959
A parametric study of part distortions in fused
deposition modelling using three-dimensional
finite element analysis
Y Zhang and K Chou*
Mechanical Engineering Department, University of Alabama, Tuscaloosa, Alabama, USA
The manuscript was received on 24 August 2007 and was accepted after revision for publication on 24 April 2008.
DOI: 10.1243/09544054JEM990
Abstract: A finite element analysis (FEA) model was previously developed by the current
authors to simulate the fused deposition modelling (FDM) process. The model considered
coupled thermal and mechanical phenomena and incorporated an element activation function
to mimic the additive nature of FDM. Due to repetitive heating and cooling in the FDM process,
residual stresses accumulate in a part during deposition. In this study, an FEA model is used to
evaluate the distortions of a part. A parametric study, with three factors and three levels, is performed to evaluate the effects of the deposition parameters on residual stresses and part distortions. Prototype models with larger sizes are fabricated, measured, and compared with the
simulations.
The simulation results are summarized as follows. First, the scan speed is the most significant
factor affecting part distortions, followed by the layer thickness. Second, the road width alone is
insignificant. However, the interaction between the road width and the layer thickness is significant as well. Third, there are other two-way and three-way interactions that are of secondary
significance. In general, residual stresses in FDM parts increase with the layer thickness. Residual stresses also increase with the road width, but to a lesser extent, although largely affected
by the layer thickness. The FDM part distortions from the experiment show a similar trend as
concluded in the simulations, but without quantitative correlation.
Keywords: fused deposition modelling, numerical simulation, parametric study, part distortion
1 INTRODUCTION
Fused deposition modelling (FDM) is one of the most
widely used solid freeform fabrication (SFF) systems
because it uses inexpensive machinery and durable
part materials. In FDM, a thermoplastic material is
heated to a semi-molten state, extruded as an ultrathin filament, and deposited to build a part. The extrusion nozzle move according to a toolpath defined
by the part cross-sectional boundary and material is
deposited on top of the existing layer. Heat is dissipated by conduction and forced convection and the
reduction in temperature caused by these processes
causes the material to quickly solidify onto the sur*Corresponding author: The University of Alabama, Mechanical Engineering Department, 290 Hardaway Hall, Tuscaloosa,
Alabama, USA. Email: kchou@[Link]
JEM990 IMechE 2008
rounding filaments. Bonding between the filaments
is caused by local re-melting of previously solidified
material and diffusion [1]. The role of the part geometry in SFF is a frequently studied subject. Recently,
Mahesh et al. [2] used a benchmark part to evaluate
several SFF processes and measured tolerances such
as flatness and symmetry were compared. The authors
noted that the residual-stress-induced-distortions
(e.g. warpage and delamination) are prominent. As in
other SFF processes, in FDM, the heating and rapid
cooling cycles of the materials result in non-uniform
thermal gradients which cause a stress build-up that
leads to distortions. Qiu and Langrana [3] studied
the toolpath effects in FDM and proposed an algorithm to match the toolpath with the extrusion speed
so as to eliminate voids and to correct overfill and
underfill defects. Pennington et al. [4] conducted an
experimental study to investigate factors that influence
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
960
Y Zhang and K Chou
dimensional accuracy in FDM. The authors reported
that the part size, location, and envelope temperature
have a significant effect. Jiang and Gu [5] studied
extrusion phenomena in FDM and reported that the
process parameters are critical to the part accuracy.
Most SFF processes involve localized energy transport, mass transfer, phase changes, and thermallyinduced mechanical loading. Due to high process
temperatures, repetitive thermal cycles, and continuous geometric changes, the process modelling and
simulation of SFF are difficult topics. Numerical
approaches have been applied to model residual
stresses in SFF processes. Dalgarno et al. [6] conducted structural analyses to model the development
of the curling of parts in a selective laser sintering
process. It was reported that applying double sintering to the first two layers of a part to relieve the
strains may significantly reduce the curling level.
Sonmez and Hahn [7] developed a thermomechanical model for the laminated object manufacturing
process, correlating process parameters with temperatures and stress distributions in the laminate
during fabrications, and further suggested that a large
roller diameter and a slower roller speed would
be favourable for laminate bonding. Chin et al. [8]
extended an earlier one-dimensional model to establish thermomechanical models for individual
layers, droplet columns, and adjacent droplets in
the shape deposition manufacturing (SDM) process.
The authors reported that while localized substrate
preheating was ineffective, process-induced preheating seemed to be effective in reducing thermal
gradients and residual stresses. Nickel et al. [9] developed a three-dimensional finite element analysis
(FEA) model to study the effect of deposition patterns
on the resulting stresses and deflections in SDM. It
was suggested that a raster pattern, with the primary
laser moving direction normal to the part major axis,
would produce the least deflections.
There are some reports in the literature on FEA
studies of part distortions in liquid-based SFF processes. Wiedemann et al. [10] developed methods to
evaluate photopolymers with respect to the dynamics
of polymerization and shrinkage. The sensitivity of
polymerization to the process conditions was also studied. A numerical simulation of part distortions was
applied to reveal the interaction between the material
properties and the process parameters. The timedependent shrinkage and increase in strength during
the polymerization were investigated. It was concluded that the deposition process must be optimized
to reduce internal stresses using a stereolithographic
apparatus (SLA). Bugeda et al. [11] modelled the
mechanical aspect of using SLA to study the influence
of different constructive and numerical parameters in
the curl distortion caused by resin shrinkage. The curl
distortion was found to increase with the volumetric
shrinkage, but was found to decrease when the layer
thickness increased. Xu et al. [12] incorporated
thermal stresses for shrinkages due to resin phase
changes into a FEA to simulate the part deformations
in SLA. Huang and Lan [13] used a dynamic finite element code to simulate the photopolymerization process in order to efficiently select process parameters
and to obtain distortion data.
A three-dimensional FEA model was developed by
Zhang and Chou [14] in order to simulate the FDM
deposition process by controlling element activations
for the involved mechanical and thermal processes.
The model could be used to predict residual stresses
and evaluate part distortions in FDM. This model
is now applied to investigate, by incorporating the
design-of-experiments approach, the effects of process parameters on the residual stresses and distortion of parts in FDM.
2 NUMERICAL SIMULATIONS
2.1 Finite element model
The commercial software, ANSYS, was utilized to
develop the simulation codes. The simulations were
conducted in a stepwise thermo-mechanical manner.
The element geometry chosen, a rectangular parallelepiped, had dual attributes (solid 45/solid 70),
compatible with the thermal and mechanical analyses. The governing equation of the thermal analysis
was a three-dimensional transient heat conduction
representation with heat generation from the phase
changes [14]. The bottom surface of the model
which was in contact with the platform was set to be
at a constant chamber temperature (75 C). The
boundary conditions of the other outer surfaces were
obtained by forced convection with an ambient
temperature of 75 C and a heat convection coefficient
of 86 W/m2 K see the appendix for further details. The
properties of ABS plastic were used in the simulations.
The initial temperature of a newly activated element
was set at the extrusion head temperature (280 C)
and for other activated elements, the initial temperatures were those obtained in the previous step of
the thermal analysis. The mechanical analysis used a
static structural analysis (elastic stress equilibrium)
approach with induced thermal strains. For the
mechanical boundary conditions, the bottom surface
of the part was fully constrained. For newly activated
elements, the initial displacement was zero. For other
elements, the result of the previous mechanical analysis step was used as the initial condition. The
element birth/death function was used to mimic
the additive feature in FDM, i.e. continuous filament
depositions. The model geometry was first specified,
meshed, and the elements were activated according
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
JEM990 IMechE 2008
A parametric study of part distortions in fused deposition modelling
to the filament deposition sequence. For each element
activated, a transient thermal analysis started with the
current temperature distributions as the initial condition. The calculation continued until the next element
was extruded, which is a function of the element size
and the linear deposition speed. The obtained temperature distribution was then used in the next-step
mechanical analysis that calculated the displacement
at each node of all activated elements. The next element following the designated path was then activated
and the convection surface was updated for the next
temperature analysis, and these cycles were repeated
until all the elements were activated. Displacements,
strains, and stresses in the constrained part were
analysed. Then a thermal analysis was executed to
change the ambient temperature from the chamber
to the laboratory. The displacement constraint of the
model was also released in the final mechanical analysis to simulate removing the part from the platform.
2.2 Parametric study
The effects of the major process parameters in FDM
such as the road width, the layer thickness, and
the scanning speed were investigated using the developed FEA model. Three levels (low, medium, and
high) of each factor were used: 0.25, 0.5, and 1.0 mm
for the road width, 0.127, 0.254, and 0.508 mm for
the layer thickness, and 32, 64, and 96 mm/s for
the scan speed. The scanning speed alters the heating
and cooling frequency during depositions and results
in different degrees of thermal gradients in the part,
which in turn affect the residual stresses. The layer
thickness and the road width affect the number of
layers and the tool paths required at each layer. A
thick layer means fewer layers, which may reduce
the number of heating and cooling cycles. On the
other hand, a smaller road width will input less heat
into the system within a specific amount of time,
but requires more loops to fill a certain area. This
will complicate the temperature and stress evolutions
and influence the residual stress distributions. Moreover, the effects due to the interactions between the
selected parameters are not clearly known.
To systematically investigate the FDM process parameter effects, a second-order approach, a central
composite design (CCD), was applied in the design
of experiments [15]. A typical CCD is made up of three
portions: the factorial, the central point, and the
axial portions. The factorial portion is a three-factor,
two-level, one-half fractional factorial design.
Together with one centre run and six axial runs, CCD
requires only 15 runs instead of 27 in a three-factor,
three-level full factorial design. Moreover, the CCD
method is able to reveal high-order interactions. In
this study, the six axial runs included either a high or
JEM990 IMechE 2008
961
low level of each factor. Due to the significant computation times required for large models, a small representative 40 10 1.016 plate, corresponding to the
longitudinal, transverse, and height dimensions, was
used in the simulations. To discretize the deposition
process, a small element, e.g. 1 0.25 0.127, was
set as a single unit for activations. In addition, the toolpath used was the long-raster pattern, i.e. the primary
direction of movement of the extrusion head is along
the longitudinal axis of the part.
3 RESULTS AND DISCUSSION
3.1 Simulation results
Table 1 lists the parameters of each case studied
as well as the corresponding simulation result, the
maximum first principal stress (s1). Figure 1 presents
a few examples of first principal stress distributions
at different deposition settings (cases 5, 11, and 4).
The shown stress contours are present at the bottom
surface of the part where high residual stresses
always occur [14]. It is noted that the process
parameters have significant effects on the residual
stresses in the prototypes. Case 5, corresponding to
a 1-mm road width, 0.127-mm layer thickness, and
32-mm/s scan speed, results in the lowest residual
stresses. Case 11, corresponding to a 1-mm road
width, 0.254-mm layer thickness, and 64-mm/s scan
speed, results in an intermediate level of residual
stresses. On the other hand, case 4, corresponding
to a 0.25-mm road width, 0.508-mm layer thickness,
and 96-mm/s scan speed, results in the most severe
stress accumulations.
All the tested cases were further analysed to investigate the situation when the part distortions imitating that the part is removed from the platform.
Figure 2 shows examples of distortion simulation
results, cases 4, 5, and 11. It is noted that the greater
the residual stresses, the more severe the part distortions. Figure 3 plots the part distortion values (flatness) against the maximum first principal stress
from 15 cases. It is observed that in general, the simulated part distortions show a monotonic increasing
trend with the maximum first principal stress in the
FDM parts. However, there are some cases that
show an opposite trend locally, i.e. a low stress with
more distortion, therefore there does not appear to
be a clear correlation between these factors. It should
be noted that, in addition to the first principal stress
that is calculated from the six stress states at a location, the FDM part distortions are affected by other
factors, e.g. different stress components and stress
distributions. Moreover, FDM part distortions show
a vaulting shape in both longitudinal and transverse
directions, indicating a complex phenomenon
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
962
Table 1
Y Zhang and K Chou
The design-of-experiments matrix and the simulation results, maximum first principal stress
Factors
Case
A (road
width)
B (layer
thickness)
C (scan
speed)
Results
max. s1 (kPa)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
1
1
1
1
1
1
1
0
1
1
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
1
1
0
0
1
1
1
1
1
1
1
1
0
0
0
0
0
1
1
6.54
2.52
1.67
2.04
5.27
8.66
5.29
6.22
8.31
2.59
4.59
1.42
8.33
7.25
1.06
101
103
102
105
101
104
101
104
103
102
103
102
103
101
105
(1: low, 0: medium, 1: high)
Fig. 2 Simulated part distortions at different deposition
conditions (unit: m)
affected by more than longitudinal and transverse
stress components. Therefore, a critical stress component, the first principal stress, was used for the following statistical analysis.
3.2 Statistical analysis
Fig. 1 First principal stress distribution (s1, Pa) at the
bottom surface of a part with different deposition
conditions
The analysis of variance (ANOVA) approach was
applied to the maximum first principle stress (s1) of
all cases to identify significant factors and interactions
between factors. The results, listed in Table 2, show
that the scan speed (C) is the most significant factor,
followed by the layer thickness (B). The road width
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
JEM990 IMechE 2008
A parametric study of part distortions in fused deposition modelling
963
0.3
0.25
Flatness (mm)
0.2
0.15
0.1
0.05
0
1.0E+00
1.0E+02
1.0E+04
1.0E+06
1.0E+08
1.0E+10
Max. first principal stress (Pa)
Fig. 3 Simulated part flatness versus the maximum first principal stress in the part
Table 2
ANOVA results of maximum first principal
stresses (in Pa)
Factor
P value
A (road width)
B (layer thickness)
C (scan speed)
AB
AC
BC
ABC
AA
BB
CC
0.2926
0.0138
0.0005
0.0046
0.2188
0.0110
0.0046
0.5779
0.7116
0.0065
(A) in isolation is insignificant. However, the interaction between the road width (A) and the layer thickness (B) is considered to be significant. In addition,
there are other high-order interactions, i.e. BC, CC,
and ABC that are of secondary significance.
After excluding the insignificant interactions, the
ANOVA process was again performed using logarithmic values, to obtain a prediction equation for the
maximum s1 (in Pascals). The regression gives a residue R of 0.94
logmax s1 6:21 0:067 6A 0:494B 1:31C
0:432AB 0:0445BC
0:0808ABC 0:168C 2
To validate the regression model, three additional
cases of different parameters, i.e. varied (A, B, C) sets,
were studied using the three-dimensional FEA model.
The three cases chosen correspond to some of the
test conditions in the experimental investigation
JEM990 IMechE 2008
reported in the following section. The simulations
were compared with the results predicted using
equation (1). The results are summarized in Table 3.
The average relative error is less than 10 per cent
with the maximum value being 12.6 per cent.
3.3 Experimental comparisons
To examine the FEA model, several prototypes were
fabricated in an FDM machine and then measured
to allow comparisons to be made. Owing to accuracy
limitations in the measurement of small parts, parts
with larger dimensions, but with the same length:
width ratio, were used: 100 25 mm blocks with a
height of 30 mm. In addition to the nominal operation condition (a 0.50-mm road width and a 0.25mm layer thickness), two road widths (0.36 and 0.76
mm), and two layer thicknesses (0.18 and 0.33 mm)
were varied to investigate the parameter effects. The
fabricated prototypes were removed from the platform and mounted on a metal flat using sealants.
The bottom surface of a part (in reference to the
deposition) was probed using a coordinate measuring machine in automatic measuring mode using a
total of 50 equally spaced points. The raw data were
processed in Matlab and two- and three-dimensional
plots of the part surface were produced. The distortion values (flatness) at different parameter settings
were compared.
Figure 4 is a typical example of the shape of a bottom
surface of a part. All the measured FDM parts showed
the same distortion pattern (vaulting shape). However,
the flatness value (about 0.63 mm in this example) is
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
964
Y Zhang and K Chou
Table 3
Validation result of stress regression equation
Log (max, s1)
Factor and level
Road width
(mm), [A]
Layer thickness
(mm), [B]
Scan speed
(mm/s), [C]
FEA
simulation
Regression
equation
Relative
error (%)
0.36, [0.55]
0.76, [0.53]
0.36, [0.55]
0.18, [0.6]
0.18, [0.6]
0.33, [0.3]
32, [1]
32, [1]
32, [1]
5.318
4.589
4.816
4.648
4.947
5.225
12.6
7.8
8.5
Fig. 4 An example of the distortion of a bottom surface of a part: (a) three-dimensional height profile
showing the vaulting shape (unit: mm); and (b) two-dimensional contour showing shifting of
the distortion centre (mm)
not necessarily comparable to the simulation results
for small parts. It is noted that the distortion core (the
small white-line oval in Fig. 4(b)) shifts away from the
geometric centre of the part along the longitudinal
axis. The shift of the distortion centre is consistent
with simulated behaviour (Fig. 2) and it is due to the
asymmetric stress distribution created by the tool
path pattern used in the deposition (long-raster in
this case). Figure 5 shows the out-of-plane distortion,
around the part centre, in the longitudinal and transverse directions. As observed in Fig. 4, the asymmetric
distortion occurs in the transverse direction.
The measured distortion results for different
parameters are compared in Fig. 6 together with the
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
JEM990 IMechE 2008
A parametric study of part distortions in fused deposition modelling
965
Height (mm)
0.20
0.00
0
20
40
60
80
100
40
50
-0.20
Longitudinal profile
-0.40
Distance (mm)
(a)
0.3
Height (mm)
0.25
0.2
0.15
Transverse profile
0.1
0
10
20
30
Distance (mm)
(b)
Fig. 5 Out-of-plane distortion in the: (a) longitudinal; and (b) transverse directions
simulated maximum residual stresses predicted using
equation (1). It is noted that the distortions increase
with both the layer thickness and the road width,
which is consistent with simulation results. However,
interactions between the layer thickness and the road
width are not significantly discernible. Note that the
simulations indicate that the stress accumulations
increase with the layer thickness, and increase with
the road width, but to a lesser extent. At small layer
thicknesses, increasing the road width will increase
the level of stress accumulations. However, at larger
layer thicknesses, the road width has little effect on
the residual stresses of the part. In this study, the FEA
simulations were only performed on small parts due
to computation time constraints. On the other hand,
the experimental investigation used thick parts and
JEM990 IMechE 2008
was only limited by measurement accuracy. Thus,
the simulations and experimental results can only be
qualitatively compared, which can clearly be demonstrated from the combined graph showing that distortions increase with the stress level.
4 CONCLUSIONS
In this study, the FDM process was simulated using a
three-dimensional FEA modelling approach that
incorporates the additive feature and thermomechanical phenomena during the material depositions.
The model was tested to evaluate the part distortions,
showing a vaulting shape and distortion-core shifting
caused by asymmetric stress distributions. The
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
966
Y Zhang and K Chou
5.5
0.90
Distortion
Stress
0.70
Stress, Log(Pa)
Distortion (mm)
0.80
0.60
4.5
0.50
(0.36,0.18)
(0.76,0.18)
(0.50,0.25)
(0.36,0.33)
(0.76,0.33)
Fig. 6 Part distortion results from the experiment compared with the stress simulations (variable pair in
abscissa is road width and layer thickness, both in mm)
model was used to study the effects of process parameters on the part distortions in FDM. The design
of experiments, using the central composite design
scheme, consisted of three factors and three levels.
ANOVA was then used to identify significant factors
and/or factor interactions, and to establish a regression model of the residual stresses. The simulations
and analysis can be summarized as follows.
1. The simulation results show that part distortions
are related to the stress accumulation during the
deposition.
2. Among the three tested parameters, the scan
speed is the most significant factor to the residual
stresses followed by the layer thickness.
3. The road width alone does not affect the residual
stress and part distortions in a statistically significant manner. However, the interaction between
the road width and the layer thickness seems
to be as significant as the layer thickness to part
distortions.
4. There are other two-way and three-way interactions that are of secondary significance.
5. The simulations indicate that the stress accumulations increase with increasing layer thickness
and increasing road width (although to a lesser
extent for the latter factor). The distortion results
from the experiment show a similar trend; distortions increase with both the road width and the
layer thickness, however, the effects between these
two factors are not discernible. A quantitative correlation is yet to be established.
10
11
REFERENCES
12
1 Sun, Q., Rizvi, G. M., Giuliani, V., Bellehumeur, C. T.,
and Gu, P. Experimental study and modeling of bond
formation between ABS filaments in the FDM Process.
In Proceedings of the Annual technical conference
ANTEC, 2004, vol. 1, pp. 11581162.
Mahesh, M., Wong, Y. S., Fuh, Y. H., and Loh, H. T.
Benchmarking for comparative evaluation of RP systems and processes. Rapid Prototyping J., 2004, 10(2),
12335.
Qiu, D. and Langrana, N. A. Void eliminating tollpath
for extrusion-based multi-material layered manufacturing. Rapid Prototyping J., 2002, 8(1), 3845.
Pennington, R. C., Hoekstra, N. L., and Newcomer, J. L.
Significant factors on the dimensional accuracy of fused
deposition modeling. Proceedings of the Annual technical conference ANTEC 2003, vol. 1, pp. 880883.
Jiang, K. Y. and Gu, Y. H. Controlling parameters for
polymer melting and extrusion in FDM. Key Engng.
Mater., 2004, 258259, 667671.
Dalgarno, K. W., Childs, T. H. C., Rowntree, I., and
Rothwell, L. Finite element analysis of curl development in the selective laser sintering process. In Proceedings of Solid freeform fabrication symposium, University
of Texas, Austin, TX, 1996, pp. 559566.
Sonmez, F. O. and Hahn, T. H. Thermomechanical analysis of the laminated object manufacturing (LOM) process. Rapid Prototyping J., 1998, 4(1), 2636.
Chin, R. K., Beuth, J. L., and Amon, C. H. Successive
deposition of metals in solid freeform fabrication processes, part 2: thermomechanical models of adjacent
droplets. J. Manuf. Sci. Engng, 2001, 123, 632638.
Nickel, A. H., Barnett, D. M., and Prinz, F. B. Thermal
stresses and deposition patterns in layered manufacturing. Mater. Sci. Engng. A, 2001, 317, 5964.
Wiedmann, B., Dusel, K. H., and Eschl, J. Investigation
into material and process on part distortion. Rapid Prototyping J., 1995, 1(3), 1722.
Bugeda, G., Cervera, M., Lombera, G., and Onate, E.
Numerical analysis of stereolithography processes using
the finite element method. Rapid Prototyping J., 1995,
1(2), 1323.
Xu, H., Zhang, Y., Lu, B., and Chen, D. Numerical simulation of solidified deformation of resin parts in stereolithography rapid prototyping. Chin. J. Mech. Engng,
2004, 40(6), 107112.
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
JEM990 IMechE 2008
A parametric study of part distortions in fused deposition modelling
13 Huang, Y.-M. and Lan, H.-Y. Dynamic reverse compensation to increase the accuracy of the rapid prototyping system. J. Mater. Process. Technol., 2005, 167(23),
167176.
14 Zhang, Y. and Chou, Y. K. 3D FEA simulations of fused
deposition modeling process. In Proceedings of ASME
International Conference on Manufacturing science
and engineering, Ypsilanti, MI, 2006, MSEC2006-21132.
15 Schmidt, S. R. and Launsby, R. G. Understanding
industrial designed experiments, 1997 (Air Force Academy Press, Colorado Springs, Colorado).
16 Incropera, F. P. and DeWitt, D. P. Introduction to heat
transfer, fourth edition, 1996 (John Wiley, New York).
Heat convection coefficient estimate.
The heat flux, from the surfaces of the FDM part to
the chamber air, is given by the following equation
q hTs T1
is the average heat convection coefficient, Ts
where h
is the surface temperature, and T1 is the bulk fluid
temperature. For forced convection, the average
is correlated by the
heat convection coefficient, h;
following equation [16]
hL
1=2
0:664ReL Pr 1=3
k
ReL
u1 L
n
and
Pr
n
x
where u1 is the air flow speed, n is the kinematic
viscosity, and x is thermal conductivity of air. In the
studied case, the following conditions were used
T1 348 K;
Ts 553 K;
u1 20 m=s
With the average temperature of 450 K
n 32:39 106 m2 =s;
x 47:22 106 m2 =s;
Thus
ReL
Pr
u1 L
20 0:04
24 700
n
32:39 106
n 32:39 106
0:686
x 47:22 106
1=2
NuL 0:664ReL Pr 1=3
0:664 24 7001=2 0:6861=3
92:04
and
where NuL is the Nusselt number over the region of
interest, L is the length of the workpiece, k is the
JEM990 IMechE 2008
thermal conductivity of the convection media, ReL is
the Reynolds number, and Pr is the Prandtl number
k 37:3 103 W=mK
APPENDIX
NuL
967
NuL k 92:04 37:3 103
86 W=m2 K
L
0:04
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Downloaded from [Link] at NANYANG TECH UNIV LIBRARY on December 30, 2014
