Professional Documents
Culture Documents
www.elsevier.com/locate/ijhmt
Abstract
This paper presents an analytical solution of the transient temperature distribution in a finite solid when heated by a moving heat
source. The analytical solution is obtained by solving the transient three-dimensional heat conduction equation in a finite domain by
the method of separation of variables (SOV). Meanwhile previous studies focus on analytical solutions for semi-infinite domains, here
an analytical solution is provided for a finite domain. The non-homogeneous equation is solved by using the Laplace transform for a unit
impulse and then convoluted with the actual heat source. Two different distributions are used: a Gaussian distribution and a spatially
uniform plane heat source.
2006 Elsevier Ltd. All rights reserved.
0017-9310/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2006.03.026
G. Araya, G. Gutierrez / International Journal of Heat and Mass Transfer 49 (2006) 4124–4131 4125
Nomenclature
by the pulse laser period and divided by the laser radius) Fig. 1). As the laser progresses with a velocity U, the heat
must be much lower than the non-dimensional velocity in from the source penetrates further into the workpiece.
order to neglect lateral conduction. The following assumptions are considered:
Komanduri and Hou [7] obtained a general solution for
the laser surface hardening process using a disk heat source • Heat losses by radiation are negligible as compared to
with a pseudo-Gaussian heat intensity distribution in a the intensity of the incident laser beam (Gutierrez and
semi-infinite steel workpiece of finite width. They inferred Araya [14], Modest and Abakians [15]).
that this exact analytical solution provided a better insight • Thermal properties are considered constant and evalu-
into the physical process of laser transformation hardening ated at an average temperature.
of steels than the semi-empirical equations developed by
Steen and Courtney [8] based on the regression analysis With the above assumptions, the governing equation is the
of the experimental data. transient heat conduction equation, that written in terms of
Stephenson et al. [9] developed a transient, three-dimen- h = T T0, yields:
sional analytical model for predicting cutting tool temper-
1 oh o2 h o2 h o2 h q000 ðx; y; z; tÞ
ature through the separation of variable method, ¼ þ þ þ ð1Þ
considering a fixed heat source on the surface of the insert. aT ot ox2 oy 2 oz2 kT
Comparison with measured temperatures showed good where kT and aT are the thermal conductivity and thermal
agreement. diffusivity of the workpiece, respectively, and q000 (x, y, z, t) is
Experimental data and numerical modeling regarding the heat generation source term.
temperature distribution in a sample of silicon nitride are
encountered [10–13] at different laser powers and cutting 2.1. Heat source modeling
speeds in a LAM process. According to these investiga-
tions, the temperature near the cutting tool location must The heat source term q000 (x, y, z, t) is considered as a laser
be in excess of approximately 1000 C to successfully beam of temporal continuous wave (CW) and spatially
machine silicon nitride material. modeled by assuming two different distributions: a spa-
Most of the analytical solutions encountered in the liter- tially uniform plane heat source and a Gaussian distribu-
ature are based on semi-infinite domains and quasi-steady tion (corresponding to the theoretical TEM00 mode of
states for a constant source speed. To the best of our the laser). TEM comes from the acronym transverse elec-
knowledge, there is not in the literature a transient, three-
dimensional analytical solution for a complete finite
domain describing the temperature evolution due to a Laser beam
space-dependent moving heat source. For this reason, the source U
present paper attempts to fill the gap above-mentioned in x
order to validate a previous numerical model, analyze the
boundary effects, investigate start-up effects and evaluate Lz
the influence of parameters involved in the LAM process,
Workpiece
mainly power and speed of the heat source. y
2. Mathematical modeling z
Lx
Ly
Consider a laser beam source used to pre-heat a parallel-
epiped workpiece of finite dimension Lx, Ly and Lz (see Fig. 1. Schematic of the laser beam and the finite workpiece.
4126 G. Araya, G. Gutierrez / International Journal of Heat and Mass Transfer 49 (2006) 4124–4131
tromagnetic mode, specifying the number of nodes gener- The Gaussian heat flux distribution is imposed as:
ated by a slight misalignment of the mirrors located in 8
kðx2 þy 2 Þ=r20
>
< I 0 e pffiffiffiffiffiffiffiffiffiffiffiffiffiffi @ z ¼ 0; Ut r0 < x < Ut þ r0 ; and
the laser cavity. Fig. 2 shows some transverse modes and 00
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the simplest mode, TEM00, is considered in this paper. q ¼ r0 x < y < þ r20 x2
2 2
>
:
The purpose is to compare the temperature profiles cal- 0 otherwise
culated by employing both heat source models with the ð4Þ
same net power and the same heating area, since a constant
In both cases, the volumetric heat generation is defined as
heat flux model would diminish enormously the calculation
q000 = q00 d(z), being d(z) the Dirac delta function.
effort.
The spatially uniform distribution takes the form of a
2.2. Initial and boundary conditions
rectangular constant energy density:
q0
@ z ¼ 0; Ut < x < Ut þ 2R; and S < y < þS Boundary conditions at the six faces of the parallelepi-
q00 ¼ 4RS
0 otherwise ped domain and an initial condition have to be provided
ð2Þ (@ t = 0, h = 0).
Insulated conditions were imposed at x = 0 and y = 0
Here, q0 is the heat flow absorbed by the workpiece, R and (condition of symmetry), meanwhile constant surface con-
S are the half-size of the laser beam along x and y, respec- ditions were considered at x = Lx, y = Ly/2 and z = Lz.
tively. Finally, U is the laser beam velocity. The main reason of selecting Dirichlet conditions at faces
The irradiance distribution of the Gaussian TEM00 x = Lx, y = Ly/2 and z = Lz can be summarized as follows:
beam can be described as follows: the numerical results obtained in a previous investigation
IðrÞ ¼ I 0 ekr
2
=r20
ð3Þ [14] showed that the above mentioned workpiece faces
2
always reached ambient temperature no matter which
where I0 is the intensity at the center of the spot [W/m ], k boundary condition was established, due to the extremely
is the constant (in general, a value of 2 is used for a concentrated heating effect of the laser source and dimen-
Gaussian model). sions of the workpiece. Two different boundary conditions
The radial coordinate r is related to the Cartesian coor- are analyzed at the top surface: Case 1 considers an insu-
dinates x and y, through: lated condition except on the heating zone powered by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ x2 þ y 2 the laser, and Case 2 regards a convection boundary condi-
rffiffiffiffiffiffiffiffiffiffiffiffiffi tion, instead.
4 The non-homogeneous boundary value problem of Eq.
r0 ¼ ðRSÞ : beam waist radiusðmÞ
p (1) can be solved by assuming that h(x, y, z, t) is expressed
In this way, the circular area for the Gaussian distribution as a series expansion of the eigenfunctions:
XXX
is the same as the rectangular area for the uniform heat flux hðx; y; z; tÞ ¼ Hijk ðtÞX i ðxÞY j ðyÞZ k ðzÞ ð5Þ
distribution. i j k
Z Lx Z Ly =2 Z Lz dHijk
q000 X i ðxÞY j ðyÞZ k ðzÞdx dy dz þ xijk Hijk ¼ Xijk ðtÞ ð18Þ
0 Ly =2 0
dt
Z Lx Z Ly =2 Z Lz With the following initial condition:
¼ /ijk ðtÞ X 2i ðxÞdx Y 2j ðyÞdy Z 2k ðzÞdz XXX
0 Ly =2 0 hðx; y; z; 0Þ ¼ Hijk ð0ÞX i Y j Z k ¼ 0 ð19Þ
Z Lz i j k
Lx Ly
¼ /ijk ðtÞ Z 2k ðzÞdz ð11Þ
4 0 Since Xi(x),Yj(y) and Zk(z) are not zero for arbitrary values
000
Since q is known, the function /ijk(t) is also known. The of x, y and z, it follows that Hijk(0) = 0. For solving Eq.
triple integral on the left hand of Eq. (11) is solved differ- (18), we consider first an impulse applied at t = 0;
ently according to the two heat source distributions consid- dHijk
ered. Also, integration of the eigenfunctions Zk(z) depends þ xijk Hijk ¼ dðtÞ ð20Þ
dt
on the case analyzed: insulated or convective conditions at
Taking the Laplace transform to Eq. (20):
the top surface.
1
sH ðsÞ þ xijk H ðsÞ ¼ 1 ) H ðsÞ ¼ ð21Þ
2.3. Spatially uniform plane heat source and insulated top s þ xijk
surface where H(s) is the transfer function. Performing the corre-
sponding inverse Laplace transform of (23),
In this case, integration of Eq. (11) is straightforward hðtÞ ¼ ‘1 ½H ðsÞ ¼ exijk t . Now, superimposing the function
and an analytical expression for /ijk(t) can be obtained: Xijk(t), the response is the convolution of h(t) with Xijk(t):
Z Lx Z Ly =2 Z Lz Z t
q 8
/ijk ðtÞ ¼ 0 X i ðxÞdx Y j ðyÞdy dðzÞZ k ð0Þdz Hijk ðtÞ ¼ hðtÞ Xijk ðtÞ ¼ hðt sÞ Xijk ðsÞds ð22Þ
4RS Lx Ly Lz 0 Ly =2 0 0
ð12Þ For the uniform heat flux distribution, the integral (22)
results:
Integrating the eigenfunctions;
Z Lx Z X s þ2R X þ2R ðxijk A þ exijk t B þ ai UCÞ
sinðai xÞ s Hijk ðtÞ ¼ 2aT q0 ð23Þ
X i ðxÞdx ¼ cosðai xÞdx ¼ ð13Þ D
0 Xs ai X s where
where Xs is the source position each instant of time, i.e., the A ¼ cosðbj S þ ai Ut þ 2ai RÞ cosðbj S ai Ut 2ai RÞ
product of the source velocity, U, by time, t. þ cosðbj S ai UtÞ þ cosðbj S þ ai UtÞ
Z Ly =2 Z S
2 B ¼xijk ½cosðbj S 2ai RÞ cosðbj S þ 2ai RÞ
Y j ðyÞdy ¼ cosðbj yÞdy ¼ sinðbj SÞ ð14Þ
Ly =2 S b j 2ai U sinðbj SÞ½cosð2ai RÞ 1
Z Lz Z Lz
h C ¼ sinðbj S þ ai Ut þ 2ai RÞ þ sinðbj S ai Ut 2ai RÞ
dðzÞZ k ð0Þdz ¼ dðzÞ cosð0Þ þ sinð0Þ dz
0 0 k T ck sinðbj S ai UtÞ sinðbj S þ ai UtÞ
Z Lz
¼ dðzÞdz ¼ 1 ð15Þ D ¼bj ai Lx Ly Lz SRk T a2T a4i þ 2a2T a2i b2j þ 2a2T a2i c2k
0
In expression (15), the same result is obtained for Cases 1 þa2T b4j þ 2a2T b2j c2k þ a2T c4k þ a2i U 2
and 2 because the sinusoidal function is null at the surface,
The temporal function (23) together with the eigenfunc-
i.e., at z = 0.
tions (6)–(9) constitutes the temperature distribution due
Finally, /ijk(t) is determined by substituting Eqs. (13)
to a moving constant heat flux dictated by Eq. (5).
and (14) into Eq. (12):
4q0 2.4. Gaussian distribution and insulated top surface
/ijk ðtÞ ¼ sinðbj SÞ½sinðai ðX s þ 2RÞÞ sinðai ðX s ÞÞ
RSLx Ly Lz bj ai
ð16Þ A similar procedure is followed; nevertheless, the heat
power now depends on the x and y coordinates, and cannot
Substituting Eq. (16) into Eq. (10) and inserting it into Eq. be removed out of the integral (11). For the insulated top
(1) together with Eq. (5), an ordinary differential equation surface case, Eq. (11) can be written as follows;
for Hijk(t) is obtained: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z xc þr0 Z r2 ðxxc Þ2 ½ðxxc Þ2 þy 2
0 k
1 dHijk /ijk I e r2
cosðai xÞ cosðbj yÞdxdy
¼ ða2i þ b2j þ c2k ÞHijk þ ð17Þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi o o
Defining xijk ¼ ða2i þ b2j þ c2k ÞaT and Xijk ðtÞ ¼ akTT /ijk ðtÞ, Eq. Lx Ly Lz
¼ /ijk ðtÞ ð24Þ
(17) results; 8
4128 G. Araya, G. Gutierrez / International Journal of Heat and Mass Transfer 49 (2006) 4124–4131
profiles a good agreement can be appreciated. Regarding 900 Net power : 50 Watts
start-up effects, it is established that a distance of approxi- Laser speed : 0.1 m/s
800 Laser position : 0.75 L x
mately three source lengths is necessary to reach the quasi- 700
steady state condition.
600
Influence of lateral boundaries (y = ±Ly/2) can be
neglected if they are located farther than, at least, one 500
source length from the laser beam path; this means that 400
300
200
Table 1 100
Thermophysical properties silicon nitride (Si3N4)
0
Melting point (C) 1900 0 0.25 0.5 0.75
Density (kg/m3) 2400
x* = x/L x
Specific heat Cp (J/kg C) @ 1000 C 1278.96
Fig. 4. Comparison of the analytical solution with numerical predictions
Thermal conductivity (W/mC) @ 1000 C 5.91
[14] along the x-direction over the surface (z = 0) and at the plane of
Thermal diffusivity (m2/s) @ 1000 C 1.9 · 106
symmetry (y = 0).
G. Araya, G. Gutierrez / International Journal of Heat and Mass Transfer 49 (2006) 4124–4131 4129
1000 1000
Temperature [ °C ]
Temperature [ °C]
900 0.1 m/s. The integral in the left-hand side of Eq. (24) is
800 Net power : 50 Watts solved numerically for each mode i, j at each time instant
Laser speed : 0.1 m/s
700 Laser position : 0.75 Lx to obtain an expression for /ijk(t). In this case, it is estab-
600
lished that 150 eigenvalues are good enough for reaching
convergence.
500
400
300 1400 0.75 Lx
1300 0.7 Lx
200 0.6 Lx
100 1200 0.5 Lx
1100 0.25 Lx
0 0.1 Lx
0 0.025 0.05 0.075 0.1 0.125 0.15 0.175 0.2
1000
z* = z/Lz
Temperature [°C]
900
Fig. 6. Comparison of the analytical solution with numerical predictions 800 Net power : 50 Watts
Laser speed : 0.1 m/s
[14] in the z-direction at the plane of symmetry (y = 0) just under the laser 700 Laser position : 0.75 Lx
spot.
600
500
400
(in z). It was necessary to increase twelve times the original
300
workpiece thickness to obtain a top surface temperature
similar to that of a semi-infinite solid by the Carslaw and 200
Jaeger [2] solution. 100
Figs. 8 and 9 show the trend of temperature profiles 0
0 0.1 0.2 0.3 0.4
behind the laser position along the y and z-directions. It
is appreciated a rapid decrease of temperatures together y* = y/(Ly/2)
with a ‘smoothing’ process of peak values. For instance, Fig. 8. Temperatures profiles along the y-direction, over and behind the
the peak temperature diminished more than three times laser spot position.
4130 G. Araya, G. Gutierrez / International Journal of Heat and Mass Transfer 49 (2006) 4124–4131
1400 Table 2
0.75 Lx Necessary parameters for reaching the softening temperature in silicon
1300 0.7 Lx
nitride
0.6 Lx
1200
0.5 Lx Net heat power (W) 50 100 200 300
1100 0.25 Lx
0.1 Lx Required speed (m/s) 0.28 0.72 2.15 3.8
1000
Temperature [ °C]
900
800 Net power : 50 Watts due to the considerable disparities obtained in peak
Laser speed : 0.1 m/s
700 Laser position : 0.75 Lx temperatures.
600 A specific application of the solution encountered in a
500
LAM process is showed in Table 2, which exhibits the
required laser speed and absorbed heat power to reach a
400
surface temperature of approximately 1000 C, considering
300 a Gaussian distribution for the laser beam. This tempera-
200 ture (1000 C) allows silicon nitride to become softer;
100 improving the production time, minimizing the damage
0 of the workpiece and resulting in a higher tool life [10–13].
0 0.1 0.2 0.3 0.4
All cases were run in a processor Apple X server with 2
z* = z/Lz GB of memory. Comparison of CPU times required by the
Fig. 9. Temperatures profiles along the z-direction, over and behind the present analytical solution and a previous numerical simu-
laser spot position. lation [14] indicated the supremacy, as expected, of the lat-
ter one. CPU time of the analytical solution with Gaussian
model is approximately ninety times longer than the CPU
time of a spatially uniform heat source model. On the other
Heat source modeling
Gaussian
hand, the ratios between CPU times of the analytical (spa-
2000 Spatially uniform tially uniform and Gaussian distributions) to numerical
solutions are, respectively, 7 · 104 and 6.5 · 106.
atures. On the other hand, surface temperature reached its [7] R. Komanduri, Z. B Hou, Thermal analysis of the laser surface
maximum value (quasi-steady state) in a very short dis- transformation hardening process, Int. J. Heat Mass Transfer 44
(2001) 2845–2862.
tance, around three source lengths. [8] W.M. Seen, C.H.G. Courtney, Surface heat treatment of EN 8 steel
Finally, forced convection effects are practically negligi- using a 2 kW continuous wave CO2 laser, Met. Technol. (1979) 456–
ble and can be ignored for the sake of simplicity without 462.
impact in the accuracy of the solution. [9] D.A. Stephenson, T.C. Jen, A.S. Lavine, Cutting tool temperatures in
contour turning: transient analysis and experimental verification, J.
Manuf. Sci. Eng. ASME Trans. 119 (1997) 494–501.
Acknowledgements [10] S. Lei, Y.C. Shin, Experimental investigation of thermo-mechanical
characteristics in laser-assisted machining of silicon nitride ceramics,
Dr. Gustavo Gutierrez and Mr. Guillermo Araya would J. Manuf. Sci. Eng. 123 (2001) 639–646.
like to thank the University of Puerto Rico-Mayagüez for [11] J.C. Rozzi, Experimental and theoretical evaluation of the laser
the support to this work. assisted machining of silicon nitride. Ph.D. Thesis. Purdue University,
West Lafayette, IN, 1997.
[12] J.C. Rozzi, F.E. Pfefferkorn, F.P. Incropera, Y.C. Shin, Experimental
References evaluation of the laser-assisted machining of silicon nitride ceramics,
ASME J. Manuf. Sci. Eng. 122 (2000) 666–670.
[1] J.C. Jaeger, Moving source of heat and the temperature at sliding [13] J.C. Rozzi, F.E. Pfefferkorn, F.P. Incropera, Y.C. Shin, Transient,
contacts, Proc. Roy. Soc. NSW 76 (1942) 203–224. three-dimensional heat transfer model for the laser assisted machining
[2] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solid, Oxford of a silicon nitride ceramic: Part I- comparison with measured surface
University Press, 1959. temperature histories, Int. J. Heat Mass Transfer 43 (2000) 1409–
[3] Z.B. Hou, R. Komanduri, General solutions for stationary/moving 1424.
plane heat source problems in manufacturing and tribology, Int. J. [14] G. Gutierrez, G. Araya, Temperature distribution in a finite solid due
Heat Mass Transfer 43 (2000) 1679–1698. to a moving laser beam, Proceedings of IMECE, ASME Congress at
[4] D. Rosenthal, The theory of moving sources of heat and its Washington, D.C., IMECE2003-42545, 2003.
application to metal treatments, Trans. ASME 80 (1946) 849–866. [15] M.F. Modest, H. Abakians, Evaporative cutting of a semi-infinite
[5] H.G. Woo, H.S. Cho, Three-dimensional temperature distribution in body with a moving CW laser, J. Heat Transfer 108 (1986) 602–607.
laser surface hardening processes, Proc. Instn. Mech. Engrs. 213 (Part [16] F.P. Incropera, D.P. DeWitt, Fundamentals of Heat and Mass
B) (1999) 695–712. Transfer, fourth ed., Wiley, 1996.
[6] M.F. Modest, H. Abakians, Heat-conduction in a moving semi- [17] Y.S. Touloukian, R.W. Powell, C.Y. Ho, P.G. Klemens, Thermo-
infinite solid subjected to pulsed laser irradiation, J. Heat Transfer physical properties of matter, TPRL Inc., West Lafayette, Report No.
Trans. ASME 108 (3) (1986) 597–601. TPRL 2128, 1970.