Professional Documents
Culture Documents
Optimisation and Prediction of The Weld Bead Geometry of A Mild Steel Metal Inert Gas Weld
Optimisation and Prediction of The Weld Bead Geometry of A Mild Steel Metal Inert Gas Weld
Optimisation and Prediction of The Weld Bead Geometry of A Mild Steel Metal Inert Gas Weld
To cite this article: Samuel Oro-Oghene Sada & Joseph Achebo (2020): Optimisation and
Prediction of the Weld Bead Geometry of a Mild Steel Metal Inert Gas Weld., Advances in Materials
and Processing Technologies
1. Introduction
Welding is a key sector in the manufacturing industry, manufactures rely on the sector
for the design of sound and efficient joints for the manufacture of mechanical compo
nents capable of meeting the reliability needs of consumers. The strength of the weld
joint (depth of penetration, weld bead height and width), the reliability of the product,
and the ease in continuity of production is determined by certain factors such as the
mechanical properties of the welded metal which in turn is dependent on the metallur
gical features of the weld [1,2]. These factors can be evaluated by means of an establish
ment mathematical model capable of specifying the input and output relationship of the
variables, along with the optimal weld qualities [3,4].
Several studies have been made on the weld bead geometry of different weld metals
using different modelling methods. Kim [5] designed and developed an intelligent system
based on multiple regressions and neural networks, which proved very effective in the
selection of input parameters based on a given bead geometry. Also the capability of the
modified taguchi method was also applied in the analysis of welding process parameters
effect as well as in determining the optimal weld pool geometry [6] in a welding experi
ment. The proposed approach along with the experimental results worked satisfactorily.
Yazdipour [7] performed a study using a novel optimisation algorithm known as
imperialist competitive algorithm (ICA) in the determination of optimal weld bead of
a tungsten inert gas (TIG) welded plate using the input parameters; weld current, arc
voltage, weld speed, and arc length. The authors developed a correlation using regression
analysis and went further to the test adequacy of the proposed model. In the study of
Benyounis and Olabi [8], they reported the emergence of a more suitable and useful
technique capable of giving the optimum value of the responses in a manufacturing
process involving process parameters settings. The response surface methodology (RSM)
as has been identified, comprises sets of statistics-based approaches designed for analys
ing experiments, building models, assessing factors and their effects and searching for
optimal solutions. Based on studies [9–11] reviewed, one unique advantage of the RSM
techniques is its ability to determine the interactive effect of the model terms.
This study investigates the capability of the response surface methodology in optimis
ing and predicting the optimal weld bead responses (depth of penetration and bead
width) of a Metal Inert Gas welded mild steel plate.
2. Methodology
A Mild steel plate, 6 mm thick cut to dimensions of 100x50mm was selected as the
specimen for the experiment along with argon gas as the inert gas, and the metal inert gas
(MIG) welding equipment for the welding process. The specimens were cleaned to
remove any impurities and held together in pairs of two in a vise for the welding
experiment. The following input process parameters and their ranges were selected for
the experiment; weld current: 60–100 Amp, Weld Voltage: 5–15 volts, Weld Speed:
50–80 mm/mins, and Root Gap: 2–3 mm. At the end of the welding process, the
responses; depth of penetration and bead width were measured using a digital vernier
caliper (Make: Starrett No. 723, American and least count = 0.01 mm), and recorded.
Table 1 shows the result of the experiment.
Preceding the experiment, thirty (30) experimental runs comprising the process
parameters was generated using the central composite design (CCD); one of the two
designs of the response surface methodology.
X
k X
k XX
k
y ¼ β0 þ βij xi 2 þ βij xi 2 þ βi xi xj þ εfor i < j (1)
i¼1 j¼1 ij
Where y and xi represents the response and input variables ranging from 1 to k, βo, βj, βjj
and βij are the intercept constant, and the quadratic coefficients of the second-order
terms, respectively, and ei is the error.
Table 2. Analysis of variance (ANOVA) test for the responses – depth of penetration.
Source Sum of Squares df Mean Square F-value p-value
Model 3.3300 14 0.2378 4.7700 0.0024 Significant
A-Weld Current 0.0315 1 0.0315 0.6327 0.0008
B-Weld Voltage 0.4455 1 0.4455 8.9400 0.0092
C-Weld Speed 0.0900 1 0.0900 1.8100 0.1989
D-Root Gap 0.8626 1 0.8626 17.3100 0.0028
AB 0.0770 1 0.0770 1.5400 0.2330
AC 0.0163 1 0.0163 0.3261 0.5764
AD 0.0095 1 0.0095 0.1907 0.6685
BC 0.3164 1 0.3164 6.3500 0.0236
BD 0.0005 1 0.0005 0.0102 0.9211
CD 0.1871 1 0.1871 3.7500 0.0718
A2 0.2503 1 0.2503 5.0200 0.0406
B2 0.6232 1 0.6232 12.5000 0.0030
C2 0.2310 1 0.2310 4.6300 0.0480
D2 0.0048 1 0.0048 0.0963 0.7606
Residual 0.7477 15 0.0498
Lack of Fit 0.2334 10 0.0233 0.2269 0.9778 not significant
Pure Error 0.5143 5 0.1029
Cor Total 4.0800 29
Table 3. Analysis of variance (ANOVA) test for the responses – bead width.
Source Sum of Squares Df Mean Square F-value p-value
Model 3.7400 14 0.2675 5.6500 0.0010 Significant
A-Weld Current 0.4895 1 0.4895 10.3400 0.0058
B-Weld Voltage 0.0085 1 0.0085 0.1799 0.6775
C-Weld Speed 0.0028 1 0.0028 0.0594 0.8107
D-Root Gap 0.4444 1 0.4444 9.3900 0.0079
AB 0.0015 1 0.0015 0.0323 0.8598
AC 0.0234 1 0.0234 0.4937 0.4930
AD 0.1263 1 0.1263 2.6700 0.1232
BC 1.1500 1 1.1500 24.2900 0.0002
BD 0.1294 1 0.1294 2.7300 0.1190
CD 0.0027 1 0.0027 0.0566 0.8152
A2 1.1400 1 1.1400 24.0400 0.0002
B2 0.0360 1 0.0360 0.7612 0.3967
C2 0.0189 1 0.0189 0.3997 0.5368
D2 0.0439 1 0.0439 0.9269 0.3509
Residual 0.7098 15 0.0473
Lack of Fit 0.2720 10 0.0272 0.3107 0.9454 not significant
Pure Error 0.4378 5 0.0876
Cor Total 4.4500 29
Figure 1. Predicted versus experimental values plot for the depth of penetration.
Figure 2. Predicted versus experimental values plot for the bead width.
As shown in Figure 6–8, the contour plots produce a visual display of the optimal
factor settings region, thereby making the prediction of the responses at any region of the
experimental domain possible.
ADVANCES IN MATERIALS AND PROCESSING TECHNOLOGIES 7
Figure 3. 3D plots showing the responses versus two process parameters (root gap and weld current).
Figure 4. 3D plots showing the responses versus process parameters (weld speed and weld current).
Figure 5. 3D plots showing the responses versus process parameters (arc voltage and weld current).
Figure 6. Contour plots showing the responses versus process parameters (weld speed and current).
Figure 7. Contour plots of the responses versus two process parameters (arc voltage and weld
current).
Figure 8. Contour plots showing the responses versus process parameters (root gap and weld current).
ADVANCES IN MATERIALS AND PROCESSING TECHNOLOGIES 9
4. Conclusion
The response surface methodology has been successfully applied in the optimisation and
prediction of the weld bead geometry of a mild steel metal inert gas (MIG) weld.
A mathematical model of the responses; depth of penetration and bead width was
obtained with weld current, arc voltage, weld speed and root gap, as input process
variables. A model adequacy test performed to validate the significance of the model as
well as evaluate the significant contributions of the variables to the responses, revealed
that the weld current has the most significant effect on the responses and that response
surface technique is a good predictor of the response haven recorded a reliability factor of
81.66%. With the use of design expert software, the optimal values of the responses
recorded a depth of penetration of 1.624 mm and bead width of 3.732 mm under the
combined input process parameters of weld current 87.86amp, weld voltage 15volts,
speed 59.02 m/s and root gap of 2 mm.
A confirmatory test performed using the optimal values shows that the response
surface technique performed accurately, as the recorded test values revealed less 3%
error.
Disclosure statement
No potential conflict of interest was reported by the authors.
References
[1] American Welding Society. Welding processes, welding Handbook. 8th ed. Miami, FL. Vol.
2; 1991.
[2] Joshi SR, Ganjigatti J,P, Kulkarni S. Application of statistical and soft computing based
modeling and optimization techniques for various welding processes” a review. Int J Latest
Trends Eng Technol. 2014;3(4):375–384.
[3] Tarng YS, Yang W. “Optimization of the weld bead geometry in gas tungsten arc welding by
the Taguchi method”. Int J Adv Manuf Technol. 1998;14:549–554.
[4] kumar P,M, Patel N,B. Parametric optimization of TIG welding on UNS S31603 using
genetic algorithm. Int J Adv Eng Res Dev. 2015;2: 256-267.
[5] Kim IS, Son J,S, Park C,E, et al. An investigation into intelligent system for predicting bead
geometry of GMA weld process. J Mater Process Technol. 2005;159(1):113–118.
[6] Juang S,C, Tarng Y,S. Process parameters selection for optimizing the weld pool geometry in
the TIG welding of stainless steel. J Mater Process Technol. 2002;122(1):33–37.
[7] Yazdipour A, Ghaderi M,G. Optimization of weld bead geometry in GTAW of CP titanium
using imperialist competitive algorithm. Int J Adv Manuf Technol Inter. 2014;74(5–8):619–
625. .
[8] Benyounis K,Y, Olabi A,G. Optimization of different welding processes using statistical and
numerical approaches. Adv Eng Software. 2008;39(6):483–496.
10 S. O. SADA AND J. ACHEBO
[9] Gunaraj V, Murugan N. Application of response surface methodology for predicting weld
bead quality in submerged arc welding of pipes. J Mater Process Technol. 1999;88
(1):266–275.
[10] Sada SO. Minimization of undercut weld defect using the response surface methodology
optimization process. Nigeria Journal of Engineering. 2018;24(2):43–52.
[11] Sada SO. Optimization of weld strength properties of tungsten inert gas mild steel welds
using the response surface methodology. Nigeria Journal of Technology. 2018;37
(2):407–415.
[12] Montgomery DC. Design and analysis of experiments: response surface method and design.
New Jersey: John Wiley and Sons Inc.; 2005.
[13] Thepsonthi T, Ozel T. Multi-objective process optimization for micro-end milling of
Ti-6Al-4V titanium alloy. Int J Adv Manuf Technol. 2012;63(9–12):903–914.
[14] Bhattacharya A, Batish A, Kumar P. Experimental investigation for multi-response optimi
zation of bead geometry in submerged arc welding using grey analysis. J Inst Eng (India) Ser
C. 2012;93(2):123–132.