.33. Accordingly, anaverage value of 0.30 is used as the di-lution in Equation 1 to calculate thesolution properties in the WP. Theproperties of the base metals are pre-sented in Table 2.

Flow and Temperature Fields

As the first step of the simulation,the electromagnetic field in the metals

was calculated. Figure 2 presents a 3Dview of the Lorentz force density fieldin the WP. A color scale is specified tothe arrows in Fig. 2 to indicate theirmagnitude. The values of electricalconductivity and magnetic permeabil-ity of the carbon steel are higher thanthose of the duplex stainless steel. Ac-cordingly, the Lorentz force field ismuch stronger in the steel side. Themaximum value of the Lorentz forcedensity in the steel side is 7.6 × 104

N/m3, which is almost three timeslarger than that in the stainless steelside with a value of 2.4 × 104 N/m3. Beside the Lorentz force, Marangonieffect plays a major role in circulationof the fluid in the WP. This effect actsat the free surface of the WP as a shearstress, for example (Ref. 3):

C DC ( D) Cfzi

bm,i

bm,i1 21 2 )(= + −

uz

ddT

Tx

ddC

Cx,

zddT

Ty

ddC

Cy

xz z

yz z

z

z(3)

=0

=0

=0

=0

τ = −μ∂∂

=γ ∂

∂+

γ ∂∂

τ = −μ∂ν∂

=γ ∂

∂+

γ ∂∂

Fig. 5 — Temperature isosurfaces and heat flux vectors in theweld pool.

Fig. 6 — Pathline of a fluid particle initially located at (0.5, 0.5,–0.5) with final location at (–1.8, 0.05, –0.72).

Table 2 — Thermophysical Properties Used in the Numerical Simulation (Ref. 24)

Property name 1018 2205

Liquidus temperature 1802 K 1680 KSolidus temperature 1770 K 1633 KHeat of fusion 240 kJ/kg 260 kJ/kg Solid specific heat 750 J/kg.K 720 J/kg.KLiquid specific heat 840 J/kg.K 830 J/kg.KSolid thermal conductivity 39.4 W/m.K 32.0 W/m.KLiquid thermal conductivity 36.5 W/m.K 29.0 W/m.KSolid density 7530 kg/m3 7250 kg/m3

Liquid density 7150 kg/m3 6850 kg/m3

Volume thermal expansion of liquid 1.20 × 104 /K 1.12 × 104/KDynamic viscosity 6.5 × 103 kg/m.s 7 × 103 kg/m.sElectrical conductivity 7 × 106 1/Ωm 1.5 × 106 1/Ωm




WELDING RESEARCH

where τxz and τyz are the shearstresses at the surface; u and v are thevelocity components in the x and y di-rections, respectively; and μ is the vis-cosity. T and C are the temperature,the concentration of the solute, anddγ/dT and dγ/dC are the temperatureand concentration derivatives of thesurface tension, respectively. The sur-face tension of the solution in the WPis a function of temperature and com-position of the surfactant elementssuch as sulfur (S) (Ref. 21). This func-tion is thermodynamically analyzedand derived for different binarysystems. For an Fe-S binary system,surface tension equation is as follows:

In this equation, γ is the surfacetension; Tl is the melting temperature(liquidus temperature); R is the gasconstant (8314.3 J/kmol.K); and as isthe sulfur activity, which can beapproximated as sulfur concentrationin the WP. The surface tension equation avail-able for Fe-S systems is proved to beapplicable for approximation of thechanges in the surface tension of fer-rous alloys (Ref. 22). In this study, it isassumed the sulfur concentration gradient at the top surface of the WP is negligible. Therefore, ∂C/∂x =∂C/∂y = 0. Accordingly, the second terms onthe right-hand sides of Equation 3 goto zero. This assumption is valid for a15-s spot GTA weld, since it will takeonly a fraction of a second to obtainuniform mixing especially at the topsurface of the WP (Ref. 17). The sulfur concentration in the1018 steel and 2205 stainless steel are0.03 and 0.002 wt-%, respectively.Based on the measured dilution of30% for the steel, the sulfurconcentration in the WP is predictedto be 0.01 wt-%. This value is appliedin Equation 4 to evaluate the surfacetension of the mixture at differenttemperatures. The temperature deriv-ative of the surface tension iscalculated by differentiation of Equa-tion 4 with respect to T. Figure 3 shows the distribution ofthe shear stresses along with the tem-

perature contours atthe top surface of theWP at 15 s of arc time.The temperature deriv-ative of the surfacetension is positive forthe temperaturesunder 1860 K;therefore, the shearstress, or Marangonistress, is negative atthe vicinity of theboundary of the WP(arrows point towardthe center). The 1860K isotherm could beconsidered as the zeroshear stress contours.The maximum temper-ature happens at thestainless side, and thegradient of the temper-ature in the x and y di-rections (∂T/∂x and∂T/∂y) are larger onthis side. As a result,the maximum value ofthe shear stress occursat the stainless steelside. The velocity field inthe WP at 15 s is pre-sented in Fig. 4. The iso-surfaces showthe magnitude of the velocity at differ-ent locations in the WP, and thearrows represent the direction of thevelocity. The velocity field in the WP isdetermined by the opposing drivingforces, Marangoni and Lorentz. Thedirection of the flow at the top surfaceis from the center toward the bound-aries. A stronger Marangoni shearfield and a weaker Lorentz force fieldin the stainless steel side lead tohigher velocities at the top surface ofthe stainless steel side. Inside the WP,the stronger Lorentz force at the steelside pushes the melt from the steelside toward the stainless steel side. Itcould be observed the velocity insidethe WP (excluding the top surface) isalways less than 0.1 m/s while at thetop surface the maximum velocity isgreater than 0.14 m/s. This meansthat the Marangoni effect is strongeron the surface than theelectromagnetic effect in the WP. In Fig. 5, the iso-surfaces show thevalues of temperature at different lo-cations in the WP, and the arrows rep-resent the total heat flux field (convec-

tive plus conductive). The maximumtemperature happens in the stainlesssteel side. The heat transfer field issimilar to the velocity field (regardlessof the magnitudes), which means thatthe convection is the dominant heattransfer in the WP. This fact isconfirmed by calculation of Pecletnumber in the WP. This dimensionlessnumber represents the ratio of convec-tion to conduction (Ref. 23), for example:

where U and L are the scale velocityand length, respectively, and k, ρ, andCp represent the thermal conductivity,density, and specific heat, respectively.If the average velocity in the WP (0.05m/s) is used for U, and the WP radius(0.004 m) is used for L, the calculatedPeclet number, Pe, will be around 32.This value confirms that theconvective heat transfer rate is muchlarger than the conductive heat trans-fer rate in the WP.

PeULk

cp

(5)=

ρ

(T T ) RT

ln ( a expRT

)

l

s

1.943–4.3 10 1.3 10

1 0.00318–1.66 10

(4)

-4 –8

8

γ = × − − ×

× +×

Fig. 7 — Initial and final locations of 99 particles in the weldpool. Steel is on the top and stainless steel is on the bottom.

A

B


WELDING RESEARCH

Motion of Liquid in the WP

To investigate the history of thefluid flow in a time-dependent system,trajectories of individual fluidparticles (very small parts of the fluid)could be tracked. These trajectories arecalled “pathlines” in fluid mechanicsterminology (Ref. 19). Figure 6 showsthe path of the particle, which isinitially located at (0.5, 0.5, –0.5), inthe stainless side (see Fig. 1 for the co-ordinate system configuration). Theparticle is incorporated into the WP1.06 s after arc initiation. As seen inFig. 6, the particle moves around andwithin the WP several times and cov-ers a relatively large distance. Eventu-ally, at 15 s, the particle ends up at (–1.8, 0.05, –0.72) in the steel side.The length of the path traversed bythe particle is 212 mm, although thefinal location is just 2.4 mm away fromthe initial location. The path length is85 times larger than the depth and 26times larger than the length of the WP. A color scale is specified to thepathline in Fig. 6, which shows themagnitude of velocities that the parti-cle experienced at different locations.The circulations near the surface ofthe WP are created by the Marangonieffect (as indicated in the figure). Ascan be seen, as a result of theMarangoni shear, the particle acceler-ates when it travels on the surface ofthe weld. The maximum velocity ofthe particle is 0.19 m/s, whichhappens at 1.7 s on the surface. As in-dicated in the figure, at center of theWP, the circulation of the particle ispredominantly driven by the Lorentzforce. The average velocity of this par-ticular particle is 0.015 m/s after13.94 s of travel.

Mixing in the WP

To understand mixing duringdissimilar weld processes, the travel ofthe particles that are initially arrangedin a 3D, uniform (cubic) lattice with aseparation distance of 0.5 mm in alldirections (see Fig. 7A) is studied. Thevolume of the WP (45.7 mm3), whichis developed after 15 s of arc time, canonly accommodate a total of 99 parti-cles (the particles located at the inter-faces are not included). Out of the 99particles, 28 (blue colors) are locatedin the steel side and 71 (red colors) are

located in the stainless steel side basedon the volume. As the welding processgoes on, the WP develops and the par-ticles gradually incorporate into themelt. Figure 7B shows the finallocations of the particles. Thepathlines of the particles generally

have similar helical structures, asshown in Fig. 6; therefore, for the sakeof visibility, the pathlines are not dis-played in Fig. 7B. Out of 28 particlesthat were initially located in the steelside, 9 particles end up in the steelside and 19 in the stainless steel side,

Fig. 8 — Weld pool geometry. A — Crosssectional view; B — joint view; C — top view; D — sectioned 3D view.

Fig. 9 — Weld geometry. A — Cross section (xzview); B — half of the joint view (yzview);C — top view (xyview).


A

A

B

B

C

C

D


WELDING JOURNAL / JULY 2014 VOL. 93268-s

WELDING RESEARCH

and out of 71 particles initially locatedin the stainless steel side, 21 particlesend up in the steel side and 50 end upin the stainless side. In other words,almost 32% of steel particles (blue)and 29% of stainless steel particles(red) are ultimately located in the steel side. The pathline length calculationsshows an average travel distance of120.3 mm for the 28 steel particles(particles initially located at the steelside), and an average travel distance of110.9 mm for the 71 stainless steelparticles (particles initially located atthe stainless steel side). The relativelylong path covered by the particles en-sures that a uniform mixing takesplace in the WP.

Weld Geometry

Figure 8 shows the WP geometry atdifferent views. Due to the highermelting temperature and thermal dif-fusivity, the size of the WP is smallerin the steel side. Accordingly, the deep-est penetration happens in the stain-less steel side (CLS = 1.21 mm). Thewhole WP has a volume of 45.7 mm3

of which 13.1 mm3 is into the steelside. The ratio of the weld volume inthe steel side to the whole volume is28.7%, which is in good agreementwith the degree of dilution previouslycalculated based on the composition ofthe weld (Ref. 18). The depth of pene-tration at the weld joint (x = 0) is 1.9mm, and the weld joint area is 8.1mm2 — Fig. 8B. To validate the geometry obtainedfrom the modeling, actual spot GTAWof 1018 steel to 2205 stainless steel isperformed with the same welding pa-rameters used in the simulation. Thewelded sample was sectioned at thecenter of the weld zone and perpendi-cular to the joint to obtain the cross-sectional view. One of the slices of thesample was sectioned again along thejoint to show the joint view — Fig. 9B.Figure 9A shows the cross section ofthe weld zone. To reveal the weldboundary without overetching thesample, different etchants wereapplied to different alloys. Nital wasswabbed on the steel side, and Glyceriawas swabbed on the stainless steelside. Figure 9B and C show the jointand the top views of the welded sam-ple, respectively. A general comparison

Fig. 10 — Weld profile for different arc locations (xzplane). Steel is on the left and stainless steel is on the right.

Table 3 — Geometric Features of the Weld Pool for Different Arc Locations

Arc Location on Depth at Joint Max. Depth CLS Joint Area Dilution of XAxis (mm) (mm) (mm) (mm) (mm2) Steel in the WP*

0 1.90 2.47 1.21 8.10 0.29–0.5 1.95 2.33 0.95 8.28 0.34–1.0 2.01 2.20 0.46 8.35 0.46–1.5 1.85 2.02 0.22 7.70 0.57–2.0 1.61 1.68 –2.12 6.48 0.71

*Based on the volume of the WP.

Fig. 11 — Top view of the weld pool for different arc locations (xyplane). Steel is on theleft and stainless steel is on the right.

A

A

B

B

C

C

D

D


WELDING RESEARCH

of Figs. 8 and 9 shows the model cap-tures the essential geometricalfeatures of the weld zone. Achieving a symmetric weld zonewith the deepest part at the joint isusually one of the factors of a goodquality weld. As previously seen in thisstudy, the deepest part of the WP is lo-cated in the stainless side rather thanbeing located at the joint (x = 0). Thiseffect is called a missed joint in thewelding terminology (Ref. 18). To achieve symmetry, a preliminarysuggestion would be to relocate the arcfrom the joint toward the alloy withthe higher melting temperature andthermal conductivity, which is the car-bon steel in the case of present study.In this section, the effect of thelocation of the arc on the geometricfeatures of the weld is investigated.The welding process for four differentlocations of the arc — 0.5, 1.0, 1.5,and 2.0 mm away from the jointtoward the carbon steel side — are nu-merically simulated. In these models,the center of the heat source and cur-rent density input is located at x = –0.5, –1, –1.5, and –2 mm. The arccurrent in all of the simulations is 150A, and the arc time is 15 s. Figure 10 shows the cross-sectionalprofiles, and Fig. 11 shows the topview profiles of the WP for theselected arc locations. As can be seen,as the arc is located farther on thesteel side, the WP shifts toward thesteel. Although the volume of the WPin the steel side increases, and that inthe stainless steel side decreases, asymmetric WP geometry is neverachieved. The geometrical features ofthe WP in different cases are listed inTable 3. It is observed the depth of weldpenetration and the weld joint area atthe weld interface (x = 0) increasesuntil the arc position of x = –1.0 mm,and then it decreases. The quality ofthe weldment depends on the weldpenetration and joint area at the weldinterface (x = 0). The depth of penetra-tion at the joint is maximum with avalue of 2.01 mm when the arc islocated at x = –1.0 mm. The weld jointarea also has its maximum value whenthe arc is located at x = –1.0. For thecase of the arc located at x = –0.5, –1.0, and –1.5 mm, the maximumdepth occurs at the stainless steel sideclose to the joint, and for the case of x

= –2.0 mm at the steel side. Therefore,the CLS approaches zero until x = –1.5mm. The CLS has a negative value forthe case of x = –2 mm, which meansthe CLS has shifted toward the steelside. From Table 3, it shows the opti-mum arc location is at x = –1.0 mm be-cause it will produce the maximumpenetration (2.01 mm) at the weldjoint, and has the maximum weld jointarea (8.35 mm2) with a steel/stainlesssteel dilution of 46/54%. According tothese results, it can be concluded thatthe best weld aspect ratio to achievethe best weld quality is obtained whenthe arc is located at x = –1.0 mm. Bycomparing the values of the joint pen-etration and joint area in Table 3, it isseen that by relocating the arc fromx = 0 to x = –1.0 mm, the penetrationat the joint increases by around 5%and the joint area increases by around2%. It is also observed the weld qualitycannot be judged solely based on thetop view of the weld zone (Fig. 11) be-cause a larger top view area does notalways guarantee a larger weldpenetration and joint area at the weldinterface (x = 0) especially in dissimi-lar welds.

Conclusion

The spot GTAW of dissimilar mag-netic ferrous alloys, 1018 carbon steelto 2205 duplex stainless steel, was in-vestigated. The velocity and tempera-ture fields in the WP were numericallymodeled. The calculated weld profile,obtained from numerical study,showed a good resemblance with thatof the experiment. The accomplish-ments of this study are summarized asfollows: • Due to different thermal physicalproperties, especially the melting tem-perature and thermal conductivity, theWP is asymmetric. A larger portion ofthe volume of the WP is located in thestainless steel side, which has thelower melting temperature. The deep-est penetration does not occur at thejoint but in the stainless steel side. • It was seen that higher electricalconductivity of the steel leads to anasymmetric Lorentz force density dis-tribution in the WP. The direction ofthe Lorentz force in either side of theWP is always toward the center of theWP, but the magnitude of this force ismuch higher in the steel side.

• The travel path of several fluidparticles in the WP during the weldingprocess was investigated. Thepathlines show the fluid generally ex-periences two types of helical motion,one at the surface under the effect ofMarangoni stresses, and one at thecenter of the WP under the effect ofLorentz force. The study of initial andfinal locations of the particles andlength of the travel pathlines ensure auniform mixing in the melt. It couldbe concluded that up to the onset ofsolidification the mixture is uniformand homogeneous. • The spot GTAW simulation wasrepeated for four arc locations alongthe x-axis toward the steel side (withhigher melting temperature) to obtaina weld profile closer to symmetry. It isobserved a perfectly symmetric profileis never achieved by moving the arc to-ward the steel. As the arc is shifted to-ward the steel, the joint area and pene-tration slightly increase, and afterreaching an optimum location, theystart decreasing. In this particularstudy, the optimum arc location is cal-culated to be 1.0 mm away from thejoint in the steel side.

1. Lippold, J. C., and Kotecki, D. J.2005. Welding Metallurgy and Weldability ofStainless Steels. New Jersey: Wiley. 2. Chakraborty, N., and Chakraborty, S.2007. Modelling of turbulent molten poolconvection in laser welding of a copper-nickel dissimilar couple. International Jour-nal of Heat and Mass Transfer 50: 1805–1822. 3. Mishra, S., Lienert, T. J., Johnson, M.Q., and DebRoy, T. 2008. An experimentaland theoretical study of gas tungsten arcwelding of stainless steel plates with differ-ent sulfur concentrations. Acta Materialia56: 2133–2146. 4. Kou, S. 2003. Welding Metallurgy.New Jersey: Wiley. 5. Traidia, A., Roger, F., and Guyot, E.Optimal parameters for pulsed gastungsten arc welding in partially and fullypenetrated weld pools. International Jour-nal of Thermal Sciences 49: 1197–1208, 10. 6. Zacharia, T., Eraslan, A., Aidun, D.,and David, S. 1989. Three-dimensionaltransient model for arc welding process.Metallurgical and Materials Transactions B20B: 645–659. 7. Fan, H., Tsai, H., and Na, S. 2000.Heat transfer and fluid flow in a partially

References



WELDING RESEARCH

or fully penetrated weld pool in gas tung-sten arc welding. International Journal ofHeat and Mass Transfer 44: 417–428. 8. Zhang, W., Roy, G. G., Elmer, J. W.,and DebRoy, T. 2003. Modeling of heattransfer and fluid flow during gas tungstenarc spot welding of low carbon steel. Jour-nal of Applied Physics 93: 3022–3033. 9. Tanaka, M., and Lowke, J. J. 2007.Predictions of weld pool profiles usingplasma physics. Journal of Physics D:Applied Physics 40: R1–R23. 10. Lu, F., Tang, X., Yu, H., and Yao, S.2006. Numerical simulation on interactionbetween TIG welding arc and weld pool.Computational Materials Science 35: 458–465. 11. Lu, F., Yao, S., Lou, S., and Li, Y.2004. Modeling and finite element analysison GTAW arc and weld pool. ComputationalMaterials Science 29: 371–378. 12. Kim, W., and Na, S. 1998. Heat andfluid flow in pulsed current GTA weld pool.International Journal of Heat and MassTransfer 41: 3213–3227. 13. Wei, P. S., and Chung, F. K. 2000. Un-steady Marangoni flow in a molten poolwhen welding dissimilar metals. Metallurgi-

cal and Materials Transactions B 31: 1387–1403. 14. Phanikumar, G., Chattopadhyay, K.,and Dutta, P. 2001. Modelling of transportphenomena in laser welding of dissimilarmetals. Int. Journal of Numerical Methodsfor Heat Fluid Flow 11: 156–171. 15. Chakraborty, N. 2009. The effectsof turbulence on molten pool transportduring melting and solidification processesin continuous conduction mode laser weld-ing of copper-nickel dissimilar couple. Ap-plied Thermal Engineering 29: 3618–3631. 16. Mukherjee, S., Chakraborty, S.,Galun, R., Estrin, Y., and Manna, I. 2010.Transport phenomena in conduction modelaser beam welding of Fe-Al dissimilar cou-ple with Ta diffusion barrier. InternationalJournal of Heat and Mass Transfer 53:5274–5282. 17. Hu, Y., He, X., Yu, G., Ge, Z., Zheng,C., and Ning, W. 2012. Heat and masstransfer in laser dissimilar welding ofstainless steel and nickel. Applied SurfaceScience 258: 5914–5922. 18. Bahrami, A., and Aidun, D. K. 2013.Modeling of transport phenomena in dis-similar welding of 2205 duplex stainless

steel to 1018 carbon steel. Trends in Weld-ing Research 2012: Proc. of the 9th Int. Con-ference. ASM International, Chicago, Ill. 19. White, F. M. 1994. Fluid Mechanics,3th ed. McGraw-Hill. 20. DuPont, J., Lippold, J., and Kiser, S.2011. Welding Metallurgy and Weldability ofNickel-Base Alloys. Wiley. 21. Sahoo, P., DebRoy, T., andMcNallan, M. 1988. Surface tension of bi-nary metal — surface active solute systemsunder conditions relevant to welding met-allurgy. Metallurgical and Materials Transac-tions B 19: 483–491. 22. McNallan, M., and DebRoy, T. 1991.Effect of temperature and composition onsurface tension in Fe-Ni-Cr alloys contain-ing sulfur. Metallurgical and MaterialsTransactions B 22: 557–560. 23. Incropera, F. P., DeWitt, D. P.,Bergman, T. S., and Lavine, A. S. 2006.Fundamentals of Heat and Mass Transfer,6th ed. Wiley. 24. Mills, K. C. 2002. Recommended Val-ues of Thermophysical Properties for SelectedCommercial Alloys. Cambridge, England:Woodhead Publishing.

WELDING JOURNAL / JULY 2014 VOL. 93270-s

CAN WE TALK? The Welding Journal staff encourages an exchange of ideas with you, our readers. If you’d like to ask a question,share an idea, or voice an opinion, you can call, write, e-mail or fax. Staff e-mail addresses are listed below, along witha guide to help you interact with the right person.

Publisher Andrew Cullison [email protected], Extension 249 Article Submissions

Editor Mary Ruth Johnsen [email protected], Extension 238 Feature Articles

Associate Editor Howard Woodward [email protected], Extension 244 Society News, Personnel

Associate Editor Kristin Campbell [email protected], Extension 257 New Products News of the Industry

Editorial Assistant/Peer Review Coordinator Melissa Gomez [email protected], Extension 475 Peer Review, International News, and Product Literature

Managing Editor Zaida Chavez [email protected], Extension 265 Design and Production

Senior Production Coordinator Brenda Flores [email protected], Extension 330 Production

Advertising Sales Director Rob Saltzstein [email protected], Extension 243 Advertising Sales

Advertising Sales & Promotion Coordinator Lea Paneca [email protected], Extension 220 Production and Promotion

Advertising Sales Representative Sandra Jorgensen [email protected], Extension 254 Advertising Sales

Advertising Production Manager Frank Wilson [email protected], Extension 465 Advertising Production

Welding Journal Dept. 8669 NW 36th St. #130 Miami, FL 33166 (800) 443-9353 FAX (305) 443-7404


Modeling of Carbon Steel Duplex Stainless Steel GTA … of Carbon Steel Duplex Stainless Steel GTA Weld Pool A study investigates the spot gas tungsten arc (GTA) welding of dissimilar

Documents