Invernational ural ef Heat ané Mass Transfer 127 (2018) 1084-1054 Contents lists available at ScienceDirect International Journal of Heat and Mass Transfer ELSEVIER journal homepage: www. elsevier.com/locate/ijhmt Fusion zone geometries, cooling rates and solidification parameters during wire arc additive manufacturing ans W, Ou*®, T, Mukherjee’, G.L. Knapp’, Y. Wei, T. DebRoy ** ® Department of Material Scene and Techno, Najing Unive of Acton: Aonau fang 210076 Choa ARTICLE INFO ABSTRACT Receive i ete form 17 Jay 2018 ‘Suuetute properties and seviceability of components mae by wite arc addiive manataccuring (WAAM) 4epend on the process parameters such 36 ate power, avel speed, wie diameter and wie fed rate However, the selection of appropriate processing conditions to fabricate defect free and structurally sound components by trial and eror is expensive and time consuming. Here we develop test and utilize 4 three-dimensional hea transfer an fad flow model of WAAM to calculate temperature snd velocity elas, deposit shape and size, cooling rates and soliification parameters. The calculated fusion zone geometries and cooling rates for various are power and travel speed and thermal cycles considering con- ‘ective lw of molten metal agreed well with the corresgoncing experimental cata for HI tool steel ‘depots. t was found that convection isthe main mechanism of heat transfer inside the molten pool. Faster travel speed enhanced the cooling rat but reduced the ratio of temperature gradient osolifca- sion growth rat indicating increase instability of plane front solidification of components. Higher depo- sition rates could be achieved by increasing the het input sing thicker wices and rapid wire feeing 15 2018 Elsevier Ll All right reserve, Keywords: ire a adive manufacunng Soliteaten parameters 1. Introduction Wire arc additive manufacturing (WAM) is developed from arc ‘welding and suitable for making large components because of high deposition rates, low equipment and feedstock costs [1], WAAM involves melting of wire by the are, transfer of molten metal dro- plets to a molten pool, convective Now of liquid metal inside the ‘molten pool driven by surface tension gradient [2-4], deformation ofthe molten poo! surface by are pressure and solidification of the molten poo! [2,3], These physical phenomena govern the tempera ture and velocity distributions, deposit shape and size, ané the structure and properties of the components. In addition, transient and spatially non-uniform temperature field results in residual Stresses and distortion [5-8], Therefore, fabrication of a struc ‘urally sound and defect free WAAM component requires precise Control of the process by appropriate selection of the process vati- ables such as arc power, travel speed, wire diameter and wire feed rate, However, selection of these variables by trial and error is expensive, time consuming and provides no guarantee of achieving the desited structure and properties. A recourse is to develop, test and utilize a mathematical framework that can serve as a basis for E-mail adres: day pote (F. DebRoy, ps atee)t01016).heammassastes 2016.08.11 937-9310/0 2018 Heeler Le A ihe reserved selecting appropriate process conditions based on scientific principles. WAAM has already been successfully applied for making com- Ponents of steels 7.9.10], aluminum alloys [11.12], titanium alloys [15.14] and nickel alloys [15] and several attempts have been made to model the process. Analytical models have been used to predict build geometry 16.17] and surface topology [15]. However, these models are based on empitical formulae and ignore the heat and mass transfer during the process. Heat conduction models have also been used to calculate temperature distribution [19.20), em- perature gradient | and residual stresses 19]. However, these models neglect the convective flow of liquid metal inside the mol- ten pool that often dominates the heat transfer mechanism inside the molten pool, Manvatkar etal. (22) and Arrizubieta etal, (22) noted that calculations neglecting the convective heat transfer sig- nificantly overestimate the peak temperature and cooling rates. Svensson et al. [24] mentioned that the heat conduction calcula- tions are inadequate to accurately calculate the cooling rates. Vol- lume of uid (VOF) based numerical models used by Silwal etal [25} consider the molten metal convection and have been used to predict the deposit geometry. However, these models are com- putationally expensive and have not been used to estimate essen- tial metallurgical variables such as temperature gradient, cooling rates and solidification parameters. What is needed and currently not available is 2 well-tested comprehensive phenomenological1-04 eat /ieratona ural of Hee end Mas Taner 127 (2018) 1086-1004 oss ‘model of WAAM capable of calculating the essential metallurgical vatiables for different process variables. Here we develop, test and use a three-dimensional heat transfer and fluid flow model of wire are additive manufacturing to calcu- late the temperature and velocity fields, deposit shape and size, cooling rates and solidification parameters for a single-track deposit. Experiments are conducted by depositing H13 tool steel at different arc powers and travel speeds to validate the model ‘The variations of fusion zone geometry with power and travel speed are compared with the corresponding experimental results. Calculated temperature variation with time is also tested against independent experimental data. After validation, the model is used to quantitatively study the effects of a wide vatiety of process vari- ables such as are power, travel speed. wire feed rate and wire diameter on different metallurgical variables 2. Theoretical model 2.4 Assumptions The following simplifying assumptions are made to make the ‘numerical calculations involving heat and fuid flow, droplet trans- fer and molten pool surface geometry tractable (1) The liquid metal is assumed to be a Newtonian fluid and its viscosity depends on temperature and pressure [26]. The Boussinesq approximation is used for the calculation of buoyancy driven flow (6) (2) Elfective thermal conductivity and viscosity of the liquid metal ate enhanced to account for turbulent convection effects [27]. No separate turbulence models are used to esti- mate turbulent components of transport properties. (3) Because the are current in the WAAM process is higher than 100 A, metal transfer mode is assumed to be globular-type [28), Due to ifficulties with the measurement, droplet tem- perature is calculated based on net heat balance |29), The droplet velocity is calculated considering arc plasma effect using the formula provided by Kumar and Bhadutt [20 22. Solution domain ‘The thiee-dimensional solution domain for a single-track deposit is shown in Fig, 1. Calculations are done in the Cartesian coordinate system, which is attached to the heat source. In other words, the ate source and the molten pool are stationary in space, and the substrate material enters and leaves the computational domain atthe scanning speed, Half of the solution domain is con- sidered in the calculations by taking advantage of symmetry. Dro- plets impinge on the molten pool to form a deposit, and their sensible heat is considered as a volumetric heat source forthe heat transfer calculations [25], The surface ofthe deposit is assumed to be flat during the calculation of the temperature and velocity fields. After the calculation, the free surface profile of the deposit is estimated by minimizing the total surface energy on the top surface of the deposit 1), Finaly, the grids are adjusted to fit the surface profile. and the temperature and velocity felds are then feassigned in the fitted grid system. 23. Governing equations ‘The heat transfer and fluid flow model solves the equations of conservation of mass, momentum and energy in three components slong the x,y, and z directions [2223]. aipu) 2% a Material entering Solution domain Tikes Substrate al wis Fig 1. Schematic ofthe solution denn consisting of olen poo, substrate ane deposi The emensons ae gen in able 2. caming deen slong negative teotue EER) - Brass 8 where pis the density, wand u are the velocity components along the i and j-directions, respectively, and x isthe distance along the direction, js the effective viscosity, ané 5, is a source term for the ‘momentum equation (2) including buoyancy force, the motion of the heat source, electromagnetic force and frictional dissipation in the mushy zone. These source terms were described in detal by ‘Zhang etal. |1| and Mundra et al. 4). The symibol his the sensible heat, Cp is the specific heat, kis the thermal conductivity, Us is the scanning speed, 5; is the source term that accounts for latent heat and Sy i the source term for the additional heat ftom metal dro- lets. The source term Sy is calculated assuming that the heat energy from the metal droplets is distributed uniformly ina eylin- rica cavity inside the work piece (28,34,25]. Detail derivation of| the source term $y is presented in Appendix A. The thermo- physical propeties ofthe alloys are provided in Table 1 [36.37] ‘able There physic properties of AISI 1040 te 12 ste and EROS tee (2627), Ac eficences ate etinated foewing aig et a [4 Properties ‘siio a fom quis temperature) 1785 1585 ma Sous temperature (K) 1800, ims 6 Thecaleanducty esol 253 aaa Bo ‘wink ‘Thermal conductviy of 340 a0 34 nhanced thera aso aio ana end of bai (wim specchesco sold yng) 6862734]. 7012 Specie eat ofliquid KEK) 700A fa Saas Density het) nso 00 re ‘scot gh 3) Gario? 57410) $7107 Enhanced wscosty (alms) 296.10 105 x10 | 264 10 at “pao ta? “bas 10% “bat 10 Neeticiensy os osr osetoss W- ue a/ntenaina ural of Het and Mass Tomer 127 (2018) 1064-1084 24, Boundary conditions On the top surface, the boundary conditions for the energy con- servation equation include the heat fux from the are source and ‘convective and radiative heat losses. The net heat fux (J) on the top surface is expressed as: IH exp (EP) — coe 13) -n- 4 ‘where the Ist, 2nd and 3rd term on the right hand side ofthe equa- tion represent the heat flux from the arc source, radiative loss and convective loss, respectively. The arc source heat flux is assumed {to have Gaussian power density distribution where lis the ac cur rent, Vis the arc voltage. 1 isthe arc efficiency, Fs the heat con tent ofthe droplet (see Appendix A) fps the are radius, x and y are ‘the distances from the arc axis. In the convective and radiative heat loss terms eis the emissivity ois the Stefan-Boltzmann constant. T isthe top surface temperature, and his the heat transfer coefficient. Convective and radiative heat losses are aso applied to the rest of the top-surface of the computational domain Spatial gradient of the surface tension on the top surface of the molten pool results in a shear stress called Marangoni stress, + [2.5]. tis expressed as of ul 6 hers ene ory esos sheep ‘etic rea tard {y=0), the boundary condition for the enthalpy calculation is 25. Geometry ofthe deposit surface ‘Total energy on the surface ofthe molten pool includes the sur= face energy, potential energy inthe gravitational field, arc pressure displacing the pool surface and work performed by the arc pres- ‘sure and droplet impact displacing the pool surface. The are pres- ssure (P.) depends on total arc force on the top surface of the molten pool and is assumed to have Gaussian distribution |) ‘Average droplet impact force on the top surface ofthe melten pool ‘depends on droplet mass, velocity and transfer frequency [28).The pressure due to the droplet impact (P,) is essentially the impact force per unit area and is assumed to have Gaussian distribution ‘on the top surface af the molten pool [39)-The molten pool surface profile is obtained by solving the following equation [21 Pop ry rin y Pein Bidrbe 14 8s oe 6) (+ éi+ea) where is the surface tension coefficient ofthe liquid isthe con- figuration function of the molten pool surface and is the Lagran- shan mrliplir, which is obtained from the following. volume continuity equation: J] say a We Solid boundary o=0 ~ Deposit Symmetry plane 2 ay wd Fig. Schematic ofthe transverse section ofthe depois Half ofthe depot i ‘oideed becase of he sel ah respet to XZ pane where w is the volume of the material deposited per unit time. The symbols gy dae and gay ate defined as: $= g4,—2% and respectively. The definitions for @, and yy are sina. ‘The boundary conditions of total surface energy and solution method of Lagrangian multiplier are described by Zhang etal (31). ‘The cross-sectional area of the deposits shown in Fiz. 2. Similar to Zhang etal, [31], the cross-sectional area (A,) considering only half of the deposit is expressed! as Ne ary ae where ry i the radius ofthe filler wire wy is wite feed rate and Usis the scanning speed. The configuration function ofthe surface profile 4s calculated by solving the overall mass conservation equations Buk 6 @ [sem Fg. (10) represents the deposit surface profile (#) by considering that amount of liquid metal distributes over a distance of W (half ofthe pool width). The boundary conditions (shown in Fig. 2) used in the calculation of solidified surface profiles are given as: ¢~ Oat the solid boundaty and S ~ 0 at the symmetry plane (y = 0, (19) 26. Solution method ‘he entte solution domain is divided into small control vol- umes. For a 120mm long deposit the total number of grids is 102 92.444. (length x width x height) A control volume method has been implemented forthe discretization af the govern~ ing equations [40 The SIMPLE algorithm 0) 1s used to solve the governing equations and the calelatons are completed using an inhouse Fortran code. Typically, around 3.5 million linear equa~ tions and 6000 iteration steps at slved forthe deposition which takes approximately 10 minutes in a personal computer with 2.8, Gigahertz 17 processor and 8 Gigabyte random access memory (RAN 3. Experimental investigation A series of 120 mm long single-track builds using H13 tool steel wire are deposited on a 250 mm x 150 mm x 4mm AISt 1040 steel substrate. The heat source of WAM was a custom-designed Pana~eration! oral of Hee end Mas Taner 1272018) 1084-1084 1a? Garena) ___Voliage (V)___ Speed foi) We fend te (emis) Wire dno fn) Subsite dimension (om. He 190 2» snr a 390% 360017 sonic digital MIG/MAG welding power source, which was inverter controlled using direct current (DC) with output current range from DC 30 A to 350 A and output voltage range from DC 12 V to 31.2 V. The specimens were Fabricated using a welding robot com- bined with control box. workbench and wite feed system. The 6 axis independent movements ofthe robot ate programed and con- trolled by a software-based control system. The deposited builds Were protected by high purity argon with 15 L/min gas flow rate. ‘The samples were ground and polished using colloidal silica and subsequently were etched using the standard Keller's reagent (25 ml HNO, +15 mL HCL+1 ml HF+95 ml. H.0) for 20s. The process conditions used in these experiments are provided in Table 2 4. Results and discussions Fig 2(a) shows the three-dimensional temperature and velocity fields during the fabrication of a single-track H13 steel deposit. Due tothe rapid scanning, the molten pool extends along the oppo~ site direction of the arc source movement. iz. 3(5) shows the mage nified view of the temperature and velocity fields close to the ‘molten pool. The two-phase region containing both liquid and solid material. commonly called the mushy zone, is bounded by the solidus temperature (1585) and liquidus temperature (1725 K) isotherms. During solidification, the characteristics of, the mushy zone play an important role in solidification morphol- ogy. The region within the liquidus isotherm is called the fusion zone. Molten metal Velocities ate shown by black arrows whose ‘magnitude can be estimated by comparing their lengths with the reference vector provided. Liquid metal motion in the molten pool is driven mainly by the surface tension gradients and to a much lesser extent by the buoy- ancy force, Due to the negative value of dy/aT, the liquid metal flows from the center to the periphery on the top surface of the ‘molten pool. The liquid metal is depressed in the middle of the pool due to the strong arc pressure and droplet impinging force. ‘This flow pattean allows more heat transfer from the heat source to the bottom of the molten pool, which. in turn, results in deep Penetration. Fig. 3 also depicts the free sutface deformation of, the molten pool. The liquid metal is pushed to the rear part of| the molten pool and forms the crown as it solidifies. ig 4{a) shows the temperature and velocity fields at different transverse sections of the deposit. Near the leading edge of the deposit, the molten pool is depressed due to are and droplet forces and the fusion zone has finger-like (ie. deep and narrow) penetra- tion in to the substrate. Towards the rear end of the molten pool, the arc and droplet forces acting on the surface decrease resulting ina crown geometry determined largely by surface tension forces Fig 4(b) and (c) show the stream traces on two transverse planes at {wo different distances along the distance of the ac travel. Under the ar. the surface depression combined with surface tension gra- Gients leads to two separate convection loops, as shown by the stream traces in section 1 of Fg. (b). Behind the arc, the circula- tion of the molten metal s slower due to the reduced temperature gradients, as shown in section 2 of fig. (0). Temperature, or Fig. 2. (a) Calculated temperate and velocity fede for & single tack HI fepostton using process parame given in table 2. (0) Magnified ews oftose W- ue aver foural of Het and Mass Tomer 127 (2018) 1064-1084 Temperature, ‘900 1100 1200 1300 1585 1725 1800 1900 2100, Temperature,K (a) ‘00. 1100 4200 1300 1585 1725 1800 1900 2100, a + 758082 84 86858032 thm Length mm 19 20 sa} (C) (©) 7 g16 B15 F “4 13 a0 82 8486. 889092 7 Length tm 0 4 Width,mm, ig 4. (a) Temperature an velocity elds 2 irene ansverse planes. Temper ire fear and stream acer of molten metal aw 3 raves fecan? Seton I (ues the are) an (secon ? etn the ae shawn gute a) All plots use the proces parameters vent abe 2 ‘Temperature and velocity fel for longitudinal sections at dif ferent distances from the are center ace showin in Fig. (a). Compas ing the two sections in Fig 5(b) and (). there litte difference in the distribution of temperatures and velocities. Stream traces for both ‘sections reveal a small convection loop in the front ofthe molten pool and a larger convection loop in the rear of the pool. Because ‘the pool is longer than it is wide, the temperature gradient from the peak temperature atthe center ofthe pool to the liquidus iso- ure and veloiy fies at iferent longitudinal planes wo longa [under the ae) ao) secon 2 (gh de the ae) show In iu a) Al ples use he proces parameters ven a Tobe 2 therm at the edge ofthe pool is larger in the width direction than the length direction, This results in a higher surface tension gradi- ent near the front of the pool and lower surface tension gradient in the rear correlating to the magnitude of the velocities. ‘Mukherjee eta. [41 used the dimensionless Péclet number (Pe) to describe the ratio of heat transfetted by convection tothe heat ttansferred by conduction during additive manufacturing w Pea ay1W- 04 ea /ieratona ural of Hee end Mas Trarfer 127 (2018) 1086-1004 ass where pis the density of the fui, ci the specific heat ofthe uid, kis the thermal conductivity, U is the characteristic velocity (max- imum velocity inside the pool). and L is the characteristic length {width of the molten pool. For the simulations shown in Figs. 3- 5, the computed values of the Péclet number are approximately 100. A Péclet number significantly greater than one indicates that convection of molten metal dominates the heat transfer inside the ‘molten pool (41). fig. 6 shows a comparison between the experi- mentally obseived transverse cross-section (in YZ-plane) of the deposit with that of the computed values considering and ignoring convective heat transfer, When the calculation i done ignoring con vection, the width ofthe pool is 13% smaller than the experimental value. This is because the radially outward flow of liquid metal car- ries heat to increase the width of the molten pool. Therefore, the deposit width is underestimated when the convective flow is ignored. In addition, the heat that should have been transported racially is now concentrated in the middle of the pool and a portion oft goes downwards. resulting in an overestimation ofthe penetra- tion depth ofthe fusion zone. Thus, the calculations that negiect the ‘mixing of hot and cold liquid inside the molten pool cannot predict the correct geometry ofthe deposi. ‘Two of the mast influential and easily controllable vatiables in 'WAAM ate the arc power and travel speed. The toles of these vari- ables on the fusion zone geometry are examined both experimen- tally and theoretically in Figs. 7 and 8. Fig. 7(a-c) shows the effects of varying travel speed. The cross-section of the fusion zone decreases in size with increasing travel speed in both the theoret- ical calculations and the experimentally observed results. Tis is because of the lower heat input per unit length at faster scanning speed. Fiz. 8(3-c) shows that the deposit sizes and fusion zone dimensions increase with arc power, which increases the heat input per unit fength, Fig. 9(2) shows the temperature variation with time monitored at the mid-height and mid-length of the deposit center while fab- ricating a single track H113 deposit. The peak temperature corre- sponds with the time the afc source reaches the monitoring location, At 11.7 mms scanning speed. the peak temperature is observed at around 2.6 s. However, ata slower taveling speed of ‘Simms the peak temperature is observed at about 6 s, because the arc source takes a longer time to reach the monitoring location. Heating takes place rapidly as the arc approaches the monitoring location. However, once heated, cooling is comparatively slower because of the time needed for the transport of heat throughout the molten pool and into the substrate. During cooling, there is a sudden change in slope between the liquidus (1725 K) and solidus Width wi 4 3 2 10 oieauea etc 218 18 a z16 16 Conduction + Convestion 14] —Condvston | 14 54 32-7012 3 4°55 Conslirng and neglecting tolen metal convection withthe cerespending fxperimentaly measured mactopraph The rests ae obtipe using he process parameters given in 15 a3 8 9 s g 2 3 3 z 6 4 2 0 Width,mm Fig.7. Comparison been the clelated transverse sections ofthe HIS epost tt the ererpondingexpecmentall reared maerograph wing 3) 5mm) £83 mms and (117 mmf caning spews. ter proses parameters are given in “ble 2 The red ron bounded by te sls temperature (1585 K) wether represents the tans rection af he depot (1585 k) temperatures because of the iquid-to-solid phase trans- formation and the release of latent heat during solidification. Fig. 9(b) compares the computed temperature variation with time during the deposition of ER70S-6 stel to the corresponding exper- mental results of Bai et al, 20), Temperatures were measured with, an IR camera during the experiment. The slight mismatch between the experimental and numerical results is likely due tothe diffcul- ties in the temperature measurement during deposition, as well as several assumptions made during modeling, ‘The agreement ofthe calculated fusion zone geometry for vari- ous travel speeds (Fig. 7) and arc powers (Fig, 8), as well as agree-tose W- ue a/nveraina ural of Het and Mass Tomer 127 (2018) 1064-1084 6 4 2 0 Width,mm ig. 8, comparison beoween ce calclted transverse scion of he 113 depot wih the comespencing experimenaly measured macrograph using) 1800 W 1000 W and (4400 W are power Utes proces parameters te geen in Toe? ‘hese fegion bounded by the seis teipeatte 1585 K)botheoytepresens ‘he ansvere section ofthe capes ment of thermal cycles (Fig. 9(b)) with corresponding experimental ddata provides confidence in using the model to investigate the roles of important process parameters on deposit shape and size ‘temperature fields, cooling rates and solidification parameters, Fig. 10(a-c) shows the effects of wite feed rate and wire diam- ‘eter on fusion zone width, penetration depth and deposit height, respectively. The fusion zone width is primarily controlled by ‘available heat energy for melting and convective flow within the molten pool. As evident from figs. 7 and 8, width depends largely ‘on arc power and scanning speed. For a constant heat input, fusion zone width does not change significantly with wire feed rate and ‘wire diameter as shown in Fig, 10(a). However, the enhanced mass Nlow rate that results from higher wire feed rate and larger wire ‘diameter increases the heat transfer from the are source to the bot- cd Phase transformation M (a) w 1800 ask! SS g Tes = 1300 i g i Esco ne 5 = 5.0mm 300) 44.7 mm/s. TPS ES ST Ss to Time, s 700 650 < 600+ ¢ $ 550 / © 500; 3 450 = 400 + = i ———_ Numerical ole Experimental 300 0 10 20 30 40 50 60 Time, s iE (a) Temperate vvition with ine monitored a the mid eight Tenth of he cept centr we abating single tack 13 depos ierent scanning speeds. The process parameters ate gen ih Tobe! () ‘jes 20) tom of the molten pool. Therefore, penetration depth increases with both wire feed rate and wire diameter as shown in Fig. 10 (b)- Fig. 10(c) shows that the deposit height increases with both ‘wire feed rate and wire diameter because of an increase in the amount of material deposited per unit time. The temperature gradient, G, and the growth rate of the soidif- cation front, R, provide important information about the soliitica- tion morphology. The value of G across the mushy zone is relevant during solidification asit drives the solidification kinetics. n steady state, R along the centerline is equivalent to the travel speed and varies spatially by the angle of the pool boundary relative to the centerline. The product of G and R, GR, represents the cooling rate, ‘which is linked to the scale of the microstructure. Tig. 11(a-c) shows the variation of cooling rate during solidification calculated at mid-length and mid-height ofthe deposit for vatious are powers, travel speeds, wire feed rates and wire diameters. Convective flow of liquid metal inside the molten pool mixes the hot and cold liquid and reduces the temperature gradient inside the pool. Therelore, hheat conduction calculations that neglect the molten metal convec~ tion overestimate the temperature gradient and cooling rate as shown in Fig, 11(a) and (b). By increasing linear heat input, higher1W- 04 ea /ieratona ural of Hee end Mas Taner 127 (2018) 1086-1004 1st Wire diameter, mm pep OO eee 20 200 Ee Wire ciameter e OF —o— Whe feed rate 3 = aot @ 2 § 70-6-—e—oo—95—0 § S col zc (a) 8.0°%0 "70-8080. 100 Wire feed rate, mm/s Wire diameter, mm. E go0,t00__140 120 "1.30 1.40 = 7 t f 7 & 275¢ g B 2sot 5 a @ 225F 2 5 © 200} Wire diameter —o— Wire feed rate 178"90-—“70-—0- 90-100 Wire feed rate, mm/s Wire diameter, mm 275,190_140_ 130’ 130 _1.40 zB Eesy {c) E asok B 225 2 200) 178 ’ Wire diameter Lo —8— Whe feed rate 1.50 Cd a a LG) Wire feed rate, mm/s Fig 10. fects of wie fed rate and wie diameter on (a sion zone width (5) proetraton dept adc) epost eight oa single ach 113 deposi Craps ate Dioted using the process parameters ive in I-be 2. Varations with respect to Wire elameter and wie lee ate se plated 380 mmewie feed ate 12m ice damete respective are power and slow scanning speed result in more molten material that cools and solidifies relatively slowly. Computed results in Fig, 11(2) and (b) support this, showing cooling rates increasing as are power increases and scanning speed decreases. Wire eed 1000.0, 200.0: —S— Conduction only ——B—— Conduction + convection (a) 3. E neg 100.0 5500) +3600 4000 4500 Arc power, W 1000.0 [-—S— Conduetion only ~ Conduction + convection, 6 7 8 98 10 11 Travel speed, mm/s Wire diameter, mm 1000,0,1:00_ 1.10 1.20’ 1.30 _1.40 0007 () ~ Wrre diameter —O— Wre feed rate (COs Ome 20 us 0 eH 100) Wire feed rate, mm/s Fig 11. Varuins of cooling ate daring seencason wit respect) arc power (by rave epee and fe) ire fre ate and wre diameter Cowling rate ate {single wack eepst raps are plated using the process params given “Shed Vanations nth respet tte ameter nd we fed ate sr plttee 3t rate and wire diameter do not significantly affect the heat transfer from the molten pool to the substrate, because they do not change the amount of heat input into the system, Therefore, the cooling rate remains almost unchanged for different wire feed rates and Wite diameters, as shown in Fig. 11(0. Dividing G by gives the solidification parameter G(R, which is necessary for determining solidification morphology. High values of G/R indicate stable plane front solidification, while low values1092 Wu ea /nveraina ural of Het and Mass Tomer 127 (2018) 1064-1084 indicate unstable growth. The constitutional supercooling criterion for plane front solidification is represented as: GIR = ATe/D, (12) Where ATs isthe range of equilibrium solidification temperatures (240K for 113 tool steel) and D; is the solute diffusion coefficient 15.00 —e— Conduction only 13.00 “5 —— Conduction + convection E1100 § (a) Z x oO Tg 88888 3000 3500 4000 Arc power, W 4500 15.00 —6— Conduction only 13.00 = Conduction + convection %11.00F = 9.00 * 700b oe 9 ss. on 3.00F (b) eee eerie 10s 6 7 88 10 tT 12 Travel speed, mm/s Wire diameter, mm 4500190 1.10 1.20" 1.30 1.40 1300 (c) -~@-- Wire diameter : —S— Wire feed rate ‘= 11.00 © goof S 7.00F g © 500- 300} G——-s—93—e-5—@ 1.00! (6g 0 ag 00 pes 00 ees 100) Wire feed rate, mm/s ig. 12. Variations of Gk ais éurng seliicason with espect (a are power rave! see and) see eed rate ane we camer, CR rato te eee i the mid phe and midength ofthe deposit enter whe facatng singe Variacns wit espe to wie eter ane whe ed Tate ate plotted 80 mms (2.x 10° mm? 5“ for carbon in liquid iron [42)). Therefore, caleu- lated value of AT;/D; for H13 tool steel is 7 « 10° Ks mm Fig. 12(a- «) shows the variation of GJR during solidification, calculated at mid- length and mid-height of the deposit, with are power, travel speed, Wire feed rate and wire diameter. Notably all the values for GIK are significantly lower than the citical value of 7 « 10" K's mmmé required for plane-tront solidification. Therefore, formation of either cellular of dendritic solidification structure is expected and observed in experimental results. Conduction calculations neglecting convec- tive flow of molten meta result in higher G/R values, because larger temperature gradients develop due to the lack of mixing of hot and cold liquids. Wider and longer molten pools at higher arc powers also result in lower temperature gradients. Therefore, G/R decreases ‘with are power as shown in Fig. 12(a). Growth rate ofthe soliifica- tion front increases with travel speed, so G/R decreases with travel speed as shown in Fig. 12(b)-Since, wire feed rate and wire diameter do not significantly affecc the heat transfer from the molten pool to the substrate, G/R remains almost unchanged with these two param- eters, as shown in Fig, 12(@) I in-depth analysis of microstructural features is desited, then the effects of convective Muid flow, travel speed. and arc power must be considered to obtain accurate solii- ation parameters. n all cases presented here, ether celislar or den- Artic solidification morphology occurs, with variations in GR atfecting the scale of the microstructure. 5. nary and conclusions A thyee-dimensional heat transfer and fluid flow model of wice arc additive manufacturing was developed, tested and used to the calculate temperature and velocity felds, deposit shape and size, cooling rates and solidification parameters for a single-track H13 tool steel deposit, Etfects of arc power, travel speed, wire feed rate and wire diameter on different metallurgical variables were exam- ined. Below are the specific findings. (1) The calculated dimensions of fusion zones and thermal pahe a7) Combining Eqs. (AS)-(A7), the total work done in creating the cav- ity, W. can be obtained as: We na(5apeti +27h.) (ns) The Kinetic eneay of opi expresed a: 1 KE = 5 mo} (as) Where m is the droplet mass depends on droplet diameter (a) and ensity (p} and vy is the velocity of droplets [43]. Therefore. by equating the total work done in creating the cavity (W) and the kinetic energy of drop (KE), the cavity height (hy) can be calculated = (22) [(22Y 2 (2-4 Eq, (Ad) is calculated by solving the acceleration equation of dro- plets [30]: (ao) 24)-£ ann) 4. le ae aph.} he where gis the acceleration due to gravity. By solving this equation, -(B)f-((6)"I} here ¢s the time that is inverse tothe droplet frequency (D}. The Aroplet frequency (D) depends on wire feed rate (w;), wire radius (ra) and droplet diameter (a) and is represented as (an) pe (3) References TY eb A we. Zak FMW Ene 0 le AM [21 SA David, DebRoy, Current issuer and problems in welding science Science BLT natty SA. ‘Davie, Physical processes in fusion welding, Rev. Mod. Phys. 67 14 We ard Dey, nln of oditatin texte dng (oF Mabbciee JS abc A De T Deby, Pata of aly ave [o) CL Xap T Mather |S Zubick Hi, Wel TA Panes, De. 1. Deby. Buiding lois for igi wn of addive manufacturing Ata Mate. 125 2017) 390-395, 17 bing Coleztove J Mednen 5. Ganguly. 5. Almeida F Wangs. ilans ‘estonedhancel anass of wie and ace acave lave manulcuring [a1 Tene, v. anvatiar A Det Debkoy. Mignon of thermal distort uring adaive manag Serta Mate. 127 Is] 5w Willan F Masting Ne Addon) ne GF dre site manufatring, Mater Se Techni 32) (2046) Ses Wo] I Xiong th tet chen, Het propagation af creat swale pats faved in dave martin ang seta ae ein} ater [141 Bong. | Ding 5 whlams, ee. Ut LE con Dig as zh We ae ine manana Tabeatin symposia, asin T2014 pp. 41158 U1 F ing Wns ARG Maipaiagy aigion om et cet dv. Manu echo 373-8) 2011) 387-00, Ina) | Zhang Wang: Fades x Zhang Fatigue tack propagation behav i tute ie sive manctictuee trsnlav eee f ilcostuctte sod Festal sess, Mater Des 90 (2016) 551-56. Ins] D-bing 2 Pan's. va Duia HU C Shen Fabien supece al bronze exnponents though wie ac zee manufacturing Maer 9 (6) (2016) 106] Geng Xone. Huang Xin} UA prediction model flyer geome foe if wre ane ates aeeve. imanctscure ng. respons itce ‘etnocloy. Ine) Adv Manat Tecnol 9 1-4) 047) 175-18, 1n71 Bebing Pan co, Hs, A mien sveianpine moe! fer robotic004 W- ue anterior ural of Het and Mass Tomer 127 (2018) 1064-1084 [ns] ¥- 1, ¥. sun, @ Han. 6. zhang t Hrvith, anced beads everapping mode fectnteané are sede manuacurng of mult-aver mul-beed fetal Dans | Mate Paces. Tecol 2522018) 838-848, Ing] binge colegrove | Mehpen, $ Willams F. Wang &.Sequeea Ames, A empucaonaly efcent fate clement tedel of wire ane ae seve tmanufsctre, Ine} Ads Mani Teco 20-4} 2048) 2372256 [20] Batt Zhang. Wane improving prediction accuracy of thera analysis fecwelebase aditve mantacsrng by eivasng Input peat SOE IRimaring int] Adv. Mant Technol 69 5-8) (2013) 1087-109, [a1] J Xone ¥ CIs nite element analysis and expecenal validation of thermal behavior for tnuwalled "pars. in CMAWaced sceie Imanufacuring wih vacous substrate. prebeating. temperate APB Ther Eng 1262017) 43-52, aa] V stones be, Debhay, Hes transfer and material ve ding er Screamer aaatve mancfacaring. | pl Phy 1 (3) 20.8) asso [aa] Jt Ariabiet, tama F Klace, Marne, K Art, Uta ulation Df he relevance af melt pool mane i Lave Material Deposition proces ‘odeing int) Heat Mats Teams 115 (2017 S081. [24] Le Sterivan, B Creo HAD. Beagesis Amana af cooing carves fhe nn ane el meld epost Sean etal 1597) (86) [255 Sal a. Sanangel, Ee af wri and hate gab tungsten are (Gta) the geomet shape) Mater frees Leena! 251 2018) 1380045 (ay RB Wt Sear EN pot asp enone 200 Jo Wey [27] A Devt Debtoy Asma model estimate effete heel conduc "Gos in the weld poo | Apt Phys 95 (92064) $230-Sa40. fae] JF Lancaster, the Riya of Weng aeration Taste of Welding, Peagaton Ofer 1985 [a9] Zang. Desay, Modeling macio-and micosuuctuces of gasmnetabare tweed HSLACIO0 see Meal Mater ane 8303) (1999, 482-823, [30] 5 Kuma, Sc whacut, Thevetialintestigaton of encom chactensues In gus ncaa welag using at elena met. Meta, Mate Tras 8 25(3) 99s) 61i-aza syand [31] Ww. zhang CH. im, Deboy, Hest and td tow in complex jints ding as etal are weling~pate nencl mage e et wel} ADDL PS 55(9) 008 S2i0-s718 [32] T ushenee. We Zhang T,Debkey. An improved prediction of cesiua Stvesses ane dsterton ih dive manufacturing, Comput Maer. Se 128 (goin aenara [p31 T Malenee, js\Zuback, W. Zhang, T. DebRos, Residual stresses and Asoruen im seGinvely manufactured Compascenay graded and silat Joints comput Mater Se 143 (2018) 3250357. [p41 Mina Deo. KM, Keli Nutercal ‘anefer in welding wth moving hes sot ppl 292311886) 1523, [351 MP Uri. 5 We Mathematica medeing of theeimensona hes and hi fowinsmanng get metal stewed pos, Metal Mater Tran B38 (3) 199 Some. [B61 KC Mil Recommended Vases of Thermenhyriel rapes for Selected fay] feo Foren Av iitani Fone, The infenee of phase ‘cantomationy on resgul eves induced by the welding race? D {nd 2 mamerel mages Model Sem Mater et Eag 142) (005) 11. las) Hs. Smare AMS Handbook ASM iterations, 1993.27 [39] Zn ca, Dong Modeng of GMA weld pols with onidrain of droplet Itnact ng Mater Teel 120 (4) (960) 212-220, ao) 5 Patankar Names Heat Taner soa ii lw, CRC Fes, 1980. Ia1] 3. Munhescev_ Marva, A De, Deby, Dinensoness umbers in ‘dive manatacring | Apt P21 (6) 2017 ooeeoa laa] V tants De. T Deby, Spt vasition of mk pool game, peak Temperature and sldiieaion putaetes duag laser asasted seeve ‘anulacttng press Mater St Teehal 31 (8) 2015) 934-930 [as] Wc ser, Walter Hea canster ane peteteaon sechans wah CMA thd psa aa welding. ele. 80 (2981) 37=42- [a4] Alten A Rusch May, Caarmetnic Anlyes ofthe comprehensive est flow for gas etl ae weldg Weld Wesie59 2035) 191-195. ican o i owe ne est ‘omer Hest Trane, Pat