Effects of Process Parameters on the Bead Shape in the Tandem Gas Metal Arc Welding of Aluminum 5083-O Alloy

Abstract

In gas metal arc welding (GMAW), the weld bead shape is an important factor that is directly related to the weld quality of welded joints. This study investigates the effects of process parameters, including welding speed (WS) and leading and trailing wire feed rates (WFR), on the weld bead shape, including the leg length and penetration depth, in the tandem GMAW of aluminum 5083-O alloy. An asynchronous direct current–direct current pulse tandem GMAW system and a tandem GMAW torch were designed and applied to improve welding productivity and welding quality. Response surface methodology was used to analyze the effects of the process parameters on the weld bead shape and to estimate regression models for predicting the weld bead shape. As a result of observing arc behavior using a high-speed camera, it was confirmed that the leading WFR affects the penetration depth and the trailing WFR affects the leg length. The coefficient of determination (R²) of the regression models was 0.9414 for the leg length and 0.9924 for the penetration depth. It was also validated that the estimated models were effective in predicting the weld bead shape (leg length and penetration depth) representative of weld quality in the tandem GMAW process.

1. Introduction

The application of aluminum in automotive parts is increasing to reduce vehicle weight. Various joining processes are applied to join aluminum parts [1,2]. Among the various joining processes, gas metal arc welding (GMAW) is an efficient welding process that is particularly suitable for welding metal plates. Whereas a general GMAW process uses a single arc, the tandem GMAW process used in this study comprises two independent welding machines, each with its own power source, welding torch, wire drive, and welding wire. Owing to its system characteristics, the tandem GMAW process has a higher deposition rate and welding speed than the conventional single-wire GMAW process, implying that the tandem GMAW process can improve welding productivity [3,4,5,6,7,8]. However, the welding system and arc behavior of tandem GMAW are more complex than those of conventional single-wire GMAW such that it is necessary to establish the effects of process parameters on the weld bead shape, which is directly related to weld quality, and welding quality in tandem with the GMAW process. Kolahan et al. [8] developed a model to analyze the influence of the wire feed rate (WFR), torch angle, welding speed, and nozzle-to-plate distance on the weld bead height, width, and penetration depth in the GMAW process through the design of experiments (DOEs). The fit of the model was confirmed through an analysis of variance (ANOVA). Saravanan et al. [9] optimized the strength according to the welding parameters (current, voltage, gas flow rate, torch angle, welding speed, wire diameter, and electrode feed rate) in aluminum GMAW using the Taguchi method. Jayaganesh et al. [10] optimized the tensile strength and mass deposition rate according to the welding current, WFR, and welding speed among the GMAW parameters using the DOE. The results confirmed that the WFR and welding current are important variables. Venkadeshwaran et al. [11] derived the optimal ultimate tensile strength through an L9 orthogonal array by optimizing the welding current, voltage, and gas flow rate among the welding parameters in the GMAW of aluminum alloy. Ramarao et al. [12] optimized the welding parameters (welding current, voltage, and bevel angle) to obtain better impact strength using the Taguchi L9 orthogonal array in GMAW and proved that the current and voltage are the main factors through ANOVA. Duan et al. [13] performed GMAW on metals with narrow gaps and optimized welding parameters, such as the rotation angular velocity, rotation angular amplitude, WFR, welding speed, and sidewall stay time, to optimize the welding quality. Tham et al. [14] performed GMAW on the shape of a T-fillet joint and developed a mathematical model to predict the bead shape according to the welding parameters (welding current, welding voltage, welding speed, and wire extension). Consequently, the deviation between the predicted and actual values was less than 1.0 mm. Yu et al. [15] investigated the effects of welding current and torch position parameters, including the torch-aiming position, travel angle, and work angle, on the bead geometry in a single-lap joint GMAW. Shen et al. [16] analyzed the effect of welding parameters on the formation of gas metal arc welding—gas tungsten arc welding (GMAW–GTAW) double-arc welding and provided optimal conditions for obtaining high-quality and good forming. Rodriguez et al. [17] compared a 6.35 mm thick 6061 aluminum alloy with the GMAW process using the hybrid GTAW–GMAW and confirmed that the GTAW–GMAW promoted grain structure refinement and reduced porosity in the welded metal. In addition, the GTAW–GMAW process reduced the degradation in the heat-affected zone under the as-weld condition and increased the welding speed by 40%. Lee et al. [18] developed a model that optimizes the parameters through a Gaussian process regression model for the bead shape in a tandem flux cored arc welding process. Wu et al. [19] studied the temperature and fluid flow fields using a three-dimensional numerical model in the tandem GMAW process and investigated the resulting molten pool formation, convection, and stability. Häßler et al. [20] investigated the arc stabilization effect of filler wires using high-speed recordings and numerical calculations in a tandem GMAW process. However, research that focused on the effects of the process parameters on the weld bead shape in tandem GMAW of aluminum alloys for automotive parts is insufficient. Wu et al. [21] investigated the welding bead shape for each pulse waveform with different phase shifts by synchronizing the welding power source in the aluminum alloy tandem GMAW. Synchronous control requires additional hardware and control software.

In this study, the effects of process parameters, including welding speed (WS) and leading and trailing WFRs, on the weld bead shape, including leg length and penetration depth, in the tandem GMAW of aluminum 5083-O alloy were investigated using an asynchronous direct current–direct current (DC–DC) pulse tandem GMAW system. Response surface methodology (RSM) was used to investigate the effects of the welding parameters on the weld bead shape and to estimate regression models for predicting the weld bead shape.

2. Materials and Methods

2.1. Materials

Aluminum 5083-O alloy, used in automotive parts, such as a cowl cross bar, was used.

Table 1. Chemical composition of aluminum 5083-O alloy (wt.%).

Si	Fe	Cu	Mn	Mg	Cr	Zn	Ti	Al
0.11	0.31	0.05	0.66	4.51	0.09	0.03	0.01	94.2

summarizes the chemical composition of the aluminum 5083-O alloy. The thickness of the aluminum 5083-O alloy used in this study is 1.5 and 2.5 mm. Figure 1 shows the configuration of the single-lap-joint. The workpiece was cut into dimensions of 150 mm × 150 mm, and an overlap width of 20 mm was used for the welding test.

2.2. Welding Equipment

A synchronous tandem GMAW system synchronizes the waveforms from the power sources of the two welding machines that have a phase difference of a certain value between the two waveforms. Additional components such as the controller and coupling module and complex control logic are required to control the two welding machines of the synchronous tandem GMAW. In contrast, an asynchronous tandem GMAW system has the advantage of being able to configure the system without additional complex components, and each welding machine of the asynchronous tandem GMAW system independently operates and outputs its own waveform. Figure 2 shows the schematic of the asynchronous tandem welding equipment used in this study, which comprises two inverter-based welding power sources: Welbee W350 (Daihen Corporation, Osaka, Japan) for the leading wire and Welbee P500L (Daihen Corporation, Osaka, Japan) for the trailing wire. The tandem control type is asynchronous, and DC pulse is selected for both leading and trailing current types. Figure 3 shows the welding current and voltage waveforms at a leading wire feed rate of 9.5 m/min and a trailing wire feed rate of 5.0 m/min. The current of the welding power source is controlled individually for the two electrodes without the hardware to control synchronization. In general, the tandem GMAW process can increase productivity such as the welding speed and high deposition rate. However, the large torch size of the tandem GMAW system may limit the applications of the tandem GMAW system because the large torch causes more interference with workpieces and jigs compared to the torch of a conventional single-wire GMAW system. To address this limitation, a tandem torch with a smaller size than the existing tandem torch was designed and fabricated in the study. The designed torch has heat resistance so that it can be used for a tandem total current of 500 A. A design process was performed to select angles and distances to prevent interference between contact tips. The outer diameter of the contact tip is 5 mm, and the distance between the wires is 7 mm. The outer diameter of the torch nozzle is 24 mm, which is smaller than the 25 mm outer diameter of the existing single-torch nozzle. A schematic of the torch used in the test is shown in Figure 4.

2.3. Welding Conditions

A synchronous

Table 2. Welding conditions.

Condition	Leading Arc	Trailing Arc
Power source	Welbee P500L	Welbee W350
Current type	DC pulse (Polarity: DCEP)
Filler wire	ER5356 (diameter: 1.2 mm)
Wire feed rate (m/min)	2.6–9.5
Welding speed (cm/min)	123–157
Work angle (°)	30
CTWD (mm)	17
Shielding gas	Ar (15 L/min)	Ar (15 L/min)

lists the welding conditions. In both GMAW machines of the tandem welding system, the current type used a DC pulse for the polarity of the direct current electrode positive. A 1.2 mm diameter ER5356 welding wire specified in AWS A5.10/A5.10M:2021 was used, and the WS was 123–157 cm/min. The work angle was 30°, and the range of WFR was 2.6–9.5 m/min. The contact tip-to-workpiece distance was set as shown in Figure 5, which is a schematic diagram of the tandem GMAW.

2.4. Analysis Method

This study investigated the effects of three welding parameters (independent variables), including WS (X₁), leading WFR (X₂), and trailing WFR (X₃), on the response variables, leg length (Y_leg), and penetration depth (Y_pen), as shown in Figure 6. Optical microscopy was used to observe the response variables. Polishing and etching were performed to measure the cross-sectional shape of the weld beads. Sodium hydroxide was used for etching. Response variables were identified using a microscope connected to image analysis software. A central composite design (CCD) and RSM were used to analyze the effects of the welding parameters. The CCD describes the relationship between the independent and response variables and estimates a second-order response surface model. CCD with three factors (X₁, X₂, and X₃ in this study) consists of three parts: eight factorial points, six center points, and eight axial points. The position $\mathsf{\alpha }$ value of the axial point was determined by the number of experiments at the origin, which is the center point of the region of interest of the CCD. Because it has three parameters, the value of $\mathsf{\alpha }$ is 1.682. The factor levels of the natural and design units used in the experiment are listed in

Table 3. Factor levels in natural and design units.

	Design Units
Factor	−1.682	−1	0	1	1.682
X1 (cm/min)	123	130	140	150	157
X2 (m/min)	2.6	4.0	6.0	8.0	9.4
X3 (m/min)	2.6	4.0	6.0	8.0	9.4

. A high-speed camera was used to investigate the effects of the welding parameters on the behavior of the molten pool, and a schematic is shown in Figure 7. To acquire the image, an illumination laser with a wavelength of 808 nm was used, and 808 nm band pass filters with full width at half maximum (FWHM) values of 10 and 5 were sequentially applied to the camera. To prevent excessive exposure to strong arc light, a neutral density filter (ND filter) was used.

3. Results and Discussion

3.1. Effect of Welding Parameters on Bead Shape

Table 4. Experimental design matrix.

Run	X1	X2	X3	Yleg			Ypen
1	−1	−1	−1	3.5	4.6	4.4	0.0	0.0	0.0
2	1	−1	−1	3.6	4.3	3.6	0.0	0.3	0.0
3	−1	1	−1	5.8	5.6	5.7	1.7	1.8	1.6
4	1	1	−1	5.0	5.2	5.0	1.3	1.3	1.3
5	−1	−1	1	6.0	6.2	6.2	2.0	1.7	1.8
6	1	−1	1	5.5	5.7	5.4	1.1	1.2	1.0
7	−1	1	1	7.5	7.6	7.6	4.8	5.0	5.0
8	1	1	1	7.7	6.7	7.4	4.0	4.3	4.3
9	−1.682	0	0	6.1	5.8	6.1	1.8	2.1	2.0
10	1.682	0	0	5.2	4.9	4.8	1.2	0.8	0.9
11	0	−1.682	0	3.8	5.0	4.1	0.0	0.3	0.2
12	0	1.682	0	7.4	7.0	7.5	4.0	4.1	3.9
13	0	0	−1.682	3.6	3.9	3.5	0.0	0.2	0.0
14	0	0	1.682	7.0	6.9	7.0	3.9	4.1	3.8
15	0	0	0	5.4	5.3	5.2	1.2	1.1	1.0
16	0	0	0	5.3	5.0	4.9	1.3	1.1	1.0
17	0	0	0	5.4	5.3	5.1	1.4	1.1	1.2
18	0	0	0	5.3	4.9	4.9	1.3	1.2	1.2
19	0	0	0	5.2	5.1	4.8	1.3	1.1	1.1
20	0	0	0	5.1	5.1	5.0	1.0	1.0	1.0

summarizes the experimental design matrix with the test results, that is, the measured response variables (Y_leg, Y_pen), according to the welding conditions. To observe the effects of each independent variable on the bead shape, macrographs of the weld cross-section at the six cubic points were compared.

Table 5. Bead shape change according to welding parameters.

Parameter	Low Level	High Level
X1	Run 5	Run 6
X2	Run 2	Run 4
X3	Run 2	Run 6

lists the effect of each parameter on the actual weld shape after fixing two of the three parameters. X₂ and X₃ were fixed (X₂: 4 m/min, X₃: 8 m/min), and when the X₁ was increased, Y_leg decreased from 6.05 to 5.34 and Y_pen decreased from 1.97 to 1.05. X₁ and X₃ were fixed (X₁: 150 cm/min, X₃: 4 m/min), and when X₂ was increased, Y_leg increased from 3.52 to 4.81 and Y_pen increased from 0.00 to 1.16. In the case of X₃ under the test condition (X₁: 150 cm/min, X₂: 4 m/min), when X₃ was increased, Y_leg increased from 3.52 to 4.81, and Y_pen increased from 0.00 to 1.16. Figure 8 shows the main effect plots for Y_leg and Y_pen. Both Y_leg and Y_pen increased as X₁ decreased and X₂ and X₃ increased. Furthermore, X₂ and X₃ exhibited strong linearity with the response variables compared with X₁.

3.2. Model Estimation for Weld Bead Shape

3.2.1. Leg Length

RSM was applied to quantitatively explain the relationship between the welding parameters and bead shape.

Table 6. ANOVA results of the reduced model for Y_leg.

Source	DF	Adj. SS	Adj. MS	F	p
Regression	5	71.40	14.28	173.37	0.00
Linear	3	68.18	22.73	275.91	0.00
X1	1	3.07	3.07	37.26	0.00
X2	1	27.12	27.12	329.31	0.00
X3	1	37.96	37.96	461.17	0.00
Square	2	3.22	1.61	19.56	0.00
X1 × X1	1	0.80	0.80	9.66	0.00
X2 × X2	1	2.66	2.66	32.33	0.00
Residual error	54	4.45	0.08	-	-
Lack of fit	39	4.14	0.11	5.22	0.06
Pure error	15	0.31	0.02	-	-
R2 = 0.9414	-	-	-	-	-
Adj R2 = 0.9359	-	-	-	-	-

summarizes the ANOVA results for Y_leg. The significance level ($\mathsf{\alpha }$) was set at 0.05. Therefore, the term with a p-value ≤ 0.05 was determined as a significant term. Because the p-values of the interaction and square terms X₁ × X₂, X₂ × X₃, X₁ × X₃, and X₁ × X₂ were over 0.05, respectively, these terms were determined to be insignificant and were excluded from the reduced model through backward elimination. In addition, F-value is an indicator for comparing sample groups and shows how far apart the mean sample group is between groups. Therefore, a high F-value means that it has a major influence on the independent variable. As a result of the ANOVA, the X₃ F-value was highest among the welding parameters. So, that means X₃ has a major influence on the Y_leg. The coefficient of determination (R²) was 0.9414, indicating that the estimated model can predict 94.14% of the data.

$$Y_{leg}=32.38-0.405X_1-0.334X_2+0.4814X_3+0.00135X_1^2+0.0617X_2^2$$

(1)

3.2.2. Penetration Depth

To analyze the penetration depth, the relationship between welding parameters and bead shape was quantitatively explained using RSM, identical to the analysis of leg length.

Table 7. ANOVA results of the reduced model for Y_pen.

Source	DF	Adj. SS	Adj. MS	F	p
Regression	9	121.08	13.45	722.24	0.00
Linear	3	108.28	36.09	1937.66	0.00
X1	1	2.61	2.61	140.24	0.00
X2	1	53.10	53.10	2850.50	0.00
X3	1	52.57	52.57	2822.24	0.00
Square	3	8.11	2.70	145.16	0.00
X1 × X1	1	0.57	0.57	30.46	0.00
X2 × X2	1	4.78	4.78	256.74	0.00
X3 × X3	1	3.97	3.97	213.27	0.00
Interaction	3	4.69	1.56	83.89	0.00
X1 × X2	1	0.09	0.09	5.03	0.03
X1 × X3	1	0.51	0.51	27.40	0.00
X2 × X3	1	4.08	4.08	219.24	0.00
Residual error	48	0.93	0.02	-	-
Lack of fit	33	0.77	0.02	2.02	0.07
Pure error	15	0.16	0.01	-	-
R2 = 0.9924	-	-	-	-	-
Adj R2 = 0.9910	-	-	-	-	-

summarizes the analysis results for Y_pen. As a result of the ANOVA the X₂ F-value was highest among the welding parameter. So that mean X₂ has the major influence on the Y_pen. The p-value of all the terms were under the 0.05. The coefficient of determination (R²) was 0.9924, indicating that the estimated model can predict 99.24% of the data.

$$\begin{array}{cc}{Y}_{pen}=& 21.00-0.2835X_1-0.610X_2+0.059X_3+0.001146X_1^2+0.08315X_2^2\\ & +0.07578X_3^2-0.00312X_1{X}_3-0.00729X_1{X}_3+0.10313X_2{X}_3\end{array}$$

(2)

3.3. Analysis of Phenomena through a High-Speed Camera

The effects of the welding parameters on the behaviors of the arc and molten pool were investigated based on the images acquired from the high-speed camera in Figure 7. Figure 9 shows high-speed images. The high-speed filming was performed with a bead on plate configuration, and the welding conditions were applied at 5 m/min for both X₂ and X₃. As shown in Figure 9 and Figure 10, the leading arc directly strikes the solid-state base metal, which forms a narrow molten pool. Therefore, the leading arc had a direct effect on the penetration depth. The trailing arc is transferred to the molten pool created by the leading arc, which expands the molten pool. Moreover, the solid surface induced by the leading arc force is filled with the molten metal generated by the trailing arc, the expansion of the molten pool remains stably maintained, and Y_leg increases [22]. A schematic of the behavior of the molten pool and arc is shown in Figure 10. The configuration of the lap fillet joint with the 2 mm thickness top plate and the 4 mm thickness bottom plate was used for high-speed filming.

Table 8. High-speed camera images of the molten pool and macrographs of the bead cross-section according to welding conditions.

Welding Condition	X2 (m/min)	X3 (m/min)	X2 (m/min)	X3 (m/min)
	9.5	5.0	5.0	9.5
High-speed camera images
Cross-section
Yleg (mm)	7.6		8.0
Ypen (mm)	3.0		2.4

shows the welding conditions, high-speed images, and cross-sections of the welding bead. WFR was fixed at 9.5 and 5.0 to observe the bead shape according to X₂ and X₃ when they have the same thermal energy. When the value of X₂ was higher than that of X₃, the Y_leg was 7.6 mm and the Y_pen was 3.0 mm. On the other hand, when X₃ was higher, the Y_leg was 8.0 mm and the Y_pen was 2.4 mm, that is, in the lap fillet joint as well, it was confirmed that the leading arc had an effect on the penetration depth formation, and the trailing arc had an effect on the leg length formation. When the value of X₂ was higher than that of X₃, a large number of porosities were observed in the weld metal. According to previous studies, porosities escape in the vertical direction [23]. When the penetration is deep, porosities become trapped because there is insufficient time to escape before solidification. In terms of material defects as well as weld bead geometry affected by process variables, the deformation of the material under the same heat input condition is less than that of single welding because the leading and trailing arcs are separated in tandem welding. Additionally, in the case of porosities, as mentioned above, the flow of the tandem molten pool is better than that of single arc welding, so trapped porosities are quickly released to the outside.

3.4. Model-Based Optimization and Validation

Through this experiment, a regression model that can predict Y_leg and Y_pen was derived, and R² was 94% for Y_leg and 99% for Y_pen. Figure 11 shows contour plots of the response variables with respect to the independent variables. Y_leg and Y_pen decreased as X₁ increased within the range of X₁ (123–157 cm/min). As X₂ and X₃ increased, the response variables exhibited the same trend. Points 1–3 in Figure 11 are the test points for the estimation models.

Table 9. Validation conditions and results.

Point	Welding Parameter	Item	Value		Deviation		Cross-Section
			Yleg (mm)	Ypen (mm)	Yleg (mm)	Ypen (mm)
1	(X1) 140 cm/min; (X2) 6.6 m/min; (X3) 5.8 m/min	Predicted values	5.9	1.2
		Actual value 1	6.0	1.4	0.1	0.2
		Actual value 2	6.1	1.3	0.2	0.1
		Actual value 3	6.1	1.3	0.2	0.1
2	(X1): 140 cm/min, (X2): 8.7 m/min, (X3): 4.1 m/min	Predicted values	6.0	1.9
		Actual value 1	5.6	1.5	0.4	0.4
		Actual value 2	5.6	1.5	0.4	0.4
		Actual value 3	5.7	1.6	0.3	0.3
3	(X1): 145 cm/min, (X2): 9.0 m/min, (X3): 3.0 m/min	Predicted values	5.5	1.5
		Actual value 1	5.2	1.1	0.3	0.1
		Actual value 2	5.2	1.2	0.3	0.3
		Actual value 3	5.1	1.2	0.4	0.3

lists the predicted and actual values at the test points. The test was repeated thrice. Consequently, in the case of point 1, the average deviations of Y_leg and Y_pen were 0.16 and 0.13 mm. For point 2, the average deviations of Y_leg and Y_pen were both 0.37 mm. Finally, at point 3, the average deviations of Y_leg and Y_pen were 0.37 and 0.23 mm.

If the optimal welding conditions are determined based on the three points shown in Table 9, in terms of productivity, the condition of point 3 is considered an optimal condition because the welding speed is the fastest and the size of the weld also satisfies the required standard for the manufacturer.

4. Conclusions

The effects of welding parameters (WS, leading WFR, and trailing WFR) on the bead shape (leg length and penetration depth) were investigated to predict and control the bead shape in tandem GMAW of aluminum 5083-O alloy. The following results and conclusions were obtained from this investigation:

• The tandem GMAW process was used in these experiments, and a torch with a 7 mm gap between the tandem torches was designed and applied;
• The bead shape (leg length and penetration depth) gradually decreased owing to the decrease in heat input as the WS increased within the WS range of 123–157 cm/min;
• The leading arc directly strikes the solid to form a narrow molten article, which has an effect on the influence of the penetration depth;
• The trailing arc is transferred to the molten pool created by the leading arc, which expanded the molten pool and influenced the leg length;
• As a result of observing arc behavior using a high-speed camera, it was confirmed that the leading WFR affects the penetration depth, and the trailing WFR affects the leg length;
• In the lap fillet joint tandem GMAW process using aluminum 5083-O alloy (thickness: 1.5 mm) for the top plate and aluminum 5083-O alloy (thickness: 2.5 mm) for the bottom plate, a regression equation was derived to predict the bead shape (leg length and penetration depth). The coefficient of determination (R²) of the regression models was 0.9414 for the leg length and 0.9924 for the penetration depth;
• It was validated that the estimated models were effective in predicting the weld bead shape of the aluminum 5083-O alloy single lap joint using the tandem GMAW process.