Prediction of element transfer due to flux and optimization of chemical composition and mechanical properties in high-strength low-alloy steel weld

Abstract

The transfer of elements C, Si, Mn, P and S from slag into the weld metal or from weld metal into the slag and microhardness has been studied using formulated fluxes. The fluxes have been formulated using extreme vertices design with an aim to develop mathematical models for change in element content and mechanical properties versus flux constituents for submerged arc welding of high-strength low-alloy steel. It is found that CaO is the most significant flux constituent and Al₂O₃ is the second most significant constituent among individual mixtures. CaO·MgO and CaO·Al₂O₃ among binary mixtures have significant effect on element transfer and microhardness. Developed mathematical models have been checked for adequacy using t-test and analysis of variance (F-test). Flux mixtures’ composition has been provided for optimum chemical composition and mechanical properties. One of the optimum flux mixture with composition, CaO 11.61, Al₂O₃ 12.33, CaF₂ 15.00 and MgO 39.06, would be providing desirable chemical composition and mechanical properties.

Keywords

High-strength low-alloy steel flux composition submerged arc welding extreme vertices design element transfer and microhardness

Introduction

Submerged arc welding (SAW) is one of the major welding processes due to its inherent attractive features such as high quality, deep penetration, smooth finish, capability to weld thicker sections and prevention of atmospheric contamination of weld pool.^1–3 Mechanical properties^4,5 of weld metal depend on element transfer in weld metal and microstructure, which further depends on flux composition for same welding parameters, electrode wire, base metal, weld joint design and weld thermal history. It was observed that in the earlier work^6–8 that increase in acicular ferrite (AF) microstructure improves mechanical properties. Kanjilal et al.⁹ in their work tried to develop mathematical model to find AF in terms of flux constituents for SAW of C–Mn steel. It was found that MgO is the most effective as an individual and CaO in interaction with other elements in controlling the amount of AF content in the weld metal.

A number of researchers have tried to study the element transfer between the weld metal and the slag on various aspects. Some researchers tried to find absolute element content of the weld metal^10,11 due to slag–metal reactions, whereas others investigated the change in element transfer due to flux.^4,12,13

Pandey et al.¹² studied the effect of welding parameters and flux basicity index (BI) on the weld chemistry and element transfer. It was concluded that both composition of the flux and BI value completely characterize a flux for studying its effect on weld chemistry. The authors also suggested elaborate optimization of the flux ingredients to be carried out in future work.

The study of Chai and Eagar¹⁰ indicates that weld metal chemistry is dependent on weld metal flux composition but insignificantly dependent on operating parameters. The study also indicated that weld bead geometry is dependent on weld parameters only, namely, welding current, voltage, travel speed, electrode diameter and electrode polarity but independent on flux composition.

The slag–metal reactions have been studied by Mitra and Eagar¹¹ to predict the transfer of elements Cr, Si, Mn, P, S, C, Ni and Mo between the slag and the weld pool for weld bead. It was concluded from the study that the transfer of chromium is strongly dependent on the type of flux used; lower BI of the flux and high chromium content of the electrode enhance the chromium content of weld metal. It was also observed that the manganese content of the weld metal depends mainly on the amount of MnO in the flux and manganese content of the electrode. Lime-based fluxes produced weld metal with a lower phosphorous content than fluxes containing MnO as flux constituent.

Kanjilal et al.⁴ in their work concluded that interaction between flux ingredient and welding parameters is more important than the individual effect of flux ingredients on weld metal chemistry and weld metal properties. Among the welding parameters, polarity was found to be most important for all responses, namely, weld metal chemical composition and mechanical properties.

Kanjilal et al.¹³ studied the change in element transfer (Δ) for the elements O₂, Mn, Si, S and C in terms of flux ingredients. It was suggested that element transfer was primarily due to chemical reactions associated with fluxes with fixed welding parameters. From the work, it was concluded that gain of Mn in the weld metal resulted due to electrochemical reaction in the weld pool, whereas gain of silicon in the weld metal was due to thermochemical dissociation and electrochemical reaction in the weld pool.

Burck et al.¹⁴ evaluated the effects of CaF₂, CaO and FeO additions on weld metal chemistry on steel plates. Carbon, oxygen, nitrogen, manganese, silicon, chromium, nickel and molybdenum contents were found in chemical analysis of the welds. The study indicated that the element transfer between the slag and the weld metal cannot be consistently explained using thermodynamic data only. Du Plessis et al.¹⁵ suggested that the diffusible hydrogen content of weld metal can be greatly reduced by flux chemistry modifications, that is, by increasing the BI of flux and increasing the fluoride-containing compounds (CaF₂, NaF, K₂TiF₆ or K₂AlF₆) in SAW electrodes.

In this study, individual flux constituents and their binary mixtures have been studied with an aim to obtain mathematical models for element transfer due to flux by varying flux composition in CaO–MgO–CaF₂–Al₂O₃ flux system using statistical mixture design for SAW of high-strength low-alloy (HSLA) steel. Mathematical models for change in element content with respect to base metal and electrode wire have been formed in terms of individual constituents and their binary interactions.

Plan of experimentation

For plan of experiments, 21 agglomerated fluxes were used to predict the effect of flux constituents on the element transfer due to flux. These fluxes were prepared by varying the composition of four flux constituents CaO, Al₂O₃, CaF₂ and MgO and keeping SiO₂, TiO₂, MnO and bentonite as constant (Table 1). The most suitable method for constrained mixtures design, extreme vertices design first suggested by Anderson and McLean,¹⁶ Cornell¹⁷ and Adeyeye and Oyawale¹⁸ was used for flux formulation. Design of experiment (DOE) technique mentioned here was also used previously by Jindal et al.¹⁹ with same limits of flux constituents and same design table¹⁷ to determine mechanical properties and H₂ content. The method suggests constrained mixture design for a mixture of q components having lower and upper limits on some or all the components may be represented mathematically as^16,17

0 \leq a_{i} \leq x_{i} \leq b_{i} \leq 100

(1)

\sum_{i = 1}^{q} x_{i} = 100

(2)

where $i = 1, 2, 3, \dots, q$ and $a_{i} and b_{i}$ are the constraints on the $x_{i}$ imposed by the experimenter.

Table 1.

Design matrix for flux formulation.

Type of point	Flux number	Varied mixture components				Basicity index
		CaO (wt%)	Al₂O₃ (wt%)	CaF₂ (wt%)	MgO (wt%)	Basicity index
Vertex	1	35.0	8.0	10.0	25.0	3.7
Vertex	2	10.0	8.0	40.0	20.0	3.7
Vertex	3	10.0	40.0	10.0	18.0	1.1
Vertex	4	20.0	8.0	10.0	40.0	3.7
Vertex	5	10.0	8.0	20.0	40.0	3.7
Vertex	6	10.0	18.0	10.0	40.0	2.5
Vertex	7	20.0	8.0	40.0	10.0	3.7
Vertex	8	35.0	8.0	25.0	10.0	3.7
Vertex	9	18.0	40.0	10.0	10.0	1.1
Vertex	10	35.0	23.0	10.0	10.0	2.1
Vertex	11	10.0	40.0	18.0	10.0	1.1
Vertex	12	10.0	18.0	40.0	10.0	2.5
Centroid	13	10.0	22.0	23.0	23.0	2.2
Centroid	14	21.7	8.0	24.2	24.2	3.7
Centroid	15	21.3	22.8	10.0	23.8	2.1
Centroid	16	21.3	22.8	23.8	10.0	2.1
Centroid	17	35.0	13.0	15.0	15.0	3.0
Centroid	18	12.7	40.0	12.7	12.7	1.1
Centroid	19	13.3	11.3	40.0	13.3	3.2
Centroid	20	13.3	11.3	13.3	40.0	3.2
Centroid	21	18.6	18.9	20.3	20.3	2.4

Upper and lower limits of the various flux constituents were decided on the basis of the phase diagrams of CaO–Al₂O₃–MgO system and performing trial experiments. The phase diagram for CaO–Al₂O₃–MgO ternary system is shown in Figure 1.²⁰ The primary requirement while deciding the percentage composition of various flux ingredients is that their melting temperature should be lower than that of base metals as the flux should melt first before the base metals to be melted and should remain in the molten state even after the solidification of weld so as to avoid the atmospheric contamination.²¹ Slag detachment, weld bead continuity and arc stability were observed during the trial experiments.

Figure 1.

Phase diagram for CaO–Al₂O₃–MgO ternary system showing different equilibrium temperatures.²⁰

Upper and lower limits of the various flux constituents are given as follows

\begin{matrix} 10 \leq CaO (x_{1}) \leq 35 \\ 8 \leq A l_{2} O_{3} (x_{2}) \leq 40 \\ 10 \leq Ca F_{2} (x_{3}) \leq 40 \\ 10 \leq MgO (x_{4}) \leq 40 \end{matrix}

(3)

\sum_{i = 1}^{4} x_{i} = 78

(4)

The whole mixture design may be represented graphically by truncated tetrahedron in three-dimensional (3D) space (Figure 2) consisting of 12 vertices of tetrahedron, 4 centroids of hexagonal faces and 4 centroids of triangular faces and 1 overall body centroid. In the first step, 12 vertices of polyhedron were selected out of 32 combinations (q.2^q−1 = 4.2⁴⁻¹) satisfying lower and upper bounds and total proportion of the mixture. Then, of the 12 vertices, combinations were grouped into three or more having either of the vertices with the same proportion $x_{i}$ , so eight such combinations were selected forming the centroids of the triangular and hexagonal faces of truncated tetrahedron and one as the overall (body) centroid. Flux numbers 1–12 represent 12 vertices, flux numbers 13–20 represent eight centroids and flux number 21 represents one overall centroid (Table 1).

Figure 2.

Graphical demonstration of flux mixture design in 3D space.

The output response characteristics are given in the form of second-order regression model as per the following equation (5)^16,17

y = \sum_{i}^{q} b_{i} x_{i} + \sum_{i < j} \sum^{q} b_{ij} x_{i} x_{j}

(5)

where y is the output response variable and $b_{i}$ and $b_{ij}$ are least square model regression coefficients. Terms $b_{i} x_{i}$ and $b_{ij} x_{i} x_{j}$ in the regression model represent individual effect and interaction effect of the mixture components, respectively. If output response variable y is increasing with regression coefficient $b_{ij}$ positive, then the excess is called the synergism of the binary mixture; opposite to that if output response variable y is decreasing with regression coefficient $b_{ij}$ positive, then the decrease is called the anti-synergism of the binary mixture and vice versa.¹⁷ So positive and negative values of regression coefficient $b_{ij}$ define synergistic and anti-synergistic effects of the binary mixtures, respectively.⁹

Experimentation

Preparation of fluxes

For preparation of fluxes, constituting elements were weighed separately on digital weighing balance (accuracy 1 mg) according to weight percentage as per Table 1 and then mixed thoroughly in a container with sodium silicate binder (20% weight) for about 30 min to get homogenous semi-solid mass. Sodium silicate was added for better arc stability and binding the individual ingredients together.^22,23 Then solid mass was dried in air for 24 h and then baked in a muffle furnace at 700 °C for nearly 1 h. After cooling, these fluxes were crushed, sieved and kept in air-tight bags.

Welding of plates

For welding, the edges of HSLA plates were prepared by making a V-groove of 60°. Twenty-one weld samples were made taking two HSLA steel (API 5L X65) plates, each of size 250 × 150 mm and thickness 18 mm. Fixtures were made to get butt weld on the base plates on SAW setup (Ador Tornado SAW M-800) at MMU, Mullana (Ambala, India). All the experiments were conducted using 3.2-mm-diameter wire electrode (EH-14) by taking optimum welding parameters: welding current of 410 A, voltage of 30.4 V and weld speed of 21.2 m/h with direct current reverse polarity (DCRP) using multi-pass welds. The optimum welding parameters were selected on the basis of trial runs and performing optimization experiments using response surface methodology (RSM) technique by varying the following parameters: welding current range of 349–651 A, arc voltage range of 24–32 V and weld speed range of 17–33 m/h keeping nozzle-to-tip distance constant at 25 mm with formulated flux number 11 (BI = 1.1) in separate work by Jindal et al.²⁴ Compositions of the flux have significant effect on chemical and mechanical properties of the welds, but the effect of flux constituents on weld bead geometry has not been reported in the earlier literature. Since weld metal chemistry is primarily dependent on weld metal flux composition and independent of operating parameters.¹⁰ Weld parameters were optimized on the basis of weld bead geometry parameters, area of penetration, weld form factor and dilution.

BI is defined as the ratio of basic oxides to acidic oxides given by equation (6)^15,25 as follows

BI = \frac{[CaO + MgO + Ca F_{2} + N a_{2} O + K_{2} O + BaO + \frac{1}{2} (MnO + FeO)]}{[Si O_{2} + \frac{1}{2} (Ti O_{2} + Zr O_{2} + A l_{2} O_{3})]}

(6)

Analysis of weld samples

Chemical composition of base metal, filler metal (electrode wire) (Table 2) and fusion zone (Figure 3) of weld samples (Tables 3 and 4) was determined by electron probe micro analysis (EPMA). Chemical analysis of all weld samples was carried out on JEOL Electron Probe Microanalyzer (JXA-8230) at NML, Jamshedpur, India, at an accelerating voltage of 15 kV at a resolution of 1000×. Compositional and secondary images of two weld samples (samples 5 and 21) are shown in Figure 4. Weld element content was determined by taking average of three randomly selected points (Figure 4). X-ray diffraction (XRD) analysis of all the slag samples was carried out to analyze the elemental pickup by slag during welding. Microhardness of weld zone (at the center of fusion zone) of all welds (Figure 3) had been taken at a load of 0.5 kg.

Table 2.

Chemical composition of base metal and electrode.

Chemical composition	C	Si	Mn	P	S	Ti	Nb	V
Base metal (API 5L X65)	0.214	0.17	1.74	0.01	0.01	0.193	0.484	0.105
Electrode wire (EH-14)	0.201	0.1	1.8	0.03	0.03	0.002	0.007	0.002

Figure 3.

Arrangement of cutting specimens for various testing.

Table 3.

Various weld metal properties: dilution (d), (1 − d) and expected weld metal content (%).

Sample number	Dilution, d	1 − d	Expected weld metal content (%)
			C	Si	Mn	P	S
1	0.37	0.63	0.2058	0.1258	1.7779	0.0226	0.0226
2	0.25	0.75	0.2042	0.1172	1.7853	0.0251	0.0251
3	0.37	0.63	0.2058	0.1258	1.7779	0.0226	0.0226
4	0.40	0.60	0.2062	0.128	1.776	0.022	0.022
5	0.31	0.69	0.2051	0.1219	1.7812	0.0237	0.0237
6	0.35	0.65	0.2056	0.1245	1.779	0.023	0.023
7	0.33	0.67	0.2053	0.1233	1.78	0.0233	0.0233
8	0.29	0.71	0.2048	0.1206	1.7823	0.0241	0.0241
9	0.26	0.74	0.2044	0.1183	1.7843	0.0248	0.0248
10	0.16	0.84	0.2031	0.1112	1.7904	0.0268	0.0268
11	0.47	0.53	0.2071	0.1326	1.772	0.0207	0.0207
12	0.40	0.60	0.2062	0.128	1.776	0.022	0.022
13	0.47	0.53	0.2072	0.1332	1.7716	0.0205	0.0205
14	0.45	0.55	0.2068	0.1315	1.773	0.021	0.021
15	0.43	0.57	0.2065	0.1299	1.7744	0.0215	0.0215
16	0.43	0.57	0.2066	0.1303	1.774	0.0213	0.0213
17	0.33	0.67	0.2053	0.1229	1.7804	0.0235	0.0235
18	0.40	0.60	0.2062	0.128	1.776	0.022	0.022
19	0.36	0.64	0.2057	0.1252	1.7784	0.0228	0.0228
20	0.37	0.63	0.2058	0.1258	1.7779	0.0226	0.0226
21	0.46	0.54	0.2069	0.132	1.7726	0.0209	0.0209

Table 4.

Various weld metal properties: observed weld metal content (%), element transfer due to flux (Δ) and microhardness (VHN).

Sample number	Weld metal content (%)					Element transfer due to flux					Microhardness (VHN)
	C	Si	Mn	P	S	ΔC	ΔSi	ΔMn	ΔP	ΔS	Microhardness (VHN)
1	0.0543	0.2426	1.0847	0.0237	0.0087	−0.1515	0.1168	−0.6933	0.0010	−0.0140	213
2	0.0311	0.1477	1.1230	0.0147	0.0080	−0.1731	0.0305	−0.6623	−0.0104	−0.0171	210
3	0.0360	0.1436	1.2533	0.0083	0.0236	−0.1698	0.0178	−0.5246	−0.0143	0.0010	215
4	0.0952	0.1954	1.1377	0.0257	0.0225	−0.1110	0.0674	−0.6383	0.0037	0.0005	215
5	0.0562	0.1985	1.1977	0.0123	0.0063	−0.1489	0.0766	−0.5835	−0.0114	−0.0174	214
6	0.0643	0.1090	1.8080	0.0067	0.0122	−0.1413	−0.0155	0.0290	−0.0163	−0.0108	221
7	0.0653	0.1266	1.3710	0.0197	0.0237	−0.1400	0.0033	−0.409	−0.0037	0.0003	216
8	0.0420	0.1775	1.3267	0.0203	0.0017	−0.1629	0.0569	−0.4557	−0.0038	−0.0224	220
9	0.0589	0.1383	1.0540	0.0217	0.0247	−0.1455	0.0200	−0.7303	−0.0031	−0.0001	208
10	0.0561	0.1157	1.4186	0.0063	0.0273	−0.1470	0.0045	−0.3718	−0.0205	0.0001	202
11	0.0248	0.2026	1.3077	0.0233	0.0060	−0.1823	0.0700	−0.4644	0.0027	−0.0147	215
12	0.0295	0.1264	1.3370	0.0117	0.0073	−0.1767	−0.0016	−0.4390	−0.0103	−0.0147	215
13	0.0181	0.1527	1.2073	0.0136	0.0026	−0.1891	0.0195	−0.5642	−0.0069	−0.0179	206
14	0.0429	0.1966	1.1673	0.0200	0.0100	−0.1640	0.0651	−0.6057	−0.0010	−0.0110	208
15	0.0587	0.1441	1.1520	0.0083	0.0214	−0.1478	0.0142	−0.6224	−0.0131	−0.0001	208
16	0.0303	0.1440	1.3420	0.0183	0.0123	−0.1763	0.0137	−0.4320	−0.0030	−0.0090	210
17	0.0439	0.1748	1.2590	0.0227	0.0125	−0.1613	0.0519	−0.5214	−0.0008	−0.0110	208
18	0.0502	0.1293	1.2330	0.0193	0.0222	−0.1560	0.0013	−0.5430	−0.0027	0.0002	210
19	0.0312	0.1280	1.3270	0.0103	0.0119	−0.1745	0.0028	−0.4514	−0.0125	−0.0109	210
20	0.0548	0.1709	1.2500	0.0123	0.0120	−0.1510	0.0451	−0.5279	−0.0103	−0.0106	215
21	0.0317	0.1373	1.2657	0.0150	0.0209	−0.1753	0.0053	−0.5069	−0.0059	0.0001	208

Figure 4.

EPMA analysis images: (a) compositional, (b) point selection and (c) secondary images for sample 5; (d) compositional, (e) point selection and (f) secondary images for sample 21.

Estimation of element transfer in weld metal

Dilution (d) due to the base plate and dilution (1 − d) due to the electrode wire in SAW were determined (equation (7)) by cutting the welded joint transversely to observe the weld profile by performing the common metallurgical polishing operations and then etching with 5% nital solution (Figure 5). Area of fused base metal and filler metal were measured using a profile projector and digital planimeter. Expected weld metal content was calculated using dilution (d) due to the base plate, dilution (1 − d) due to the electrode wire, base metal and electrode wire element contents as per equation (8). A quantity delta (Δ) was introduced to find the element transfer due to flux, which is calculated by subtracting expected weld metal content from observed weld metal content. Positive and negative values of delta (Δ) indicate gain of element due to transfer of element from slag to weld metal and loss of element due to transfer from weld metal to slag, respectively. The results of the dilution (d), (1 − d), expected weld metal content (%), observed weld metal content (%) and element transfer due to flux (Δ) quantities have been presented in Tables 3 and 4.

Dilution, d (%) = \frac{Area of base metal fused}{Area of base metal fused + Area of filler metal fused}

(7)

\begin{matrix} Estimated weld metal content \\ = d \times base metal element content \\ + (1 - d) filler metal element content \end{matrix}

(8)

Figure 5.

Typical weldment and exploded weld zone.

Development of regression models for change in element transfer and microhardness

Using the observed values of chemical composition from experimentation, least square regression equations were developed in terms of percentage composition of individual flux constituents (CaO, Al₂O₃, CaF₂ and MgO). Second-order quadratic regression models are formed in terms of predictors of individual effect of mixture components (CaO, Al₂O₃, CaF₂ and MgO) and interaction effect of binary mixtures (CaO·Al₂O₃, CaO·CaF₂, CaO·MgO, Al₂O₃·CaF₂, Al₂O₃·MgO and CaF₂·MgO)

\begin{matrix} Δ C & = - 0.00380 CaO - 0.00073 A l_{2} O_{3} \\ + 0.00033 Ca F_{2} + 0.00100 MgO \\ + 0.00006 CaO \cdot A l_{2} O_{3} + 0.00001 CaO \cdot Ca F_{2} \\ + 0.00001 CaO \cdot MgO \\ - 0.00013 A l_{2} O_{3} \cdot Ca F_{2} - 0.00010 A l_{2} O_{3} \cdot MgO \\ - 0.00014 Ca F_{2} \cdot MgO \end{matrix}

(9)

\begin{matrix} Δ Si & = + 0.00465 CaO + 0.00538 A l_{2} O_{3} \\ - 0.00281 Ca F_{2} + 0.00025 MgO \\ - 0.00027 CaO \cdot A l_{2} O_{3} - 0.00004 CaO \cdot Ca F_{2} \\ + 0.00003 CaO \cdot MgO + 0.00002 A l_{2} O_{3} \cdot Ca F_{2} \\ - 0.00021 A l_{2} O_{3} \cdot MgO + 0.00014 Ca F_{2} \cdot MgO \end{matrix}

(10)

\begin{matrix} Δ Mn & = + 0.01454 CaO - 0.02373 A l_{2} O_{3} \\ + 0.00108 Ca F_{2} + 0.01849 MgO \\ - 0.00015 CaO \cdot A l_{2} O_{3} - 0.00022 CaO \cdot Ca F_{2} \\ - 0.00144 CaO \cdot MgO \\ + 0.00038 A l_{2} O_{3} \cdot Ca F_{2} + 0.00051 A l_{2} O_{3} \cdot MgO \\ - 0.00107 Ca F_{2} \cdot MgO \end{matrix}

(11)

\begin{matrix} Δ P = - 0.00106 CaO + 0.00066 A l_{2} O_{3} \\ - 0.00096 Ca F_{2} - 0.00030 MgO \\ - 0.00002 CaO \cdot A l_{2} O_{3} + 0.00005 CaO \cdot Ca F_{2} \\ + 0.00005 CaO \cdot MgO \\ + 0.00002 A l_{2} O_{3} \cdot Ca F_{2} - 0.000042 A l_{2} O_{3} \cdot MgO \\ + 0.00001 Ca F_{2} \cdot MgO \end{matrix}

(12)

\begin{matrix} Δ S & = - 0.00243 CaO - 0.00019 A l_{2} O_{3} \\ + 0.00053 Ca F_{2} - 0.00037 MgO \\ + 0.00007 CaO \cdot A l_{2} O_{3} + 0.00003 CaO \cdot Ca F_{2} \\ + 0.00006 CaO \cdot MgO \\ - 0.00004 A l_{2} O_{3} \cdot Ca F_{2} + 0.00001 A l_{2} O_{3} \cdot MgO \\ - 0.00003 Ca F_{2} \cdot MgO \end{matrix}

(13)

\begin{matrix} VHN & = + 3.3290 CaO + 3.2701 A l_{2} O_{3} \\ + 3.0573 Ca F_{2} + 3.6445 MgO \\ - 0.0424 CaO \cdot A l_{2} O_{3} - 0.0019 CaO \cdot Ca F_{2} \\ - 0.0303 CaO \cdot MgO - 0.0104 A l_{2} O_{3} \cdot Ca F_{2} \\ - 0.0162 A l_{2} O_{3} \cdot MgO - 0.0381 Ca F_{2} \cdot MgO \end{matrix}

(14)

Analysis of mathematical models

Prediction equations (9)–(14) have been checked for adequacy using t-test and analysis of variance (ANOVA) (F-test). Individual regression coefficients ( $b_{i}$ , $b_{j}$ and $b_{ij}$ ) have been checked by comparing t₀ value with tabulated t-value (at 95% significance level for dof_error)^16,17 (Appendix 1). Whole mixture model has been checked separately for linear and quadratic least square models as well as for full model tested by calculating F-value and p-value at 95% of significance level. R² values for all the mathematical models are nearly 80% or more except ΔMn, indicating that the predicted values are in close agreement with the actual/observed values of chemical composition for weld metal (Figure 6). Summary of regression equations indicating individual and interaction effects of binary mixtures on the chemical composition for weld metal is represented in Table 5.

Figure 6.

Predicted values versus actual/observed values of chemical composition for weld metal: (a) ΔC, (b) ΔSi, (c) ΔMn, (d) ΔP, (e) ΔS and (f) VHN.

Table 5.

Summary indicating the effect of mixtures on the change in element transfer and microhardness of weld metal.

Type of effect	Term	ΔC	ΔSi	ΔMn	ΔP	ΔS	Microhardness
Individual effect	CaO	Decrease	Increase	−	Decrease	Decrease	Decrease
	Al₂O₃	−	Increase	Decrease	Increase	−	−
	CaF₂	−	Decrease	−	Decrease	−	−
	MgO	−	−	Increase	−	−	−
Binary mixture effect	CaO·Al₂O₃	Synergistic	Anti-synergistic	−	−	Synergistic	Synergistic
	CaO·CaF₂	−	−	−	Synergistic	−	−
	CaO·MgO	−	−	Anti-synergistic	Synergistic	Synergistic	Synergistic
	Al₂O₃·CaF₂	Anti-synergistic	−	−	−	−	−
	Al₂O₃·MgO	Anti-synergistic	Anti-synergistic	−	Anti-synergistic	−	−
	CaF₂·MgO	Anti-synergistic	Synergistic	Anti-synergistic	−	−	−

Discussion

The effect of different mixtures on the element transfer of weld metal is indicated in Table 5 and Appendix 1. Table 5 indicates the type of effect (synergistic/anti-synergistic) of different mixtures on the change in element transfer of weld metal, and t-table (Appendix 1) shows the significance of various flux mixtures according to t-value of the corresponding term (flux mixture).

Effect of flux constituents on chemical composition of weld metal

Weld metal has lower carbon content than expected from base metal and filler metal as shown by results, which is in accordance with the previous literature,¹¹ which may be due to oxidation of carbon to its oxides during slag–metal reactions. Among the individual flux mixtures, CaO, MgO and Al₂O₃ affect ΔC. CaO has significant decreasing effect and Al₂O₃ has insignificant decreasing effect, whereas MgO has insignificant increasing effect on ΔC.

Binary mixture of CaO with Al₂O₃ has synergistic effect, whereas binary mixtures of Al₂O₃ with CaF₂ and MgO have anti-synergistic effect on ΔC. In the high-temperature environment near the welding plasma, CaO decomposes to release oxygen. The oxygen released from flux reacts with carbon of base metal and filler metal to form its oxides, thereby reducing weld metal carbon.

From the results, it seems that weld metal silicon content does not have any relationship with initial Si content of flux (SiO₂ content is constant) but dependent on Si content of base metal and filler metal up to some extent²⁶ but mainly dependent on slag–metal reactions. Increase in ΔSi content occurs in all the experiments except experiments 6 and 12 where it decreases, as shown in Table 4.

Gain of Si in all samples is due to dissociation of SiO₂ present in the flux to Si and O₂ as per equation (15)

Si O_{2} \to Si + O_{2}

(15)

Individual flux ingredients CaO and Al₂O₃ tend to increase weld metal silicon content, whereas CaF₂ tends to decrease Si content; binary mixtures CaO·Al₂O₃ and Al₂O₃·MgO have anti-synergistic (decreasing) effect whereas CaF₂·MgO has synergistic (increasing) effect on ΔSi, which is in agreement with the previous literature¹³ (Table 5).

MgO has insignificant increasing effect on Si content, which may be due to dissociation of MgO to Mg, in turn Si will form SiO₂, which will again react with Mg in the weld pool to form MgO again. So Si content is expected to increase due to flux.

Binary mixture CaO·MgO has insignificant anti-synergistic effect on ΔSi, which may be due to reduction of SiO₂ to complex silicates also verified in XRD analysis of slags in section “XRD analysis of slag.” Binary mixture CaF₂·MgO increases density of the molten metal, which hinders the transfer of Si across the slag–metal interface.

In general, manganese transfer depends on manganese content of electrode and flux composition.¹¹ Mn content in all samples is decreased except sample 6, which shows gain of Mn content. Individual flux ingredient Al₂O₃ tends to decrease whereas MgO tends to increase Mn content and binary mixtures; CaO·MgO and CaF₂·MgO have anti-synergistic effect on ΔMn.

Increase in Mn in weld metal may be due to the electrochemical reaction occurring at the base metal (cathode in DCRP). During this reduction reaction, Mg dissociates from MgO. Manganese will form MnO with oxygen in the molten metal and Mg will react with this MnO to form MgO again, which increases Mn content of weld metal. Addition of MgO to CaO and CaF₂ increases density and decreases conductivity of the molten metal, which increases the resistance to Mn movement at the slag–metal interface, thereby decreasing Mn content of weld metal and also lost by the vaporization of Mn from electrode/base metal.

Phosphorus is assumed to be an impurity, which increases strength of low-alloy steels and increases crack tendency during welding at the same time so should be kept minimum. Samples 1, 4 and 11 show gain of P content in weld metal and all the rest show loss of P. Individual flux ingredients CaO and CaF₂ tend to decrease, whereas Al₂O₃ tends to increase P content and binary mixtures; CaO·CaF₂ and CaO·MgO have synergistic effect on ΔP and Al₂O₃·MgO has anti-synergistic effect on phosphorus.

CaO and CaF₂ oxidize phosphorus to its oxides, thereby decreasing its content. Substitution of flux ingredient by CaO and CaF₂ lowers activity coefficient of phosphorous pentoxide, $γ_{2 P_{2} O_{5}}$ in the slag promoting the reaction given by equation (16) and thereby decreasing phosphorus content¹⁰

4 P + 5 O_{2} = 2 (P_{2} O_{5})

(16)

Samples 3, 4, 7, 10, 18 and 21 show gain of S content in weld metal and all the rest show loss of sulfur. Only CaO among individual flux ingredients has significant decreasing effect on ΔS as seen from equation (12) and table (t coefficients in Appendix 1), and binary mixtures CaO·Al₂O₃ and CaO·MgO have synergistic effect on ΔS. Lime fluxes are well known to remove impurities during welding processes. CaO combines with sulfur of weld metal forming CaS and releasing O₂ as per reaction mentioned in equation (17), thereby lowering sulfur content from weld metal contrary to previous results by Kanjilal et al.¹³ and in agreement with the results by Chai and Eager¹⁰

CaO + S \to CaS + \frac{1}{2} O_{2}

(17)

MnO from flux also reacts with weld metal S to form MnS and releasing O₂, thereby decreasing S content in all other samples having lesser CaO.

Effect of flux constituents on microhardness (VHN)

The effect of flux constituents on mechanical properties has been studied and mathematical models have been developed in the earlier work done by the authors using same DOE and same design table.¹⁹ The prediction equations for these mechanical properties are given in Appendix 2. Only CaO among individual flux ingredients has significant decreasing effect and binary mixtures; CaO·Al₂O₃ and CaO·MgO have synergistic effect on microhardness as seen from equation (14) and table t coefficients (Appendix 1). CaO has decreasing effect on phosphorus and sulfur, which is also justified in same section, thus improving the microhardness. CaO·Al₂O₃ and CaO·MgO tend to increase impurities, as CaO·Al₂O₃ also has synergistic effect on S content and CaO·MgO has synergistic effect on S and P, thereby increasing microhardness.

XRD analysis of slag

XRD analysis of all slag samples was carried out after welding. For the XRD measurements, the slag specimens were analyzed using 2θ diffraction mode ranging from 10° to 120°. The XRD patterns of slags 1, 10, 16 and 19 are shown in Figure 7. The different crystalline compounds are formed in the slags, which are in agreement with the previous literature^27–29 such as gehlenite (calcium magnesium aluminum silicate, Ca₂(Mg_0.25Al_0.75)(Si_1.25Al_0.75O₇)), melilite (calcium magnesium aluminum silicate, Ca₂(Mg_0.5Al_0.5)(Si_1.5Al_0.5O₇)), merwinite (calcium magnesium silicate, Ca₃Mg(SiO₄)₂), joesmithite (calcium lead aluminum iron magnesium beryllium silicate hydroxide, Ca_5.2Pb.₈Al.₃Fe_4.8Mg_4.9Be₄Si₁₂O₄₄(OH)₄), gismondine (calcium aluminum silicate hydrate, CaAl₂Si₂O₈·4H₂O), garronite (sodium calcium aluminum silicate hydrate, Na.₈Ca_2.82(Al₆Si₁₀O₃₂)(H₂O)_12.08), ungarettiite (sodium potassium calcium magnesium manganese silicate, (Na_0.80K_0.15)(Na_1.97Ca_0.03)(Mn_0.93Mg_0.07)(Mn_1.78Mg_0.22). Mn₂(Si₈O₂₂)O₂) and margarosanite (lead calcium manganese silicate, Pb(Ca Mn)₂(SiO₃)₃).

Figure 7.

XRD analysis of some slag samples.

It may be seen from the phase identification database (PDF) index names and chemical formula of the above compounds that silicates and silicate hydroxide have been observed in most of the high basicity slags and lesser in low basicity slags. Silicates are formed due to oxidation of different elements such as Mn, Al and Mg in the presence of Si, whereas silicate hydroxides have been formed due to oxidation of elements and moisture pickup from atmosphere. Sodium element is being observed in about all the slag compounds, which may be due to sodium silicate binder added during the agglomeration process in the formulation of fluxes.

Contour surface plots for various properties

Contour plots indicating predicted values of various properties are shown in Figure 8. Different regions on contour surface³⁰ show variation in change in element transfer (Δ) so every contour curve marked on the surface gives constant value of change in element transfer and each dotted point on plot indicates one of the flux mixture combination. Figure 8 shows contour plots for ΔC, ΔSi, ΔMn, ΔP, ΔS and microhardness for different proportions of flux components CaO, CaF₂ and Al₂O₃ with constant MgO = 20% content.

Figure 8.

Contour plots for (a) ΔC, (b) ΔSi, (c) ΔMn, (d) ΔP, (e) ΔS and (f) microhardness at different proportions of flux components CaO, CaF₂ and Al₂O₃ with constant MgO = 20% content.

Optimization of chemical composition and mechanical properties

In the multi-response optimization; attempt has been made to optimize the chemical composition of the weld metal equivalent to that of base,³¹ and mechanical properties are maximized with microhardness to a target value of VHN = 215 (base metal). Mechanical properties, ultimate tensile strength and impact strength values have been used from the previous work done by Jindal et al.¹⁹ given in Appendix 2. Simultaneous optimization of these output responses has been done using composite desirability optimization method suggested by Derringer and Suich.³² This method makes use of an objective function D(x), called desirability function, which transforms an estimated response into a value called composite desirability. Composite desirability is the weighted geometric mean of individual desirability for the responses. The factor settings with maximum total desirability are considered to be the optimal parameter combinations.³³

The composite desirability is given by the following equations³⁴

D = {[d_{1}^{w_{1}} \cdot d_{2}^{w_{2}} \dots d_{n}^{w_{n}}]}^{^{\frac{1}{(w_{1} + w_{2} + \dots + w_{n})}}}

(18)

D = {[Π_{i}^{n} d_{i}^{w_{i}}]}^{^{\frac{1}{(\sum_{i}^{n} w_{i})}}}

(19)

where n is the number of responses, $d_{i}$ the desirability of particular response and $w_{i}$ is the weights satisfying 0 < w_i < 1 and $(w_{1} + w_{2} + \dots + w_{n}) = 1$ .

Three optimum solutions with equal weights for all responses were determined at different levels of desirability as presented in Table 6.

Table 6.

Optimum flux mixtures for chemical composition and mechanical properties.

Number	CaO	Al₂O₃	CaF₂	MgO	C	Si	Mn	P	S	UTS	Impact	VHN	Desirability
1	11.61	12.33	15.00	39.06	0.06	0.16	1.37	0.01	0.01	552.55	109.56	215.00	0.71
2	31.44	11.07	25.49	10.00	0.04	0.16	1.40	0.02	0.01	603.03	104.78	215.00	0.63
3	10.00	37.89	12.58	17.53	0.03	0.15	1.30	0.01	0.01	557.79	125.68	213.83	0.57

UTS: ultimate tensile stress.

Model validation

Confirmatory experiments were conducted using randomly selected flux mixtures to validate the regression model and to ensure the repeatability and reliability of the predicted values (Table 7). Six random flux mixtures were selected; weld samples were chemically and mechanically analyzed. From the experimental values, it has been observed that error (%) for each response is nearly 5% in almost all the samples.

Table 7.

Percentage error of predicted values and experimental values.

Flux constituents				Predicted values						Experimental values						Error (%)
CaO	Al₂O₃	CaF₂	MgO	ΔC	ΔSi	ΔMn	ΔP	ΔS	VHN	ΔC	ΔSi	ΔMn	ΔP	ΔS	VHN	ΔC	ΔSi	ΔMn	ΔP	ΔS	VHN
10.0	8.0	20.0	40.0	−0.1489	0.0766	−0.5835	−0.0114	−0.0174	214	−0.1509	0.0746	−0.6014	−0.0119	−0.0168	208	1.34	2.57	3.05	4.39	3.22	2.80
35.0	8.0	25.0	10.0	−0.1629	0.0569	−0.4557	−0.0038	−0.0224	220	−0.1589	0.0587	−0.4557	−0.0037	−0.0215	229	2.43	3.11	0.00	1.85	4.37	4.09
18.0	40.0	10.0	10.0	−0.1455	0.0200	−0.7303	−0.0031	−0.0001	208	−0.1496	0.0205	−0.6989	−0.0032	−0.0001	216	2.78	2.36	4.30	3.23	1.50	3.85
21.7	8.0	24.2	24.2	−0.1640	0.0651	−0.6057	−0.0010	−0.0110	208	−0.1560	0.0681	−0.5901	−0.0010	−0.0116	200	4.86	4.62	2.58	4.64	4.68	3.85
21.3	22.8	23.8	10.0	−0.1763	0.0137	−0.4320	−0.0030	−0.0090	210	−0.1850	0.0142	−0.4502	−0.0029	−0.0086	220	4.93	4.12	4.21	3.33	4.89	4.76
18.6	18.9	20.3	20.3	−0.1753	0.0053	−0.5069	−0.0059	0.0001	208	−0.1698	00052	−0.4888	−0.0061	0.0001	217	3.12	2.35	3.57	4.10	3.64	4.33

$Error (% age) = \frac{Actual Value - Predicted Value}{Predicted Value} \times 100$ .

Conclusion

The following conclusions are drawn:

Statistically designed flux mixtures have been formulated by mixing eight flux components, varying four components CaO, Al₂O₃, CaF₂ and MgO and keeping SiO₂, TiO₂ and MnO constant using the extreme vertices design method.

The mathematical models have been developed for the element transfer and microhardness, in terms of individual and binary flux mixtures for welding of HSLA steel. These models may be readily used to determine element transfer in weld metal for the same set of welding variables.

CaO is the most significant flux constituent and Al₂O₃ is the second most significant constituent among individual mixtures.

CaO·MgO and CaO·Al₂O₃ among binary mixtures have significant effect on element transfer and microhardness.

CaO is the important flux constituent, which tends to decrease ΔC, ΔP, ΔS and microhardness, but tends to increase ΔSi content.

Only MgO flux mixture has increasing effect on ΔMn, whereas no effect on any element transfer or microhardness.

Individual effect of Al₂O₃ in the flux has unfavorable effect on weld metal properties, whereas its binary mixtures have favorable effect on weld metal properties.

Optimum flux mixtures provided here may be used practically in the industry providing weld metal composition equivalent to base metal (HSLA steel).

The methodology adopted in this work may be applied to other base metal materials and electrode wire combinations to get optimum mixtures.

Footnotes

Appendix 1

Table 10.

t-Value and significance testing of regression coefficients for element transfer and microhardness.

Predictor	ΔP		ΔS		VHN
	t₀	Whether significant or not^*	t₀	Whether significant or not^*	t₀	Whether significant or not^*
CaO	−1.90	Significant	−2.60	Significant	12.95	Significant
Al₂O₃	2.01	Significant	−0.34	Not significant	21.67	Significant
CaF₂	−2.67	Significant	0.87	Not significant	18.45	Significant
MgO	−0.83	Not significant	−0.61	Not significant	21.99	Significant
CaO·Al₂O₃	−0.88	Not significant	2.51	Significant	−5.26	Significant
CaO·CaF₂	2.57	Significant	1.16	Not significant	−0.23	Not significant
CaO·MgO	2.69	Significant	2.08	Significant	−3.68	Significant
Al₂O₃·CaF₂	1.29	Not significant	−1.58	Not significant	−1.55	Not significant
Al₂O₃·MgO	−2.65	Significant	0.37	Not significant	−2.42	Significant
CaF₂·MgO	0.79	Not significant	−1.20	Not significant	−5.51	Significant

Predictor is significant if |t₀|> t_α, dof error (t_{0.05, 11} = 1.796).

Appendix 2

Table 11.

Mechanical properties of weld metal samples.

Experiment number	Ultimate tensile stress (MPa)			Impact strength (N m)
	Sample 1	Sample 2	Mean	Sample 1	Sample 2	Mean
1	565.7	510.4	538.1	63.8	157.0	110.4
2	535.6	511.5	523.6	49.1	98.1	73.6
3	528.4	591.84	560.1	107.9	112.8	110.4
4	582.7	571.55	577.1	127.5	112.8	120.2
5	520.2	572.04	546.1	98.1	89.8	94.0
6	527.1	527.1	527.1	100.0	110.0	105.0
7	647	630	638.5	101.9	98.1	100.0
8	614.8	583.5	599.2	88.3	88.3	88.3
9	422.2	600	511.1	39.2	117.7	78.5
10	460	416	438.0	39.2	56.4	47.8
11	600.1	589.64	594.9	132.4	147.2	139.8
12	573.4	584.6	579.0	107.9	132.4	120.2
13	540.1	567.9	554.0	98.1	186.4	142.2
14	635.3	574.7	605.0	132.4	137.3	134.9
15	520.6	579.5	550.1	148.1	107.9	128.0
16	610.3	581	595.7	157.0	103.0	130.0
17	496.4	507.3	501.9	93.2	103.0	98.1
18	545.9	544.1	545.0	122.3	117.7	120.0
19	625.7	559.9	592.8	93.2	122.6	107.9
20	575.4	573.5	574.5	98.1	122.6	110.4
21	600.5	585.9	593.2	132.4	141.8	137.1

Ultimate tensile strength = −4.9531 CaO + 7.6824 Al₂O₃+ 2.7335 CaF₂+ 6.0239 MgO + 0.0344 CaO·Al₂O₃+ 0.5013 CaO·CaF₂+ 0.3182 CaO·MgO + 0.1326 Al₂O₃·CaF₂− 0.0218 Al₂O₃·MgO − 0.0197 CaF₂·MgO.

Impact strength = −4.6949 CaO − 0.8874 Al₂O₃− 3.1084 CaF₂− 1.9637 MgO + 0.0484 CaO·Al₂O₃+ 0.2148 CaO·CaF₂+ 0.2447 CaO·MgO + 0.2145 Al₂O₃·CaF₂+ 0.1044 Al₂O₃·MgO + 0.0906 CaF₂·MgO.

Declaration of conflicting interests

The authors declare that there is no conflict of interest.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

References

Houldcroft

. Submerged-arc welding. Cambridge: Abington Publishing, 1989.

Murugan

Gunaraj

. Prediction and control of weld bead geometry and shape relationships in submerged arc welding of pipes. J Mater Process Tech2005; 168: 478–487.

Parmar

. Welding processes and technology. New Delhi, India: Khanna Publishers, 1992.

Kanjilal

Pal

Majumdar

. Combined effect of flux and welding parameters on chemical composition and mechanical properties of submerged arc weld metal. J Mater Process Tech2006; 171: 223–231.

O’Brien

Guzman

. Welding handbook: welding processes, part 2. Miami, FL: American Welding Society, 2004.

Olson

Frost

. Welding and joining processes. Presented at the winter annual meeting of the ASME, Atlanta, GA, 1–6 December 1991.

Cochrane

Ward

Keville

. The influence of deoxidation and/or desulphurisation practice on the weld metal toughness of high dilution welds. In: The effects of residual, impurity, and microalloying elements on weldability and weld properties, London, 15–17 November1983, p.16. Cambridge: The Welding Institute.

Barritte

Edmonds

. Proceedings conference on advances in the physical metallurgy of steel. London: The Metals Society, 1981.

Kanjilal

Majumdar

Pal

. Prediction of acicular ferrite from flux ingredients in submerged arc weld metal of C-Mn steel. ISIJ Int2005; 45: 876–885.

10.

Chai

Eagar

. The effect of SAW parameters on weld metal chemistry. Weld J1980; 59: 93s–98s.

11.

Mitra

Eagar

. Slag metal reactions during submerged arc welding of alloy steels. Metall Trans A1984; 15: 217–227.

12.

Pandey

Bharti

Gupta

. Effect of submerged arc welding parameters and fluxes on element transfer behaviour and weld-metal chemistry. J Mater Process Tech1994; 40: 195–211.

13.

Kanjilal

Pal

Majumdar

. Prediction of element transfer in submerged arc welding. Weld J2007; 86: 135s–146s.

14.

Burck

Indacochea

Olson

. Effects of welding flux additions on 4340 steel weld metal composition. Weld J1990; 3: 115–122.

15.

Du Plessis

du Toit

Pistorius

. Control of diffusible weld metal hydrogen through flux chemistry modification. Weld J2007; 86: 273s–280s.

16.

Anderson

McLean

. Design of experiments: a realistic approach. New York: Marcel Dekker, Inc., 1974.

17.

Cornell

. Experiments with mixtures: designs, models, and the analysis of mixture data. New York: John Wiley & Sons, 2002.

18.

Adeyeye

Oyawale

. Mixture experiments and their applications in welding flux design. J Braz Soc Mech Sci Eng2008; 30(4): 319–326.

19.

Jindal

Chhibber

Mehta

. Investigation on flux design for submerged arc welding of high-strength low-alloy steel. Proc IMechE, Part B: J Engineering Manufacture2013; 227: 383–395.

20.

Zhang

Lee

. Use of phase diagrams in studies of refractories corrosion. Int Mater Rev2000; 45: 41–58.

21.

Bhandari

Chhibber

Arora

. Effect of electrode coatings on diffusible hydrogen content, hardness and microstructures of the ferritic heat affected zones in bimetallic welds. Adv Mater Res2012; 383: 4697–4701.

22.

Houldcroft

. Welding process technology. Cambridge: Cambridge University Press, 1977.

23.

Singh

Pandey

. Economics of recycling submerged arc welding slag as a flux. In: Proceedings of the 8th international conference on trends in welding research, Pine Mountain, GA, 2–6 June 2008, pp.806–810. Materials Park, OH: ASM International.

24.

Jindal

Chhibber

Mehta

. Effect of welding parameters on bead profile, microhardness and H₂ content in submerged arc welding of high-strength low-alloy steel. Proc IMechE, Part B: J Engineering Manufacture2014; 228: 82–94.

25.

Linnert

. Welding metallurgy. Miami, FL: American Welding Society, 1994.

26.

Blondeau

. Metallurgy and mechanics of welding. Hoboken, NJ: John Wiley & Sons, 2013.

27.

Baune

Bonnet

Liu

. Reconsidering the basicity of a FCAW consumable—part 1: solidified slag composition of a FCAW consumable as a basicity indicator. Weld J2000; 79: 57s–65s.

28.

Chang

K-L

Huang

C-T

Huang

. Investigations of microstructure and phosphorus distribution in BOF slag. China Steel Tech Rep2008; 21: 1–6.

29.

Paniagua-Mercado

Lopez-Hirata

. Chemical and physical properties of fluxes for SAW of low-carbon steels. Arc Weld2011; 13: 281–298.

30.

Hummel

. Introduction to phase equilibria in ceramics systems. Boca Raton, FL: CRC Press Ltd, 1984.

31.

Ramirez

. Characterization of high-strength steel weld metals: chemical composition, microstructure, and nonmetallic inclusions. Weld J2008; 87: 65s–75s.

32.

Derringer

Suich

. Simultaneous optimization of several response variables. J Qual Technol1980; 12: 214–219.

33.

Harrington

. The desirability function. Ind Qual Control1965; 21: 494–498.

34.

Del Castillo

Montgomery

McCarville

. Modified desirability functions for multiple response optimization. J Qual Technol1996; 28: 337–345.