Robotic welding study finds optimal settings for medium-carbon steel joints

Shielded metal arc welding, or SMAW, remains one of the most widely used welding processes in industry because of its versatility, portability, low equipment cost and independence from external shielding gas, particularly in maintenance, repair, construction and field applications. Coated electrodes account for roughly 29 percent of the global welding consumables market. Yet the quality of the joints it produces depends strongly on the right choice of welding parameters, which have traditionally been selected on the basis of empirical knowledge and welder experience.

The experiments focused on double-V groove butt joints made from 6-millimeter-thick AISI/SAE 1045 steel plates, a medium-carbon steel containing about 0.45 percent carbon that is prized for its combination of strength, toughness, wear resistance and machinability. The filler material was rutile E6013 covered electrodes of 3.2 millimeters in diameter, welded with direct current electrode positive polarity. A single welding pass was performed per side of the plates, with a constant electrode travel angle of 20 degrees relative to the vertical. To eliminate the human factor inherent to the process and ensure high reproducibility, the joints were fabricated using a specially adapted two-degree-of-freedom XZ Cartesian robot rather than a human welder.

The experimental framework was defined by a three-level Box-Behnken design with three input factors: bevel angle, welding current and welding speed. The design comprised 17 experimental runs, including five center-point replicates, with bevel angles spanning 0 to 45 degrees, currents from 105 to 140 amperes and speeds from 4.16 to 5.97 millimeters per second. Four additional randomly selected conditions, not used in model fitting, were reserved to independently test the predictive performance of the models. In total, 21 welded joints were fabricated, and each experimental condition was produced in triplicate, with reported values corresponding to the average of three specimens.

Before any destructive testing, each joint was screened with non-destructive techniques. Liquid penetrant inspection revealed no surface defects, including cracks or open porosity, in any of the welded joints. Ultrasonic inspection did indicate lack of fusion in joints produced with a 0-degree bevel angle, consistent with the limited penetration associated with that configuration and the 6-millimeter plate thickness. Those joints were deliberately retained because the 0-degree bevel was a predefined level of the experimental design needed to evaluate its effect on the response variables. No other critical internal discontinuities were detected in joints produced with bevel angles of 22.5 and 45 degrees.

The welded joints were characterized in terms of 11 response variables grouped into four categories: mechanical behavior, including ultimate force, elastic force, impact toughness and hardness; heat-affected zone metrics covering width, depth and area; surface roughness variables, namely the arithmetic mean surface height and the primary orientation angle of the surface texture; and macro-geometrical weld descriptors, including weld area and the weld toe stress concentration factor. Surface topography was measured with a confocal microscope and processed through a filtering procedure that separated roughness from waviness. The researchers note that the orientation of surface texture matters because when it lies roughly perpendicular to the loading direction its detrimental effect is limited, whereas orientations approaching parallel alignment produce significantly higher stress concentrations and reduced fatigue resistance.

Temperature monitoring added an important metallurgical safeguard. Twelve mineral-insulated thermocouples were connected to data loggers, with the four sensors closest to the weld centreline positioned about 6 millimeters from the weld axis. The specimen with the highest current-to-speed ratio, and therefore the highest estimated heat input per unit length, reached a maximum recorded temperature of only about 280 degrees Celsius, well below the austenite start temperature of 758 degrees Celsius reported for a medium-carbon steel. That means austenite formation followed by martensitic transformation could be ruled out at the monitored locations, and the observed hardness variations were instead attributed to differences in local thermal exposure and cooling conditions. Fractographic examination of the specimens with the lowest impact toughness and highest hardness revealed predominantly ductile fracture surfaces with regularly distributed dimples, alongside cleavage facets in localized regions indicating a mixed-mode fracture mechanism.

Second-order polynomial response surface models were fitted to all 11 variables and evaluated with mean absolute percentage error and normalized root mean square error on both training and testing data. Most responses were predicted accurately, with testing errors generally low, but surface roughness proved clearly the most challenging response to model, reaching a testing mean absolute percentage error of 46.33 percent and a testing normalized root mean square error of 49.36 percent. The authors caution that although that model remains statistically valid, its quantitative predictions should be interpreted with greater caution than those of the remaining responses. The ultimate force model, by contrast, achieved a coefficient of determination of 0.9784 with a non-significant lack of fit.

Five optimization scenarios were then defined using the desirability function approach, each targeting a different manufacturing priority. The first sought joints with the highest resistance under static and dynamic loading while minimizing heat-affected zone area, hardness, surface roughness and the weld toe stress concentration factor. The second aimed to minimize electrode filler consumption by minimizing weld area, the third to minimize manufacturing time through higher welding speed, the fourth to minimize energy consumption through a low current combined with a high speed, and the fifth to obtain the lowest possible surface roughness with the highest texture orientation angle on the weld bead. All five scenarios converged within a narrow process window: bevel angles between 29.30 and 33.29 degrees, currents between 105 and 107.89 amperes, and speeds spanning 4.16 to 5.97 millimeters per second. The researchers emphasize that the optimized values represent balanced compromises among competing objectives rather than the individual optimum of each response, which is why, for example, the optimized ultimate force in the first scenario, 90.47 kilonewtons, fell below the experimental maximum of 111.18 kilonewtons.

To validate the approach, five new joints were fabricated using the optimized parameter values and characterized under the same conditions as the rest of the samples. All variables except surface roughness exhibited low relative deviations, generally below 7 percent, confirming the robustness of the optimization strategy. From an industrial perspective, the authors identify the fourth scenario, which achieved the highest overall desirability of 0.74 alongside a reduced heat-affected zone area and lower manufacturing time and energy consumption, as the most balanced solution for general applications. They also stress the limits of the findings: the optimum parameter combinations are specific to double-V groove SMAW joints made from 6-millimeter AISI/SAE 1045 steel plates using E6013 electrodes, and should not be directly generalized to other thicknesses, electrode classifications, base materials or welding configurations without further experimental validation.

Subject of Research: Multi-objective optimization of shielded metal arc welding parameters for double-V groove joints in medium-carbon steel

Article Title: Comprehensive analysis and multi-objective optimization of shielded metal arc welding process in double-V groove welds of medium-carbon steel

Article References: Estremera Carrera, V., Lostado Lorza, R., Berlanga Labari, C., & Fernandez Martinez, R. (2026). Comprehensive analysis and multi-objective optimization of shielded metal arc welding process in double-V groove welds of medium-carbon steel. Results in Engineering, 32, Article 113686. https://doi.org/10.1016/j.rineng.2026.113686

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113686

Keywords: shielded metal arc welding, SMAW, medium-carbon steel, AISI 1045, double-V groove welds, Box-Behnken design, response surface methodology, desirability function, multi-objective optimization, heat-affected zone, surface roughness, welding automation