1 Introduction

Production of heavy gauge (> 16 mm wall thickness) X70 pipe is a recent focus in pipeline manufacturing to achieve the requirement of increased operating pressures [1]. As one of the manufacturing processes, submerged arc welding (SAW) offers substantial benefits, including high deposition rate, deep penetration, and reduced welding times. These attributes are beneficial for welding heavy gauge plates [1, 2]. To increase the productivity of heavy gauge pipes, tandem submerged arc welding (TSAW) containing two or more electrodes has been developed for the welding procedure of line pipes to achieve a high deposition rate [3]. The use of multi-electrode TSAW may, however, result in an increase in overall heat input which affects the metallurgy/properties of the weld metal (WM) and heat-affected zone (HAZ), especially in the coarse-grained heat-affected zone (CGHAZ). Consequently, the quality of weldments, such as toughness and microstructure, may deteriorate [4, 5]. To increase productivity while maintaining the quality of TSAW pipeline products, cold wire tandem submerged arc welding (CWTSAW) was developed [6, 7].

CWTSAW is defined as a TSAW process where an additional electrode with no arc is introduced within the molten pool. The primary objective of involving a cold wire is to increase the deposition rate and decrease the overall heat input into the weld. This method was first developed by Mruczek and Parker [6] in the SAW process for enhancing welding capabilities and improving weld productivity by increasing deposition rate. Ramakrishnan et al. [8] utilized three wires and cold wire addition in SAW welds, and these contributed to a 30% increase in weld toughness relative to a conventional three-wire process. Júnior et al. [9] reported that feeding a cold wire in the multiple wires SAW process of a duplex stainless steel with a heat input of 2.7 kJ/mm resulted in increased corrosion resistance due to a decreased HAZ width and increased austenite phase fraction. Previous studies [7, 10, 11] applied CWTSAW to an intermediate gauge X70 line pipe (13.4 mm thick) and showed a 17% reduction in overall heat input and a 12% increase in deposition rate with improved fracture toughness, which enhanced the productivity of intermediate gauge X70 line pipe welding. In the pursuit of promising welding techniques with cold wire addition, the levels of heat input (HI) and cold wire feed speed (CWFS), in contrast to previous work [7], need to be increased for the production of heavy gauge X70 line pipe (19.1 mm thick). Adjustments to the levels of welding process parameters and bevel geometry can cause a significant change in the resulting geometry for the weld and HAZ [2]. Furthermore, the microstructure and properties of the weld are influenced by the composition of the WM and weld shape [12]. In the pipeline industry, a qualified weld geometry can be described as a shape with deep penetration and a smaller HAZ area at lower heat inputs [7, 12]. Therefore, it is necessary to design an experiment to study the effect of CWTSAW process parameters and bevel design on the change in weld geometry and micro-hardness profiles of heavy gauge X70 steel.

The Taguchi design is a statistical technique used to design an experiment and optimize manufacturing process parameters [13]. This method has been widely used in numerous studies to optimize the welding process. Tarng and Yang [14] determined the optimal welding parameters of the SAW process with a small number of tests utilizing the Taguchi design. Sarkar et al. [15] employed Taguchi design to analyze the effective contribution of bead geometry on tensile strength and validated the results with confirmatory tests in the SAW process. The weld geometry and micro-hardness measurements from the Taguchi design were analyzed using three statistical methods; i.e., analysis of variance (ANOVA), three-order multiple regression (TOMR) analysis, and signal-to-noise (S/N) ratio. ANOVA was used to determine the significant weld parameters by analyzing the variability of the data using the variance ratio (F-value), the sum of squares (SS), and level of significance (P-value) [16]. TOMR analysis is a nonlinear model used to develop empirical equations for response characteristics and mechanical properties, such as the CGHAZ area and micro-hardness of the CGHAZ and WM [17]. S/N ratio is the method for optimizing the levels of weld parameters to improve weld geometry [18].

In this study, the CWTSAW process was applied to heavy gauge (19.1 mm thick) X70 line pipe steel. The effects of heat input of the lead (HIL) and trail (HIT) electrodes, the voltage of the lead (VL) and trail (VT) electrodes, travel speed (TS), bevel design (BD), and cold wire feed speed (CWFS) on the weld geometry, hardness of the HAZ and microstructural modification in the CGHAZ were studied. In total, 16 test welds were conducted. The specific test welds were designed using the Taguchi method. Following welding, the weld geometrical characteristics were measured, including the reinforcement area (RA), the height of the reinforcement area (HRA), bead toe angle (BTA), aspect ratio (AR), semi-penetration ratio (SPR), and coarse-grained heat-affected zone (CGHAZ) area. In addition, the amount of weld dilution (DIL) and the micro-hardness of the CGHAZ and WM were measured. Finally, a comparison of martensite-austenite (MA) constituents in the CGHAZ of the conventional TSAW and CWTSAW processes was undertaken using optical microscopy (OM) and scanning electron microscopy (SEM).

2 Experimental methods

2.1 Base and electrode material

The CWTSAW weld tests were conducted on heavy gauge (19.1 mm thick) X70 microalloyed steel. The composition of the X70 steel is shown in Table 1. The studied X70 microalloyed steel was fabricated through thermo-mechanical controlled processing (TMCP) [1].

Table 1 Composition of X70 microalloyed steel and electrode (wt. %)
Full size table

The electrodes used in the study for both hot wires and the cold wire were selected based on EN14295/EN 756. A 4 mm diameter S2Mo solid wire was selected for the electrodes and cold wire. The compositions are shown in Table 1.

2.2 Weld bevel specifications

Weld samples were fabricated and machined with two different bevel designs. Figure 1 compares the two bevel types used. Each bevel is generated on the same thickness of skelp, but the cross-sectional areas are different. For instance, the 60° bevel with a 4.5 mm depth has a 12 mm2 cross-sectional area, while the 90° bevel with a 5 mm depth has a larger cross-sectional area of 25 mm2. The quantitative description of the bevel specification is convenient for the representation of the two bevel types in statistical analysis.

Fig. 1
figure 1

Schematic view of bevel specifications

Full size image

2.3 CWTSAW setup

Three electrodes (lead, trail, and cold electrode) were fed into the molten pool, as illustrated in Fig. 2a. The electrically cold electrode is located at a lagging position relative to the trailing electrode. The electrode setup is shown in Fig. 2b, including the stick-out length (25 mm), electrode separation (13 mm), and the angular position of each electrode relative to a normal skelp surface. The constant power sources of the lead and trail electrodes were direct current electrode positive (DCEP) polarity and square wave alternating current (ACSQ) polarity, respectively. There was no power source for the cold electrode. As a basic feature in SAW, a consumable granular flux is needed to shield the welding pool and fill the bevel area. According to EN 760, BF6.5 consumable flux was used in the CWTSAW process.

Fig. 2
figure 2

Setup for CWTSAW process: a overview of welding setup, b schematic view of fixed welding variables

Full size image

2.4 Welding parameters

The four fundamental welding parameters are current, voltage, travel speed, and feed speed of the electrodes. For the CWTSAW process, there are additional welding parameters based on the use of multiple electrodes. However, in order to understand the effect of heat input on the weldments and to keep the number of welding tests manageable, the current of the lead and trail electrodes was not varied independently and the feed speed of the trail and lead electrodes was kept constant. The nominal heat input (HI) and voltage (V) are key welding parameters and the current was calculated and set during welding according to Eq. (1) [7].

$$\mathrm{HI }\left(\frac{\mathrm{kJ}}{\mathrm{mm}}\right)=\frac{\eta \cdot V\cdot I}{1000\cdot \mathrm{TS}}$$
(1)

where HI, V, I, and TS represent nominal heat input, voltage, current, and travel speed. The arc efficiency (ƞ) is in the range of 0.9–1.0 for SAW [7].

Seven welding parameters with mixed levels, including five main welding parameters such as heat input, the voltage of lead, trail electrodes, and travel speed, were selected for CWTSAW tests. These are shown in Table 2. The heat input, voltage, and travel speed are fundamental and crucial parameters for a welding process and need to be included. Moreover, the cold wire parameter and bevel specification parameter are CWFS and BD, respectively. CWFS and BD were selected to study the effect of the cold wire addition and varied bevel types on weld characteristics and hardness. Only CWFS is a four-level welding parameter, while the others are two-level.

Table 2 CWTSAW tests parameters and input levels
Full size table

2.5 Experimental table

Taguchi analysis was employed for the parametric study of CWTSAW. The main advantage of Taguchi analysis is the use of a small number of welding tests, which is more economical and effective than using a factorial design [13]. Based on six parameters with two levels and one parameter with four levels, an L16 orthogonal array was designed using Taguchi analysis and is shown in Table 3. The L16 array comprised 16 weld tests, and each level of the parameters appears the same number of times in each column. The weld geometrical values were measured for all 16 weld tests and analyzed using statistical methods. Additionally, a welding table of three validation tests is shown in Table 4. These tests were used to validate the linear trend between the measured and predicted results in the TOMR analysis.

Table 3 L16 orthogonal array based on Taguchi analysis
Full size table
Table 4 Welding conditions for validation tests
Full size table

2.6 Weld characteristics

Three weld samples extracted from each weldment, as shown in Fig. 3a, were used to measure geometry characteristics. A vertical band saw, assisted with coolant, was employed in sectioning. A total of 48 specimens were examined. Each sample was mounted and polished following the ASTM E3-11 standard [19] and then was macro-etched using 4% nital to reveal the HAZ and WM boundaries (Fig. 3b). A stereomicroscope image was obtained from each section and analyzed using Image J 1.52a software to obtain the weld geometry values, including bead width (BW), penetration depth (PD), penetration area (PA), bead width at half penetration depth (BW1/2), HRA, RA, bead toe angle (BTA), and CGHAZ area, as per Fig. 3b.

Fig. 3
figure 3

a Schematic showing sectioning position of welds. b Macrograph of sample test 1–1

Full size image

The weld geometry values obtained from the stereomicroscope images were utilized to calculate AR, SPR, and the amount of DIL of the weld, as shown in Eqs. (2) to (4) [7], respectively.

$$\mathrm{AR}=\frac{\mathrm{PD}}{\mathrm{BW}}$$
(2)
$$\mathrm{SPR}= \frac{{\mathrm{BW}}_{1/2}}{\mathrm{BW}}$$
(3)
$$\mathrm{DIL}= \frac{\mathrm{PA}-\mathrm{BA}}{\mathrm{PA}+\mathrm{RA}}$$
(4)

The bevel areas (BA) in the two bevel specifications were 12 and 25 mm2 (Fig. 1).

2.7 Micro-hardness

Vickers micro-hardness measurements were obtained from the BM, HAZ, and WM, as illustrated in Fig. 4 [20]. Two indentation lines (5 mm below the surface) contained a total of 50 indents from each weld sample to guarantee 10 to 12 indents in each of the WM and CGHAZ. A 500 g load and a dwell time of 14 s were used. The distance between two indents was three times the size of an indent, as indicated in the optical images in Fig. 4.

Fig. 4
figure 4

Schematic of the micro-hardness mapping along the HAZ and WM of sample T1-1

Full size image

2.8 Microstructural characterization

Optical microscopy (OM) was conducted using an Olympus BX61 microscope and OLYMPUS Stream Motion software. Scanning electron microscopy (SEM) was done using a Zeiss EVO M10 SEM operating at 20 kV accelerating voltage. The weld specimens were micro-etched with modified LePera’s solution to reveal MA constituents in the HAZ [21]. Then, three optical and three SEM secondary electron (SE) micrographs were taken from a location 5 mm below the weld surface and 50–200 µm from the fusion line. The phase fractions of MA constituents in the CGHAZ of TSAW and CWTSAW welds were measured from both optical and SEM micrographs using the color threshold feature of Image J 1.52a software.

3 Results

The BW, HRA, BTA, RA, CGHAZ area, CGHAZ hardness, WM hardness, PA, PD, and BW1/2 were measured for all 16 test welds. The values for each characteristic are shown in Table 5. Of particular note is the largest value of CGHAZ area (30.2 mm2) observed for test 8. This test had the highest heat input of all the welds. Maximum values for other characteristics are indicated in bold, and minimum values are underlined in each column.

Table 5 Measured weld characteristics
Full size table

The calculated values for AR, SPR, and DIL are shown in Table 6. Maximum values for each calculated characteristic are indicated in bold and minimum values are underlined in each column. The largest values of AR and SPR are for test 8 (highest heat input). The lowest amount of DIL is observed in test 13. In general, a lower amount of DIL leads to better weld properties, as reported in references [22, 23].

Table 6 Calculated weld characteristics
Full size table

The relationship between different weld characteristics can be made to help understand the welding process. Figure 5a shows a plot of PA vs. CGHAZ area, where CGHAZ area is shown to increase as PA increases. The trend line shown is for illustrative purposes only. This indicates that as weld metal size increases (via heat input and bevel design), an accompanying increase in CGHAZ area will occur. Both test 4 and test 8 appear to show significant increases in PA relative to the size of the CGHAZ area. Test 8 has the highest heat input, and test 4 appears to be an outlier.

Fig. 5
figure 5

Plot of CGHAZ area vs. a PA and b RA

Full size image

The relationship between RA and CGHAZ area is shown in Fig. 5b. The CGHAZ area increases as RA decreases. The decreased RA may facilitate full penetration in the weld, which may influence the size of the CGHAZ area. However, test 7 and test 8 do not follow the trend, which may be related to their high heat input. The bevel design (BD) is also believed to influence the observed correlation between the RA and CGHAZ area. This will be discussed in Sect. 4.1. Additional analysis of the correlation between welding parameters and measured weld geometry is presented in Sect. 4.2.

4 Discussion

4.1 Significance of welding parameters

The analysis of variance (ANOVA) was carried out on the measured weld characteristics to determine the statistical significance of the welding parameters. All ANOVA tables are presented in Appendix and include probabilities of significance (P), degrees of freedom (DF), the sum of squares (SS), variance ratios (F), and coefficient of determination (R2) values for each welding parameter. According to the statistical analysis reported by Mruczek and Parker [6] and Shahverdi et al. [24], P-values are used to determine the statistical significance of each of the welding parameters. A P-value equal to or less than 0.05 indicates the parameter is statistically significant with 95% confidence. A P-value of 0.25 corresponds to a 75% confidence level that the parameter is statistically significant for the welding characteristic. For this work, the significant parameters for the weld geometry results and the amount of DIL were selected based on a 95% confidence level, while a 75% confidence level was used to analyze the miro-hardness values of the CGHAZ and WM. The selected significant parameters for each weld characteristic, the amount of DIL, and micro-hardness profiles are shown in Fig. 6. For example, the significant welding parameters affecting RA are HIT, VT, and BD, all with a confidence level of 95%. Of particular note is that BD has a significant effect on the CGAHZ area and reinforcement size (RA and BTA), but BD has only a minor influence on the micro-hardness profiles. In addition, CWFS has a dominant influence on the micro-hardness profiles in the CGHAZ and WM, but the effect of CWFS on the DIL, bead ratio values, and reinforcement size is insignificant. The dominant effect of BD and CWFS on reinforcement size and micro-hardness profiles will be discussed when the analysis of effective distribution is considered.

Fig. 6
figure 6

Significant welding parameters for weld geometry, DIL, and the micro-hardness of WM and CGHAZ

Full size image

The most significant welding parameter was decided for each weld geometry result and micro-hardness profile using the effective contribution (\(\uprho\)). The effective contribution of each parameter depends on the sum of squares, which is the deviation from the total average value of population. The concept of effective contribution is a fundamental term in ANOVA analysis and can be calculated by Eq. (5) [16].

$$\rho (\mathrm{\%})=\frac{{\mathrm{SS}}_{i}}{{\mathrm{SS}}_{t}}\cdot 100\mathrm{\%}$$
(5)

where \(\rho\) is the effective contribution of each parameter to the response characteristics and \({\mathrm{SS}}_{i}\) and \({\mathrm{SS}}_{t}\) are the sum of squares for each parameter and the total sum of squares, respectively. The contribution evaluates the importance of parameters on each weld characteristic and the WM and CGHAZ micro-hardness. The significant contributions for BW, AR, DIL, and SPR are shown in Fig. 7.

Fig. 7
figure 7

Effective contributions of CWTSAW process parameters for BW, AR, DIL, and SPR

Full size image

Overall voltage (lead + trail electrodes) and TS significantly influence BW, AR, and SPR, as shown in Fig. 7. It is generally accepted that a higher arc voltage leads to a wider arc length promoting the formation of a wide BW [2, 25, 26]. Pepin et al. [26] correlated TS with the penetration profile of intermediate gauge strip and reported that a high TS reduces the filler metal per unit length of weld, leading to a narrow weld. Specifically, a faster TS and lower V result in a shorter arc length and, as such, a smaller BW. The AR and SPR were calculated using Eqs. (2) and (3). Both geometric ratio results are significantly affected by VL and TS. An increasing TS results in a smaller BW due to the reduced heat input and reduced melted metal per unit length [25]. Therefore, both AR and SPR are affected by voltage and TS due to the change in BW.

DIL is defined as the ratio of the amount of adjacent metal melted to the total amount of fused metal. As such, the amount of dilution can be calculated by using two geometric results, PA and RA, as expressed in Eq. (4). In Fig. 7, the maximum effective contribution for welding parameters on DIL is 39.1% for HIL. This finding is correlated with the polarity of the electrode and weld penetration depth. The use of positive polarity can increase penetration depth and area since additional base metal is melted [25, 26]. In this study, the polarity of the HIL is selected as DCEP in the welding process. Therefore, a higher HIL increases the penetration depth, leading to more DIL. Finally, it is common to minimize the amount of dilution since the amount of dilution affects the composition of the molten pool and the resultant mechanical properties in the welds [7, 27].

Figure 8 shows that the CWFS has the most dominant effect on the CGHAZ and WM micro-hardness profiles since cold wire addition alters the local thermal cycle by consuming heat from the molten pool. In addition, the heat input and voltage may contribute to the CGHAZ hardness profile since they can change the local thermal cycle and size of the CGHAZ area. It is difficult to compare the effects of CWFS on hardness when other parameters are varied as well. Therefore, nonlinear relationship analysis between interactions of welding parameters and micro-hardness profiles in the CGHAZ and WM is necessary and is discussed in Subsect. 4.2.

Fig. 8
figure 8

Effective contribution of CWTSAW process parameters for the WM and CGHAZ micro-hardness

Full size image

In Fig. 9, HIL is less effective for reinforcement regions (HRA, RA, and BTA) since the positive polarity of HIL is more dominant for the penetration profiles. However, BD showed the greatest effective contribution to HRA, BTA, CGHAZ area, and RA. In order to understand the effect of BD, the 16 measurements of RA, HRA, BA, and CGHAZ area were plotted against the two bevel specifications separately, as shown in Fig. 10. The BD with a wider bevel area had lower RA and HRA, and increased BTA and CGHAZ area. This means that smaller and shallower reinforcement regions were produced for the larger bevel angle. It is generally accepted that a larger bevel angle and depth lead to a larger bevel area promoting more molten metal flowing downward. The heat transformation from the top to the bottom of the weld resulted in full penetration. Therefore, this achievement of full penetration caused a smaller RA in the wider BD. These findings are consistent with the results in the simulation of Chen et al. [28] and the tungsten inert gas welding results of Huang et al. [29].

Fig. 9
figure 9

Effective contribution of CWTSAW process parameters for HRA, RA, CGHAZ area, and BTA

Full size image
Fig. 10
figure 10

Measurements of RA, BTA, HRA, and CGHAZ area for the two bevel specifications

Full size image

Of particular note is the formation of a larger CGHAZ area in the deep BD than in the shallow BD. The reason for this phenomenon is that the arc and molten metal contact a larger area when the bevel angle is larger so that the solid metal phase easily conducts more heat to promote the formation of a coarse-grained structure. These statements are consistent with the results in Chen et al. [28]. In addition, results from a previous study of intermediate gauge X70 steel [7] showed that TS was dominant in the CGHAZ area. In the current study, BD provided a more effective contribution to the CGHAZ area and not TS.

The BD had a smaller effect on the weld geometry ratio values (SPR and AR) and DIL than HI, V, and TS, as shown in Fig. 7. The values of SPR and AR were calculated from the BW. The smaller effect of BD on the BW can be explained in terms of the main liquid metal flow pattern and surface tension in the molten pool. The main flow pattern of liquid metal in the molten pool is liquid metal flowing upward along the boundary of the molten pool and colliding at the top of the weld, whereupon the liquid metal changes direction and descends into the pool [28].

The main flow pattern of liquid metal and surface tension are governed by the BW and they are affected by active elements, such as O, S, Si, and Ni, dissolved in the liquid mixture metal [28, 30]. The BW is controlled by the surface tension force since it pulls liquid metal toward the center of the weld pool, which is varied by the concentration of the active elements. Another study reported that bead width depends on the concentration of surface-active elements and the local temperature profile [31]. In this study, the composition of the base metal X70 steel, the electrodes, and the granular flux is uniform and identical for all 16 fabricated weldments. This means that the main flow pattern, surface tension, and BW were not altered even for different bevel angles and bevel depths. Therefore, the BD had little influence on the weld geometry ratio.

4.2 Nonlinear relationship of CWTSAW

Three-order multiple regression (TOMR) was used to analyze the nonlinear relationship between controllable variables (welding parameters) and the response factors (geometry results and micro-hardness profiles). In addition, the interactions of welding parameters and micro-hardness profiles in the CGHAZ and WM are considered in the nonlinear regression analysis. The empirical equations were developed using Minitab 18 with the form of Eq. (6) [7]:

$$\begin{aligned}y&= {C}_{0}+{\sum}_{i=1}^{7}\left({C}_{i}\cdot {x}_{i}\right)+{\sum}_{i=1}^{7}\left({C}_{ii}\cdot {x}_{i}^{2}\right)+{\sum}_{i=1}^{7}{\sum}_{i>j}^{7}\left({C}_{ij}\cdot {x}_{i}\cdot {x}_{j}\right)\\&+{\sum}_{i=1}^{7}\left({C}_{iii}\cdot {x}_{i}^{3}\right)+{\sum}_{i=1}^{7}{\sum}_{j>i}^{7}{\sum}_{k>j}^{7}\left({C}_{ijk}\cdot {x}_{i}\cdot {x}_{j}\cdot {x}_{k}\right)\end{aligned}$$
(6)

where y is the response factor (geometry characteristics and micro-hardness profile) which was predicted by the controllable variable \({x}_{i}\)(welding parameters and different interactive combinations); \({C}_{i}\), \({C}_{ii}\), \({C}_{ij}\), \({C}_{iii},\) and \({C}_{ijk}\) are the coefficients. In this study, the confidence level of TOMR in Minitab 18 was set at 90% which means that any controllable variables with a P-value less than or equal to 0.1 are statistically significant and considered predictors in empirical equations. For example, 8 predictors in the CGHAZ micro-hardness TOMR equation from a total of 72 possible predictors (individual welding parameters and interactions) were considered, which resulted in a good fit.

$$\begin{aligned}\mathrm{CGHAZ\;micro}-\mathrm{hardness } & = 112 + 1.86\cdot \mathrm{CWFS }+ 298\cdot \mathrm{HIL }- 13.31\cdot \mathrm{VL } \\ & + 1.44\cdot \mathrm{TS }-0.0327\cdot {\mathrm{CWFS}}^{2} - 5.89\cdot \mathrm{HIL}\cdot \mathrm{TS } \\ & + 0.264\cdot \mathrm{VL}\cdot \mathrm{TS }+ 0.000169\cdot {\mathrm{CWFS}}^{3}\end{aligned}$$
(7)

Of particular note is that squared and cubed predictors are only associated with CWFS in the CGHAZ micro-hardness equation. This appears to be the dominant effect contributed by cold wire addition in comparison with other TOMR equations. The other TOMR equations for HRA, RA, CGHAZ area, SPR, and WM micro-hardness are shown in Appendix.

The calculated values for HRA, RA, CGHAZ area, micro-hardness of the CGHAZ, and WM are plotted against the observed values in Fig. 11. To validate each equation, three (3) complementary tests were conducted and are included in Fig. 11 (triangles). The welding table for these complementary tests is shown in Table 4. TOMR equations have been developed to correlate the weld geometry results and CGHAZ and WM micro-hardness with weld parameters for heavy gauge X70 steels in comparison with the previous study of intermediate gauge X70 steels [7]. In addition, the range of R2 values is from 81.6 to 97.9% and the geometric characteristics of the three complementary tests with the varied welding parameters levels show a good correlation with the observed values (Fig. 11).

Fig. 11
figure 11

Observed and calculated values for the a HRA, b RA, c CGHAZ area, d micro-hardness of the CGHAZ, and e micro-hardness of the WM. Complementary tests results are also shown (triangles)

Full size image

4.3 Optimized levels of CWTSAW

The S/N ratio was utilized to determine the optimized levels for each welding parameter. The welding parameters were categorized by two quality requirements, which are “lower-the-better” and “higher-the-better”, respectively. AR, BTA, APR, and BW are included in the “higher-the-better” quality requirements. DIL, CGHAZ area, HRA, and RA are included in the “lower-the-better” quality requirements. The S/N ratio analysis was not conducted on the micro-hardness profiles since there was no agreement on which approach is better. The S/N ratio was calculated using Eqs. (8) and (9) [18].

$${\mathrm{S}/\mathrm{N}}_{ \left(\mathrm{lower}-\mathrm{the}-\mathrm{bettter}\right)}= -10{\mathrm{log}}_{10}\left(\frac{1}{n}{\sum}_{i=1}^{n}{y}_{ij}^{2}\right)$$
(8)
$${\mathrm{S}/\mathrm{N}}_{ \left(\mathrm{higher}-\mathrm{the}-\mathrm{bettter}\right)}= -10{\mathrm{log}}_{10}(\frac{1}{n}{\sum}_{i=1}^{n}\frac{1}{{y}_{ij}^{2}})$$
(9)

where S/N is the signal-to-noise ratio, \({y}_{ij}\) is the experimental value of the \(i\) th response characteristic in the \(j\) th test, and \(n\) is the number of tests.

A higher average S/N ratio value representing a given level of the weld parameter resulted in an optimal effect on the geometric characteristics since higher S/N values mean lower noise effects [17, 32]. A weld parameter level with a higher S/N value is considered the optimized parameter level, which results in an optimal effect on the geometric characteristics. The calculated S/N ratio values for the weld characteristics are shown in Appendix. Based on the calculated S/N ratio values, the optimized levels for the CWTSAW parameters are summarized (Table 7). Overall, the optimal geometric characteristics are achieved using optimized levels of CWTSAW parameters; i.e., 1.6 kJ/mm for HIL (level 1), 1.3 kJ/mm for HIT (level 1), 21.2 mm/s for TS (level 1), and 25 mm2 for BD (level 2).

Table 7 Optimized levels of CWTSAW parameters
Full size table

4.4 Comparison of TSAW and CWTSAW

Two heavy gauges X70 welds were produced by conventional TSAW and CWTSAW processes. Then, a comparison in terms of average micro-hardness and the phase fraction of MA constituents in the CGHAZ of TSAW and CWTSAW weld was undertaken. The weld testing conditions are shown in Table 8. The HIL, HIT, VL, VT, TS, and BD for both the TSAW and CWTSAW weld are identical, and only the CWFS is varied.

Table 8 Welding condition of TSAW and CWTSAW
Full size table

The approach to micro-hardness measurements followed the schematic description in Fig. 4. Figure 12 shows the average micro-hardness values measured in the CGHAZ of the TSAW and CWTSAW welds. The hardness is higher for the CGHAZ of the TSAW weld than for the CWTSAW weld. The lower hardness distribution in the CGHAZ of the CWTSAW weld is closely related to the microstructure modification due to heat reduction by heat consumption of the cold wire addition.

Fig. 12
figure 12

Average micro-hardness values in the CGHAZ for TSAW and CWTSAW

Full size image

Optical and SEM SE micrographs of MA constituents in the CGHAZ of TSAW and CWTSAW samples are shown in Fig. 13. The MA constituents appear as shiny white features in the optical and SEM images after etching with the modified LePera’s etchant [21]. White linear segments are visible in the SEM images since some of the MA constituents were formed at the grain boundaries. The MA fractions from the optical micrographs (Fig. 13a, c) are 5.3% (0.2%) and 2.8% (0.1%) for the TSAW and CWTSAW welds, respectively. The MA fractions in the CGHAZ determined from the SEM micrographs (Figs. 13b, d) are 5.7% (0.2%) and 3.3% (0.2%) for TSAW and CWTSAW samples, respectively. The values in the brackets represent one standard deviation.

Fig. 13
figure 13

Weld samples fabricated by TSAW (a, b) and CWTSAW (c, d) showing MA constituents (shiny white features) in the CGHAZ. Images a and c are optical micrographs, while images b and d are SEM SE micrographs

Full size image

In terms of the morphology of MA constituents, the MA regions in the TSAW sample from both optical and SEM SE micrographs are mainly massive, and the MA constituents of the TSAW sample are more elongated and larger than those in the CWTSAW sample. The MA features in the CWTSAW sample are finer and more dispersed (Fig. 13c, d) than those in the TSAW sample (Fig. 13a, b). The average micro-hardness in the CGHAZ of TSAW samples is higher than that in the CWTSAW sample, which can be correlated with the higher MA constituents fraction and the different MA morphology (blocky and elongated) in the TSAW sample.

Denser MA regions with elongated MA constituents that formed in the CGHAZ resulted in localized brittle zones (LBZs). Luo et al. [33] and Mohammadijoo et al. [10, 34] reported that the formation of LBZs can cause initiation and propagation of cleavage fracture at the PAG boundaries in the HAZ, deteriorating the fracture toughness. There are elongated MA constituents and a higher overall MA fraction in the CGHAZ of the TSAW samples than in the CWTSAW samples. This can be interpreted as the reason for the higher hardness in the CGHAZ of the TSAW samples. In addition, fine and more dispersed MA constituents in the CGHAZ of the CWTSAW samples inhibit the formation of LBZs. This change in MA morphology is strongly related to the cold wire addition, reducing the overall heat input by consuming heat from the molten pool.

5 Conclusions

To meet the essential upgrade of X70 pipeline steel in heavy gauge pipes, the effect of cold wire tandem submerged arc welding (CWTSAW) parameters and bevel design on the change in weld geometry and micro-hardness of heavy gauge X70 (19.1 mm) was investigated, and the input levels of welding parameters were optimized. The following general conclusions can be made:

  1. 1.

    Cold wire feed speed had the most dominant effect on micro-hardness profiles since the cold wire addition altered the local thermal cycle by consuming heat from the molten pool. Cold wire addition can restrain the increase in overall heat input of conventional tandem submerged arc welding in heavy gauge pipe making;

  2. 2.

    Geometry changes, including the reinforcement size and CGHAZ area, are sensitive to the bevel design due to the achievement of full penetration, but the bevel design only had a minor effect on the weld metal and CGHAZ micro-hardness;

  3. 3.

    Predictive equations for micro-hardness values of the CWTSAW sample were developed using three-order multiple regression analysis and showed a good correlation in contrast to previous work;

  4. 4.

    A CWTSAW process, with an overall nominal heat input of 2.9 kJ/mm (lead + trail electrode), a travel speed of 21.2 mm/s, a bevel angle of 90°, and a bevel depth of 5 mm, reduced the CGHAZ area, dilution, and reinforcement size;

  5. 5.

    The CWTSAW samples had lower martensite-austenite (MA) fractions with fine and dispersed MA constituents, resulting in lower micro-hardness values in the CGHAZ. This type of MA morphology, which is due to the lower actual heat input introduced to the weld pool, inhibits the formation of the local brittle zone in the CGHAZ of the CWTSAW sample.