Sie sind auf Seite 1von 8

PIERS Proceedings, Beijing, China, March 2327, 2009

946

Modeling of Thermal-metallurgical Behavior during Hybrid


Plasma-laser Deposition Manufacturing
Fanrong Kong1, 2 , Haiou Zhang1 , and Guilan Wang3
1

College of Mechanical Science and Engineering


Huazhong University of Science and Technology, Wuhan 430074, China
2
Research Center for Advanced Manufacturing, Southern Methodist University
3101 Dyer Street, Dallas, Texas 75205, USA
3
College of Material Science and Engineering
Huazhong University of Science and Technology, Wuhan 430074, China

Abstract A three-dimensional nonlinear finite element method combining with Monte Carlo
model was developed to investigate the temperature field and grain growth in the heat affected
zone (HAZ) during the plasma-laser hybrid deposition manufacturing (PLDM) thin wall metal
parts. The numerical study shows that the temperature gradient directly decides the grain growth
speed in the HAZ of deposited wall. However, the effect of thermal impact due to continuous
scanning of laser and plasma arc on the microstructure in the substrate material is negligible.
This thermal-microstructure model could be further applied to study the variable high energy
forming processes.
1. INTRODUCTION

Plasma-laser hybrid direct rapid manufacturing (PLDM) [13], consisting of fusion deposition forming based on plasma arc combined with laser beam, fine machining and surface treatment, belongs
to dieless direct rapid manufacturing technology. PLDM could obviously reduce development cost
and cycle of novelty product such as high temperature parts of aerospace engine. As compared
with laser engineering net shape (LENS) technique, PLDM has many advantages, including high
forming efficiency, low investment cost, facilitating to acquire full dense metal parts and finish the
formed surface as the deposition forming. Moreover, it is available for controlling of concentration distribution of material composition in PLDM [2]. Therefore, PLDM has a huge potential
of development in aerospace, energy, biological engineering, and environment protection fields for
manufacturing difficulty-to-machine parts and functionally graded materials, even long-life mould
with complex geometry.
It is noted that complicated thermal dynamic phenomena exist in the rapid forming metal
parts process by PLDM; especially, when a layer of material is deposited, one or more previously
deposited layers can be reheated or remelted, which under unfavorable process conditions may lead
to undesired effects. For example, during the PLDM of steel, martensite can be reheated above
the martensite start temperature, resulting in retaining of tempered martensite after finally cooling
down. However, according to the previous literature [4], if the process parameters are controlled
such that most of the part remains at temperature higher than the martensite start temperature,
after cooling down this will lead to a uniform microstructure consisting of non-tempered martensite
with minor proportions of retained austenite and carbides. Costa et al. [5] presented that short idle
time and small substrate size can reduce the proportion of tempered martensite and result in
a more uniform microstructure and mechanical property. This is because that both short idle
time and small substrate size are more probable to keep most of the part at temperature higher
than the martensite state temperature [4]. Wang et al. [4] also presented that control of the
molten pool determines the properties of the resulting solidified product. J. Choi et al. [6] studied
the influence of preheating and power input on the microstructure and mechanical properties of
deposited layer by using laser cladding. PLDM, as a recently developing forming technique, has
a little bit difference with other ones. The addition of plasma arc make the energy input mode
become more complicated, of which it is necessary being fully understood to make the simulation
results reasonable.
Recently, the development of advanced computational science gives us an efficient way to simulate those grain growth by developing a thermal microstructure model. Great achievements have
been made in the simulation of the grain growth or crystal evolution in the casting, laser welding,
and electric arc welding process by applying the Monte Carlo model in cooperation with thermal

Progress In Electromagnetics Research Symposium, Beijing, China, March 2327, 2009

947

analysis that uses the finite element method or finite difference method [79]. The advantage of this
thermal-microstructure model is that the temperature field analysis and the coupled microstructure
simulation can be performed simultaneously.
In the proposed study, an experimentally based finite element model is used for studying the
temperature distribution and grain growth in the HAZ during the PLDM processes, in which a
new heat source model is developed. And the processing parameters influencing the thermal and
metallurgical behaviors are also investigated in detail.
2. MATHEMATICAL MODELING

PLDM is accompanied with thermal-mechanical-metallurgical behaviors, including the interaction


between laser and arc induced plasma, powder, shielding gas and substrate material. Some features
in this process are shown as following:
(1) Effective energy distribution of PLDM depends on the plasma zone and varies with the powers
and space location between the two ones.
(2) Rapid melting and cooling is followed by the motion of hybrid heat sources and induce a
higher temperature and stress concentration as well as microstructure evolution around and
in the molten pool zone.
(3) Free surface profile and fluid flow in the molten pool are related to the plasma arc pressure
and surface tension due to temperature which act on the molten pool surface, and buoyancy
force in the molten zone.
As shown in Fig. 1(a), in the deposition process, moved heat source radiates on the metal
surface to form a melt pool and directionally conducting temperature field. Powder is melted and
deposited on the substrate point by point, and then rapidly solidifies and metallurgically bonded
with substrate and/or previous deposited layers.
Plasma Torch

Laser

Work plane

(a) Experimental photo

(b) Schematic view of PLDM

Figure 1: Representation of plasma laser hybrid deposition manufacturing (PLDM).

An uncoupled thermal-metallurgical analysis is implemented in this study. A thermal analysis


is first performed to acquire temperature field data, which are then applied to the microstructure
analysis by using Mont Carlo (MC) model. All calculations are realized based on ANSYS Parametric Design Language (APDL). And a birth and death technique is used to model the formed clad.
The whole simulation zone consists of substrate and powder materials. To accurately simulate the
thermal behavior of PLDM, temperature dependent material properties are used. In this model,
the grain growth behavior in the melt pool has not been involved due to its much more complexity.
However, it is also noted the heat affected zone is higher danger zone for thermal fracture induced.
Therefore, it is of importance to study the grain growth behavior in HAZ for further understanding
the forming mechanism in the PLDM. The whole microstructure analysis is performed by using
Matlab code.
2.1. Finite Element Thermal Analysis

The governing equation used to analyze the temperature field can be specialized to a differential
control volume and shown as follows:

...
T
T
c
+ {V } {L} T + {L}T {q} = q
(1)
t

PIERS Proceedings, Beijing, China, March 2327, 2009

948

where, is the density, c is the specific


heat,
and T is the location-time dependent temperature.

{L} is the vector operator, {L} =


; {V } is the velocity vector for mass transport of heat,
y

vx ...
vy . q is the heat generation rate per unit volume. {q} is the heat flux vector, and
{V } =

vz
Fouriers law is used to relate the heat flux vector to the thermal gradient:
{q} = [D] {L} T

(2)

where conductivity matrix [D] is given by:

Kxx
0
0
Kyy
0
[D] = 0
0
0
Kzz

(3)

Kxx , Kyy , Kzz are conductivity in the element x, y, and z direction, respectively.
Substituting Eq. (2) into Eq. (1) and ignoring the effect of fluid flow in the weld pool on the
temperature field of the welded joint, i.e., letting {V } = 0, the governing equation of temperature
field solution can be further shown below:

...
T
c
= {L}T ([D]{L}T ) + q
(4)
t
In addition, the boundary conditions are considered.
(1) Specified heat flow input acting over the weld bead surface S1 :
{q}T {} = q

(5)

where {} is unit outward normal vector; q is input heat flow due to laser/arc hybrid heat
sources and given by q = qarc (x, y, z) (x, y, z) + qlaser (x, y, z). The whole schematic view
of hybrid heat sources in the welding system is shown in Fig. 1(b). And the plasma arc and
laser hybrid heat source model in Ref. [3] is referenced in this study.
(2) Specified convection and radiation heat loss acting over surface S2 (Newtons law of cooling):
4
{q}T {} = hf (TS T ) +  (TS4 T
)

(6)

where, hf is the heat convection coefficient acting on the surfaces of deposited wall, T is
the bulk temperature of the surrounding air, and TS is the surface temperature. is StefanBoltzmann constant, and  denotes the effective emissivity.
Combining Eq. (4) with Eqs. (2), (5) and (6), and integrating them over the volume of the
element, the governing equation of temperature field solution in this study can further yield
as below:

Z
T
T
c
T + {L} ([D]{L})T d(vol)
t
vol
Z
Z
Z

...

4
=
T q dS1
T hf (T T ) + T  T T dS2 +
T q d(vol) (7)
S1

S2

vol

where vol denotes volume of the element, T denotes an allowable virtual temperature dependent of the location x, y, z, and time t.

Progress In Electromagnetics Research Symposium, Beijing, China, March 2327, 2009

949

2.2. Grain Growth Simulation of Heat Affected Zone Based on Monte Carlo Model

The concept behind the Monte Carlo method in grain growth simulation is both simple and fascinating: its only basis is the thermodynamic of atomic interactions. There are no other experimental
or theoretical inferences, nor mathematical approximations. The first step is to represent the material as a 2D or 3D matrix, in which each site corresponds to a surface or volume element. The
content of each element represents its crystallographic orientation. Contiguous regions (containing
the same number) represent the grains as shown in Fig. 2. The grain boundaries are fictitious
surfaces that separate volumes with different orientations. After choosing the kind of matrix and
filling it with an initial random content, the simulation itself begins. These are the four main steps
of the algorithm:
1) Calculation of the free energy of an element of the matrix (Gi ) with its present crystallographic
orientation (Qi ) based on its neighborhood.
2) Random choice of a new crystallographic orientation for that element (Qf ).
3) New calculation of the free energy of the same element (Gf ), but with the new crystallographic
orientation (Qf ).
4) Comparison of the two values (Gf Gi ). The orientation that minimizes the energy is chosen
with transition probability W .
These four steps are repeated millions of times in random positions of the matrix. The overall
result is a microscopic simulation of the free energy decay in the system, which is actually the main
impelling force for grain growth. The Hamiltonian that describes the interaction among the closest
neighbors, which represents the grain boundary energy, is:
X
H = J
(si sj 1)
(8)
nn

where, Si is one of the Q possible orientations in the i element of the matrix and ab is the Kroneckerdelta, which is 1 when the two elements are equal and 0 otherwise. As a result, neighbors with a
different orientation contribute J to the system energy and 0 when equal. The transition probability
W is given by:

exp G G > 0
kb T
W =
(9)
1 G 0
where, G is the change in free energy due to the orientation alteration, kb is the Boltzman
constant, and T is the temperature. Thus the speed of the moving segment is given by:

G
vi = C 1 exp
(10)
kb T
where, C is the boundary mobility.

Figure 2: The grain structure represented by a 2D square matrix.

PIERS Proceedings, Beijing, China, March 2327, 2009

950

A continuous grain growth model is shown as follows [6]:


Ln Ln0 = k(T )t

(11)

where L and L0 are the initial and final mean grain sizes calculated with the linear intercept
method, n is the grain growth exponent, and k(T ) generally given as an Arrhenius-type equation
and its expression is shown as follows:

Q
k(T ) = k0 exp
(12)
RT
where, k0 is the pre-exponential coefficient, R is the gas constant, and Q is the activation energy
for grain growth. The equation performs well in the simulation when the effect of grain boundary
precipitates is small [7].
Monte Carlo simulation is an efficient way to model the grain growth under constant temperature
or slow and uniform temperature evolution such as the casting process [7]. However, the PLDM
process is a dynamic thermal process with a rapid melting and solidification evolution. In other
words, there exist abrupt temperature gradient in the fusion zone and heat affected zone. In order
to simulate microstructure evolution in HAZ of deposited wall, an experimentally data based (EDB)
model [8] is used in this study to relate time and tMCS in PLDM process.
L = K1 (tMCS )n1

(13)

where is the grid point spacing, while K1 and n1 are constants. Through the regression calculation
of tMCS and the Monte Carlo simulating grain size, the value of K1 and n1 are respectively 0.715
and 0.477 [7]. In the EDB model, the relationship between the tMCS and the real time temperature
T (t) is expressed as follows [8]:

L0 n
K
Q
2n
(tMCS ) =
t
(14)
+
exp
K1
(K1 )n
RT (t)
where n is the grain growth exponent and K is the model constant. The tMCS varies with the
thermal cycle as shown in Eq. (14). In order to unify the tMCS in the simulation domain, a site
selected probability P is employed here [7]:
P =

tMCS

(15)

tMCSMAX

here, tMCSMAX is the maximum of tMCS in the simulation domain.


Table 1: Material physical parameters used in this study.
Temperature ( C)
0
500
1000
1200
1350
1400
1500
1800
2000

Density (kg/m3 )
7900
7700
7450
7350
7240
7200
7130
6950
6840

Specific heat (J/(kg K))


465
618
780
610
550
460
390
400
410

Thermal conductivity (W/(m K))


50
39
25
26
27
28
29
29.5
30

3. NUMERICAL EXPERIMENT OF PLDM FORMING THIN WALL METAL PARTS

Figure 3 shows the finite element model. The material of substrate and powder used in this study
are both nickel base high-temperature alloy. The computational zone used is half of solid model

Progress In Electromagnetics Research Symposium, Beijing, China, March 2327, 2009

951

according to the symmetry of geometry and boundary condition. The related physical parameters
of material are shown in Table 1. A double-peak point distribution exists in the temperature &
time curve during PLDM process; one is due to the laser radiation, the other is from the plasma
arc. The temperature distribution is obviously influenced by the coupling of laser beam and plasma
arc as shown in Fig. 4.
Location at the root of deposited wall
Location at the bottom of 2nd layer
Location at the bottom of 4th layer
Location at the bottom of 6th layer
Location at the bottom of 8th layer

A
B

Figure 3: Finite element mesh.

Figure 4: Temperature distribution of PLDM.

Figure 5: Effect of power ratio of laser to arc on the temperature distribution during PLDM (3000 W plasma
arc +800 W laser, scanning speed is 8 mm/s).

t1
t2

t3
t4
Martensite start temperature

(a) Position A located at base metal

(b) Position B located at roof of deposited wall

Figure 6: The relation between the Monte Carlo step and real time & temperature during PLDM process.

PIERS Proceedings, Beijing, China, March 2327, 2009

952

The temperature evolution at locations at the different deposited layer is shown in Fig. 5.
Numerical results clearly indicate that the double thermal impact exists in the deposited layer
during PLDM process. Monte Carlo model then use the temperature data from the thermal analysis
as loading. The relation between the Monte Carlo step and real time and temperature is shown
in Fig. 6. According to the Ref. [5], the grain growth behavior exists when the temperature is
higher than the martensite start point. From the results shown in Fig. 6(a), it is clearly seen
that the influence of thermal impact due to deposition manufacturing on the microstructure of
substrate is negligible. To further study the whole grain evolution in the PLDM, a characteristic
location at the root of deposited layer is chosen (see Fig. 6(a)). And several moments in deposition
manufacturing process are also picked, the relation between real time & temperature and MC step
is listed in Table 2. Microstructure simulation results at location A during PLDM are shown in
Fig. 7. Numerical results indicate that the grain growth speed depends on the temperature gradient
and is inclined to slow down with the increasing of deposition time.
Table 2: Relation between real time & temperature and MC step.
Mark in Fig. 6
t1
t2
t3
t4

(a) t = t1

(c) t = t 3

Real time (s)


1.01
1.56
3.88
4.44

Temperature ( C)
895.12
573.45
695.34
627.41

Monte Carlo step


305
476
1181
1352

(b) t = t2

(d) t = t 4

Figure 7: Microstructure simulation result at location A during PLDM.


4. CONCLUSIONS

(1) A three-dimensional thermal and metallurgical analysis based on finite element method combining with Monte Carlo model is developed to study the temperature field and grain growth
in HAZ during the PLDM process.
(2) The thermal impact directly influences the grain size in the microstructure of deposited layers,
but almost has no effect on the substrate material in the PLDM process. And the grain growth

Progress In Electromagnetics Research Symposium, Beijing, China, March 2327, 2009

953

speed in the HAZ depends on the temperature gradient and is inclined to slow down with the
increasing of deposition time.
(3) The thermal-microstructure model could be further used to study the relation between temperature curve and the grain growth in the deposition manufacturing process by variable high
energy beam techniques.
REFERENCES

1. Zhang, H., Y. Qian, G. Wang, and Q. Zheng, The characteristics of arc beam shaping in
hybrid plasma and laser deposition manufacturing, Science in China: Series E Technological
Sciences, Vol. 49, No. 2, 238247, 2006.
2. Zhang, H., Y. Qian, and G. Wang, Study of rapid and direct thick coating deposition by
hybrid plasma-laser manufacturing, Surface and Coating Technology, Vol. 201, 17391744,
2006.
3. Kong, F., H. Zhang, and G. Wang, Numerical simulation of transient multi-phase field during
hybrid plasma-laser deposition manufacturing, ASME Journal of Heat Transfer, Vol. 130,
No. 112101, 17, 2008.
4. Wang, L. and S. Felicelli, Influence of process parameters on the phase transformation and
consequent hardness induced by the LENS process, Material Processing Fundamentals, Edited
by P. Anyalebechi, TMS (The Minerals, Metals & Materials Society), 6372, 2007.
5. Costa, L., R. Vilar, T. Reti, and A. M. Deus, Rapid tooling by laser powder deposition:
Process simulation using finite element analysis, Acta Materialia, Vol. 53, 39873999, 2005.
6. Choi, J., S. K. Choudhuri, and J. Mazumder, Role of preheating and specific energy input on
the evolution of microstructure and wear properties of laser clad Fe-Cr-C-W alloys, Journal
of Materials Science, Vol. 35, 32133219, 2000.
7. Mishra, S. and T. Debroy, Measurements and Monte Carlo simulation of grain growth in the
heat-affected zone of Ti-6Al-4V welds, Acta Materialia, Vol. 52, 11831192, 2004.
8. Gao, J. H. and R. G. Thompson, Real time-temperature models for Monte Carlo simulations
of normal grain growth, Acta Materialia, Vol. 44, 45654575, 1996.
9. Kong, F., D. Lin, S. Santhanakrishnan, and R. Kovacevic, A thermal-microstructure model
to predict the grain growth of a dual-phase steel DP980 in laser heat-treatment, 2009 TMS
Annual Meeting Exhibition, San Francisco, California, USA, February 1519, 2009.

Das könnte Ihnen auch gefallen