Professional Documents
Culture Documents
1 s2.0 S0030399212005762 Main
1 s2.0 S0030399212005762 Main
a r t i c l e i n f o abstract
Article history: This paper investigates laser beam welding of dissimilar AISI430F and AISI440C stainless steels.
Received 13 August 2012 A combined welding and pre-and-postweld treatment technique was developed and used successfully
Received in revised form to avoid micro-crack formation. This paper also examined the effects of laser welding parameters and
10 November 2012
line energy on weld bead geometry and tried to obtain an optimized laser-welded joint using a full
Accepted 17 December 2012
Available online 24 January 2013
factorial design of experiment technique. The models developed were used to find optimal parameters
for the desired geometric criteria. All the bead characteristics varied positively as laser power increased
Keywords: or welding speed decreased. Penetration size factor decreased rapidly due to keyhole formation for line
Laser-welding energy input in the range of 15–20 kJ/m. Laser power of 790–810 W and welding speed of 3.6–4.0
Steels
m/min were the optimal parameters providing an excellent welded component. Whatever the
Butt joint
optimization criteria, beam incident angle was around its limiting value of 151 to achieve optimal
geometrical features of the weld.
& 2012 Elsevier Ltd. All rights reserved.
0030-3992/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.optlastec.2012.12.025
126 M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136
Fig. 1.1. butt welded surface (a) before and (b) after applying the combined preheat-and-postheat welding technique for the line energy, LE¼ 15 kJ/m.
’ Finally, determination of optimal range of laser welding head mounted on Z motion stage. The laser head is then set
parameters, using the developed models with numerical tangentially in the direction of rotation to an intended beam
optimization, to minimize the weld width and maximize the incident angle. Finally, the positions of the X–Y–Z motion stages
weld penetration depth. are adjusted in such a way that pointed tool tip touches the
planned point of focus. Laser beam is focused on this located point
through the laser head at the specified angle, and the necessary
rotary motion is provided to the specimen through specimen
2. Materials and experimental procedures holder mounted on an X–Y motion stage. Computer control panel
is interfaced with the linear X–Y–Z as well as rotary motion
2.1. Materials systems to regulate the aforesaid movements.
The experiment is initially planned based on statistical factorial
Two cylindrical shells made of ferritic AISI430F (cold drawn, experimental design with full replication. During experimenta-
annealed and centerless ground) and martensitic AISI440C (pre- tion, laser power, (P), welding speed (S), and beam incident angle
hardened and tempered) stainless steels are welded circularly to (A) are selected as process input variables for laser welding.
make a butt joint. This dissimilar joint is selected based on both Table 2.2 shows the experimental condition, laser welding input
technical and economical aspects, because they can provide variables, and design levels used at a glance. Each of the input
satisfactory service performance and considerable savings. More- variables and its working range is selected based on industrially
over, in automotive industries, these materials are frequently recommended laser-welding parameters used in automotive
used in welded form for making different types of fuel injectors. industries.
The chemical compositions of base metals available in as-received General Full Factorial Design is used as a statistical design of
condition and the weld seam characteristics are shown in Fig. 2.1 experiment technique to develop statistical models relating the
and Table 2.1, respectively. The inside diameter of the outer shell welding input parameters to each of the two output responses
and the outside diameter of the inner shell are machined to of the weld (weld width and penetration depth). The adequacies
Ø7.570.025 mm and Ø7.458 70.015 mm, respectively, to have a of the models developed and their significant linear and interac-
clearance fit between them when the shells are assembled. tion model terms are measured by analyzing variance and other
adequacy measures. Finally, these mathematical models are used
2.2. Experimental procedures to determine the optimal settings of welding parameters to
ensure the desired weld quality. In this study, the quality criteria
Specimens are welded circularly in a fillet joint configuration defined for the weld to determine the optimal settings of welding
using a 1.1 kW continuous wave Nd:YAG laser (Rofin DY011). parameters are the minimization of weld width and the
The optical system consisted of a 300 mm fiber and two lenses maximization of weld penetration depth.
of 200 mm focal and collimate lengths are used to deliver the Besides, the energy delivered per unit length of weld line is
laser with a minimum focal spot diameter of 300 mm. A three-step referred to as line energy (LE), which is frequently used in various
procedure is followed to locate the focal point. First, an excep- laser-processing techniques and termed as a key-parameter when
tionally sharp-nosed tool of 200 mm in height is attached to laser continuous-wave laser is used. This term is calculated as the ratio
Ø9.56±0.03
AISI AISI
440C
440C 430F
430F Outer
Dp Shell
Rm
Inner
Shell
Inner Shell
Inner Shell Outer Shell
Outer Shell
Ø7.458±0.015
Fig. 2.1. Characterization of welding cross-section (W: Weld width, Dp: Weld penetration depth, Rm: Minimum crack-path).
Table 2.1
Chemical compositions of base metals of the weld.
C Cr Ni Mn P S Si Mo Se Fe
of laser power over the welding speed as shown in Eq. (1): 2.3. Weld bead characterization
Table 2.2
Experimental conditions and response factors.
Table 2.3
Design matrix with actual factors and measured mean responses.
and after the cut using the optical microscope. Hermetic weld is the effects of energy input on weld profile characteristics, and to
ensured by performing leak test in vacuum for each of welded explain different laser welding phenomena as well.
specimens. During leak test, nitrogen is inflated into the
assembled part at a pulsed pressure in the range 10–150 bar for 3.1. Combined welding and pre-and-postweld treatment concept
the expected life cycles. This method also guarantees that the and its effects
weld will not fail during its service life. In case of failure, the
internal cracks generated during the welding process propagate The main concept adopted to modify the experimental set up
up to the free surface and N2 leakage is detected by a loss of has been described schematically in Fig. 3.1. When a moving laser
vacuum into the chamber. beam hits the surface of a metal in perpendicular direction, an
asymmetric melt pool is generated (Fig. 3.1(a)). Heat conduction
3. Results and discussion in the surroundings is not isotropic and generates different
thermal gradients in frontal (GF), lateral (GL) and back (GB)
Various weld profile characteristics are measured with axially directions. The entity of such difference can be calculated follow-
cut specimens using an optical microscope and are recorded for ing [10]. The lengths of the arrows represent a realistic proportion
further analyses described in the succeeding sections. among the modules of the thermal gradients considering
Perturbation plots are used to illustrate the effects of indivi- the jump between vaporization temperature TV and ambient
dual process parameter such as laser power (P), welding speed (S), temperature TA, and the distance in which it is achieved [3].
and beam incident angle (A) on geometrical features of the weld The frontal gradient is extremely higher with respect to the back
e.g., weld width, weld penetration depth, and minimum crack- due to the deformation of isotherms caused by the welding speed.
path. Contour plots are used to show the two-factor interaction The basic idea is to reduce this shock by decreasing GF: this effect
effects on the same weld bead geometry. is obtained displacing and widening the laser spot (dot line),
The line energy is plotted against weld width, weld penetra- as reported in Fig. 3.2(b).
tion depth, and minimum crack-path with a view to demonstrate From the experimental point of view, this is obtained by
inclining the beam on the tangential plane while the specimen
moves against the inclined beam as shown in Fig. 3.2(b). For any
GL Ta generic position of the butt specimen welded with an inclined
beam shown in Fig. 1.2, the cross sectional view of the melt pool
Tm is found to be also asymmetric with respect to the beam axis as is
GF GB Ta reported in Fig. 3.2(a) [16]. This is because the portion of melt
Ta Tv Melt pool
Spot pool surface exposed directly to the laser beam is depleted more
as compared to other portion of the keyhole surface.
The top view shows an elliptical print in which the left side is
GL wider than the right one thus generating a thermal gradient of
Ta
lower entity in the cross sectional plane. This phenomenon
causing the enlargement of the melt pool and the unsymmetrical
GL
distribution of energy inside the print is adopted in the present
research to generate a preheating effect before fresh material
reaches the beam axis
GF Melt pool Tv GB During welding, the fresh material is first exposed to the laser
Spot
radiation in a region with positive defocus which determines the
reduction in the thermal gradient. It then reaches the highest
GL intensity zone of the laser beam. At the exit (right) side, the beam
acting on welded material results in slight postheat treatment
Fig. 3.1. Top views of melt pools and basic draft of thermal gradients (G) in of the weld. No relevant changes occur in the lateral gradient.
different directions with: (a) laser beam perpendicular to the surface (b) laser The ultimate result is a welded seam due to continuous rotation
beam inclined on the tangential plane. of the unsymmetrical weld pool over 3601 that reduces the
n
io
La
ct
se
re
rb
di
ea
m
e nt
id
Fo A
nc
pl cal
an
ri
e
se
La
Melt pool
+ defocus
Solidified seam
AISI 430F
AISI 440C
Fig. 3.2. (a) Top and cross sectional views of an instant during the welding process, (b) draft of the basic concept to obtain a less severe heating rate on the fresh material.
130 M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136
Perturbation Perturbation
1220 1300
-1.00 -0.50 0.00 0.50 1.00 - 1.00 -0.50 0.00 0.50 1.00
Deviation from Reference Point (Coded Units) Deviation from Reference Point (Coded Units)
Perturbation
1200
Minimum Crack Path, µm
925
A
650 P
S
S
375 A
P
100
Fig. 3.3. Perturbation plot showing effect of all factors on (a) weld width, (b) weld penetration depth, (c) minimum crack path.
occurrence of crack formation as can be seen in Fig. 1.1. This 3.3. Effects of line energy
combined process also decreases the penetration depth of the
weld since welding with inclined laser beam causes reflection Fig 3.5(a)–(c) show the effects of line energy input on the weld
losses and a more complex heat transmission through refraction penetration depth (DP), minimum crack path (Rm), and weld
described by Snell’s law. width (W), respectively, whereas variation in penetration size
factor with line energy is illustrated in Fig 3.5(d).
For line energy in the range of 9.0–15 kJ/m, as illustrated in
3.2. Effects of process parameters Fig 3.5(c)–(d), there is a rapid growth in weld width (W) with
energy input, whereas change in penetration size factor (W/Dp) is
The Fig 3.3(a)–(c) show the perturbation plots to compare the insignificant. Slight negative variations in penetration size factor
effects of all the process parameters at the center point in the prove that the laser welding is mainly conduction limited. Since
design space. The results suggest that laser power has the most the melt pool geometry depends on energy intensity, uniform
significant positive impact on the weld width; weld penetration conduction occurring in all directions usually results in semi-
depth; and minimum crack-path. The opposite phenomena are circular weld profile. However, the heat conduction along the
observed for the welding speed. This is because higher laser beam axis becomes dominant with the increase in energy input
power and slower welding speed result in higher energy deposi- and weld shape changes from semi-circular to parabolic. Similar
tion on the weld area, and longer irradiation time for the parametric effects on the welding pool width at surface [17] and
deposited energy to diffuse into material. the upper width [18] are also observed for welding dissimilar low
These figures also illustrate that increase in beam incident carbon and austenitic stainless steels in a butt joint configuration
angle results in shallower weld penetration, and shorter mini- using the CO2 laser.
mum crack-path, whereas the larger beam incident angle causes Again, a small variation in weld width is observed for the line
the wider weld width. These are due to following consequences: energy input in the range of 15–20 kJ/m. Nonetheless, as shown
(i) the larger the angle of incidence, the higher the energy loss to in Fig 3.5(a) and (d), a sharp decrease (starting from 15 kJ/m) in
the surrounding through reflection; and (ii) dominance of uni- W/Dp demonstrates the fact that the weld penetration depth
form heat conduction in all directions over the z-preferential or increases at a faster rate than the weld width in this range and
axial heat conduction for the lower energy input to the materials establishes a keyhole formation regime. As a result, the weld bead
to be welded. becomes almost cylindrical. Penetration size factor increases with
The contour plots shown in Fig. 3.4(a)–(c) demonstrate the further increase in line energy. This is due to the creation of upper
facts that interactions of higher laser power and slower welding keyhole plasma plume that acts as a point heat source above weld
speed cause wider weld width; deeper weld penetration; and plane. This generated plasma plume acts in the keyhole and forms
longer minimum crack-path. a ‘chalice’ shaped weld bead profile when energy input is more
M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136 131
882
950
2.5 2.5
1030
1110
2.0 2.0
600 700 800 900 1000 600 700 800 900 1000
Laser Power, W Laser Power, W
3.5
307 504
406
3.0
603
2.5 701
2.0
600 700 800 900 1000
Laser Power, W
Fig. 3.4. Contour graphs to show the interaction effects of P and S on (a) weld width, (b) weld penetration depth, and (c) minimum crack path at A¼ 301.
than 20 kJ/m, which is quite similar to the result obtained from an The fit summary includes sequential model sum squares to select
experimental study conducted by Khan et al. [5]. the highest order polynomial where additional terms are signifi-
Besides, as illustrated in Fig 3.5(b), variation in minimum cant and the model is not aliased. In addition, model summary
crack-path with the line energy input shows the same trend as statistics of the fit summary focuses on the model that maximizes
the weld penetration depth. This is because of the existing linear, adjusted R-squared and predicted R-squared values. The sequen-
positive relationship between them as can be seen in Fig. 3.6. tial F-test is carried out using the same statistical software
package to check if the regression model is significant and find
out the significant model terms of the developed models as well.
3.4. Process parameter optimization
The step-wise regression method is also applied to eliminate the
insignificant model terms automatically.
The optimization part in Design-Expert software V7 searches
Suitable response models for the response factors are selected
for a combination of factor levels that simultaneously satisfy the
based on the fit summaries. From fit summary output of the
requirements placed (i.e., optimization criteria) on each of the
measured responses shown in Tables 3.1–3.4, it is evident that
responses and process input factors (i.e., multiple-response opti-
quadratic model is statistically significant for the weld width,
mization). Numerical and graphical optimization methods are
whereas for weld penetration depth; two-factor interaction (2FI)
used in this work by selecting the desired goals for each factor
models are statistically recommended for further analyses.
and response. As mentioned before, the numerical optimization
The test for significance of the regression models and the test
process involves combining the goals into an overall desirability
for significance on individual model coefficients are performed
function (D). The numerical optimization feature in the design-
using the same statistical package. By selecting the step-wise
expert package finds one point or more in the factors domain that
regression method that eliminates the insignificant model terms
maximizes this objective function. In this study, the objective is to
automatically, the resulting ANOVA Tables 3.5 and 3.6 for the
optimize the autogenous laser-welded joints subjected to mini-
selected models summarize the analysis of variance of each
mize the weld width and maximize weld penetration depth a
response and illustrate its significant model terms as well. The
characterizing factor that determines the minimum crack-path as
aforestated tables demonstrate that calculated Fisher’s ‘Model-F’
shown in Fig. 3.6. In order to achieve these objectives, mathema-
and ‘Model-P’ values are, respectively, 57.17 and o0.0001 for
tical models are developed to relate the aforesaid geometrical
weld width quadratic model; and 107.74 and o0.0001 for weld
features and the selected laser welding input variables.
penetration depth 2FI model. These ‘Model-F’ and ‘Model-P’
values imply that the selected models are highly significant and
3.4.1. Development of mathematical models there is only a less than 0.01% chance that these large ‘Model-F’
At this stage, the fit summary in the design-expert software is values could occur due to noise. The associated P value is also
used to select the models that best describe the response factors. used to estimate whether F is large enough to indicate statistical
132 M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136
1400 1400
1000 1000
800 800
600 600
400 400
15°
15° 15°
200 30°
30° 200 30°
45°
45° 45°
0 0
5 10 15 20 25 30 35 5 10 15 20 25 30 35
Line Energy, LE (kJ/m) Line Energy, LE (kJ/m)
1300 3.5
1100
2.5
1000
2.0
900
1.5
800
1.0
700
15°
600 30° 0.5
45°
500 0.0
5 10 15 20 25 30 35 5 10 15 20 25 30 35
Line Energy, LE (kJ/m) Line Energy, LE (kJ/m)
Fig. 3.5. Effect of line energy on (a) weld penetration depth, (b) minimum crack-path, (c) weld width, and (d) penetration size factor for different beam incident angle.
Table 3.1
900 Sequential model sum of squares for weld width model.
700
Mean 4.372E þ 007 1 4.372Eþ 007
Linear 1.325E þ 006 3 4.417Eþ 005 29.28 o0.0001
600 2FI 3042.62 3 1014.21 0.059 0.9808
Quadratic 2.340E þ005 3 77994.39 13.15 0.0001 Suggested
500 Cubic 53388.46 7 7626.92 1.65 0.2363 Aliased
Residual 41493.34 9 4610.37
400 Total 4.538E þ 007 26 1.745Eþ 006
300
Table 3.2
200
30° Model summary statistics for weld width model.
significance. If P value is lower than 0.05, it indicates that the Again, the ANOVA table for the weld width model shows that
model is statistically significant as stated by Zulkali et al. [19]. there is a quadratic relationship between weld width and welding
The same tables also show other adequacy measures e.g., parameters. The linear terms of laser power (P) and welding speed
R-squared, adjusted R-squared, and predicted R-squared values. All (S); and the quadratic terms of welding speed and incident angle are
the adequacy measures are in logical agreement and indicate the significant model terms associated with the weld width. How-
significant relationships. Moreover, adequate precision compares ever, linear term of beam incident angle is added to support hierarchy
range of predicted value at the design points to average prediction of weld width model. For the weld penetration depth model, ANOVA
error. The adequate precision ratios in all cases are dramatically table demonstrates that all three linear terms i.e., laser power (P),
greater than 4 indicating adequate models discrimination. welding speed (S) and beam incident angle; and two-factor
M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136 133
interactions (2FI) of laser power-welding speed (P–S), are the sig- Normality of residual data and amount of residuals in predic-
nificant model terms. tion are then checked to ensure statistical validation of the
From the results shown in Tables 3.1–3.6, it is, therefore, developed models. The normality of data is verified by plotting
apparent that the developed statistical models for predicting weld the normal probability plot (NPP) of residuals. The residual is the
width and penetration depth are fairly accurate and can be of difference between observed and predicted values (or fitted
following forms: value) obtained from the regression model. The data set is
normally distributed if the points on the plot fall fairly close to
(i) Weld width the straight line. The normal probability plots of residual values
for weld width, and penetration depth are illustrated in Fig
ðWÞ1:06 ¼ 2163:33 þ0:76P1154:6S þ 36:97A
3.7(a)–(b), respectively. The experimental points are reasonably
þ154:87S2 0:58006A2 aligned with predicted or fitted points suggesting the normality of
data. This is an implication that empirical distribution of residual
(ii) Weld penetration depth data is well-compared with a normal distribution having the
same mean and variance
Dp ¼ 446:02 þ2:16 P106:24 S þ 7:674 A0:024 P x A
Fig 3.8(a)–(b) are showing the relationships between the
actual and predicted values of weld width and penetration depth.
Since the points plotted are close to and around the diagonal line,
Table 3.3 the difference between the predicted and actual value for each
Sequential model sum of squares for weld penetration depth model. point can be considered to be minimal. It is also an indication that
the statistical models for prediction are adequate and predicted
Source Sum of df Mean F p-value
results are in good agreement with the measured data.
squares square value prob4 F
Table 3.5
ANOVA table for weld width quadratic model.
Table 3.6
ANOVA table for weld penetration depth 2FI model.
99 99
95 95
Normal % Probability
Normal % Probability
90 90
80 80
70 70
50 50
30 30
20 20
10 10
5 5
1 1
-1.70 -0.82 0.07 0.95 1.84 -1.98 -0.85 0.27 1.40 2.53
Studentized Residuals Studentized Residuals
Fig. 3.7. Normal probability plot for weld (a) width, and (b) penetration depth.
1625 1000
Predicted
Predicted
1350 700
1075 400
800 100
856 1108 1360 1613 1865 124.7 407.6 690.5 973.4 1256.3
Actual Actual
Fig. 3.8. Scatter diagrams of weld (a) width, and (b) penetration depth.
Parameters or Limits Importance Criterion Soln no. P (W) S (m/min) F (lm) DP (lm) W (lm)
responses
Lower Upper First Second 1 998.6 2.02 15.2 1248.1 1084.6
2 999.3 2.16 15.1 1235.9 1039.7
P (W) 600 1000 3 Is in range Minimize 3 998.9 2.03 15 1250 1078.5
S (m/min) 2.0 4.0 3 Is in range Maximize 4 997.9 2.00 15 1251.7 1088.6
A (1) 15 45 3 Is in range Is in range 5 999 2.02 15.5 1243.8 1088
W (mm) 584 1218 5 Is in range Minimize 6 999.5 2.04 15.6 1240.4 1082.3
DP (mm) 206 1235 5 Maximize Maximize 7 998.9 2.03 15.2 1247.4 1081.1
8 998.5 2.14 15.2 1235.3 1048.7
9 999.6 2.18 15 1236 1033.9
optimization criteria, the beam incident angle has to be its 10 999.7 2.01 15.8 1240.8 1095
limiting value of 151 to achieve maximum weld penetration
depth and minimum weld width. This result also supports the
discussion made earlier on the effect of beam incident angle on In this case, as shown in Table 3.9, the weld width and penetration
the geometrical features of the weld. depth would be of 665 mm and 706 mm, respectively, which are
Again, Table 3.8 demonstrating the optimal welding conditions much higher than the prerequisite values for the weld. These results
according to the first criterion, it is found that maximum weld also indicate the fact that laser welding ought to be conduction-
penetration depth of 1250 mm is obtained when laser power and limited for this particular joint type and laser-material combinations
welding speed are set to their respective highest and lowest limits. in order to obtain the optimal geometrical features of the weld.
However, with acceptable weld penetration depth, the laser power Since optimal range of laser power and welding speed selected
can be minimized to around 790 W and the highest limit of the based on second criterion is, respectively, much lower and higher
welding speed can be used instead of its lowest limit of 2 m/min. than that obtained for first set of criterion, any combination of process
M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136 135
parameters of the second optimal set would cause less energy input formation of cracks, blow holes, and spatter with a consequent
to constrained butt joints to be made. This reduced energy input to improve in the weld quality. The results given in Table 3.10 also
weld materials would, ultimately, result in less distortion, and support the improvement of weld quality at lower laser power and
higher welding speed.
Table 3.9
Optimal solutions as obtained based on second criterion.
4. Weld microstructure and microhardness
Soln no. P (W) S (m/min) F (lm) DP (lm) W (lm)
A B
Fig. 4.1. (a) typical micrograph of laser welding of AISI440C and AISI430F stainless steels, and base metal (BM), heat affected zone (HAZ) and Fusion zone (FZ) of (b) inner
shell and (c) outer shell.
136 M.M.A. Khan et al. / Optics & Laser Technology 49 (2013) 125–136
Intersection
500 Line weld.
| Whatever the optimization criteria, the beam incident angle
has to be around its limiting value of 151.
400
’ Efficient and low-cost weld joints could be achieved using the
AISI AISI welding conditions drawn from the numerical optimization
440C 430F
300
Acknowledgements