Introduction

In recent years, the research field of nanotechnology has attracted considerable attention of scientists and engineers. Nanostructure materials often are characterized by a grain size or the particulate size of up to about 100 nm. These materials are suitable candidates for different applications including aerospace, electronics, and automotive and chemical industries because of the reduction of the grain size in the nanometer scale and large surface area of nanoparticles offer unique mechanical, elevated temperature, electrical, magnetic, and optical properties.[1,2]

Among the new technologies in the field of materials joining together, the issue is related to nanotechnology and welding. Recently, researchers investigating the possibility of using nanomaterials in welding processes at the nanoscale that a new chapter in the field of nanotechnology associated with welding is open.[3]

Fattahi et al.[4] have reported the improvement of impact toughness of the AWS E6010 weld metal, specifically when a medium TiO2 nanoparticles content was added to the electrode coating. Also, Pal et al.[5] reported that the addition of nanoparticles of TiO2 on coating of the coated AWS E11018M electrode was improved tensile and charpy impact property. Furthermore, Aghakhani et al.[6–9] have reported that the depth of penetration, heat affected zone width, and hardness of the melted zone were affected by the addition of TiO2 and Cr2O3 nanoparticles to the weld pool.

Submerged arc welding (SAW) is an arc welding process widely applications in various industries due to its inherent advantages including ease of control of process parameters, depth of penetration, high metal deposition rate, smooth bead, ability to weld thick plates, and prevention of atmospheric contamination of the weld pool.[10,11] In this process similar to any other arc welding process, the high quality weld is mainly influenced by the independent input parameter including welding current, welding speed, arc voltage, and nozzle-to-plate distance since they are closely related to the geometry of weld bead, a relationship which is thought to be complicated because of the nonlinear characteristics. In order to attain the desired weld bead quality, the input parameters ought to be selected in appropriate combination and the correct choice of these parameters require adequate details about the effects of various input parameters on weld bead characteristic[12].

Response surface methodology (RSM) is commonly used in different fields such as chemistry, engineering, physics, and so on[13–16]. Box and Wilson developed central composite rotatable design (CCRD) as a design of RSM.[17] Box and Behnken offered another design of RSM.[18] A lot of attempts happen to be modeled the relationship between the important welding input parameters and weldment characteristics in SAW process.

Yang et al.[19] reported the application of linear regression model for computing the weld bead geometrical features of the SAW process parameters. The results shown that the linear regression model was useful for computed the various geometrical features of the SAW process.

Murugan et al.[20–22] were developed mathematical models for SAW of pipes using RSM to predict and control of the weld bead parameters (penetration, reinforcement, width, and dilution) as affected by the welding input parameters (arc voltage, wire feed rate, welding speed, and nozzle-to-plate distance).

In the present work, first aim is to use RSM to relate the SAW welding input parameters current (A), arc voltage (B), nozzle-to-plate distance (C), welding speed (D), and MgO nanoparticles (E) to the prediction of the depth of penetration of St37 steel. The second aim is to find the optimal welding combination that would maximize the depth of penetration, keeping the cost relatively low.

Materials and Methods

In the present paper, the effect of welding input parameters on the depth of penetration based on CCRD having five-level-five-factor each was examined. The welding current (A), arc voltage (B), welding speed (D), nozzle-to-plate distance (C), and thickness of MgO nanoparticles (E) were considered as the input parameters, respectively. These parameters and their levels are shown in Table I. The values of parameters were coded using the formula below.

Table I Input Parameters and Their Levels
Full size table
$$ x_{i} = \frac{{X_{i} - \frac{{(X_{\text{high}} + X_{\text{low}} )}}{2}}}{{\frac{{X_{\text{high}} - X_{\text{low}} }}{2}}}, $$
(1)

where x i is the coded value of ith factor, X i is the real value (uncoded) of ith factor, and X high and X low are the high and low levels of ith factor in uncoded form, respectively.

The relationship between depended parameter (depth of penetration) and independent parameters (welding input parameters) were obtained using a quadratic polynomial model. This model is as follows:

$$ Y = b_{0} + \sum\limits_{i = 1}^{k} {b_{i} x_{i} + \sum\limits_{i = 1}^{k - 1} {\sum\limits_{j = 2}^{k} {b_{ij} x_{i} x_{j} } } } + \sum\limits_{i = 1}^{k} {b_{ii} x_{i}^{2} + e}, $$
(2)

where Y is the predicted response, b 0 is the constant term, b i is the coefficient of the main effect of ith factor, b ij is the coefficient of interaction effect between ith and jth factors, b ii is the squared effect of ith factor, and e is the error.

The indices of quality of the model are R 2, R 2 (pred), R 2 (adj), and p value of regression. These indices are between zero and one. First to third indices should be big enough and p value closed to zero. Analysis of variance (ANOVA) is used to determine significant terms and p value of the model. All calculations were done using Minitab 16 software.

MgO nanoparticles supplied by Neutrino Co, Iran, were used as the input welding parameters. Figure 1 shows the TEM image of MgO nanoparticles. It can be seen that the morphologies of MgO nanoparticles are spherical. Also, the average particle size and specific surface are about 40 nm and 50 m2/g, respectively. The characteristics of the nanoparticles are shown in Table II.

Fig. 1
figure 1

TEM image of morphology MgO nanoparticles[23]

Full size image
Table II Characteristic of MgO Nanoparticles[23]
Full size table

Prior to welding operation, workpiece dimensions of size 15 × 50 × 15mm3 of St37 steel plates were cut and their surfaces were cleaned with acetone and coated with layers of MgO nanoparticles. Also, flux of agglomerated aluminate‐rutile (AWS A5.17: F7AZ- EL12) type with the basicity index of 0.4 (Al2O3 + MnO55 pct, SiO2 + TiO2 30 pct and CaF2 5 pct), copper-coated electrode wire with diameter of 3.2 mm in coil form with the type of DIN S1 (AMA Co. Trade Name 50-11) and the angle of 90 deg between electrode and the workpiece were used in the process.

The experiments were performed by PARS CAT P2310 semi-automatic SAW equipment using direct current reverse polarity (DCRP) and the bead-on-plate weld was carried out in random order. In this research work, thirty two experiments were carried out as per the design matrix given in Table III. The specimens were cut perpendicular to welding direction and finishing with emery papers of grade 240, 320, 400, and 600 and etched with 2 pct Nital solution. Finally, the weld bead profile was measured by means of Olympus optical microscope and obtained the depth of penetration for each run are shown in Table III.

Table III Design of Matrix Based on CCRD
Full size table

Results and Discussion

The used design matrix is shown in Table III. In this matrix, standard order, run order, and levels of each factor are illustrated. ANOVA was carried out for the quadratic polynomial model. This model is used to determine five input parameters and called full model. The p value of each term is shown in Table IV. This model is as follows:

Table IV p Value of Each Term in Full Model
Full size table
$$ \begin{aligned} {\text{Weld}}\;{\text{bead}}\;{\text{penetration}} = 3.487 - 0.0879A + 0.6263B - 0.2271C - 0.1454D - 0.2513E + 0.0216A^{2} + 0.1516B^{2} + 0.0353C^{2} + 0.0716D^{2} - 0.1209E^{2} + 0.2969AB + 0.2881AC - 0.0381AD - 0.3119AE - 0.1956BC + 0.0931BD - 0.3281BE - 0.3956CD + 0.0356CE + 0.0469DE\end{aligned} $$
(3)

However, the p value of this model is great; some terms do not have acceptable p value. These terms are A*A, C*C, D*D, A*D, B*D, C*E, and D*E. Thus, ANOVA was done for the model without these terms, therefore this model is called reduced model. The p value of each term is illustrated in Table IV. This model is as follows:

$$ \begin{aligned} {\text{Weld}}\;{\text{bead}}\;{\text{penetration}} = 3.5973 - 0.0879{A}\,{+}\,0.6263{B} - 0.2271C \hfill \\ - \,0.1454D - 0.2513E + 0.1424B^{2} - 0.1301E^{2} + 0.2969AB\, \hfill \\ +\,0.2881AC - 0.3119AE - 0.1956BC - 0.3281BE - 0.3956CD \hfill \\ \end{aligned} $$
(4)

In this model, the factors A, B, and C have main effect, but there are no quadratic terms of these factors. It means that variation of them affects the depth of penetration linearly. Since there are some interaction terms in this model, the level of factors should be considered in the existence of the other factors.

Adequacy of model was examined using analysis of residuals that includes normality assumption, independence assumption and constant variance assumption. Figure 2 shows normality assumption which is satisfied because all points are centered on straight line. Figure 3 shows the residuals vs the run order. Independence assumption is accepted by this figure because there isn’t any trend among all points. Figure 4 exhibits residuals vs fitted values. There is no pattern in this plot, which is why constant variance assumption is satisfied.

Fig. 2
figure 2

Normal probability plot of the residuals

Full size image
Fig. 3
figure 3

Residuals vs the order of the data

Full size image
Fig. 4
figure 4

Residuals vs the fitted value

Full size image

Table 5 compares full and reduced models with regard to R 2, R 2 (pred), R 2 (adj), and p value. As can be seen, the reduced model is better than the full model based on mentioned criteria. R 2 (pred) value of full model is zero. However R 2 value of full model is better than reduced model, but there is no significant difference between R 2 values of these two models. Thus, the reduced model is used for predicting values of depth of penetration in different levels of factors. The values of predicted depth of penetration are shown in Table III.

Table V Comparing Full and Reduced of Model
Full size table

The optimization of factor levels was carried out using reduced model and the results of optimization are shown in Table VI. Welding current, arc voltage and nozzle-to-plate distance should be adjusted in high level, whereas the welding speed and thickness of nanoparticles should be adjusted in low level. The value of depth of penetration is 10.7624 in this condition. The maximum value of the depth of penetration corresponds to the arc voltage of 32 V, welding current of 700 A, nozzle-to-plate distance of 40 mm, welding speed of 300 mm/min, and minimum thickness of the MgO nanoparticles is 0.0 mm.

Table VI Optimum Levels of Factors
Full size table

The results can be used in companies which use submerged arc welding. Based on existing literature, it seems that researchers did not examine the effect of MgO nanoparticles as the additional flux on depth of penetration. The originality of this paper is to use MgO nanoparticles in submerged arc welding which provides suitable depth of penetration.

Conclusions

From the results of the present research work, the following conclusions are summarized as follows:

  • The five-level-five-factor CCRD techniques can be effectively used for predicting depth of penetration in the submerged arc welding within the range of parameters.

  • The optimized quadratic polynomial model is shown that welding current, arc voltage, and nozzle-to-plate distance should be adjusted in high level, whereas the welding speed and thickness of nanoparticles should be adjusted in low level.