Int J Adv Manuf Technol (2016) 86:1339–1346

DOI 10.1007/s00170-015-8298-1

ORIGINAL ARTICLE

A FEM model to study the fiber laser welding of Ti6Al4V

thin sheets
Giuseppe Casalino 1 & Michelangelo Mortello 1

Received: 8 September 2015 / Accepted: 21 December 2015 / Published online: 9 January 2016
# Springer-Verlag London 2016

Abstract This paper describes a finite element model for fi- temperature properties. Peters et al. pointed out that the out-
ber laser beam welding of Ti6Al4V in butt configuration. Full standing properties of Ti alloys are exploited for many aero-
penetration keyhole welding of 2-mm-thick plates was per- nautic and aerospace applications, including airframe, engine,
formed by adopting a Yb-YAG fiber source. Neither filler wire and helicopter components [1]. Moreover, thanks to its bio-
nor groove preparation was adopted to perform the seam. compatibility, Ti is also widely demanded for bio-engineering
Various welding conditions were considered in order to devel- applications, mostly in dentistry and orthodontics [2]. As
op a versatile and simple model for the heat source. Thus, the discussed by Veiga et al., the possibility of varying the prop-
study was carried out by varying the modeling strategy of the erties by either adding alloying elements or thermo-
thermal source as a function of the calculated expected seam mechanical processing has made α-β titanium alloys the most
geometry. ANSYS parametric design language (APDL) pro- desirable materials in numerous fields [3]. Henriques present-
gramming was used for the generation of the numerical code. ed a review on titanium metallurgy, including ingot
The calibration of the model involved both cross-section ge- manufacturing, re-melting, and powder technology [4].
ometry and thermal cycle beside the weld seam. The accuracy Laser welding is the most suitable fusion technique for joining
was improved by simulating the thermal contact conductance metals. This is due to both production and quality aspects. In
at the interface. Transient temperatures and geometric charac- fact, on one hand, the process presents high productivity and
teristics of the weld zone were calculated. Numerical results flexibility for various weld geometries and configurations,
were in good agreement with experimental ones. while, on the other hand, narrow bead, deep penetration, high
precisions, and low distortions are favored thanks to the high
Keywords Laser welding . FEM . Fiber laser . Titanium alloy energy density in a small spot size. Gao et al. demonstrated
that, in comparison with traditional GTA welding, a laser
source both increases the strength and ductility and reduces
1 Introduction the overall residual distortions, the grain size, and the width of
altered zones [5]. In particular, fiber source boasts desirable
Titanium and its alloys are currently employed for many in- optical properties in comparison with other commercially
dustrial purposes, for which more and more stringent chemi- available lasers, such as high beam quality, low divergence,
cal, physical, and mechanical requirements are demanded. In and wavelength. Hence, energy is highly absorbed, and deep
fact, Ti exhibits high strength-to-weight ratio, fracture tough- penetration is reached in many metal alloys. Many works have
ness, corrosion resistance, good fatigue behavior, and high been conducted on the optimization of laser-welded Ti6Al4V.
Schneider e al. improved the stability of the process by opti-
mizing the shielding gas flow rate leading to the suppression
* Michelangelo Mortello of the vapor plume [6]. Akman et al. improved the weld ge-
[email protected]
ometry by controlling the laser output parameters [7]. Also,
Squillace et al. investigated on geometrical defects and me-
1
Politecnico di Bari, Dipartimento di Meccanica Matematica e chanical properties for different welding conditions [8]. Gao
Management (DMMM), Viale Japigia 182, Bari 70126, Italy et al. promoted the evacuation of pores from the weld pool by
1340 Int J Adv Manuf Technol (2016) 86:1339–1346

Table 1 Chemical composition

of the as-received material (wt%) Ti Al H Fe O N C V W Other

Ti6Al4V Balance 6.10 0.01 0.05 0.20 0.05 0.10 4.00 0.30 0.40

optimizing the pulse parameters of the laser source [9]. observations with good accuracy, which allows to drawing
Another work by Casalino et al. aimed to compute the signif- inference on the thermal cycle and metallurgical transforma-
icance of several parameters and their interactions on bead tion inside the weld.
characteristics by artificial neural network (ANN) [10].
Full comprehension of a technological process can benefit
from a numerical model. The simulation reduces times for
analyzing, cost of experiments, and material consumption. 2 Experimental investigation
The numerical simulation of welding process is not an easy
task. It involves the interaction of thermal, mechanical, and 2.1 Evaluation of the expected seam shape
metallurgical phenomena and enables the estimation of weld
pool geometry, transient temperatures, residual stresses, and Experimental investigation was performed to calibrate and to
others. An essential presentation on the theoretical founda- verify the numerical model. Two millimeter thick Ti6Al4V
tions of the finite element method (FEM) of weld phenomena α-β titanium alloy plates were autogenously fiber laser
is in Olabi’s book chapter [11]. This aims to find quantitative welded in butt configuration. The geometry of the plates
solutions through the analysis of complex theoretical phenom- (width × length × thickness) was 50 × 100 × 2 mm. Chemical
ena. Zhang et al. adopted a numerical simulation to explain in composition, mechanical behavior, and physical properties
details the molten pool behavior and keyhole dynamics during of the as-received alloy are reported in Tables 1, 2, and 3,
laser full penetration welding [12]. Tsirkas et al. exploited the respectively. A 200-μm diameter optical Yb fiber laser, with
temperature-dependent material properties and the continuous a maximum average output power of 4000 W, was used in
cooling transformation (CCT) diagram to predict metallurgi- continuous wave regime. A 400-μm diameter (1/e2 width)
cal transformations [13]. Thus, optimal process parameters are near-Gaussian distribution laser spot was produced on the
obtained with a reduced number of experiments in order to top surface of the workpiece. The beam was delivered by
save materials, time, and resource. Kazemi et al. carried out an collimating lens and focusing lens in which focal lengths were
accurate description of the heat source and indicated the most 120 and 250 mm, respectively.
sensitive parameters that influence the bead shape [14]. Welding conditions determine the morphology of the weld
Aloraier and Joshi examined the temperature distribution, zone and its geometric characteristics, which mostly depend
residual stresses, and distortions through the simulation of of four factors:
the flux current arc welding using the software Sysweld
[15]. Another study was also conducted by Akbari et al., & Process parameters adopted
which determined the temperature distribution in the seam & Joint preparation
with the purpose to predict the geometric characteristics of & Geometry and configuration
the bead for various welding speed values [16]. Mackwood & Environmental conditions
and Crafer published a thermal simulation overview of
welding processes [17]. In particular, this study was aimed to predict the seam
In this work, a finite element model was implemented to shape for assigned combinations of process parameters and
simulate the autogenous fiber laser welding of 2-mm-thick joint geometry. Although it is difficult to establish precisely
Ti6Al4V plates in butt configuration. 3D FEM model was the bead geometry without experiments, the type of seam
developed by using the parametric design language (APDL) shape is easily pre-determinable. The modeling strategy,
available in the ANSYS finite element code. It consists of a which depended on the expected type of seam shape, resulted
scripting language that encompasses a wide range of features, simple and versatile. Since the heat transfer in the weld pool
such as if-then-else constructs, do-loops, and algebraic opera- mostly occurs by convection, the bead shape is determined by
tions with vectors and matrices. The heat source model
Table 2 Mechanical properties of the as-received material: ultimate
depended on the preliminary evaluation of the expected type tensile strength (UTS), yield stress (YS), Young’s modulus (E),
of seam shape. Thermal fields and temperature cycles were elongation to fracture % (A%), Vickers microhardness (HV)
recorded for the validation of the numerical model, which was
UTS (MPa) YS (MPa) E (GPa) A% HV
completed with the comparison between experimental mor-
phology and numerical temperature fields in the cross section. Ti6Al4V 950 880 114 14 349
It was found that numerical results approached experimental
Int J Adv Manuf Technol (2016) 86:1339–1346 1341

Table 3 Thermo-physical properties of the as-received material:

thermal conductivity (K), melting temperature (Tm), density (ρ)

K (W/(m K)) Tm (°C) ρ (g/cm3)

Ti6Al4V 6.7 1650 4.43

the synergy of different physical phenomena related to weld

pool fluid dynamics. As demonstrated by Squillace et al. [8],
the main influent physical phenomena are the surface tension,
vapor pressure, volume reduction, gravity, and phase transfor-
mation. The heat input is the most significant parameter that Fig. 2 The most significant geometric features of the seam geometry
determines the bead morphology; it represents the energy sup-
ply to the workpiece. In general, the higher the heat input, the
diffusivity). The results of preliminary tests provided a more
deeper is the penetration. If the heat input does not exceed the
detailed knowledge of how the seam shape changes as a func-
threshold limit beyond which the excessive penetration oc-
tion of the heat input. The following geometric features (also
curs, the keyhole can provide different bead shapes, as a func-
shown in Fig. 2) were considered the most significant for the
tion of the dynamic behavior of the weld pool. As the thermal
simulation:
input rises, temperature in the lower part of the weld pool
increases leading to a lower surface tension of the liquid wall
& Thickness of the sheet th.
of the keyhole. Since the plasma-gas pressure in the keyhole is
& Width of the molten zone at the upper surface W1.
more weakly hindered by the surface tension, the lower part of
& Minimum width of the molten zone W2.
the fusion zone extends in width. Moreover, during the source
& Distance between the upper surface and the minimum
motion, the overheating of the rear bottom part of the weld
width p.
pool involves clockwise vortex, which favors the growth of
the lower zone. Casalino et al. demonstrated and explained the
Aspect ratio parameters ARi were defined by the Eqs. 1, 2,
same effect [18]. Therefore, as the thermal input rises, the
and 3.
weld undergoes a root enlargement. Differently, as the laser
energy decreases, the keyhole size stays stable and the surface AR1 ¼ W 1=th ð1Þ
tension counterbalances the recoil pressure force at the bottom
of the weld pool, which tends to the V-shaped geometry. AR2 ¼ W 2=th ð2Þ

Figure 1 shows the 2D cross-section geometries of the fused AR3 ¼ h th ð3Þ
zone at a glance, as a function of the thermal input.
The picture in Fig. 1 is only aimed to provide a schematic The heat input influences the seam shape and the aspect
representation without taking into account quantitative param- ratios ARi. Since these parameters do not define the seam
eters, which are useful to define the seam geometry in details. geometry univocally, the level of accuracy will be dependent
Thus, preliminary calibration tests were conducted to evaluate on the number of geometric parameters taken into account. In
the thermal behavior of the material, which depends on its the present study, it was considered that 3 is the suitable num-
physical properties (mainly boiling point, specific heat, and ber for a reasonable level of accuracy.

Fig. 1 Variation of the expected

type of seam shape with
increasing the thermal input
1342 Int J Adv Manuf Technol (2016) 86:1339–1346

Table 4 Process parameters

adopted Sample Laser power (kW) Welding speed (m/min) Linear energy (J/mm) UTI (J/mm4)

1 1.20 2.50 28.8 90.0

2 1.20 2.00 36.0 112.5
3 1.20 1.50 48.0 150.0
4 1.20 1.00 72.0 225.0

2.2 Experimental procedure 3 Numerical model

An experimental analysis was conducted to validate and 3.1 Assumptions and boundary conditions
verify the goodness of the numerical model. The choice
of the values of process parameters was made in such a The following hypotheses and boundary conditions were ap-
way that all the types of seam presented in Fig. 1 could plied for the model.
be investigated. Preliminary bead-on-plate trial tests de-
fined a suitable range of variability of the process param- & Thermo-physical properties of the materials were isotropic
eters, which lead satisfactory penetration and appearance. and temperature-dependent according to standard values
Thus, the experimental plan was developed at constant la- found in literature. The behavior at the molten state was
ser power P by varying the welding speed v at four levels. estimated by experimental results.
Table 4 reports the values of the process parameters & Energy losses due to slow air convection were equal to
adopted. The table also indicates the linear energy LE 20 W/m2 K.
and the unit thermal input UTI, as defined in Eqs. 1 and & A convective heat transfer coefficient of 200 W/m2 K was
2, respectively. As will be mentioned later, they are strictly associated to the thermal exchange between the bottom
related to both the seam shape and modeling strategy. surface of the workpiece and the support. This value was
determined according to experimental and numerical pre-
LE ¼ P=v ð4Þ liminary trials.
& The starting temperature was assumed equal to 20 °C.
LE
UTI ¼ ð5Þ
D2 *th
3.2 Geometry and meshing
where D is the laser spot diameter and th is the sheet thick-
ness. Therefore, LE and UTI are technological parameters The analysis was carried out by adopting an eight-node qua-
that aim to express the amount of energy supplied during dratic 3D solid element SOLID70. Convergence test was con-
the welding and to predict the seam shape. In fact, LE ducted to find a suitable number of elements, resulting from a
stands for the energy transmitted to the workpiece (i.e., compromise between computing time and accuracy. The do-
the thermal input), whereas UTI takes into account the main partition had a nonuniform grid system, which was com-
thickness of the sheet and the spot area, since they signif- posed of three different zones. In fact, a finer mesh was
icantly influence the bead geometry. adopted close to the welding centerline (the less the distance

Fig. 3 Mesh of the model and

element size
Int J Adv Manuf Technol (2016) 86:1339–1346 1343

Fig. 4 Node selection in the

modeling strategy as a function of
the expected type of seam shape
and aspect ratios

from the welding centerline, the more the element size), in load was the ratio of the total power ideally transmitted Ptot
order to ensure the desirable precision (Fig. 3). The thickness (without considering energy losses) to the volume of each
was divided into eight elements. element.
:
Q ¼ Ptot=V ð7Þ
3.3 Thermal source
UTI parameter defined the modeling strategy. Figure 4
The laser welding process worked in the keyhole regime. shows how the modeling of the beam worked as a function
The modeling strategy of the source can be summarized by of UTI and aspect ratio Ri. The node selection varied with
the following steps: them.
The source motion was simulated by adopting DO* cycle
& Determination of the aspect ratio parameters ARi. loop in a finite number of subsequent steps with the aim to
& Calculation of UTI. provide a nodal temperature history for each node. The ther-
& Application of the thermal load. mal load was associated at the beginning of each step, and
temperature fields were re-calculated at every step starting in
As mentioned before, the definition of ARi and UTI is function of the previous step. The ratio of the step increment to
required for evaluating how to apply the source action, as a the welding speed determined the time required for each step.
function of thermal and geometric aspects. The geometric The thermal cycle and the temperature field in the whole
characterization of the welding heat source can be also found workpiece were determined for four different welding condi-
in Goldak’s double ellipsoidal model [19]. However, in this tions by combining the experimental result with the numerical
work, the power distribution was defined in function of the computation. Since the bead shape depended mostly on
expected seam shape and aspect ratios.
An internal production of thermal load was associated to
several specific elements close to the welding centerline. The
classic governing partial differential equation expresses the
transient heat conduction (Eq. 6).
      
∂2 T ∂2 T ∂2 T ̇ ∂T
k ðT Þ þ þ þ Q ¼ ρðT ÞC p ðT Þ
∂x2 ∂y2 ∂z2 ∂t
ð6Þ

where T is the punctual temperature, k is thermal conductivity,

Q is the heat input, and Cp is the specific heat.
As concerns the element SOLID70, the numerical code
expresses the heat input in terms of heat generation rate, which
is the input to the node of the element. As shown by Eq. 7, the Fig. 5 Position of thermocouples for calibrating the numerical model
1344 Int J Adv Manuf Technol (2016) 86:1339–1346

Fig. 6 Temperature in the cross section of sample 1 (expected seam Fig. 8 Temperature in the cross section of sample 3 (expected seam
shape A) shape C)

temperature, comparing the calculated and the experimental thermal behavior in the workpiece, since it only gives the
thermal field was one of the criteria for model validation. general feature of the weld. Therefore, the validation was
completed by the experimental calibration of the model.
3.4 Thermal contact resistance This was carried out by measuring the thermal cycles in the
workpiece. Both the geometry of the sheet and the system for
The contact between the sheets is an element of major impor- temperature detection are represented in Fig. 5.
tance to an accurate investigation. Thus, 3D thermal surface- As shown, thermal cycle for validation purpose was detect-
to-surface contact pairs (CONTA174 and TARGE170) were ed by positioning three thermocouples onto each sheet at
adopted to mesh the interface layer and simulate the contact 1.5 mm from the welding centerline. The detected values were
between target surfaces and a deformable surface. These ele- used to calibrate the model and provide the thermal cycles in
ments were selected to consider the large deflection and non- the workpieces. After comparing experimental results with
linear behavior of contact asperities. The conductive heat numerical ones, a coefficient Kcal was introduced to take into
transfer between two contacting surfaces is defined by Eq. 8. account energy losses and actual effects (Eq. 9).
: :
Qcont ¼ T CC*ðT T −T C Þ ð8Þ Qcorr ¼ Q *K cal ð9Þ
:
where Qcont is the thermal flow for a unity of contact area, Finally, the corrected value of heat generation rate Qcorr ,
TCC is the thermal contact conductance coefficient, and which takes into account the energy losses, was validated
(TT −TC) is the difference between the temperatures of the experimentally and then input in the model.
contact points on the target and contact surfaces
(TARGE170 and CONTA174).
4 Results and discussion
3.5 Model calibration
The numerical model was validated by comparison between
The preliminary analysis on the expected type of seam shape experimental and numerical results. Calibration factors were
is not sufficient for providing an accurate simulation of the determined with the aim to correct the amount of load

Fig. 7 Temperature in the cross section of sample 2 (expected seam Fig. 9 Temperature in the cross section of sample 4 (expected seam
shape B) shape D)
Int J Adv Manuf Technol (2016) 86:1339–1346 1345

analysis of the numerical curves shows that heating and

cooling rates, peak temperatures, and starting time required
for a sensitive increase in temperature were in accordance with
the experimental conditions. The slight disagreement between
the curves can depend on incorrect assumptions, boundary
conditions, and source modeling. However, the simulation
aimed to reproduce a very complex behavior into a simple
and versatile representation. Since temperatures presented a
similar trend when they were compared to each other, the
simulation was in accordance with the experiment. The model
Fig. 10 Experimental thermal cycles at a distance of 1.5 mm from the implemented was highly efficient thanks to its versatility, sim-
welding line for various LE
plicity, and short computing time.
associated to each element of the model. As shown in Figs. 6,
7, 8, and 9, the cross sections of specimens were compared to
5 Conclusions
the corresponding temperature fields. The legend on the left
side indicates the scale of temperature. Both the experimental
This paper contains a FEM study of fiber laser beam welding
and numerical distributions were taken from the middle length
of Ti6Al4V alloy. The cross section for 2-mm-thick butt weld
of the joints.
in different welding condition was calculated. The following
FZ and HAZ temperature were compared with those ob-
list summarizes the main achievements of the investigation.
tained by calculation. The results were in good agreement.
The cross-sectional side view exhibited large temperature gra-
& The heat source as a time-dependent internal heat genera-
dients in the weld zone. A slight asymmetry about the mid
tion provided a simple and rapid tool. The model had the
thickness of the cross section formed.
capability to simulate various conditions, by varying di-
The numerical and the experimental thermal cycles were
mensions, material properties, and welding parameters.
plotted in Figs. 10 and 11, respectively. They were detected at
& A new technological measurable parameter (UTI) permit-
a distance of 1.5 mm from the welding centerline at the middle
ted to select the most suitable model for the thermal
length (50 mm). The representation of cooling phase was bro-
source. The determination of aspect ratios improved the
ken off at a time of 10 s. The following steps were not relevant
accuracy of the modeling strategy.
for both the model calibration and the microstructural
& Transient temperatures and geometric characteristics of
investigation.
the weld zone were calculated. Since the numerical results
The peak temperatures increased with the LE. As the ther-
were in good agreement with experimental observations
mal load increased, the source over-heated the material sur-
for both thermal cycles and cross-section geometry, the
rounding the welding zone and the heat disposals get reduced.
model was capable to predict correctly the temperature
Also, the time to reach the peak increases with the LE certain-
during the fiber laser welding of titanium.
ly because the laser source needed more time to arrive in
& The knowledge of the right thermal cycle enhances the
proximity to the middle length. On the other hand, the cooling
possibility to predict the metallurgy of the weld before
rates were slightly influenced by the LE.
welding.
The experimentally detected curves were in good agree-
ment with those computed by the numerical model. The
Acknowledgments This work was performed at the “Intelligent Com-
puting for Manufacturing Processes and Systems” laboratory of DMMM
Politecnico of Bari (Italy).

References

1. Peters M, Kumpfert J, Ward CH, Leyens C (2003) Titanium alloys

for aerospace applications. Adv Eng Mater 5(6)
2. Elias CN, Lima JHC, Valiev R, Meyers MA (2008) Biomedical
applications of titanium and its alloys. Biol Mater Sci Overview
3. Veiga C, Davim JP, Loureiro AJR (2012) Properties and applica-
tions of titanium alloys: a brief review. Adv Mater Sci 32:133–148
Fig. 11 Numerical thermal cycles at a distance of 1.5 mm from the 4. Henriques VAR (2009) Titanium production for aerospace applica-
welding line for various LE tions. J Aerosp Technol Manag 1
1346 Int J Adv Manuf Technol (2016) 86:1339–1346

5. Gao XL, Zhang L-J, Liu J, Zhang J-X (2013) A comparative study plate with high power high brightness laser. J Mater Process
of pulsed Nd:YAG laser welding and TIG welding of thin Ti6Al4V Technol 214:1710–1720
titanium alloy plate. Mater Sci Eng 559:14–21 13. Tsrikas SA, Papanikos P, Kermanidis T (2003) Numerical simula-
6. Schneider A, Gumenyuk A, Lammers M, Malletschek A (2014) tion of the laser welding process in butt-joints specimens. J Mater
Laser beam welding of thick titanium sheets in the field of marine Process Technol 134:59–69
technology. Phys Procedia 56:582–590 14. Kazemi L, Goldak JA (2009) Numerical simulation of laser full
7. Akman E, Demir A, Canel T, Sınmazcelik T (2009) Laser welding penetration welding. Comput Mater Sci 44:841–849
of Ti6Al4V titanium alloys. J Mater Process Technol 209:3705– 15. Aloraier AS, Joshi S (2012) Residual stresses in flux cored arc
3713 welding process in bead-on-plate specimens. Mater Sci Eng 534:
8. Squillace A, Prisco U, Ciliberto S, Astarita A (2012) Effect of 13–21
welding parameters on morphology and mechanical properties of 16. Akbari M, Saedodin S, Toghraie D, Shoja-Razavi R, Kowsari F
Ti-6Al-4V laser beam welded butt joints. J Mater Process Technol (2014) Experimental and numerical investigation of temperature
212:427–436 distribution and melt pool geometry during pulsed laser welding
9. Gao X-L, Zhang L-J, Liu J, Zhang J-X (2014) Porosity and micro- of Ti6Al4V alloy. Opt Laser Technol 59:52–59
structure in pulsed Nd:YAG laser welded Ti6Al4V sheet. J Mater
17. Mackwood AP, Craferb RC (2005) Thermal modelling of laser
Process Technol 214:1316–1325
welding and related processes: a literature review. Opt Laser
10. Casalino G, Curcio F, Memola Capece Minutolo F (2005)
Technol 37:99–115
Investigation on Ti6Al4V laser welding using statistical and
Taguchi approaches. J Mater Process Technol 167:422–428 18. Casalino G, Mortello M, Campanelli SL (2015) Ytterbium fiber
11. Olabi AG, Casalino G (2014) Mathematical modeling of weld phe- laser welding of Ti6Al4V alloy. J Manuf Process 20:250–256
nomena, part 1: finite-element modeling. Compr Mater Process 6: 19. Joshi S, Hildebrand J, Aloraier AS, Rabczuk T (2013)
101–109 Characterization of material properties and heat source parameters
12. Zhang LJ, Zhang JX, Gumenyuk A, Rethmeier M, Na SJ (2014) in welding simulation of two overlapping beads on a substrate
Numerical simulation of full penetration laser welding of thick steel plate. Comput Mater Sci 69:559–565