Decreasing the surface roughness of aluminum alloy welds fabricated by a dual beam laser

Accepted Manuscript Decreasing the surface roughness of aluminum alloy welds fabricated by a dual beam laser

Guang Yang, Junjie Ma, Blair E. Carlson, Hui-Ping Wang, Mehdi M. Atabaki, Radovan Kovacevic PII: DOI: Reference:

S0264-1275(17)30442-2 doi: 10.1016/j.matdes.2017.04.085 JMADE 3007

To appear in:

Materials & Design

Received date: Revised date: Accepted date:

11 February 2017 6 April 2017 24 April 2017

Please cite this article as: Guang Yang, Junjie Ma, Blair E. Carlson, Hui-Ping Wang, Mehdi M. Atabaki, Radovan Kovacevic , Decreasing the surface roughness of aluminum alloy welds fabricated by a dual beam laser. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Jmade(2017), doi: 10.1016/j.matdes.2017.04.085

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Decreasing the surface roughness of aluminum alloy welds fabricated by a dual beam laser Guang Yang, Junjie Ma, Blair E Carlson, Hui-Ping Wang, Mehdi M Atabaki, Radovan Kovacevic Highlights 1. Practical methods were provided to mitigate humping and rippling on the weld surface.

T

2. The jet flow was experimentally visualized.

IP

3. The effects of nozzle shape, gas flow rate, and inclination angle of the gas tube on the surface

CR

roughness of aluminum welds were discussed. Abstract:

US

Decreasing the surface roughness of visible laser welds on aluminum automotive closure panels is of critical importance in the automotive industry. During welding, surface roughness is strongly influenced

AN

by the interaction of filler wire, laser, and gas. Observation by CCD camera revealed that placement of the filler wire tip at the front wire-feeding position could effectively mitigate humping phenomenon on

M

the weld surface. Visualization of the gas flow and detection of the plasma intensity demonstrated that

ED

optimizing the shielding gas parameters could further reduce the rippling phenomenon by affecting the molten pool surface tension and the pressure differential acting upon the molten pool. The surface quality

PT

could be improved via optimization of nozzle outlet geometry, gas flow rate, inclination angle of gas tube,

CE

and distance between the nozzle and laser beam. After shrinking the processing parameter window through the single factor investigation, a Taguchi L9 orthogonal array was designed to optimize the

AC

shielding gas parameters. The weld surface roughness, Ra, could be effectively reduced to below 1 μm when a circular gas nozzle was positioned at 5 mm behind the laser beam, delivering pure argon gas at 30 SCFH with an inclination angle of 45° to the horizontal plane. Keywords: Dual beam laser, surface roughness, aluminum alloy

1

ACCEPTED MANUSCRIPT 1 Introduction In the premium car market, the need for a flawless and harmoniously painted automotive body is driving more stringent requirements for the surface quality of laser welded aluminum joints. After painting with various layers of coatings, the surface roughness of aluminum welds

T

could be reduced by ~1 µm [1, 2]. Although these layers served the function of smoothing out

IP

irregularities of the surface morphology [2], an inconsistent surface after painting could be

CR

clearly seen when the weld surface roughness was high. If a smooth weld surface could be achieved directly close to the panel surface roughness of 0.3~1 µm, some surface processing

US

procedures would be simplified and the manufacturing cost would be reduced greatly. Our prior

AN

work showed that adjusting the temperature field was mainly effective in reducing the surface roughness from 2 µm to 1.5 µm during laser joining of AA 6111 coach-peel panels [3]. Obtaining

M

a weld surface roughness below 1.5 µm remains a challenge, as the deviation may still be large

ED

(~ 0.3 µm) even if identical parameters were adopted. Troubleshooting this instability requires a

processing parameters.

PT

fundamental understanding of the formation of weld surface patterns and a further control of

CE

The formation mechanisms of surface patterns [4-13] could be categorized according to solidification and hydrodynamic behaviors. Trivedi and Kurz [14] pointed out that the

AC

solidification rate influenced the morphological stability of an interface during solidifying alloys. The extraction of the latent heat of fusion could periodically retard the grain growth until the heat dissipated [15]. When materials, groove geometry, welding speed, wire feed rate, laser power, and stand-off distance were constant, the difference in surface roughness caused by solidification instabilities could be negligible if contamination were not a consideration. Another factor that affected the surface patterns was hydrodynamic behaviors [6]. According to Kotecki et al. [16], 2

ACCEPTED MANUSCRIPT the molten pool surface could be rendered as a membrane that oscillated periodically. Perturbations that were induced by Kelvin-Helmholtz instability [11], Rayleigh-Taylor instability [12], thermocapillary instability [13], and hydraulic jump [5] were stretched out by surface tension force. Decreasing the surface roughness of aluminum alloy welds in the current work

T

focused mainly on reducing perturbations of the molten pool such as the humping and rippling

IP

phenomena. The above mechanisms illustrate that any physical or chemical factor that affects

CR

surface tension or causes a pressure difference could contribute to oscillations at the molten pool surface. Surface tension depends largely on the chemical elements and their compositions. input, the values of surface tension

US

Without considering the influence of heat

AN

𝜎𝑠𝑜𝑙−𝑔 , 𝜎𝑠𝑜𝑙−𝑙 𝑎𝑛𝑑 𝜎𝑙−𝑔 could be altered during the interaction of a gas and the molten metal [17]. In addition, the disturbances could be caused by the hydraulic pressure difference when the

M

drop of filler wire falls into the molten pool and by the plasma plume pressure and the shielding

ED

gas pressure impinging on the molten weld pool surface [18]. Consequently, the idea to improve

PT

the weld surface quality by controlling the flow of filler wire into the molten pool and optimizing the shielding gas parameters comes naturally.

CE

Although studies of shielding gases are abundant, most of the literature sets the evaluation

AC

criteria of the weld quality as the strength of welds [19-21], the porosity [19, 22-24], and the weld geometry [20, 22, 23, 25, 26]. Chung et al. [27] and Reisgen et al. [28] reported that pure argon was the worst shielding gas for CO2 laser welding of steels from the viewpoint of surface appearance and penetration. In contrast, the performance of argon was better than that of helium in Nd:YAG or ytterbium-doped fiber laser welding of steel [25] and nickel alloys [22]. The study of optimizing gas parameters in laser welding of aluminum is inadequate. The effect of shielding gas on the surface roughness of aluminum alloy welds that are used as the non-load-bearing 3

ACCEPTED MANUSCRIPT components on the exterior of automotive bodies has not yet been studied. Campana et al. [29] simulated the hybrid Nd:YAG laser-MIG welding of aluminum alloys using a computational fluid dynamics (CFD) method without considering the presence of the plasma plume and the arc. Their optimization is not instructive because the first ionization potentials of argon (15.67 eV)

T

and helium (24.5 eV) exceeded that of aluminum (5.98 eV) [27], indicating that ionization of

IP

argon /helium is unavoidable. The nozzle could be quickly burned by the plasma, if the nozzle is

CR

positioned at a distance within 3 mm from the laser beam. Therefore, detailed shielding gas parameters for fiber laser joining of aluminum alloys need to be scrutinized.

US

The primary purpose of this study was to optimize the processing parameters that could

AN

provide a stable surface roughness of the top weld surface below 1 µm during the fiber laser welding of aluminum panels. The mitigation of humping by altering the hydraulic pressure was

M

described first. Reducing the rippling was explored by adjusting the shielding gas parameters

ED

such as nozzle geometry, flow rate, inclination angle of the gas nozzle, and distance between laser and nozzle. After shrinking the parameter window, the design of experiments was

PT

conducted in consideration of multi-factorial interaction by using the Taguchi method [30].

CE

2 Experiments

The material for welding was a pair of AA 6111 coach-peel panels with an average surface

AC

roughness (Ra) of 0.275 µm, a thickness of 1.2 mm, and a length of 300 mm (Fig. 1a). The AA 4047 filler wire with a diameter of 1.6 mm was selected as a filler material. Before welding, the surfaces were cleaned thoroughly with acetone to remove any grease and dirt. All edges of the panels were tightly clamped to achieve a zero-gap condition at the bottom of the groove. A 4-kW continuous wave fiber laser (IPG) with a wavelength of 1070 nm was delivered through a 400μm optical fiber and split into a dual-beam with an energy ratio of 50/50. The dual beam was 4

ACCEPTED MANUSCRIPT positioned side by side on the centerline of the groove. The axis of the dual-beam was inclined to 5° with regard to the vertical axis in order to prevent any damage to the optics caused by the back-reflected laser beam (Fig. 1b). All experiments were conducted at a scanning speed of 60 mm/s, a wire feed rate of 70 mm/s, and a laser power of 4 kW. AA 4047 filler wire with a

T

diameter of 1.6 mm was fed by a wire feeder (ABICOR BINZEL) at an angle of 45° towards the

AN

US

CR

IP

molten pool (Fig. 1b).

M

Fig. 1 (a) Experimental setup and the configuration of the dual-beam laser; (b) gas-laser-

ED

wire position; (c) extraction of the weld surface roughness from the surface morphology

CE

PT

obtained by a 3D optical profilometer.

Table 2 provides the detailed shielding gas parameters. Flowmeters (Praxair, Inc.) were used

AC

to control the flow rates of pure argon and helium. The effects of flow rate and nozzle outlet geometry on a gas flow behavior were studied by mixing a fume (Backwood bay-polyfunctional long lasting fog) with a shielding gas in a chamber. A gas nozzle in Fig. 2c was connected to the outlet of the mixing chamber shown in Fig. 2a. Visualization of a gas flow was achieved by tracking the flow of fume by using a machine vision system and applying a green laser with a wavelength of 532 nm as an illumination source. The nozzle was positioned above the groove of the coach-peel panels simulating the welding condition (Fig. 2b). To mitigate the reflection of the 5

ACCEPTED MANUSCRIPT green laser and achieve a better visualization, the groove region for observation was replaced by a plastic insert of the same geometry as the cut-out piece of AA 6111 panel (Fig. 2b). The surface roughness, Ra, was measured by a Nanovea 3D non-contact profilometer at the acquisition frequency of 300 Hz in a dual frequency mode. Surface roughness profiles were

T

extracted along the centerline of the weld surface (Fig. 2c). During joining, the molten pool was

IP

observed using a high-speed charge-coupled device (CCD) camera with a maximum of 4000

CR

frames per second (fps). The emission spectrum of the plasma zone was detected in real time by an SD2000 Ocean Optics spectrometer (integration time: 3 ms; wavelength resolution: 0.364 nm).

US

The collimator of the spectrometer was at a distance of 200 mm from the molten pool. The

AN

microstructures of welds were observed by an optical microscope and a Leo-Zeiss 1450VPSE scanning electron microscope in the backscattered electron imaging mode (beam current is 10

M

mA; working distance is 15 mm; acceleration voltage is 25 kV).

ED

Table 2 Gas parameters Nozzle shape

Circular nozzle with multiple openings, regular circular

PT

nozzle, small rectangular nozzle, and large rectangular

AC

Gas flow rate

CE

nozzle (Fig. 2c) 10, 20, 30, and 40 SCFH (4.721, 9.442, 14.163, 18.884 L/min)

Inclination angle of the gas 5°, 15°, 25°, 35°, 45°, 55° tube Shielding gas

Ar, He

6

IP

T

ACCEPTED MANUSCRIPT

CR

Fig. 2 (a) Equipment used in the visualization of the gas flow; (b) position of the nozzle in correspondence to the panels; (c) different shapes of nozzle: nozzle with multiple openings,

AN

US

circular nozzle, and small and large rectangular nozzles.

3 Results and discussion

M

3.1 Minimizing humping by adjusting the laser-wire position

ED

Two weld surfaces prepared with the same parameters except for different wire-feeding positions were compared in Fig. 3. Fig. 3a presents a bunch of swelled protuberances with a large

PT

depression between each of the two “waves,” a typical topography of humping. In contrast, the

CE

smooth weld surface in Fig. 3b exhibits slight chevron-shaped ripples. The comparison of these two surface patterns indicates that perturbations were impacted dramatically by the liquid filler

AC

material flowing into the groove. Thus, the melting, dropping, and spreading of filler wire at different laser-wire positions was observed using a high-speed CCD camera. It could be seen from the CCD images that the liquid filler wire seemed not to flow from the bottom of the groove upward to the surface of the workpiece as described by Berger et al. [31]. Instead, the transfer of a drop from the tip of wire went through a stable growth and spread over the groove surface. The observed spreading process could be described as consisting of three stages (Fig. 3c). First, the rolling downward of a liquid drop started at a certain angle of surface 7

ACCEPTED MANUSCRIPT inclination. Once the liquid filler material contacted the panel, it spread under capillary and gravitational force. The second stage corresponded to the melt in motion, and the free-surface deformation occurred mainly at this stage. Viscous force affected the flow of molten filler material, and the surface tension was influenced by the geometrical issues of the Raleigh

T

instability that determined the growth time of disturbance [32]. Finally, the spreading rate rapidly

M

AN

US

CR

IP

decreased until the spreading stopped.

ED

Fig. 3 (a) A weld surface with humps showing a series of swelled protuberance; (b) a smooth and flat weld surface with slight ripples; (c) schematic illustration of the spreading

CE

PT

of the melt to form the weld bead.

The adjustment of the laser-wire position was made by placing the wire tip according to the

AC

location of the laser spot. Using this method, the effect of laser-wire position is shown in Fig. 4. When the wire was placed at the front-wire-feeding position, the melted wire flowed smoothly into the groove. The 3D topography of the weld exhibited only shallow surface ripples with a corresponding measured surface roughness of 0.702 ± 0.033 μm. In comparison, when the center-wire-feeding position was used, the “tides” on the surface of the molten pool altered the molten pool size periodically, and the surface roughness increased to 2.45 ± 0.101 μm. The formation of the roughened surface was attributed to the potential energy of the droplet being 8

ACCEPTED MANUSCRIPT transformed into kinetic energy, and the dissipation of this kinetic energy further resulted in strong oscillations within the molten pool. This phenomenon is also called the hydraulic jump, a primary mechanism of humping demonstrated by Wei [5]. When the tip of the wire was moved further towards the rear-wire-feeding position, splashing of the molten filler material occurred,

T

and the weld bead was discontinuous. The CCD camera captured the instant when the molten tip

IP

of the wire grew quickly into a large droplet rolling over the groove without wetting (t+0.347 s in

CR

Fig. 4), and two keyholes appeared on the rolling droplet. Sometimes, the dual-beam laser broke the droplet (t+1.313 s), and the weld was formed by the splashing of the liquid into the groove

US

(t+1.405 s). But the phenomenon of the molten droplet balling up occurred repetitively for the

AC

CE

PT

ED

M

AN

rear-wire-feeding position.

Fig. 4 Effect of the filler wire position on the surface quality. Gas parameters: Ar, 30 SCFH, and gas nozzle of 55° from the horizontal direction

9

ACCEPTED MANUSCRIPT A good wetting during the laser welding process requires that the liquid metal has a low incidence angle by spreading onto the surface and adhering to the substrate rather than balling up. Deyev [32] concluded that the factors influencing wetting include material composition, gas atmosphere, surface roughness, intermetallic phases, and temperature. The significant difference

T

in resulting surface roughness among the filler wire positions could be related to the temperature

IP

of substrate heated by the laser beam. Given that the laser beam power was constant, the amount

CR

of substrate being heated was a function of the amount of thermal shielding caused by the filler wire. For example, the front-wire-feeding mode exposed the largest substrate area and heated the

US

smallest amount of the filler wire so that the molten drop would fall upon the substrate having

AN

the greatest temperature, which aided in wetting. Conversely, the rear-wire-feeding position had the greatest amount of thermal shielding with a higher proportion of heat going into the filler

M

wire that created a large molten drop to fall upon a relatively cooler substrate. Furthermore, since

ED

a larger drop would mean a longer time to build up, there would be a greater distance between drops that could account for the lack of a continuous weld in Fig. 4 for the rear-wire-feeding

PT

mode. In summary, one wants a balance between heating of the substrate and melting of the filler

CE

wire in order to achieve a good wetting. Under the current processing parameter settings, a smooth weld surface could be generated by adjusting the laser-wire position to the front-wire-

AC

feeding position that minimized humping of the molten pool surface. 3.2 Decreasing rippling by optimized gas parameters 3.2.1 Gas nozzle design and gas flow behavior The investigation of the shielding gas effect on surface ripples began with the gas flow behavior. Inspired by the coaxial gas delivery tube with multiple openings in the laser head, a circular nozzle with multiple openings was designed as shown in Fig. 2c. For comparison, a 10

ACCEPTED MANUSCRIPT normal circular tube and two rectangular tubes with different-sized cross sections were used to deliver the shielding gas. The flow behavior was strongly dependent on whether the flow was laminar or turbulent. In laminar flow, fluid particles stayed in the laminar layers. In contrast, in turbulent flow, the flow fluctuations were randomly distributed in space and time; the structures,

ρVD μ

CR

Re =

IP

and transitional flow is usually described by the Reynolds number,

T

such as eddies, were not predictable and reproducible. The flow type such as laminar, turbulent,

(1)

US

where D is the inner diameter of the circular tube, V is the mean velocity of the flow, and the dynamic viscosity (μ) and density (ρ) of argon at 20 ℃ are 2.2294×10-5 kgˑm-1ˑs-1 and 1.6617

AN

kgˑm3, respectively. For noncircular pipes, the hydraulic diameter Dh instead of D was used to

M

calculate the Reynolds number, 4(cross−sectional area)

(2)

ED

Dh = (wetted circumference)

Laminar flow occurred when Reynolds number R e < 2300, and turbulent flow occurred

PT

when R e > 4000 [33]. Between 2300 and 4000, the transition from laminar to turbulent flows

CE

depended on many factors, such as pipe roughness and the flow uniformity [34]. At a constant flow rate of 30 SCFH, the Reynolds number of argon is calculated in Table 3. Based upon the

AC

calculation of Re, the large rectangular nozzle, the regular circular nozzle, and the other two nozzles were expected to generate laminar, turbulent, and transitional flow, respectively. A visualization of shielding gas also was applied to understand the flow behavior at the exits of the four types of gas tube while in contact with the weld surface (Fig. 5a). Since flow rate 𝑞̇ was defined as a product of the cross-sectional area and jet velocity, a smaller cross-sectional area meant a higher jet velocity and a better shielding effect. Because the cross-sectional area of a rectangular tube was larger than those of the two circular tubes, the jet that flowed from the 11

ACCEPTED MANUSCRIPT rectangular tubes was weaker in observation, and a rougher weld surface was obtained. Admittedly, once the small molten pool was covered, it was not productive to deliver an inert gas by a wide nozzle, such as in a rectangular shape. Hence, special attention was paid to the two circular nozzles having small cross sections.

T

Table 3 Effect of the nozzle shape on the weld surface roughness.

IP

(The shielding gas was pure argon; the inclination angle of the tube was 55°; the horizontal

Small rectangular

Large rectangular

nozzle (4.5 mm ×9

nozzle (5 mm×14

Multiple-

mm)

mm)

5.84

CE

(m/s)

PT

6

Average Velocity

2613.4

AC

Reynolds number Re

ED

diameter Dh (mm)

opening nozzle nozzle

7.37

1.67

4.4

3.38

26.36

15.57

1857.3

3280.7

5106.3

Transitional

Turbulent

flow

flow

Flow type based on Re

circular

M

Hydraulic

Regular

AN

Type of nozzle

US

SCFH.)

CR

distance between the center of the nozzle tip and the laser was 5 mm; the argon flow rate was 30

Transitional flow

Laminar flow

Surface roughness Ra (µm)

0.937 ± 1.617 ± 0.13

1.79 ± 0.11

2.137 ± 0.087

0.049 12

AN

US

CR

IP

T

ACCEPTED MANUSCRIPT

Fig.5 (a) Visualization of the flow jets with parameters listed in Table 3; (b) Gas flows from

M

a circular nozzle and a multiple-opening nozzle without impinging on the substrate. (c)

ED

Surface roughness vs. argon flow rate using two circular nozzles.

PT

The visualization of the jet from the circular tube in Fig. 5b shows that the flow was

CE

dissipated from the centerline to infinity in the transverse (y) direction, but the fume intensity did not diminish in the horizontal (x) direction. The results indicated that the flow had a high mean

AC

velocity gradient in the transverse direction and weak gradient in the longitudinal direction, a typical characteristic of a free-shear layer. No three-dimensional eddies could be observed by the machine vision system. Therefore, the jets at the flow rate of 30 SCFH did not fully turn to turbulent flow within the distance of 15 mm, as might have been thought given the relatively high Reynolds number. The inconsistency between the observed and the predicted flow behavior could be attributed to two reasons: the major problem of using the calculated R e to evaluate the flow behavior was that the flow might not have been developed fully in the pipe, so the 13

ACCEPTED MANUSCRIPT identification of flow type might have differed from reality. Furthermore, after the jet exited the circular nozzle, the downstream continuation of the boundary layer inside the pipe surrounded the core region, which was not fully developed [35]. After the flow jet impinged on the substrate, the reflected gas with a relatively higher Reynolds number was more likely to behave turbulent

T

(Fig. 5a).

IP

Compared to a circular tube, the multiple-opening tube had several disadvantages. First,

CR

because the multiple tubes inside the outer tube blocked the gas flow, the maximum flow rate was limited to 30 SCFH. Second, when argon came out of this nozzle, the jets with steady mean

US

flow were laminar, similar to the water streams flowing out of a shower head (Fig. 5b). When the

AN

two streams collided, the flow became fuzzy. The distance between the tube and the position of the molten pool during welding was 8.72 mm. The streams started to become a transitional flow

M

due to the gas viscosity. However, there were still multiple gaps among these streams. The

ED

molten pool could not be uniformly covered using this tube, which explained the overall greater surface roughness compared to the single circular tube (Fig. 5c). From the results in Fig. 5c one

PT

can see that with an increase of gas flow rate, the surface roughness first decreases, reaches a

CE

minimum value, and then increases for both types of tubes, albeit less for the multiple-opening nozzle given its higher roughness values overall. The effect of gas flow rate on surface roughness

AC

is explained next. Based on the above results, the circular tube was selected to conduct the following experiment.

3.2.2 Effect of gas flow rate The function of a shielding gas is not just to protect the molten pool from contamination; a shielding gas can also minimize the laser beam attenuation by plasma plume [36, 37]. Fig. 6a shows that the surface roughness decreased quickly with an increase in gas flow rate from 0 to 14

ACCEPTED MANUSCRIPT 30 SCFH. The surface oscillation, plasma intensity, and microstructure of the weld were changed with the gas flow rate. Welding in air created a keyhole welding condition while the presence of the Ar at 30 SCFH created a conduction welding mode (Fig. 6b). The keyhole welding condition also exhibited a relatively high plasma intensity accompanied by strong oscillations of the

T

molten pool (Fig. 6c). Arata [38] and Wei et al. [8] demonstrated that the plasma plume is mainly

IP

generated and then impinges at the beginning of the molten pool beneath the laser beam; the

CR

pressure gradient of the plasma plume breaks the equilibrium position of the surface kept by the surface tension, resulting in an unbalanced surface. The reduced surface tension by oxidation [39]

US

even facilitates the surface movement in the form of oscillation. At the argon flow rate of 30

AN

SCFH, the plasma plume intensity was reduced greatly, and the melt flowed into the groove like a slow-moving stream with a quiescent surface. The better smoothness of molten pool surface

M

was also born out in the lower weld surface roughness measurements along the length of the

ED

weld. It is recognized that keyholes are formed above a threshold laser power density [22]. Therefore, it was possible that the transition between the two welding modes resulted from the

PT

differences in energy absorption. The molten pool with a smooth surface might be more

CE

reflective than the oscillated surface with oxidation; the welding process underwent a conduction mode at a high flow rate of argon, because a large portion of the laser was reflected off the

AC

relatively flat molten pool surface. The polished cross sections in Fig. 6d show that the keyhole mode appears to have a larger fusion zone filled with α-Al dendrites while the conduction mode generated a shallow fusion zone and a bottom capillary brazing zone distributed with Si particles. Our prior work demonstrated that the dendritic growth is promoted from spherical morphology only when the actual temperature gradient exceeds the degree of constitutional undercooling [40]. The Si phase of the shielded coupon still kept the spherical morphology in the bottom brazing 15

ACCEPTED MANUSCRIPT zone, indicating that the turbulent convection in the shielded molten pool was not as strong as in

US

CR

IP

T

the oxidized coupon.

Fig. 6 (a) Effect of gas flow rate on surface roughness; (b) images of the molten pools

AN

captured by a CCD camera; (c) spectrums of the plasma at the same scale; (d) polished

ED

M

cross sections of the welds.

PT

Fig. 7a is a series of images taken of the jets emitted from the circular tube at different gas flow rates. When the argon jet stroked the substrate, the reflected argon was split in two

CE

directions. At the flow rate of 10 SCFH, the reflected argon delivered by the circular tube had eddies around the nozzle (Fig. 7a). With an increase of argon flow rate, the reflected argon

AC

quickly dispersed into the ambient environment, and eddies were no longer captured by the CCD camera. Eddies surrounding the nozzle were avoided, as vorticities could involve oxygen from the ambient environment to the shielding gas. With an increase in argon flow rate, the plasma intensity was suppressed from 600 counts at 10 SCFH to below 350 counts at 30 SCFH (Fig. 7b). Argon is approximately 1.4 times as heavy as air, and 10 times as heavy as helium [37]; therefore, argon provided better shielding than helium, and reduced the surface roughness more effectively 16

ACCEPTED MANUSCRIPT as Fig. 6a presents. When the flow rate of argon and helium reached 40 SCFH, however, the weld surfaces were more concave and rougher than those shielded at 30 SCFH. One possible explanation is that the force generated by the shielding gas was large enough to break the balance maintained by the surface tension force. To verify this assumption, the net force exerted by a gas

PT

ED

M

AN

US

CR

IP

T

was explored.

CE

Fig.7 (a) Visualization of argon jets at different flow rates; (b) spectral intensities of plume shielded by argon; (c) flow pattern when the flow past a cylindrical wire at certain

AC

Reynolds number [41]; (d) schematic illustration of a shielding gas impinged on the molten pool surface.

When a partial amount of argon passes the wire, the net force exerted on the root of the melting wire separates into lift and drag, as illustrated in Fig. 7c. Lift (L) is the component in a direction perpendicular to the free stream velocity vector, while drag (D) is the component in a 17

ACCEPTED MANUSCRIPT direction parallel to the free stream velocity vector [33]. The solution of the potential flow illustrates the pressure distribution up to the separation points. 𝑑 𝐿 = −𝑃𝑐𝑦𝑙 (sin 𝜑)𝑏𝑟0 𝑑𝜑

(3-1)

d D= −𝑃𝑐𝑦𝑙 (cos 𝜑)𝑏𝑟0 𝑑𝜑

(3-2)

2

) (1 − 4𝑠𝑖𝑛2 𝜑)

T

𝜌𝑉𝑗 2

(3-3)

IP

𝑃𝑐𝑦𝑙 = 𝑃0 + (

CR

where 𝑃𝑐𝑦𝑙 is the argon pressure on the cylindrical wire; 𝑃0 is the pressure of the uniform flow that is considered as the ambient pressure; 𝜑 as shown in Fig. 7c is the angle in the range starting

US

from one separation point to another ( 𝜑1 ≤ 𝜑 ≤ 𝜑2 ) ; b is the length of filler wire to be

AN

calculated; 𝑟0 is the radius of filler wire; and 𝜌 is the density of argon. Eq. (3-2) reveals that increasing the flow velocity enhances the drag, and the root of the cylindrical wire is pressed

M

toward the liquid surface by the net force. If the separation points are symmetrical, the lift in Eq.

ED

(3-3) is cancelled out during the integration from 𝜑1 to 𝜑2 . Otherwise, the shear effect of lift might lead to the detachment and re-attachment of the molten wire to one side of the panel

PT

randomly when wetting is insufficient inside the groove. The anchoring force, FA, which is

CE

required to hold the surface stationary, shown in Fig. 7d, can be calculated by Reynolds transport theorem and the conversation of mass.

AC

𝐹𝐴 = 𝜌𝑞𝑗 2 sin 𝜃 /𝐴

(4)

where 𝜌 is the density of argon; 𝑞𝑗 is the flow rate of the gas; A is the cross-sectional area of the nozzle; and 𝜃 is the smaller angle between the tube and the molten pool surface. Eq. 4 also proves that the gas flow rate (𝑞𝑗 ) does have an effect on the anchoring force, which is the normal component of the capillary force at the boundary of molten pool. Moreover, there is a balance among the hydrostatic pressure, the effect of surface tension, and the applied gas pressure. The 18

ACCEPTED MANUSCRIPT pressure jump at a curved surface in the top or bottom follows the Young-Laplace equation given by [42]: 1

1

∆ p = γ (𝑅 + 𝑅 ) = 2γH 1

(5)

2

where ∆ p is the pressure difference among the liquid, gas, and hydrostatic pressures across the

IP

T

fluid interface; γ is the surface tension; 𝑅1 and 𝑅2 are two principal radii of a surface; and H is

CR

the mean curvature. The Young-Laplace equation shows that if the surface tension is considered constant, the surface becomes more concave with the increase in gas pressure, which can be

US

controlled by the gas flow rate. When the liquid wire was pressed into the curved groove, the surface of the molten pool suffered from instability due to the enhanced free-surface deformation

AN

[5]. In summary, the tradeoff between the promoted oscillation induced by the surface

M

deformation and the reduced oscillation by preventing oxidation influences the surface roughness.

rate.

PT

3.2.3 Effect of nozzle position

ED

This explains why surface roughness did not decrease monotonically with the increase in flow

If the nozzle was too close to the molten pool, it would be burnt by the plasma plume. For

CE

this reason, the minimum horizontal direction of the nozzle as presented in Fig. 8a was 5 mm

AC

from the center of the laser beam. When argon flowed at 30 SCFH towards the molten pool, the surface roughness increased dramatically with the increase in the horizontal distance (Fig. 8a). The flow field at the exit of a circular jet could be defined by three regions: a core region, a transition region, and a fully developed region [35]. After the jet exited the circular nozzle, the convection speed diminished, and the shear layer thickness spread linearly with the increase of the downstream distance in the core region. According to Durbin [35], significant entrainment takes place in the core region that covers an axial distance of 4 to 5 nozzle diameter (17.6 ~22 19

ACCEPTED MANUSCRIPT mm for the current nozzle). The transition region is from 10 to 15 diameters (44 ~66 mm). Therefore, the flow of shielding gas at the exit of the circular tube during welding was in the core region. As the jet evolved downstream, the entrainment rate in the near field increased from zero at the nozzle to a constant rate in the far field [43]. This increased rate meant that a greater

T

amount of oxygen was involved in the flow. A horizontal distance between the center of the

IP

nozzle and the laser beam of 5 mm was implemented when the inclination angle of gas tube was

AC

CE

PT

ED

M

AN

US

CR

adjusted.

Fig. 8 (a) Effect of the distance between nozzle and laser beam on surface roughness; (b) schematic illustration of a jet flow; (c) spectral intensities of plume shielded by argon with the inclination angles (θ) of 15° and 55° to the horizontal plane; (d) effect of inclination angle of gas tube on surface roughness.

20

ACCEPTED MANUSCRIPT The intensity of the spectrum in Fig. 8c at 55° is lower than at 15°, indicating that an increase in the angle of argon flow suppresses the plasma plume more efficiently. Accordingly, the surface roughness decreased quickly when the inclination angle of the argon tube increased from 5° to 35° (Fig. 8d). In contrast, increasing the inclination angle could have also enlarged the

T

vertical component of flow rate and the force on the liquid surface as revealed by Eq. 4. As for

IP

argon, the values of surface roughness became stable in the angle range of 35° to 55° (Fig. 8d). A

CR

balance between the surface deformation due to the vertical force and the suppression of the plasma plume might have contributed to this stability. Because helium is lighter than argon, the

US

surface roughness decreased with an increase in the inclination angle at the flow rate of 30 SCFH

AN

(Fig. 8d).

3.2.4 Parameter optimization using Taguchi Method

M

Continuous efforts to achieve stable and predictable process results are of vital importance

ED

to the high-volume production of non-load-bearing aluminum joints. Taguchi method [30, 44] was used to determine the optimal set of shielding gas parameters in consideration of the

PT

simultaneous effects of many parameters. The above single-factor study showed that front-wire-

CE

feeding method was effective to mitigate humping; argon generated lower surface roughness than helium at the same gas parameters, and the optimum nozzle was in a circular shape in the

AC

cross section. Therefore, only the tube angle (TA), flow rate (FR), and horizontal distance between laser and nozzle (D) were selected as the independent control factors (see Table 4). Experiments were carried out as the orthogonal array L9 shown in Table 5, and surface roughness of four coupons in each experiment was measured. In the Taguchi design, signals are defined as the effect of selected factors on the average responses while the noises are the deviations from the average responses [45]. The calculation of signal-to-noise ratio (SNR) intended to identify 21

ACCEPTED MANUSCRIPT the response to the variability of noise effects. As the goal was to minimize the surface roughness, the-smaller-the-better criterion was used for SNR analysis. The SNR is determined by the following expression [30, 44]: 1

SNR = -10 log(n ∑ni=1 yi2 )

(6)

T

where yi is the experiment response for the ith performance characteristic.

(°)

(SCFH)

1

30

20

2

45

30

3

60

AN

(mm) 5 10 15

CE

PT

ED

M

40

Factor C: Horizontal distance

CR

Factor B: Flow rate

US

Factor A: Tube angle

AC

Level

IP

Table 4 Gas parameters and their levels in the experimental design

22

ACCEPTED MANUSCRIPT

Factor A:

Factor B:

Factor C:

Ra 1

Ra 2

Ra 3

Ra 4

Aver.

Exp.

Tube

Flow rate

Distance

(µm)

(µm)

(µm)

(µm)

Ra

No.

angle (°)

(SCFH)

(mm)

1

35

20

5

1.57

1.72

1.7

1.49

1.62

-3.670

2

45

30

10

1.58

1.21

1.23

T

Table 5 Taguchi design matrix, measured response and the calculated SNR SNR

1.58

1.4

-2.091

3

55

40

15

2.11

2.6

2.07

2.22

2.25

-6.153

4

35

30

15

1.9

1.8

1.77

1.73

1.8

-4.387

5

45

40

5

0.893

0.808

0.935

0.871

1.013

6

55

20

10

1.25

1.24

1.28

1.48

1.313

-1.567

7

35

40

10

8

45

20

15

9

55

30

5

US

1.37

1.45

1.26

1.33

1.353

-2.153

1.91

1.76

1.8

1.74

1.803

-4.388

0.904

0.915

0.974

0.993

0.947

0.454

PT

ED

M

0.847

AN

CR

IP

(µm)

The calculated SNR values of three control factors according to each level in Fig. 9 reveal

CE

that the horizontal distance between nozzle and laser was more significant in affecting the surface roughness than the other two parameters. SNR values agreed with the above analysis that

AC

increasing the gas flow rate and tube angle would lead to better protection against oxidation while at the same time promote the surface deformation. The balance between these two factors influenced the weld surface roughness. Since a larger SNR value represented a lower sensitivity of response to the noise, the process operation with the highest SNR corresponded to the optimum quality with minimum deviation [30]. According to Fig. 9, the optimal processing parameters were the tube angle of 45°, gas flow rate of 30 SCFH, and horizontal distance of 5 23

ACCEPTED MANUSCRIPT mm between the laser and the nozzle. A weld surface roughness of 0.816± 0.068 µm could be

CR

IP

T

achieved in these optimal parameters.

US

Fig. 9 Effects of processing parameters on the SNR for surface roughness

AN

4 Conclusions

Controlling the flow of filler wire into the molten pool and optimizing the shielding gas

M

parameters were effective methods to reduce humping and rippling phenomena on the molten

ED

pool surface. The weld surface roughness below 1 μm could be achieved as follows.  The front-wire-feeding method could mitigate the humping on the molten pool surface and

PT

improve wetting and spreading of the liquid filler wire.

CE

 During fiber laser joining of aluminum alloys, argon achieved a lower surface roughness than helium at the same parameters.

AC

 Compared to the nozzles with rectangular cross sections and a circular nozzle with multiple openings, a normal circular nozzle with an inner diameter of 4.4 mm generated the best weld surface quality.  Taguchi experiments showed that the optimal gas parameters for a low surface roughness of aluminum welds were the argon flow rate of 30 SCFH, a nozzle inclination angle of 45°, and a horizontal distance of 5 mm between the center of nozzle and laser beam. 24

ACCEPTED MANUSCRIPT

Acknowledgements The financial supports by NSF Grant No. IIP-1539853 and General Motors are acknowledged. The authors would like to thank Mr. Andrew Socha, a research engineer, at the Research Center

T

for Advanced Manufacturing, Southern Methodist University, for his help during experiments.

AC

CE

PT

ED

M

AN

US

CR

IP

No conflicts of interest exist.

25

ACCEPTED MANUSCRIPT REFERENCES [1] Association EA. The aluminium automotive manual. EAA, available at. 2013. [2] Xuesong Wang H-PW, Blair E. Carlson, Jianping Lin, Michael Poss, Joshua Solomon. Development of Surface Evaluation Methods for Class A Laser Welds. AWS Detroit Sheet Metal

T

Welding Conference XVII. 2016.

IP

[3] Leggett P, Carlson, Blair, Wang, Hui-Ping, and Kovacevic, Radovan. Optimization of Split-

CR

Beam laser Braze-Welding of AA 6111 in a Coach-Peel Joint Configuration using a Hybrid

US

Taguchi-Grey Relation Grade Method. Proc of the Sheet Metal Welding Conference XVI 2014.

manufacturing. Scientific reports. 2015;5.

AN

[4] Wei H, Mazumder J, DebRoy T. Evolution of solidification texture during additive

M

[5] Wei P. The physics of weld bead defects: INTECH Open Access Publisher; 2012.

heat transfer. 1996;118:960-9.

ED

[6] Wei P, Chang C, Chen C. Surface ripple in electron-beam welding solidification. Journal of

PT

[7] Wei P, Chuang K, DebRoy T, Ku J. Scaling of spiking and humping in keyhole welding. Journal

CE

of Physics D: Applied Physics. 2011;44:245501. [8] Wei P, Chuang K, Ku J, DebRoy T. Mechanisms of spiking and humping in keyhole welding.

AC

IEEE transactions on components, packaging and manufacturing technology. 2012;2:383-94. [9] Aziz M. Model for solute redistribution during rapid solidification. Journal of Applied Physics. 1982;53:1158-68. [10] Siegman AE, Fauchet PM. Stimulated Wood's anomalies on laser-illuminated surfaces. Quantum Electronics, IEEE Journal of. 1986;22:1384-403.

26

ACCEPTED MANUSCRIPT [11] Ang L, Lau Y, Gilgenbach R, Spindler H, Lash J, Kovaleski S. Surface instability of multipulse laser ablation on a metallic target. Journal of applied physics. 1998;83:4466-71. [12] Zhou X, Wang D, Liu X, Zhang D, Qu S, Ma J, et al. 3D-imaging of selective laser melting defects in a Co–Cr–Mo alloy by synchrotron radiation micro-CT. Acta Materialia. 2015;98:1-16.

IP

T

[13] Fujimura K, Ogawa M, Seki M. Possible mechanism of the roughness formation on a liquid

CR

layer caused by a high heat flux. Fusion engineering and design. 1992;19:183-91. [14] Trivedi R, Kurz W. Morphological stability of a planar interface under rapid solidification

US

conditions. Acta metallurgica. 1986;34:1663-70.

[15] Brooks J, Mahin K. Solidification and Structure of Welds'. Welding-Theory and Practice.

AN

1990:35-78.

M

[16] Kotecki D, Cheever D, Howden D. Mechanism of ripple formation during weld solidification.

ED

WELD J. 1972;51:368.

[17] Grigorenko G, Pomarin YM, Orlovsky VY. Theoretical and experimental investigation of the

PT

thermodynamic and kinetic nitrogen absorption by liquid alloys. Materials science forum: Trans

CE

Tech Publ; 1999. p. 25-30.

[18] Le H, Vuillon J, Zeitoun D, Marine W, Sentis M, Dreyfus R. 2D modeling of laser-induced

AC

plume expansion near the plasma ignition threshold. Applied surface science. 1996;96:76-81. [19] Kuo T-Y, Lin Y-T. Effects of Shielding Gas Flow Rate and Power Waveform on Nd: YAG Laser Welding of A5754-O Aluminum Alloy. Materials transactions. 2006;47:1365-73. [20] Sathiya P, Mishra MK, Shanmugarajan B. Effect of shielding gases on microstructure and mechanical properties of super austenitic stainless steel by hybrid welding. Materials & Design. 2012;33:203-12. 27

ACCEPTED MANUSCRIPT [21] Pan LK, Wang CC, Hsiao YC, Ho KC. Optimization of Nd: YAG laser welding onto magnesium alloy via Taguchi analysis. Optics & Laser Technology. 2005;37:33-42. [22] Kuo T-Y, Lin Y-D. Effects of different shielding gases and power waveforms on penetration characteristics and porosity formation in laser welding of Inconel 690 alloy. Materials

IP

T

transactions. 2007;48:219-26.

CR

[23] Katayama S, Naito Y, Uchiumi S, Mizutani M. Physical phenomena and porosity prevention mechanism in laser-arc hybrid welding. TRANSACTIONS-JWRI. 2006;35:13.

US

[24] Seto N, Katayama S, Matsunawa A. High-speed simultaneous observation of plasma and keyhole behavior during high power CO 2 laser welding: effect of shielding gas on porosity

AN

formation. Journal of Laser Applications. 2000;12:245-50.

M

[25] Quintino L, Miranda R, Williams S, Kong C. Gas shielding in fibre laser welding of high

ED

strength pipeline steel. Science and Technology of Welding and Joining. 2011;16:399-404. [26] Motlagh NH, Parvin P, Jandaghi M, Torkamany M. The influence of different volume ratios

PT

of He and Ar in shielding gas mixture on the power waste parameters for Nd: YAG and CO 2

CE

laser welding. Optics & Laser Technology. 2013;54:191-8. [27] Chung B, Rhee S, Lee C. The effect of shielding gas types on CO 2 laser tailored blank

AC

weldability of low carbon automotive galvanized steel. Materials Science and Engineering: A. 1999;272:357-62.

[28] Reisgen U, Schleser M, Mokrov O, Ahmed E. Shielding gas influences on laser weldability of tailored blanks of advanced automotive steels. Applied Surface Science. 2010;257:1401-6. [29] Campana G, Ascari A, Fortunato A, Tani G. Hybrid laser-MIG welding of aluminum alloys: the influence of shielding gases. Applied surface science. 2009;255:5588-90. 28

ACCEPTED MANUSCRIPT [30] Xiansheng N, Zhenggan Z, Xiongwei W, Luming L. The use of Taguchi method to optimize the laser welding of sealing neuro-stimulator. Optics and Lasers in Engineering. 2011;49:297304. [31] Berger P, Hügel H, Hess A, Weber R, Graf T. Understanding of humping based on

IP

T

conservation of volume flow. Physics Procedia. 2011;12:232-40.

CR

[32] Deyev GF. Surface phenomena in fusion welding processes: CRC Press; 2005. [33] Fox RW, McDonald AT, Pritchard PJ. Introduction to fluid mechanics: John Wiley & Sons

US

New York; 1985. [34] Holman J. Heat transfer, 9th. McGraw-Hill; 2002.

AN

[35] Durbin PA, Reif BP. Statistical theory and modeling for turbulent flows: John Wiley & Sons;

M

2011.

ED

[36] Walsh C. Laser welding–literature review. Materials Science and Metallurgy Department, University of Cambridge, England. 2002.

PT

[37] Praxair. Shielding gases selection manual. 2011.

processing. 1986.

CE

[38] Arata Y. Plasma, electron and laser beam technology: development and use in materials

AC

[39] Backhaus-Ricoult M. A model of oxygen-activity-dependent adsorption (desorption) to metal–oxide interfaces. Acta materialia. 2000;48:4365-74. [40] Yang G, Ma J, Carlson B, Wang H-P, Kovacevic R. Effect of laser beam configuration on microstructure evolution and joint performance in laser joining AA 6111 panels. Materials & Design. 2017;123:197-210. [41] Munson BR, Young DF, Okiishi TH. Fundamentals of fluid mechanics: New York; 1990. 29

ACCEPTED MANUSCRIPT [42] Adamson AW, Gast AP. Physical chemistry of surfaces. 1967. [43] Liepmann D, Gharib M. The role of streamwise vorticity in the near-field entrainment of round jets. Journal of Fluid Mechanics. 1992;245:643-68.

IP

Relational Analysis in the Taguchi Method. JOM. 2016;68:1762-73.

T

[44] Zhang Z, Kovacevic R. Multiresponse Optimization of Laser Cladding Steel+ VC Using Grey

CR

[45] Dongxia Y, Xiaoyan L, Dingyong H, Zuoren N, Hui H. Optimization of weld bead geometry in laser welding with filler wire process using Taguchi’s approach. Optics and Laser Technology.

AC

CE

PT

ED

M

AN

US

2012;44:2020-5.

30

US

CR

IP

T

ACCEPTED MANUSCRIPT

AC

CE

PT

ED

M

AN

Graphical Abstract

31