Particle Swarm Method for Optimization of ATIG Welding Process to Joint Mild Steel to 316L Stainless Steel

Abstract

316L stainless steel joined to mild steel is widespread in several applications to reach a requested good association of mechanical properties at a lower cost. The activating tungsten inert gas (ATIG) weld was carried out using a modified flux composed of 76.63% SiO₂ + 13.37% Cr₂O₃ + 10% NaF to meet standard recommendations in terms of limiting the root penetration. Modified optimal flux gave a depth of penetration 1.84 times greater than that of conventional tungsten inert gas (TIG) welds and a root penetration of up to 0.8 mm. The microstructure of the dissimilar joints was investigated using a scanning electron microscope and EDS analysis. The mechanical properties of the weld were not affected by the modified flux. The results show that the energy absorbed in the fusion zone in the case of ATIG weld (239 J/cm²) is greater than that of TIG weld (216 J/cm²). It was found that the weld bead obtained with the optimal flux combination in ATIG welding can better withstand sudden loads. The obtained UTS value (377 MPa) for ATIG welding was close to that of TIG welding (376 MPa). The average Vickers hardness readings for ATIG welds in the fusion zone are up to 277 HV, compared to 252 HV for conventional TIG welding.

1. Introduction

Stainless steel is a widespread material used in various industries such as automotive, construction, rolling, and chemical processing. Tungsten inert gas (TIG) is largely used for joining parts. TIG welding is one of the most widely used welding processes for stainless steel due to its excellent quality and sound weld [1,2]. However, the main drawback of TIG welding is its limitation to weld thicknesses less than 3 mm. Thus, it is highly required of the welders to use the filler rod, edge preparation, and multi-passes, and it is synonymous with less productivity and an increase in the product cost [3]. Activating tungsten inert gas (ATIG) welding is a genius and alternative technique allowing to achieve a full penetrated weld with only one pass, square edges, and without filler rod [4,5,6]. In the ATIG technique, the used equipment and welding condition parameters are the same as those of the TIG process, except that, prior to the welding process, a flux is deposited. This flux, in the form of a paste, is applied to the joints that will be welded by a brush, sprayed, or conveyed to the pieces to be joined. A shielding gas covers the coating with a density range between 5 and 6 mg/cm² [7].

Three main mechanisms have been proposed to explain the phenomenon occurring in the ATIG weld pool. The first mechanism, proposed by Heiple et al. [8], says that the surfactant elements such as O, Se, S, and Te present in the weld pool contribute to reversing the Marangoni circulation of molten metal, leading to inward convection and hence a deeper weld bead. The second mechanism, proposed by Howse et al. [9], suggests that the weld depth is ascribed to an arc constriction linked to the migration of elements with high electronegativity, such as halides such as fluorine and elements such as oxygen. These elements react with the outer arc electrons, leading to a constriction of the weld arc. The latter enhances the current density at the anode arc root, enabling deeper penetration in weld metal compared to that of TIG welding. The third mechanism stipulates that the contraction of the arc is ascribed to the insulation effect of the powders used, particularly silicon dioxide [10]. By means of the insulating effect of the high electrical resistivity of flux, the anode spot at the workpiece diminishes, and the heat density at this region increases.

Compared to popular TIG welding, ATIG welding can significantly enhance the welding efficiency and reduce welding costs without altering the mechanical properties or corrosion resistance under the same welding conditions [11,12,13]. Several dissimilar materials have been welded by ATIG welding, such as carbon steel, stainless steel, and light materials such as aluminum, magnesium alloys, titanium alloys, etc. The effects of single-component activating fluxes on the morphology and weld mechanical properties of ATIG welding were extensively investigated [14,15,16]. Austenitic stainless steel metals are considerably used in many industries, such as chemical processing, aerospace oil and petrol plants, pharmaceutical manufacturing, and food processing. The weldability of these steels is usually very good. Austenitic stainless steel 304 grade is extensively used in day-to-day life applications [17,18]. Mild steels, well known as low-carbon steel, contain mainly iron, ranging from 0.05 to 0.2% carbon (C) and 0.60 to 0.90% manganese (Mn), and are also mostly used in various structural applications for their strength, weldability, and formability properties [19]. Patel et al. [20] reported in their study on 316LN that the maximum depth of penetration was obtained with the use of flux Co₃O₄ and TiO₂. The authors noticed that the enhancement in depth of penetration is attributed to the reverse Marangoni effect and arc constriction. Meanwhile, dissimilar welding is becoming a more common technique in order to meet the requirements of many industries [21,22,23]. However, achieving efficient welding posed a major challenge compared to welding similar materials due to differences in the thermo-mechanical and chemical properties of the materials to be joined under common welding conditions [24,25,26].

Dissimilar welding of stainless steel and low carbon steel is a competitive alternative widely used in power plants, oil industries, and infrastructure structure building owing to its combination of structural performance and cost-efficient designs [27,28,29]. Sanjay et al. [30] investigated the effect of current, welding speed, joint gap, and electrode diameter on dissimilar ATIG welds between carbon steel (CS) and stainless steel (SS). They tested the mono-oxide flux, such as TiO₂, ZnO, and MnO₂. They observed the benefits of oxides in increasing penetration without altering the mechanical properties of the joint. They noticed the migration of carbon from CS to SS. On the other hand, many optimization methods were used to optimize the welding parameters to enhance the mechanical properties of the joint including the Taguchi approach [31,32], the Response Surface Method [33,34], the Jaya Algorithm Method [35], and even the Particle Swarm Method for Optimization (PSMO) [36]. Some other works were interested in improving weld bead geometry by using the Support Vector Machine (SVM) method [37] or the Artificial Neural Network (ANN) method [38]. Several works [39,40] were dedicated to investigating the effect of single powder oxide on the morphology, mechanical properties, and corrosion resistance of the weld bead. Other studies focused on optimizing the proportions of powders in the flux to be deposited prior to welding. Mixing method design is the main method applied when two or more mixed powders are used [41]. The flux can be a binary flux composed of two different types of powders [42,43] or a tri-component flux [44,45].

This study uses PSMO for dissimilar tri-component flux welding in order to optimize the best flux composition and achieve a penetrated sound weld without altering the mechanical properties of the joint consisting of 316L stainless steel (316L) and mild steel (MS). The weld line carried out with the ATIG technique is compared to another one performed with the conventional TIG process in terms of bead morphology, SEM-EDS analysis, and mechanical properties. The obtained results will enrich the database relating to fusion welding and will help manufacturers and researchers develop innovative methods and accomplish performant welds at a low cost. The present study demonstrates that the novel idea to associate oxides and fluorides ensures a sound weld that meets the standards recommendations.

2. Materials and Methods

2.1. Materials

The materials to be joined are austenitic stainless steel (316L grade) and mild steel.

Table 1. Chemical composition of 316L stainless steel and mild steel (weight %).

Elements	C	Mn	Si	P	S	Cr	Ni	Mo	N	Cu	Al	Fe
316L	0.026	1.47	0.42	0.034	0.0016	16.60	10.08	2.14	0.044	0.50	-	Balance
Mild steel	0.0521	0.165	0.009	0.0098	0.0137	0.0234	0.0277	0.00647	0.0053	0.0985	0.0245	Balance

shows the chemical compositions of both alloys.

2.2. Welding Procedure

Experiments consist of welding a line of about 20 cm on a rectangular plate of 6 mm thickness. Prior to welding, the plates were cleaned using acetone, and the powders were heated in a furnace at 100 °C for a period of one hour to eliminate humidity. After that, the powders were mixed with acetone in a (1 ÷ 1) ratio, and a layer of paste of about 0.3–0.4 mm was applied with a brush to the edges of the plates up to 10 mm wide, as shown in Figure 1. The joints were accomplished with a square butt weld design without any edge preparation on clamped plates with zero clearance distance. The deposited used powders and their characteristics are listed in

Table 2. Powders, melting, and evaporation temperatures.

Powders	Melting Temperature (°C)	Evaporation Temperature (°C)	Free Enthalpy of Formation |∆dH°|(kJ/mol)	Oxygen Percentages in Oxide %
SiO2	1722	2950	902	68.35
TiO2	1830	2972	941	63.26
Fe2O3	1540	1987	826	68.35
Cr2O3	2435	300	1128	61.71
V2O5	681	1750	1550.6	49.05
MoO3	802	1155	745.1	67.73
NaF	933	1704	573.6	-

, where the oxygen percentage in oxides was determined by XPS tool analysis.

Table 3. Welding parameters.

Parameters	Range
Welding speed	150 mm/min
Welding current	150 A
Arc Length	2 mm
Electrode tip angle	45°
Shielding gas on the workpiece	Argon with flow rate 8 L/min
Shielding gas on the backside	Argon with flow rate 4 L/min
Welding mode	Negative direct current electrode

represents the welding parameters after several trials to adjust the operational welding parameters.

After welding, the samples were cut far from the welding starting point to be sure that the arc welding was stabilized, as shown in Figure 2.

The tensile tests were performed with a computerized universal testing machine. The samples were prepared in accordance with ASTM E8M-04 and shown in Figure 2. The tests were carried out on 3 samples for each TIG and ATIG welded using the modified optimal flux. Vickers hardness tests were carried out according to ASTM E-384-99. Figure 2 shows the hardness reading position and tracks micro-indentation. Impact tests were carried out using the Charpy “V” notch impact testing machine on 3 samples for each TIG and ATIG weld according to ASTM E23.

2.3. Design of the Experiment Methodology

The design of the experiment will be applied to the depth and ratio as shown in Figure 3. The optimal flux obtained from the highest depth penetration and weld aspect ratio resulting from macrograph analysis will be used to investigate the mechanical properties. Note that W_f is the face weld width, W_b is the back weld width (root width), D is the depth weld penetration till the plate thickness, b is the excess depth penetration, and R is the weld aspect ratio.

The fusion zone microstructure of both TIG and ATIG welds was analyzed. Micrographs, chemical composition, and elemental distribution (mapping) were characterized by a field-emission scanning electron microscope. To avoid sample charging under the electron beam, samples were coated with a layer of platinum with a 25 nm thickness for 35 s. An accelerating voltage of 20 kV was used, with a working distance of 8 mm for the best signal intensity.

A mathematical model was developed, where D (depth penetration) is written in terms of selected oxides’ percentages, as will be explained in the discussion. In the second step, the optimal combination that maximizes D is determined. In the third step, an equation relating D to the proportions of the selected oxides was developed using the particle swarm optimization (PSO) method. Finally, the Matlab R2020 software module(Mathworks, Natick, Massachusetts · USA) was used to obtain the optimal combination of oxides that permits maximizing the depth of weld by a constrained optimization algorithm. Using the optimal flux obtained by the mathematical model, ATIG weld lines were performed to join mild steel to 316L.

3. Results and Discussions

3.1. Weld Bead Aspect

3.1.1. Selection of Appropriate Fluxes

Six oxide fluxes were used for dissimilar ATIG welding. The weld lines were performed using a square butt join design.

Table 4. Weld aspect of a single oxide flux of dissimilar ATIG welds.

Oxides	SiO2	TiO2	Fe2O3	Cr2O3	V2O5	MoO3
Depth (mm)	5.96	3.28	4.59	4.33	3.26	3.95
Width (mm)	12.22	11.28	11.40	11.08	11.54	11.16
Ratio	0.49	0.29	0.40	0.39	0.28	0.35

shows that the sample welded with the flux SiO₂ has the highest values of depth (5.96 mm) and ratio (0.49), followed by the samples welded with the fluxes Fe₂O₃ and Cr₂O₃. Based on these results, the oxides SiO₂, Fe₂O_3, and Cr₂O₃ were selected.

3.1.2. Mathematical Modeling

In this study, the three chosen fluxes (Fe₂O₃, Cr₂O₃, and SiO₂) with different percentages, varying within a range of 100%, were mixed to form a ternary flux. In

Table 5. Actual (experimental) depth vs. predicted depth (in mm) for the different ternary mixtures.

Order	X1 (en%)	X2 (en%)	X3 (en%)	D Actual	D Predicted
1	75	25	0	6.4625	6.2868
2	75	0	25	6.4575	6.0203
3	0	75	25	6.2875	6.0868
4	25	75	0	6.2375	6.0402
5	25	0	75	6.7733	7.3305
6	0	25	75	8.2467	7.6437
7	0	50	50	6.7867	7.1309
8	50	0	50	7.1200	6.7955
9	50	50	0	6.1275	6.632
10	50	25	25	7.5467	6.9794
11	25	50	25	7.4600	7.0016
12	25	25	50	7.4700	7.4317
13	33.33	33.33	33.33	6.4125	7.2325
14	66.67	16.67	16.67	6.7033	6.5238
15	16.67	66.67	16.67	6.6600	6.4711
16	16.67	16.67	66.67	6.6900	7.5636
17	100	0	0	4.5850	5.0048
18	0	100	0	4.3250	4.5114
19	0	0	100	7.9600	7.6253

, depth penetration (D) is expressed as a function of stream mixes as follows: (D = f(X₁,X₂,X₃)), where X₁, X₂, and X₃ indicate, respectively, the proportions of the fluxes Fe₂O₃, Cr₂O_3, and SiO₂ (in terms of percentages, forming a variable mix of 100%).

The selected mathematical model that depicts the effect of mixing flux on the depth penetration (depth) of the weld bead is the coupling of a second-order model as proposed in [46] with classical linear regression as in [47]. This mathematical model involves three components: (i) the linear effect of the proportions, (ii) their quadratic effects, and (iii) the interactions between these proportions. The starting model formulation is given in Equation (1):

$$\begin{gathered} \mathrm{D}={\mathsf{\alpha }}_1{(\mathrm{X}}_1)+{\mathsf{\alpha }}_2{(\mathrm{X}}_2)+{\mathsf{\alpha }}_3{(\mathrm{X}}_3)+{\mathsf{\alpha }}_4{(\mathrm{X}}_1){}^2+{\mathsf{\alpha }}_5{(\mathrm{X}}_2){}^2+{\mathsf{\alpha }}_6{(\mathrm{X}}_3){}^2+{\mathsf{\alpha }}_7{(\mathrm{X}}_1\left){(\mathrm{X}}_2\right) \\ +{\mathsf{\alpha }}_8{(\mathrm{X}}_1\left){(\mathrm{X}}_3\right)+{\mathsf{\alpha }}_9{(\mathrm{X}}_2\left){(\mathrm{X}}_3\right) \end{gathered}$$

(1)

where: X₁ = % Fe₂O₃, X₂ = % Cr₂O₃, X₃ = % SiO₂.

3.1.3. First Step of the Modeling Process

The first step of the modeling process is to find out the optimal parameters that minimize the squared error between actual D (experimental) and predicted D using the proposed model.

Therefore, the modeling problem is transformed into an optimization problem comprising decision variables and quadratic error into an objective function to be minimized.

Knowing that the problem can have irregularities such as the non-convexity of the criterion as well as numerous feasibility constraints, the meta-heuristic method known as particle swarm optimization (PSO) was used. A set of coefficients was randomly initialized within the bounds of the search space to solve the problem in PSO, and gradually a sub-optimal solution that combines three components was found. Three components consist of (i) tracking their speeds, (ii) returning to their best positions, and (iii) going to the position of the best neighbor. Four vectors were assigned to the swarm as follows:

-

The position;

-

A speed;

-

The best personal position;

-

The best common position.

To optimize the process, the equations of motion at the kth iteration were developed as follows [48,49]:

$${\mathrm{V}}_{\mathrm{k}+1}^{\mathrm{i}}={\mathrm{w}}_{\mathrm{k}}{\mathrm{V}}_{\mathrm{k}}^{\mathrm{i}}+{\mathrm{c}}_1{\mathrm{r}}_1\left({\mathrm{P}}^{\mathrm{i}}-{\mathsf{\alpha }}^{\mathrm{i}}\right)+{\mathrm{c}}_2{\mathrm{r}}_2\left({\mathrm{G}}^{\mathrm{i}}-{\mathsf{\alpha }}^{\mathrm{i}}\right)$$

(2)

$${\mathsf{\alpha }}_{\mathrm{k}+1}^{\mathrm{i}}={\mathsf{\alpha }}_{\mathrm{k}}^{\mathrm{i}}+{\mathrm{V}}_{\mathrm{k}+1}^{\mathrm{i}}$$

(3)

$${\mathrm{w}}_{\mathrm{k}}={\mathrm{w}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-\frac{{\mathrm{w}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{w}}_{\mathrm{m}\mathrm{i}\mathrm{n}}}{{\mathrm{k}}_{\mathrm{m}\mathrm{a}\mathrm{x}}}\times \mathrm{k}$$

(4)

As reported in the literature [48,49], the inertial weight decreases linearly from {0.9 to 0.4}. Where c₁ = c₂ = 0.75 are the cognitive and social factors, and r₁ and r₂ are two random numbers generated between {0 and 1}. The search limits of the model parameters are fixed, respectively, at −1 and +1 for the global stability of the model [48]. After a certain number of iterations, the optimization process will stop (k_max = 5000), and then a local search will be performed to obtain better solutions around the sub-optimal global solution obtained by the algorithm of PSO. In order to select model D, the optimization process will be executed several times. This model is expressed by X₁, X₂, and X₃ in Equation (5). To evaluate the efficiency of the model, three performance measures were used as follows [48]:

$$\begin{array}{l}D=5.8818\left(\%\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3\right)+6.2719\left(\%\mathrm{C}{\mathrm{r}}_2{\mathrm{O}}_3\right)+6.4349\left(\%\mathrm{S}\mathrm{i}{\mathrm{O}}_2\right)\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}-0.877\left(\%\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3\right){}^2-1.7605\left(\%\mathrm{C}{\mathrm{r}}_2{\mathrm{O}}_3\right){}^2+1.1904\left(\%\mathrm{S}\mathrm{i}{\mathrm{O}}_2\right){}^2\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+4.858\left(\%\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3\right)\left(\%\mathrm{C}{\mathrm{r}}_2{\mathrm{O}}_3\right)+2.2355\left(\%\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3\right)\left(\%\mathrm{S}\mathrm{i}{\mathrm{O}}_2\right)\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+3.6801(\%\mathrm{C}{\mathrm{r}}_2{\mathrm{O}}_3)(\%\mathrm{S}\mathrm{i}{\mathrm{O}}_2)\end{array}$$

(5)

$$\mathrm{M}\mathrm{A}\mathrm{P}\mathrm{E}=\frac{100}{19}\sum _{t=1}^{19}\frac{\left|\mathrm{D}\left(\mathrm{t}\right)-\mathrm{D} \mathrm{p}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}\left(\mathrm{t}\right)\right|}{\mathrm{D} \mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}$$

(6)

$${\mathrm{R}}^2=100\times (1-\frac{\frac{1}{19}\sum _{t=1}^{19}{\left(\mathrm{D}\left(\mathrm{t}\right)-\mathrm{D} \mathrm{p}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}\left(\mathrm{t}\right)\right)}^2}{\frac{1}{19}\sum _{t=1}^{19}{\left(\mathrm{D}\left(\mathrm{t}\right)-\mathrm{D} \mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}\right)}^2})$$

(7)

$$RMSE=\sqrt{\frac{1}{19}\sum _{t=1}^{19}{(\mathrm{D}\left(\mathrm{t}\right)-\mathrm{D} \mathrm{p}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}\left(\mathrm{t}\right))}^2}$$

(8)

The precision of this model is measured by the calculation indicators mentioned above as follows: MAPE = 5.8674%, R² = 71.15% and RMSE = 0.448 mm.

3.1.4. Second-Step Modeling Process

The second step to consider in the optimization process is to determine the optimal percentages of flux to obtain the maximum D.

To reach this target, a constrained optimization algorithm (Matlab 2020 Optimization Toolkit, Mathworks, Natick, MA, USA) was used. This algorithm permits to obtain the optimal combination presented by Equation 9 which allows to predict the value of (D = 7.7012 mm) with a root mean square error (RSME = 0.448 mm).

$$\mathrm{F}\mathrm{l}\mathrm{u}\mathrm{x} \mathrm{O}\mathrm{p}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{l}=0\%{\mathrm{F}\mathrm{e}}_2{\mathrm{O}}_3+13.368\%{\mathrm{C}\mathrm{r}}_2{\mathrm{O}}_3+86.632\%{\mathrm{S}\mathrm{i}\mathrm{O}}_2$$

(9)

3.1.5. Validation Test

The optimal flux is tested in the validation step by comparing the weld beads obtained by ATIG and conventional TIG.

Table 6. Depth, widths, and ratios for TIG and ATIG.

Welds	D	WF	WB	(D + WB)/WF
TIG Weld [50]	3.7	12.02	0	0.31
ATIG Weld (optimal flux)	8.2	9.94	7.3	1.56

shows the depth bead aspect data for ATIG and TIG. It can be seen clearly that for ATIG, the penetration depth D increased by 2.2 times and the aspect ratio (D + W_B)/W_F was enhanced by 5 times. Moreover, the obtained depth of weld (8.2 mm) was higher than the predicted one (7.7 mm).

Figure 4 shows macrographs of TIG and ATIG weld beads. It is clearly seen that the ATIG weld bead is fully penetrated, unlike the TIG weld bead, which is not.

According to ISD 341-2 [51] Engineering Standards, excessive root penetration should be less than 25% of the nominal thickness of the base material of the thinnest component to be joined. In other words, the root penetration in our study should be less than 1.5 mm, knowing that the workpiece thickness is 6 mm. The full penetration of the ATIG weld depth, using the optimal flux, is 8.2 mm; consequently, the excess of penetration is 2.2 mm beyond the back side of the base material, which exceeds 1.5 mm, as shown in Figure 4.

A temporary copper support plate is usually used to support the molten weld metal (WM) and prevent excessive root penetration. Unfortunately, the latter can be a source of stress concentration in addition to the appearance of rust-promoting crevices and additional manufacturing costs. For these reasons, a support plate as a solution is discarded. To decrease the root weld penetration to an acceptable range to meet the requirements of BS EN ISO 15614-1:2017 industrials and standards [52], the optimal combination obtained was modified by adding 10% NaF. The modified optimal combination obtained is 13.368% Cr₂O₃ + 76.632% SiO₂ + 10% NaF. The addition of NaF to the optimal flux is aimed to constrict the arc, leading to an increase in the arc density and temperature, as reported by several studies [53,54].

Another weld line using modified optimal flux was carried out and compared to the welds performed with unmodified optimal flux and with conventional TIG welding.

Figure 5 shows the benefit of adding 10% NaF at the optimal flux to reduce root penetration to an acceptable range without weld bed collapse. The upper surface of the modified flux weld is almost flat and at the same level as the materials to be joined, as shown in Figure 5b.

The results depicted in

Table 7. Morphology of TIG, ATIG with optimal flux, and weld with modified optimal flux.

Welds	D	WF	WB	(D + WB)/WF
Conventional TIG Weld [50]	3.7	12.02	0	0.31
ATIG Weld (optimal flux)	8.2	9.9	7.3	1.56
ATIG Weld (with modified optimal flux)	6.8	8.8	3.5	1.17

show that the modified optimal flux meets industrial requirements as long as penetration is 6.8 mm with an excess penetration of 0.8 mm. Ion fluoride is released from NaF and is characterized by a relatively low melting point (682 °C) and a low free enthalpy of formation (573.6 kJ/mol). The migration of fluoride ions into the arc react with the outer free electrons, leading to the constriction of the arc as mentioned above. In a further study of this work, we will focus on the microstructure and mechanical properties in comparison between double-sided TIG and the modified optimal flux. Also, the fluoride ions migration in the arc welding decreases the anode spot and contributes to increasing the energy density of both the heat source and the electromagnetic force in the weld pool. Consequently, the weld morphology is relatively narrow and deep [55].

3.2. SEM-EDS investigation

3.2.1. ATIG Weld

In

Table 8. EDS analyzes regions across the ATIG weldment from 316 L to MS base metal (weight %).

Elements	Reg.1	Reg.2	Reg.3	Reg.4	Reg.5	Reg.6	Reg.7	Reg.8	Reg.9	Reg.10
C	2.51	2.30	2.17	2.62	2.44	2.45	0	0	0	0
Cr	0	0	0	4.49	9.81	11.25	10.75	10.56	18.82	18.63
Fe	97.49	97.70	97.83	90.73	83.29	81.82	84.42	83.40	70.82	71.21
Ni	0	0	0	2.16	4.46	4.48	4.82	4.92	8.29	8.13

, the chemical composition obtained for dissimilar ATIG butt joints along a horizontal line from the MS side to the 316L side throughout the ATIG weld region is summarized. The weld chemistry composition analysis is performed using an EDS surface area scan from the MS side to the 316L side for the dissimilar ATIG weld presented in

Table 9. EDS analyses across the ATIG weldment from 316L to MS base metal.

Regions	EDS Spectrum for Dissimilar 316-MS ATIG Weld
Inside MS region 1
Inside MS region 2
MS/WZ border (MS side) region 3
MS/WZ border (WZ side) region 4
MS/WZ border (WZ side) Region 5
MS/WZ border (WZ side) region 6
Inside WZ region 7
WZ/316L border (WZ side) region 8
{49}sWZ/316L border (316L side) region 9
Inside 316L region 10

.

The analysis reveals a decrease in chromium and nickel content from the 316L side to the M.S. side throughout the weld zone. On the other hand, we notice an increase in iron content in the same direction.

Close to the boundary located between the MS side and the ATIG weld zone, the weight percentages of carbon are low, as revealed by Table 8 at region 3 (2.17%). In this region, the depletion of carbon is suspect. The percentage of carbon at the weld zone border is up to 2.62%. The migration of carbon from MS to the weld fusion zone is altered due to the fast cooling rate characterized by the ATIG weld.

Figure 6 depicts the mapping of the principal chemistry elements available at the MS base metal and the weld zone (WZ) border. The mapping of C content across the boundary fusion line shows a gradient of C elements, that is marked in favor of the MS region. However, the density of elements such as Cr, Ni, and Mn is more pronounced in WZ, and Fe content decreases from MS to WZ.

At the ATIG 316L-WZ border, the diffusion of elements such as Cr, Ni, and Mn from 316 L SS to WZ is pronounced. However, the migration of carbon is not obvious, as shown in Figure 6 and Figure 7.

3.2.2. TIG Weld

The results collected in

Table 10. EDS analyzes regions across the ATIG weldment from 316L to MS base metal (weight %).

Elements	Reg.1	Reg.2	Reg.3	Reg.4	Reg.5	Reg.6	Reg.7	Reg.8	Reg.9	Reg.10
C	5.03	4.99	1.82	4.67	3.58	1.92	0	0	0	0
Cr	0	0	0	4.98	4.86	14.41	11.47	16.09	18.55	18.50
Fe	94.97	95.01	98.18	87.46	89.51	76.01	82.69	75.08	70.89	71.17
Ni	0	0	0	2.89	2.06	6.41	4.83	6.53	8.06	8.30

are extracted from the

Table 11. EDS analyses across the TIG weldment from 316 L to MS base metal.

Regions	EDS Spectrum for Dissimilar 316-MS Weld
Inside MS region 1
Inside MS region 2
MS/WZ border (MS side) region 3
MS/WZ border (WZ side) region 4
MS/WZ border (WZ side) Region 5
MS/WZ border (WZ side) region 6
Inside WZ region 7
WZ/316L border (WZ side) region 8
WZ/316L border (316L side) region 9
Inside 316L region 10

and tell us about the evolution of the content of principal elements such as C, Cr, Fe, and Ni throughout the TIG dissimilar weld.

The weight percentages of carbon elements revealed by EDS scan analysis at WZ border region 4 are up to 4.67% against 1.82% at region 3 at the closest location to the boundary MS/WZ. The carbon content in region 1, far from weld solidification, is up to 5.03 % and steeply decreases to 1.82% in region 3, near the border line, which reveals the migration of the carbon element towards the weld zone. The carbon migration from MS base metal to WZ leads to a carbon depletion zone at the heat-affected zone (HAZ)-MS side. On the other hand, the WZ zone bordering the MS side is enriched by carbon, which favors the formation of carbides, as revealed by Wenyong et al. [56]. Chromium and nickel contents decrease from the 316L side to the MS side throughout the weld zone. We also notice an increase in iron content in the same direction.

The mapping scan depicted in Figure 8 shows the decrease in Cr, Ni, and Mn content, whereas Fe content gradually increases from WZ to MS base metal across the WZ-MS interface. We can remark that the chromium content gained a gradual decrease across the WM-MS interface, while the ferrous content had no notable change. Regarding the carbon content, the scan map shows a decarburized layer in MS near the fusion line and a carburized layer in WM. This can be explained by the carbon diffusion from MS to WM, and the same phenomenon was reported in some studies [30,50].

Figure 9 obviously shows a gradient in Cr, Ni, and Fe contents across the border 316L-WZ; an increase in Cr and Ni contents and a decrease in Fe content from 316L to WZ. However, the migration of carbon apparently has not occurred.

3.3. Tensile Test

Table 12. Measurements of tensile strength and standard deviations of TIG and ATIG (modified optimal flux).

Sample	Number of Tests	UTS Max. (MPa)	UTS Min. (MPa)	UTS Average (MPa)	Standard Deviations σ
As received MS base metal	3	364	361	362	2.02
As received 316L base metal	3	626	623	624	1.4
TIG MS/316L [50]	3	380	372	376	4.00
ATIG MS/316L	3	382	370	377	4.04

shows that the average value of UTS for ATIG dissimilar welds is 377 MPa, which is almost equal to that of conventional TIG welding (376 MPa).

We remark that the fracture for both TIG and ATIG dissimilar welds occurs on the mild steel side, far from the weld joint, which attests that this region is the weakest location in comparison to the entire specimen, as shown in Figure 10. This phenomenon shows, firstly, that the MS base metal is weaker than the weld metal and, secondly, that the weld joint quality is good. We also notice that the necking for ATIG is more pronounced than that of TIG welds, which confirms the enhanced ability of plastic deformation until fracture occurs.

3.4. Hardness Test

Figure 11 shows the variations of Vickers micro-hardness as compared to the distance from MS to 316 L SS.

Table 13. Measurements of hardness and standard deviations of TIG and ATIG (with modified optimal flux) at FZ.

Sample	Zone of Tests	HV Max.	HV Min.	HV Average	Standard Deviations σ
TIG [50]	FZ	287	235	252	12.75
ATIG	FZ	283	270	277	4.84

shows the hardness measurements and standard deviations of ATIG and TIG at FZ. The hardness of TIG and ATIG weld regions is higher than that of both 316L and MS base metals, as reported in Figure 11.

The highest hardness value in the TIG weld is situated in the weld zone, close to the fusion boundary beneath the mild steel side. This aspect is due to the formation of harder micro-constituents in this region by the migration of carbon into weld metal from the mild steel side, as previously shown in the SEM-EDS analyses as a decarbonized zone. On the other hand, there is no migration of carbon in the ATIG weld, which explains the homogenous values of hardness in the weld. The tendency of carbon diffusion from the MS side to the weld zone is due to the presence of elements like chromium in the weld zone and the high cooling rate characterized by ATIG weld decreases. The average hardness measurements in FZ are higher for ATIG than for TIG welds, which can also be explained by the rapid cooling rate in ATIG.

In the ATIG weld zone, as shown in Table 13, the obtained hardness values are uniform (hardness standard deviation is less than 5 HV), which means that there is a homogenization of hardness. Contrarily, in the TIG weld zone, the obtained hardness values are non-uniform (hardness standard deviation is more than 5 HV), as reported by Osoba et al. [57]. Moreover, we can observe a decrease in the hardness values from FZ to base metal in both MS and SS.

3.5. Impact Test

Table 14. Measurements of absorbed energy and standard deviations of TIG and ATIG (with optimal flux) at the fusion zone for dissimilar 316L/MS welds.

Sample	Number of Tests	Absorbed Energy (J/cm2) Min	Absorbed Energy (J/cm2) Max	Absorbed Energy (J/cm2) Average	Standard Deviations σ
TIG—316L SS/MS [50]	3	215	238	216	15.56
ATIG—316L SS/MS	3	234	243	239	4.58

presents the obtained results of the absorbed energy in impact tests and the standard deviations of TIG and ATIG at the fusion zone. We can remark that the average absorbed energy in ATIG weld (239 J/cm²) is higher than that of TIG weld (216 J/cm²) by 23 J/cm², which means that the ATIG welds withstand more sudden loads.

One set of fractured impact specimens obtained after V-notch testing at room temperature is illustrated in Figure 12.

Figure 13b shows the fractography of an ATIG impact test specimen. We can see small and deep dimples with spreading small voids, which indicate a fully ductile fracture mode. On the other hand, Figure 13a shows the morphology of the fractured surface of a TIG specimen. The fractography exhibits mixed fracture, which is composed of a large number of islets of fine dimples separated by the facies of quasi-cleavage and indicates less resistance to sudden impact loads.

4. Conclusions

PSMO was used to obtain the optimum combination of flux in order to weld dissimilar MS and austenitic 316L SS. The resultant optimal flux ensured full penetration of the weld bead with an excess root. A modification of optimal flux by adding fluorine (NaF fluorine) avoids excessive root penetration. The morphologies and mechanical properties of the weld beads were compared for ATIG with optimal flux and conventional TIG. The following main conclusions can be deduced:

-

The results of this investigation show that 316L can be joined with MS using a tri-component flux composed of 76.63% SiO₂, 13.37% Cr₂O₃, and 10% NaF.

-

A fully penetrated bead in a single pass without edge preparation is achieved. The excess root penetration can reach 0.8 mm, which meets the requirements of industrial standards. The obtained depth (D) is 6.8 mm, the bead face width (W_F) is 8.8 mm, and the back face width (W_B) is 3.5 mm, leading to an aspect ratio (D + W_B)/WF of 1.17. Hence, compared to TIG welding, the depth and the ratio increased by 1.83 and 3.77 times, respectively.

-

The fluoride in the form of ions present in the modified flux migrates to the welding arc and contributes to an arc constriction. Thus, a phenomenon of reduction of the weld bead is observed, leading to an increase in the penetration of the weld compared to conventional TIG. Moreover, the surfactant elements, such as oxygen liberated in the weld pool, contribute to the reversal of Marangoni convection, resulting in a fully penetrated weld.

-

SEM-EDS analysis shows carbon depletion closest to the weld zone at the MS side border. The migration of carbon from MS to the weld zone is suspected to occur in cases of TIG weldment. However, this situation is very limited or inexistent in ATIG welding due to its high cooling rate.

-

The tensile test reveals that the strength has almost the same value, and the fracture happens at MS base metal for both ATIG and TIG welding. This indicates that the welding zone is stronger than that of the MS parent metal.

-

The hardness of ATIG welds exhibits homogenous values (≈277 HV), and the average value is higher than that of TIG welds (≈255 HV). The disparities in hardness between ATIG weld readings are negligible.

-

The fractography in the impact test shows a fully ductile fracture with many fine dimples overall on the fractured surface in ATIG welding, compared to gatherings of dimples separated by quasi-cleavage facies in TIG welding. ATIG welding exhibits more resistance to sudden loads (239 J/cm²) than TIG welding (216 J/cm²).