Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 64 (2013) 1030 – 1039

International Conference On DESIGN AND MANUFACTURING, IConDM 2013 &RPPHQW>6@Elsevier to

volume and page numbers.

Sensitivity analysis of flux cored arc welding process variables in

super duplex stainless steel claddings.
B. Senthilkumara* and T.Kannanb
a
Assistant Professor, Department of Mechanical Engineeirng, Kumaraguru College of Technology, Coimbatore – 641 049, Tamilnadu, India.
b
Principal, SVS College of Engineering, J P Nagar,Arasampalayam, Coimbatore – 641 109, Tamilnadu, India

Abstract

Weld surfacing with super duplex grade stainless steel found to improve corrosion resistance and functional life of the mild
steel components used in the process industries. The properties of the deposited layer were influenced by the process variables
that affect the heat input to the process. The influence exerted by the process variables on the responses of super duplex
stainless steel claddings were modeled using the response surface models. The response surface models developed by the
regression techniques using the data collected from central composite rotatable design of experiments. The data extracted from
32 single bead on the plate welds were deposited by flux cored arc welding process. The developed models can be used to
predict and simulate the influence of the process variables on the responses. The insignificant variables found in the full models
were removed by the backward elimination technique. Sensitivity analysis performed on the reduced models helps to identify
and rank the process variables based on their extent of influence on the responses. Then the ranked variables are closely
regulated to tailor the properties of the surfaced layer.
© 2013
© 2013 The
TheAuthors.
Authors.Published by Elsevier
Published Ltd.
by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and
Selection andpeer-review underunder
peer-review responsibility of the organizing
responsibility and review committee
of the organizing and reviewof IConDM 2013.
committee of IConDM 2013
Keywords: FCAW; super duplex stainless steel; cladding; response surface; sensitivity; regression;

1. Introduction

The duplex grade stainless steel possesses two-phase microstructure of austenite and ferrite at equal proportions
found to improve the strength and corrosion resistance properties. The stainless steels are difficult to process with

* Corresponding author. Tel.: +91 9486632722; fax: +91 422 2669402.

E-mail address: bsk_senthilkumar@hotmail.com.

1877-7058 © 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of the organizing and review committee of IConDM 2013
doi:10.1016/j.proeng.2013.09.180
 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039 1031

conventional manufacturing techniques because of their properties increasingly dependent on the composition and
microstructure [1]. The variation in the composition decreases the resistance against pitting and stress corrosion
cracking in the aggressive environments exists in petrochemical, chemical, food processing, paper industries and
desalination plants. Presence of chlorine ions in these environments accelerates the pitting and stress corrosion
cracking. Weld surfacing is used in the fabrication and repair industries to increase the durability of the functional
components used in the process industries [2]. The property of the surfaced layer depends on the percentage
dilution, which is the amount of the base metal melted and mixed with the deposited electrode. The percentage
dilution found to be influenced by the heat input which in-turn affects the composition, strength and integrity of the
fused layer [3]. The adaptability of the weld surfacing stems from the fact that it can process semi-finished products
and wide range of processes available to cater the needs of different industries. The flux cored arc welding
(FCAW) process is a versatile process because of its portability, smooth bead appearance, all position and
automation capabilities [4]. The influence of welding variables on the percentage dilution can be simulated using
the second order mathematical models developed by using response surface methodology [5]. The developed
models represent the important responses like percentage dilution, reinforcement height, etc., as a function of
independently controllable process variables like welding voltage, wire feed rate, welding speed, etc., The validated
models can be used to simulate direct and interaction effect of process variables on the responses as well as to
study the sensitivity of the process variables on the responses [6–9]. The simulation of the weld cladding process
provides significant savings in the man, machine and materials. The sensitivity analysis is used to assess and rank
the effectiveness of the process variables on the responses [7, 8, 10]. The information gathered from the sensitivity
analysis helps to identify the process variables to tailor made the properties of the fused layer to suit the
requirements of the end user.

Nomenclature

X1 welding voltage (Coded)

X2 wire feed rate (Coded)
X3 welding speed (Coded)
X4 nozzle to plate distance (Coded)
X5 welding gun angle (Coded)
R response or dependent variable
β0 , β1 , β2 … coefficients of the second order mathematical model
D Percentage dilution in percentage
H Reinforcement height in mm
W Bead width in mm
SS Sum of squares
DF Degree of Freedom

2 Experimental investigation

The mathematical models were developed using the data collected from the five factors five level experiments
based on the central composite rotatable design. The experimental investigation was carried out in the following
steps.
• Identification of process variables and responses
• Establishing the range of process variables and coding
• Development of the design matrix
• Experimentation and data collection
• Mathematical model development
• Validation of the developed mathematical models
1032 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039

Table 1 Composition of electrode and base metal

Elements (% wt.)
Material
C Mn Si S P Cr Ni Mo Cu N Fe
Electrode 0.027 0.64 0.47 0.005 0.018 25.8 8.62 4.36 0.06 0.28 Balance
Base metal 0.196 1.12 0.293 0.011 0.0044 0.128 0.336 0.275 0.0963 -- Balance

i. Identification process variables and responses

The process variables were identified based on their influence on the responses and relative ease of
controllability. The selected process variables to construct the mathematical models are welding voltage (X1), wire
feed rate (X2), welding speed (X3), nozzle to plate distance (X4) and welding gun angle (X5). The identified
responses have extensive influence on the bead shape and corrosion resistance of the deposited layer are
reinforcement height, bead width and percentage dilution.

ii. Establishing the range of the process variables and coding.

The working ranges of the process variables were established by conducting trial runs and inspecting the weld
bead for any visual defects such as cracking, porosity and discontinuity. The maximum and minimum values of the
process variables were coded as +2 and -2 respectively. The intermediate levels were calculated by using the
formula given by the equation 1. The process variables with their levels and corresponding values along with
symbol and units were presented in the table 2.
2[2 X − ( X max + X min )]
Xi = (1)
( X max − X min )
iii. Development of design matrix

The design used for conducting experiments was based on the five factors and five levels central composite
rotatable design. The design matrix consists of 32 experimental combinations with sixteen factorial combinations,
six centre points and ten star points [11]. The design matrix with recorded responses was presented in the table 3.

Table 2 Process variables levels and coding

Factor levels
Variables Units Symbol
-2 -1 0 +1 +2

Welding voltage Volts X1 22 24 26 28 30

Wire feed rate mm/min X2 5080 5715 6350 6985 7620
Welding speed mm/min X3 120 140 160 180 200
Nozzle to plate distance mm X4 15 17 19 21 23
Welding gun angle deg. X5 70 80 90 100 110

iv. Experimentation and data collection

The experiments were conducted based on the factorial combinations from the table 3 were deposited at
random. After depositing each bead the plates were allowed to cool to the room temperature in the open air. The
 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039 1033

parameters were intentionally disturbed and reset at the beginning of each experiment to introduce the variance in
the experimental setting errors. Then mid portions of the weld beads were sectioned using the power hacksaw, and
parallel end faces were polished and etched. The bead profiles were constructed on the transparencies from the
etched specimens using the reflective type optical profile projector with 10X magnification. The bead profiles were
imported into AutoCAD software to measure responses and table 3 lists the collected responses.

v. Mathematical model development and its validation

The equation 2 represents the responses as a function of the identified process variables.
R = f (X1, X 2 , X 3 , X 4 , X 5 ) (2)
The equation 3 presents the second order mathematical equations from response surface methodology [12].
5 5 5
R = β 0 + ¦ β i X i + ¦ β ii X i + ¦ β ij X i X j (3)
i =1 i =1 i =1
i< j
The expand equation 4 representing all the coefficients of the developed model are
R = β 0 + β1 X 1 + β 2 X 2 + β 3 X 3 + β 4 X 4 + β 5 X 5 + β11 X 12 + β 22 X 22 + β 33 X 32 + β 44 X 42 + β 55 X 52
+ β12 X 1 X 2 + β13 X 1 X 3 + β14 X 1 X 4 + β15 X 1 X 5 + β 23 X 2 X 3 + β 24 X 2 X 4 + β 25 X 2 X 5 + β 34 X 3 X 4
+ β 35 X 3 X 5 + β 45 X 4 X 5
(4)
The coefficients of the second order mathematical model equation were calculated using the least square method.
Table 4 presents full model coefficients for the responses.

Table 3 Design matrix with responses

Process variables (Coded) Responses

Exp. Nozzle to
Welding Wire feed Welding Welding Percentage Reinforcement Bead
No. plate
voltage rate speed gun angle dilution (D) height (H) width (W)
distance

X1 X2 X3 X4 X5 % mm mm
1 -1 -1 -1 -1 -1 1.216 4.272 8.910
2 -1 -1 -1 +1 +1 3.319 4.391 9.088
3 -1 -1 +1 -1 +1 8.629 3.615 8.237
4 -1 -1 +1 +1 -1 4.615 3.843 7.596
5 -1 +1 -1 -1 +1 4.058 4.699 10.717
6 -1 +1 -1 +1 -1 3.527 4.718 9.121
7 -1 +1 +1 -1 -1 6.082 4.116 9.241
8 -1 +1 +1 +1 +1 5.304 4.304 9.324
9 +1 -1 -1 -1 +1 1.352 4.192 9.999
10. +1 -1 -1 +1 -1 1.749 4.601 8.530
11. +1 -1 +1 -1 -1 5.560 3.740 8.038
12. +1 -1 +1 +1 +1 4.453 3.747 8.759
13. +1 +1 -1 -1 -1 5.081 3.898 9.268
14. +1 +1 -1 +1 +1 1.177 4.753 10.652
15. +1 +1 +1 -1 +1 4.328 3.882 10.530
1034 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039

16. +1 +1 +1 +1 -1 4.387 4.086 9.059

17. -2 0 0 0 0 2.764 4.913 7.084
18. +2 0 0 0 0 2.859 4.167 9.341
19. 0 -2 0 0 0 4.689 3.826 7.503
20. 0 +2 0 0 0 4.628 4.311 10.677
21. 0 0 -2 0 0 1.286 4.984 10.804
22. 0 0 +2 0 0 8.390 3.518 8.500
23. 0 0 0 -2 0 5.436 3.680 9.932
24. 0 0 0 +2 0 4.430 4.130 9.144
25. 0 0 0 0 -2 3.025 4.280 8.843
26. 0 0 0 0 +2 4.128 4.150 9.748
27. 0 0 0 0 0 5.031 4.192 9.031
28. 0 0 0 0 0 5.783 3.857 9.442
29. 0 0 0 0 0 4.406 4.223 8.985
30. 0 0 0 0 0 6.059 4.147 9.724
31. 0 0 0 0 0 6.039 3.879 9.777
32. 0 0 0 0 0 4.424 4.012 9.757

Table 4 Full model coefficients

Responses
Percentage dilution D Reinforcement height H
Bead width W (mm)
Coefficients (%) (mm)
Reduced Full Reduced Full Reduced
Full Model
Model Model Model Model Model
β0 5.259 5.030 4.061 4.056 9.433 9.449
β1 -0.353 -0.353 -0.106 -0.106 0.296 0.296
β2 0.122 -- 0.126 0.126 0.629 0.629
β3 1.504 1.504 -0.297 -0.297 -0.421 -0.421
β4 -0.408 -0.408 0.122 0.122 -0.183 -0.183
β5 0.109 -- 0.002 -- 0.390 0.390
β 11 -0.589 -0.570 0.113 0.113 -0.292 -0.292
β 22 -0.127 -- -0.006 -- -0.071 --
β 33 -0.082 -- 0.041 0.041 0.069 --
β 44 -0.058 -- -0.046 -0.046 0.041 --
β 55 -0.397 -0.378 0.032 0.032 -0.020 --
β 12 0.042 -- -0.086 -0.086 0.024 --
β 13 -0.196 -0.196 0.013 -- 0.086 --
β 14 -0.083 -- 0.058 0.058 0.071 --
β 15 -0.709 -0.709 0.012 -- 0.159 0.159
β 23 -0.585 -0.585 0.052 0.052 0.143 0.143
β 24 -0.158 -- 0.031 -- -0.024 --
β 25 -0.551 -0.551 0.083 0.083 0.095 --
β 34 -0.244 -0.244 -0.048 -0.048 0.012 --
β 35 0.234 0.234 -0.049 -0.049 -0.107 --
β 45 -0.028 -- -0.026 -- -0.032 --
 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039 1035

vi. Validation of the developed models

The developed models were checked for statistical and experimental validity. The criterion for experimental
validity is that the calculated F – ratio should be less that critical value and calculated R – ratio of the model should
exceed the critical value at 5% level of significance. The developed models were found to satisfy these require
hence are not rejected at 5% level of significance [3, 13]. The developed models also checked for experimental
validity and found that the models predicted the values falls well within the acceptable levels [14, 15]. Table 5
presents the statistical validity of the developed models.
Table 5 Statistical validity of the developed models

First order terms Second order terms Lack of Fit Error terms Adequacy of
Response F – ratio* R – ratio**
SS DF SS DF SS DF SS DF the model

Percentage
61.89 5 36.94 15 3.58 6 2.99 5 1.00 8.11 Adequate
dilution (D)
Reinforcement
3.12 5 0.94 15 0.16 6 0.13 5 1.07 8.00 Adequate
height (H)

Bead width (W) 20.32 5 4.12 15 1.84 6 0.67 5 2.30 9.19 Adequate

F ratio = MS of lack of fit / MS of error term, R ratio = MS of first order and second order term / MS of error terms,
*
Critical value of F ratio F(6, 5, 0.05) = 4.95, ** Critical value of R-ratio (20, 5, 0.05) = 4.56

3. Sensitivity analysis

The reduced models were developed by eliminating the insignificant coefficients using the backward
elimination technique without sacrificing the accuracy of the model. Table 4 presents the reduced mathematical
models. The mathematical models with significant variables were differentiated with the process variable of
interest like welding voltage (X1), wire feed rate (X2), welding speed (X3), etc.,

3.1. Sensitivity of welding voltage

The sensitivity of welding voltage on the responses was presented by the equations 5, 6 and 7.
∂ ( D)
= −0.353 − 1.140 X 1 − 0.196 X 3 − 0.709 X 5
∂( X 1 ) (5)
∂(H )
= −0.106 + 0.226 X 1 − 0.086 X 2 + 0.058 X 4
∂( X 1 ) (6)
∂(W )
= 0.296 − 0.584 X 1 + 0.159 X 5
∂( X 1 ) (7)
From the Fig. 1 (a), it has been found that the sensitivity of the welding voltage on the responses percentage
dilution and bead width decreased with increasing welding voltage. On the other hand, sensitivity of welding
voltage on reinforcement height found to exhibit positive relationship with increasing welding voltage. This effect
is observed due to the increased arc length with increasing welding voltage decreased the arc force acting on the
weld pool.

3.2. Sensitivity of wire feed rate

The sensitivity of the wire feed rate on the responses was presented in the form of equations 8, 9 and 10
1036 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039

∂ ( D)
= −0.585 X 3 − 0.551X 5
∂( X 2 ) (8)
∂(H )
= 0.126 − 0.086 X 1 + 0.052 X 3 + 0.083 X 5
∂( X 2 ) (9)
∂ (W )
= 0.629 + 0.143 X 3
∂( X 2 ) (10)

Fig. 1. (a) Sensitivity of the welding voltage (X1) on the responses; (b) Sensitivity of the wire feed rate (X2) on the responses.

Fig. 1 (b) shows the sensitivity of wire feed rate is high when the welding speed is at the level –2. This effect is
better explained with the low welding speed and relatively smaller weld pool produces deeper penetration and
hence percentage dilution exhibits a positive relationship. Wire feed rate found to influence the welding current
and weld pool size, relatively larger weld pool and reduced heat input per unit length with increased welding speed
reduces the ability of the arc to penetrate the base metal and hence percentage dilution decreases with increasing
welding speed. The sensitivity of wire feed rate on the responses reinforcement height and bead width increases
with increased welding speed.

3.3. Sensitivity of welding speed

The sensitivity of the welding speed on the responses was presented in the form of equations 11, 12 and 13.

∂ ( D)
= 1.504 − 0.196 X 1 − 0.585 X 2 − 0.244 X 4 + 0.234 X 5
∂( X 3 ) (11)
∂( H )
= −0.297 + 0.082 X 3 + 0.052 X 2 − 0.048 X 4 − 0.049 X 5
∂( X 3 ) (12)
∂ (W )
= −0.421 + 0.143 X 2 
∂( X 3 )     (13)
Fig. 2 (a) shows the sensitivity of welding speed on the response percentage dilution decreases with increasing
wire feed rate. This effect is attributed to the effect of increased size of the weld pool effectively reduced the
ability of arc to penetrate the base metal. The sensitivity of welding speed on the responses increased with
increasing wire feed rate. The influence of wire feed rate is attributed to the effect of increased rate at which
electrode is melted and deposited.
 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039 1037

3.4. Sensitivity of nozzle to plate distance

The sensitivity of the nozzle to plate distance on the responses was presented in the form of equations 14, 15 and
16

Fig. 2. (a) Sensitivity of the welding speed (X3) on the responses; (b) Sensitivity of nozzle to plate distance (X4) on the responses
∂ ( D)
= −0.408 − 0.244 X 3
∂( X 4 ) (14)
∂(H )
= 0.122 − 0.092 X 4 + 0.058 X 1 − 0.048 X 3
∂( X 4 ) (15)
∂ (W )
= −0.183
∂( X 4 ) (16)
From the Fig. 2 (b) it has been found that the sensitivity of the nozzle to plate distance on the response percentage
dilution decreases with increasing welding speed. The welding speed was found to influence the heat input per unit
length and also the rate at which fresh base metal exposed under the arc. The sensitivity of the nozzle to plate on
reinforcement height decreases with increasing welding speed. The sensitivity of the nozzle to plate distance on the
response bead width is constant with increasing welding speed.

3.5. Sensitivity of welding gun angle

The sensitivity of the welding gun angle on the responses was presented in the form of equations 17, 18 and 19.

∂ ( D)
= −0.756 X 5 − 0.709 X 1 − 0.551X 2 + 0.234 X 3
∂( X 5 ) (17)
∂(H )
= 0.064 X 5 + 0.083 X 2 − 0.049 X 3
∂( X 5 ) (18)
∂ (W )
= 0.390 + 0.159 X 1
∂( X 5 ) (19)
From the Fig. 3 it has been found that welding gun angle exhibits negative sensitivity on the response when
welding speed is at the level -2. This effect is attributed to the inverse relationship between the welding speed and
weld pool size. The size of the weld pool decreases with increasing welding speed. It increases the rate at which
base metal exposed under the arc and hence percentage dilution exhibits increased the sensitivity of welding gun
angle, when the welding speed is at the level +2. The volume of the electrode melted and deposited per unit length
1038 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039

decreased with increasing welding speed is attributed to the decrease in the sensitivity of welding gun angle on the
response reinforcement height. At the same time, sensitivity of welding gun angle on the response bead width
remains unchanged.

Fig. 3 Sensitivity of the welding gun angle(X5) on the responses

4. Conclusion

• Five factors five levels central composite rotatable design was used to collect the required data to
construct the second order mathematical models. The developed models were validated both statistically
and experimentally. The insignificant coefficients were removed to generate reduced models. The reduced
models were then used to study the sensitivity of the process variables on the responses.
• The sensitivity analysis carried out on the validated models helps to study the effect influential variables
on the responses. The response percentage dilution found to have maximum positive sensitivity to the
welding speed when the wire feed rate at the level –2. The same response exhibits negative sensitivity to
the welding voltage when the welding voltage is at the level +2.
• The response reinforcement height found to show positive sensitivity to the welding voltage when the
welding voltage is at the level +2. It shows maximum negative sensitivity with welding voltage when the
welding voltage is at the level –2.
• The response bead width found to exhibit the maximum positive and negative sensitivity with the welding
voltage at levels –2 and +2 respectively.
• The process variable welding speed found to influence the sensitivities of the wire feed rate, nozzle to
plate distance and welding gun angle. The sensitivity of the welding speed varies with wire feed rate. The
sensitivity analysis found to be useful in identifying important process variables so that the properties of
the deposited layer can be tailored to suit the requirements of the process industries.

Acknowledgement
The authors wish to thank M/S Metrode welding consumables for providing flux cored super duplex stainless
steel welding wire for this work. The authors also wish to thank the management of Kumaraguru College of
Technology and SVS College of Engineering for providing the required facilities to carry out this research work.

References

[1] Nowacki, J., Rybicki, P.: The influence of welding heat input on submerged arc welded duplex steel joints imperfections. Journal of
Materials Processing Technology. 164-165, 1082–1088 (2005).
 B. Senthilkumar and T.Kannan / Procedia Engineering 64 (2013) 1030 – 1039 1039

[2] Aljoboury, A.I., Mourad, A.-H.I., Alawar, A., Abou Zour, M., Abuzeid, O.A.: Stress corrosion cracking of stainless steels recommended for
building brine recirculation pumps. Engineering Failure Analysis. 17, 1337–1344 (2010).

[3] Murugan, N., Parmar, R., Sud, S.K.: Effect of submerged arc process variables on dilution and bead geometry in single wire surfacing.
Journal of Materials Processing Technology. 37, 767 – 780 (1993).

[4] Kannan, T., Murugan, N.: Effect of flux cored arc welding process parameters on duplex stainless steel clad quality. Journal of Materials
Processing Technology. 176, 230–239 (2006).

[5] Gunaraj, V., Murugan, N.: Application of response surface methodology for predicting weld bead quality in submerged arc welding of
pipes. Journal of Materials Processing Technology. 88, 266 – 275 (1999).

[6] Karao÷lu, S., Seçgin, A.: Sensitivity analysis of submerged arc welding. journal of materials processing technology. 202, 500–507 (2008).
[7] Palani, P.K., Murugan, N.: Sensitivity Analysis for Process Parameters in Cladding of Stainless Steel by Flux Cored Arc Welding. Journal
of Manufacturing Processes. 8, 90–100 (2006).
[8] Kim, I.S., Son, K.J., Yang, Y.S., Yaragada, P.K.D. V: Sensitivity analysis for process parameters in GMA welding processes using a
factorial design method. International Journal of Machine Tools & Manufacture. 43, 763–769 (2003).
[9] Shanmugam, R., Murugan, N.: Optimization and Sensitivity Analysis of Wear Model of Stellite – 6 Hardfacing by Gas Tungsten Arc
Welding Process. IE(I) Journal-MM. 85, 65–70 (2004).
[10] Kim, I.S., Jeong, Y.J., Son, I.J., Kim, I.J., Kim, J.Y., Kim, I.K., Yaragada, P.K.D. V: Sensitivity analysis for process parameters
influencing weld quality in robotic GMA welding process. Journal of Materials Processing Technology. 140, 676–681 (2003).
[11] Montgomery, D.C.: Design and Analysis of Experiments. John Wiley & Sons Inc (2001).
[12] Montgomery, D.C., Runger, G.C.: Applied Statistics and Probability for Engineers Third Edition. John Wiley & Sons, Inc. (2003).
[13] Kannan, T., Yoganandh, J.: Effect of process parameters on clad bead geometry and its shape relationships of stainless steel claddings
deposited by GMAW. The International Journal of Advanced Manufacturing Technology. 47, 1083–1095 (2009).
[14] Palani, P.K., Murugan, N.: Development of mathematical models for prediction of weld bead geometry in cladding by flux cored arc
welding. The International Journal of Advanced Manufacturing Technology. 30, 669–676 (2006).

[15] Palanivel, R., KoshyMathews, P., Murugan, N.: Development of mathematical model to predict the mechanical properties of friction stir
welded AA6351 aluminum alloy. Journal of Engineering Science and Technology Review. 4, 25–31 (2011).