**Authors**

Carpenedo, Marcelo^{a}. *mcarpenedo@gmail.com*

Gonzalez, Arnaldo R^{b}. *ruben@mecanica.ufrgs.br*

Reguly, Afonso^{a}. *reguly@ufrgs.br*

^{a }Department of Materials Engineering, Federal University of Rio Grande do Sul. Av. Bento Gonçalves, 9500 – Sector 4 – Building 74 – Room 211, 91501-970, RS, Brazil.

^{b }Department of Mechanical Engineering, Federal University of Rio Grande do Sul. Rua Sarmento Leite, 425 – Room 202 – 2º floor, 90046-902, RS, Brazil.

Corresponding author: Tel.: +55 51-982178461; *E-mail address: mcarpenedo@gmail.com*

**Abstract**

Semi-automatic welding process is widely used in industry, but has limitations for the correct positioning of weld beads regarding length and spacing when applied intermittently. This paper provides a statistical research that reports the existence of important dimensional variations in the final product of this welding process, analyzing the main factors that have influence over these variances and the lack of resources to mitigate their undesirable effects on the welded joint. To mitigate this gap on manufacturing process is proposed the application of the welding markings, which are physical reliefs applied on the welding edge of the joint aimed to orient the welder to apply the weld beads in the correct position, precisely. The experimental results showed a solution that provides significant gains to the welding process, presenting a welding technique that mitigates most of the process variables with influence over the dimensional precision of the weld beads.

**Keywords**

Semi-automatic weld, Weld length, Dimensional precision, Process improvement

**Abbreviations**

L_{p} – Specified weld bead length

L_{pm} – Verified weld bead length applied with joint markings usage

L_{psm} – Verified weld bead length applied without joint markings usage

L_{e} – specified weld bead spacing

MAG – Metal Active Gas

**1-Introduction**

The semi-automatic welding process, despite widely used in the industry, has an intrinsic characteristic which is the movement control of the welding gun strictly dependent by the operator’s skills. This condition generates restrictions regarding the achievement of desired dimensions of the weld beads, which has relation to their length and spacing.

Currently there is a lack of continued studies about the resulting structural effects of the dimensional precision of the weld beads for the semi-automatic welding process applied intermittently, having no references about the state of art and solutions to improve this manufacturing condition. Welding standards provide some degrees of variation about weld beads length and spacing, but for some applications these range of tolerance can bring some structural negative effects for the joint.

This scenario shows the need for improvements in this welding process, conducing to the development and evaluation of new techniques to apply intermittent weld beads under semi-automatic handling. The weld markings, which are physical reliefs applied on the weld edge of the joints, are the solution proposed to orientate the welder about the correct weld bead positioning, eliminating variances from the process that causes imprecise weld start/end placing.

This study has been based in a statistical investigation about the improvements obtained by using the weld markings compared to the state of art through ANOVA assessment, applying experiment project factorial 3^{2} with three replications. The study was carried out evaluating the welding markings effects on the weld bead length and positioning, comparing the same analysis applied to the state of art of semi-automatic welding technique, both using conventional MAG – Metal Active Gas – and intermittent weld beads.

ANOVA methodology was carried out to statistically determine the significance of each factor evaluated, weld bead length – L_{p} and weld bead spacing – L_{e}, by *p* value analysis. Once welds applied with markings and by state of art have no relation between their results, inference statistic method was carried out to compare the dimensional accuracy of the welds applied by both process regarding its dimensional variations.

**2-Revision**

Along welding process evolution regarding consumables and equipment, especially considering the guidance to the process automation and robotization [3-4], the semi-automatic MAG welding process continues to be widely used in the industry [5], mainly when applied for intermittent weld beads. However, one characteristic remained unchanged since the beginning of the application of this welding process in the industry, which is the weld edge shape.

Weld quality regarding dimensional precision, when applied by semi-automatic process, is a direct result of the welder’s skills [6], once the weld bead positioning is defined by superficial references, tooling or just by the welder ability and experience to handling the weld gun. These characteristics have direct relation to the joint quality, reported by [7] and [8] as an aspect with increasing industrial concerning in the current economic scenario.

Complementing the welding quality concerning, [9] and [10] carried out studies to verify the influence of the weld bead positioning on the structural rigidity and mechanical resistance of the component, which demonstrated direct influence on these characteristics on the mechanical resistance of the welded joint.

Following the recommendations of welding standards [11] and [12] the weld sizes and lengths should not be lower than that specified in project, reinforcing the need to achieve the correct weld length and, consequently, the mechanical requirements designed for the joint. These criteria conduces designers and engineers to search for solutions, on shop floor or design departments, to achieve the standardized recommendations for a safe, but more related to a standardized, product delivering.

The weld marking is a solution developed to fix the issues about weld length and positioning accuracy, designed to be applied to intermittent welds in sheet metal parts, thin or thick, cut by laser process. It is a relief added on the welding edge of the parts during the component cutting process, which serves as a guidance for the welder about the correct places to start and end the weld bead.

This paper provides data about a welding solution to prepare and use sheet metal parts to be welded in order to improve the dimensional accuracy of the weld bead length and positioning, applying physical markings directly on the design phase, enabling the welder to follow visible features to ensure the weld bead positioning during welding process.

In this work the three-level factorial experimental project was carried out to investigate the influence of the factors L_{p} and L_{e} on the weld bead body produced by GMAW, followed by RSM analysis, applied to predict the weld bead length with respect to the tested factors (L_{p} and L_{e}), followed by a statistical inference analysis, aiming to verify the dimensional variation among both welding techniques.

**3-Experimental Procedure**

The current research considers the analysis of two factors regarding the weld beads, L_{p} (weld bead length) and L_{e} (weld bead spacing) and three dimensional levels for each factor, evaluating the interaction between them and the effects of each combination on the responses for the weld bead positioning accuracy based on its technical specifications. Both factor’s evaluations were performed using two welding techniques for semi-automatic MAG process:

1 – The proposed welding technique, where the positioning of the intermittent welds are defined by the presence of markings, orienting the welder about the exact places to start and end the weld beads, as show in the figure 1.a;

2 – The current welding technique used on the industry, applying intermittent welds direct on the welding edge of the sample, as shown in the figure 1.b.

The analysis has the aim of compares the results to quantify statistically how the joint markings can improve the current welding process regarding its dimensional precision.

Figure 1: Weld edges characteristics for both welding techniques: a) proposed, using the joint markings and b) current, with flat weld edge.

Figure 1.a shows the welding edge with the presence of the markings, conferring references for the welding gun positioning about weld start and end positions. Figure 1.b shows the weld edge in the current shape, with a flat and simple edge.

For the statistical analysis was performed a factorial design at three-levels with three replications 3^{2}, submitting the results to analysis of variance – ANOVA to analyse the relevance of each factor assessed and concluding by statistical inference method, observing how the factors are influenced by itself, without influence of neighbouring. The factor’s levels selected for the experimental procedure were defined based on common values for intermittent weld bead dimensions regarding length and spacing applied in the industry, applied in plates with thickness of 3 mm, defining the weld bead lengths of 30, 40 and 50 mm and spacing of 30, 50 and 70 mm, as show in Table 1.

Table 1. Coded and natural levels of the design factors and levels.

Table 2 shows the factors and levels combination and their respective responses, where L_{pm} is the response for weld bead length with pre-markings and L_{psm} is the response for weld bead length for the current technique. Both L_{pm} and L_{psm} responses are results from L_{p} and L_{e} combination, according each run step.

Table 2: Summary of the 3^{k} factorial design combinations and the experimental results of L_{pm} and L_{psm}.

The samples were welded using cold rolled sheet metal plates based on standard ASTM A1011 Grade 50 (American Society for Testing and Materials, 2008), obtained as strips with width, length and thickness by 25,0 mm, 400,0 mm and 3,17 mm, respectively. Material chemical composition and mechanical properties are shown in tables 3 and 4, respectively, as specified in the standard.

Table 3: Base metal chemical composition for the steel Grade 50. Values in weight %. Source: ASTM A1011 (2008).

Table 4: Base metal mechanical properties the steel Grade 50. Source: ASTM A1011 (2008).

The samples were cut by a laser cutting machine brand Cincinnati, model n° CL-7A and welded by semi-automatic method using an ESAB source of power, model LAI 400 with nominal power by 14,6 KVA.

Figure 2 shows the geometry of the samples prior welding, where Figure 2.a shows the basic strip intended for the current welding technique, with flat welding edge. The strip 2.a double combined results in the final sample for the current technique evaluation, without markings. The strip 2.b in combination with 2.a result in the final sample for the proposed technique evaluation, with the joint markings. The strip 2.c shows a strip with markings resulted from the cutting process.

Figure 2: Basic samples resulted by the laser cutting process.

Welding parameters applied on the samples welding are shown in Table 5.

Table 5: Welding parameters used for the samples welding.

The final welded samples resulted with width and length by 50,0 mm and 400,0 mm, respectively, as shown in the figure 3, where can be seen a sample for the weld done using the markings.

Figure 3: a) Example of sample specification for welding process; b) Example of final welded sample.

In figure 3.a is shown the design specification and dimensions for one of the samples assessed, while in figure 3.b is shown a sample welded, both for the proposed welding technique, with joint markings.

The weld beads were measured using callipers with resolution of 0,05 mm and the results are shown in table 2. The measurements considered the length between the two visible extremities of the weld beads, as shows the figure 4.

Figure 4: Weld bead length measured.

**4-Results and Discussions**

From results shown in Table 2, applying ANOVA, were obtained the following data about the current and proposal welding techniques concerning weld bead length and spacing accuracy.

**4.1-New Welding Technique Proposed.**

An ANOVA was carried out to quantify the relative importance of each factor (L_{p} and L_{e}) and also its interaction on L_{pm} results. The contribution percentage was calculated by the ratio of the sum of squares (SS) of the selected factors and total sum of squares. Table 6 shows the results for ANOVA performed with the acquired data considering an interval of confidence by 95%. The factors are physically significant when its percentage of contribution is greater than the error associated, which is by 5% I this assessment.

The results showed in this table are in agreement with the results obtained in figure 7 in which L_{p} (mm) is the most important factor affecting L_{pm} (84,1%). L_{e} (2,5%) itself has no significant influence on the L_{pm} response for the selected range of factors, followed by L_{p} x L_{e} interaction (1,2%). The total contribution rate of L_{p} and L_{e} factors over L_{pm} was 87,8%.

Table **6**. ANOVA of L_{pm} values.

The significance of L_{p} indicates the effectiveness of welding markings about weld beads dimensional control, where both length and spacing are not affected one by other. The only visible influence on weld beads length variation occurs regarding its own length L_{p}, which is defined by design.

Figure 5 shows L_{pm} results according to L_{p} specification, showing the accuracy level of the proposed welding technique regarding L_{p}.

Figure 5. Main factor L_{p} effects over L_{pm}.

On Figure 5 is observed that in all three levels (L_{p}) evaluated L_{pm} remained slightly above L_{p} specifications, with the responses keeping the variances increasing inversely proportional to the levels increasing. This means that as greater is L_{p} greater is the weld bead dimensional accuracy.

Figure 6 shows the average variation of L_{pm} as function of L_{e} values.

Figure 6. Main effects plot for L_{pm} x L_{e}.

On Figure 6 can be observed that L_{pm} tends to keep its dimensionally constancy as L_{e} increases, remaining without influence from external factors. A subtle characteristic can be observed due a smooth L_{pm} decreasing with L_{e} increasing, following the trend in figure 5 and reinforcing that no external factors affects the weld bead variation in length when applied using markings.

Figure 7 shows Lpm variations according its specification L_{p} in combination to L_{e}.

Figure 7. Interaction factors (L_{p} x L_{e}) plot for L_{pm}.

Figure 7 shows that L_{pm} follow a dimensional variation proportionally constant to L_{p} increasing, independently of L_{e} specification, with L_{pm} remaining lightly above L_{p} except for a specific point where L_{p} = 50 mm combined with L_{e} = 70 mm resulted in L_{pm} beneath L_{p}, representing 33,3% of sampling for L_{p} = 50 mm (runs 9, 16 and 17 from table 2).

Figures 5, 6 and 7 showed that the responses for each L_{pm} level evaluated follow a constant dimensional variation according to L_{p} specification, not affected by L_{e} and remaining above L_{p} specified values predominantly in most of cases.

**4.2-Current Welding Technique.**

An ANOVA was also carried out to quantify the relative importance of each factor (L_{p} and L_{e}) and also its interaction on L_{psm} results for the current wending process, without markings.

The results showed in this table are in agreement with the results obtained in figure 10 in which L_{p} (mm) is the most important factor affecting L_{pm} (86,3%). L_{p} x L_{e} interaction has an relevant significance (4,4%) and L_{e} (1,3%) itself has no significant influence on the response L_{pm} for the selected range of factors.

Table 7 shows the analysis of variance (ANOVA) for L_{p} data by the current welding techqnique, without markings usage.

Table **7**. Analysis of variance for L_{psm} data.

Can be observed that, for the welds done without markings (current welding process), L_{p} factor is also significant for L_{psm} control. However L_{p} x L_{e}, despite has no statistical significance for the responses evaluated, is considerably near to the significance border control. This condition takes to the consideration that its statistical no significance can be revised, considering it for this situation as significant due the results obtained for the comparisons L_{psm} x L_{e} and L_{psm} x L_{p} x L_{e} shown on Figures 8 and 9, respectively, where L_{p} presents a tendency to influences L_{psm} dimensions as function of L_{e}.

The significance of L_{p} x L_{e}, besides L_{p}, indicates the influence of external factors related to the dimensional variations not only from L_{p}, but also L_{e}. In this case, as there is no physical limitations to define the weld bead start and end, its positioning is defined regarding nearby geometries as weld beads or other joint components, besides the welder skills.

Figure 8 represents L_{psm} positioning according to L_{p} specified, showing the lack of dimensional precision of current welding technique regarding L_{p} values specified. These results reflect the current scenario of the semi-automatic welding process applied on intermittent welds.

Figure 8. Main factor L_{p} effects plot for L_{psm}.

On Figure 8 can be observed that all weld beads – L_{psm} – resulted in lengths below that specified for L_{p}, showing that the current welding technique has a consistent tendency to not achieve project specifications for the weld bead length. Only in one case (run 22 from table 2) L_{psm} remained above Lp specification, representing 11,1% of cases for L_{p} = 30 mm. Is observed also the responses keeping the variance increasing directly proportional to the levels increasing (inverse of observed in figure 5), what means that as greater is L_{p} greater is the weld bead dimensional variation.

Figure 9 shows the average L_{psm} values as function of L_{e} values.

Figure 9. Main effects plot for L_{psm} x L_{e}.

Can be observed on Figure 9 that L_{psm} average values increase as L_{e} increases, showing a direct relation between the weld bead length and its neighbor elements. Compared to the figure 6 L_{psm} averages are consistently lower than L_{pm}, showing an evidence of the dimensional accuracy discrepancy between the two welding techniques.

Figure 10 shows L_{psm} variation according its reference L_{p} in combination to L_{e}.

Figure 10. Interaction plot for L_{psm} x L_{p} and L_{e}.

Figure 10 shows that, for L_{e} values by 50 mm and 70 mm, L_{psm} follows a proportional variation regarding L_{p}, however for L_{e} value of 30 mm is observed a dispersion on L_{psm} values, what indicates the influence of external factors on L_{psm} length.

The most important observation in Figure 10 is that 100% of L_{psm} cases resulted below L_{p} specifications, meaning 100% of welds out of specification.

The influences of external factors are connected to L_{psm} regarding welder’s perception during weld application, once the presence of geometrical characteristics near the next weld bead to be applied has an important influence on the welder decisions regarding start/end weld bead points.

**4.3-Statistical Inference**

Despite the promissory results with the joint markings usage is important take in count that the significance of the results are dependent of the welding methods evaluated, having no relation between them. To analyse this relation and compare its influence above the efficiency of the proposed welding technique the data results was assessed under inference statistic approach, looking for the effective difference between the two methods evaluated. The analysed numbers are shown in Table 8, derived from table 2.

Table 8: Data for statistical inference analysis.

Is observed that the averages µ are significantly greater for L_{psm} regarding L_{pm}, what confirms the superior accuracy of the welding technique using the pre-marking.

Figure 11 shows the graphic analysis for both factor levels evaluated, comparing L_{pm} x L_{psm} for L_{p} = 30 mm, 40 mm and 50 mm on figures 11.a, 11.b and 11.c, respectively.

Figure **11**. L_{pm} and L_{psm} comparison for the three levels analysed.

Can be seen a predominant data pattern for both weld techniques assessed and no dependent of the level evaluated, where the weld beads applied with the pre-markings remained above the specified line in most cases, meaning the predominant achievement of the project specifications. For the three levels evaluated was observed that in a confidence level of 95% the averages from L_{pm} exceed L_{psm}.

This analysis corroborates the superiority of the weld beads applied with the joint markings usage, granting a higher dimensional precision to the component, which invariably means a quality increasing in process costs and mechanical properties of the component.

**5-Conclusion**

The welds applied without the joint markings resulted in 3,7% of the samples with L_{psm} values above L_{p}, which means that 96,3% of the samples not reached the design specifications. The opposite situation was observed for the welds applied with the markings, where 92,6% of the samples reached the design specifications.

The dimensional variation of the weld length, when evaluated without the markings usage, resulted in numbers significantly influenced by weld bead extension and its distance to the next joint elements. L_{p} and L_{e} specification affects directly L_{psm} values, which are dependent of the length, position and weld edge extension. When the dimensional variation is evaluated with the joint marking usage L_{e} don’t affect L_{pm} values, what makes L_{pm} variation independent of the weld bead distance with the next element.

The usage of the joint markings allows the welder to have references about physical limitations dedicated exclusively to orient him to the correct weld gun start and end positioning, ensuring that the weld start point be positioned lightly before the weld marking and the end point be lightly after the marking. These characteristics offer to the semiautomatic welding process a higher dimensional precision to the dimensional accuracy of the weld bead length and spacing when compared to the current process, resulted from the controlling methods currently available.

**Authors’ contributions**

MC: formal analysis, writing – original draft, methodology. ARG and AR: data curation, supervision, writing – review & editing.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

**6-References**

[1] KHEDMATI, M.; RASTANI, M.; GHAVAMI, K. Numerical study on the permissible gap of intermittent fillet welds of longitudinally stiffened plates under in plane axial compression. Journal of Constructional Steel Research, v.63, p. 1415-1428, 2007.

[2] RASTANI, M.; REZA, K. M.; GHAVAMI, K. A comparative study on three different construction methods of stiffened plates-strength behaviour and ductility characteristics. Rem: Revista Escola de Minas, v. 60, p. 365-379, 2007.

[3] THAM, G. et al. Predicting the GMAW 3F T-Fillet Geometry and Its Welding Parameter. Procedia Engineering, v. 41, p. 1794-1799, 2012.

[4] RAMOS, D.; LOPEZ,; J. I.; Ramos, D., Lopez, J.I., Perez, P., (2013). Effect of Process Parameters on Robotic GMAW Bead Area Estimation. Procedia Technology. 7. 398-405. 10.1016/j.protcy.2013.04.050.

[5] Aini, I.I., Mohamat, S., Amir, A., Ghalib, A., (2012). The Effect of Gas Metal Arc Welding (GMAW) Processes on Different Welding Parameters. Procedia Engineering. 41. 1502-1506. 10.1016/j.proeng.2012.07.342.

[6] Kim, I.s., Son, J.S., Park, C.E., Kim, I.J., Kim, H.H., (2005). An investigation into an intelligent system for predicting bead geometry in GMA welding process. Journal of Materials Processing Technology,Vol. 159, Issue 1, Pages 113-118. Journal of Materials Processing Technology – J MATER PROCESS TECHNOL. 159. 113-118. 10.1016/j.jmatprotec.2004.04.415.

[7] Metternich, J., Böllhoff, J., Seifermann, S., Beck, S., (2013). Volume and Mix Flexibility Evaluation of Lean Production Systems. Procedia CIRP. 9. 79-84. 10.1016/j.procir.2013.06.172.

[8] Stenberg, T., Barsoum, Z., Åstrand, E., Öberg, A.E., Schneider, C., Hedegard, J., (2017). Quality control and assurance in fabrication of welded structures subjected to fatigue loading. Welding in the World, Le Soudage Dans Le Monde. 61. 10.1007/s40194-017-0490-5.

[9] Khedmati, M., Rastani, M., Ghavami, K., (2009). Ultimate strength and ductility characteristics of intermittently welded stiffened plates. Journal of Constructional Steel Research. 65. 599-610. 10.1016/j.jcsr.2008.07.029.

[11] American Society for Testing and Materials. Standard Specification for Steel, Sheet and Strip, Hot-Rolled, Carbon, Structural, High-Strength Low-Alloy and High-Strength Low-Alloy with Improved Formability. ASTM A 1011/A 1011M – 02, 2008.

[10] Dimitrakis, D., Lawrence, F.V., (2001). Improving the fatigue performance of fillet weld terminations. Fatigue & Fracture of Engineering Materials & Structures. 24. 429 – 438. 10.1046/j.1460-2695.2001.00418.x.

[12] European Prestandard. Eurocode 3: Design of steel structures – Part 1.1: General rules and rules for buildings. ENV 1993-1-1, 1992.