You are on page 1of 8

Effects of Surface Roughness on Microinjection

Molding

H.L. Zhang, N.S. Ong, Y.C. Lam


School of Mechanical and Aerospace Engineering, Nanyang Technological University, Nanyang Avenue,
Singapore 639798, Republic of Singapore

Injection molding is one of the most common processes for cost-effective mass production of microplastic parts. When the dimensions of the part, and
thus the cavity of the mold, are small, microscale factors which are normally neglected in the analysis of
conventional injection molding may play an important
role. This investigation addresses the effects of mold
surface roughness on the injection of polymer melt,
which is a non-Newtonian uid, during the lling stage
of microinjection molding. The surface roughness
effect on the volume of the mold cavity is discussed. A
simple, but effective model, to describe the conductivity and the specic heat of the surface roughness is
proposed. Subsequently, by employing the nite volume method and the level set method, a numerical
procedure incorporating the proposed surface roughness model to describe the ow behavior of the polymer melt in the cavity is implemented. Finally, simulation on the melt ow injected into a microdisk cavity is
performed using the proposed model and the results
are found to be in good agreement with experiment.
POLYM. ENG. SCI., 47:20122019, 2007. 2007 Society of
Plastics Engineers

INTRODUCTION
Injection molding is one of the most common processes for cost-effective mass production of microplastic
parts. Many researchers have carried out extensive numerical and experimental analysis on the optimization of the
injection molding process parameters. However, not many
of these investigations [14] were focused on the effects
of microscale factors such as wall slip and mold surface
roughness. When the dimensions of a part, and thus the
cavity of the mold, are small, these microscale factors
which are normally neglected in the analysis of conventional injection molding may play an important role in the
ow behavior [58]. This article addresses the effects of
mold surface roughness on the cavity lling of polymer
melt in microinjection molding.
Correspondence to: N.S. Ong; e-mail: mnsong@ntu.edu.sg
Contract grant sponsor: Nanyang Technological University, Singapore.
DOI 10.1002/pen.20904
Published online in Wiley InterScience (www.interscience.wiley.com).
C 2007 Society of Plastics Engineers
V

The following three methods were proposed by


researchers for modeling the surface roughness effects on
uid ow in microchannels, conduits, and tubes:
The use of some regular (normally rectangle) roughness
elements [9] or irregular and randomly distributed
roughness elements [10] in the computation domain.
The roughness layer was considered as a porous medium layer (PML) [5, 11, 12] and analysis was performed based on the available PML theory.
The roughness effect was included into the viscosity
model [13, 14].

On the basis of these prior investigations, the following


conclusions can be drawn:
Higher roughness will enhance heat transfer while
keeping the other conditions constant [10, 12, 15].
The roughness effect on heat transfer is dependent on
the Reynolds number of the ow [11, 12, 15].
For uid ow with low Reynolds number ranging from
0.001 to 10, the roughness has a negligible effect on
pressure drop under isothermal conditions [9].
When Reynolds number is above 100, the roughness affects
the friction factor and hence the ow parameters [5].

However, the existing models of surface roughness


have the following limitations:
A number of researchers used the height of the roughness elements or the relative roughness to describe the
magnitude of the roughness. However, in practice, the
measurement of roughness is often expressed in Ra,
Rq, or Rz, etc. There is no explicit relation between the
relative roughness and these measurable values. This
limits the applicability of their models from a practical
standpoint.
Although the PML was used to model the roughness
effects by some researchers, the volume porosity and
permeability of the roughness layer are difcult to
determine in practical applications, and signicant modication will be required for numerical implementation
into the existing algorithms for injection molding.
The uids used in their investigations were assumed to
be Newtonian and isothermal, without interaction
POLYMER ENGINEERING AND SCIENCE-2007

between the ow eld and the thermal eld, and that


the ow was assumed to be in a steady state. These
assumptions are not suitable for microinjection molding.
The Reynolds numbers of the ow were typically above
100. Only a few studies considered Reynolds number
as low as 1 or even 0.01, which is normally encountered in microinjection molding.
Although the roughnessviscosity model was used to
account for the roughness effects in some studies, it
was not sensitive to roughness at low Reynolds number.
Moreover, the contribution of temperature effects was
not taken into consideration.

As such, there are deciencies in the existing approaches to account for the effects of surface roughness
for microinjection molding, where the non-Newtonian nature of a polymer melt, and the heat transfer between the
melt and the mold are important. In this investigation, a
new model will be proposed which will take into consideration the conductivity and specic heat of the roughness
layer, and the volume of the mold cavity as a function of
surface roughness. A numerical procedure incorporating
the surface roughness model is implemented by employing the nite volume and level set methods. Simulation
on the melt ow injected into a microdisk cavity was performed and experimental investigations were conducted to
verify the simulation results.
MATHEMATICAL MODELING OF
SURFACE ROUGHNESS
A cavity having two parallel plates is used as an illustration (see Fig. 1a) to model surface roughness effect on
the volume of the mold cavity. In practice, the distance
between the roughness peaks of the lower and the upper
walls (i.e., Hm) is usually measured to represent the
height of the cavity, as normal measurements are determined by the peaks of the roughness. Indeed, as shown in
Fig. 1a, the values of Hm are different at different measurement locations because of the randomness of surface
roughness. The apparent height, Happ, of the cavity as
shown in Fig. 1b, is often determined as the average of
the values of Hm at different measurement locations. In
the conventional analysis of injection molding, Happ is
usually used to represent the height of the cavity and
mold surface roughness is neglected. However, when
mold surface roughness is comparable to Happ, the volume
of the melt for lling the rough region (i.e. the region
consisting of roughness peaks and valleys) will become
signicant and cannot be neglected. In such situation,
ignoring surface roughness by employing Happ could
result in signicant inaccuracy in the calculation of the
cavity volume.
To consider the surface roughness effect on the volume
of the mold cavity, the effective mold wall should be at
the mean line of the surface roughness (see Fig. 1a). The
mean line is generated by calculating a weighted average
DOI 10.1002/pen

FIG. 1. Physical and modeled mold cavities: (a) Physical cavity with
surface roughness. (b) Modeled cavity without consideration of surface
roughness in conventional analysis. (c) Modeled cavity with consideration of surface roughness in this work.

for each data point resulting in equal volume above and


below the line. In this case, the volume of the cavity as
dened by the mean line will correspond to the volume
of the physical cavity. Thus, instead of Happ, the effective
height, Heff, which is dened as the distance between the
mean line of the roughness elements at the lower and the
upper walls (see Fig. 1c), should be employed in the analysis. Conceptually, Heff can be considered as the sum of
Happ and the lower and upper effective heights of the
roughness layers, i.e., the average distance from the
roughness peaks to the mean line at the lower wall
and the upper wall, namely, dlower and dupper as shown in
Fig. 1c.
To model the surface roughness effect on heat transfer,
surface roughness can be considered as a homogenous
roughness layer characterized by its thermal conductivity,
specic heat, and effective height, d. On the basis of a
simple rule of mixture, the thermal conductivity within
POLYMER ENGINEERING AND SCIENCE-2007 2013

TABLE 1. The values of d, l0, and l0 for some simple roughness


proles.
Roughness prole

l0

l1

0.5

0.5y/d 0.5

0.75

0.25y/d 0.75

0.5

0.5

0.5

where, l1 is the ratio between the area that will be occupied by the melt and the total area in the top plane of the
roughness layer (i.e., the plane of y dlower or the plane
of y Heff  dupper). Similarly, l0 is the ratio between
the area that will be occupied by the melt and the total
area in the central plane (i.e., the plane of y 0).
By using these three parameters, d, l0, and l1, the proposed model takes into account the heights, the spacing,
the shape, and the irregularity of the roughness prole.
Table 1 provides the values of d, l0, and l1 for some simple roughness proles.

MICROFILLING SIMULATION
During the lling stage, the melt and air can be
assumed to be incompressible, and the effects of gravity
and surface tension can be neglected. Thus, the governing
equations for uid ow in the cavity can be written as:
Continuity equation

0.75

0.75

0.75

5/6

1/6

2y/3d 1/6

r ~
v0
Momentum equation


q~
v
r
~
v  r~
v rp r  Zg;
qt

(4)

g r~
v r~
v t

The average roughness (Ra) is dened as the average deviation from


the mean line or arithmetical average.

(5)
Energy equation

the roughness layer (either the lower or the upper) is


modeled as:
K yk1 1  yk2

(1)

where, k1 and k2 are the thermal conductivities of the melt


and the mold, respectively, and y is the volume fraction
that will be occupied by the melt, which is a function of
the specic roughness prole.
Similarly, the heat capacity (in term of density and the
specic heat) of the roughness layer is modeled as:
rCp yr1 Cp1 1  yr2 Cp2

(2)

where, r1Cp1 and r2Cp2 are the heat capacities of the melt
and the mold, respectively.
Assuming a linear distribution on the relative volume
between the mold and polymer melt materials, which is
applicable for the various roughness proles as depicted
in Table 1 the volume fraction, y, can be expressed in a
general form as:
8
y
>
< l1  l0  dlower l0 ; within the lower
roughness layer
y
Heff y
>
l

l

l
;
within
the upper
1
0
0
:
dupper
roughness layer
2014 POLYMER ENGINEERING AND SCIENCE-2007



qT
~
v  rT r  krT Zg2 ;
rCp
qt

r
1
g
g : g
2
(6)

where, r, Z, Cp, and k are the density, viscosity, specic


heat, and conductivity either of the polymer melt in the
melt lled region or of the air in the unlled region.
On the basis of the proposed model, the effects of surface roughness on the conductivity of the roughness layer
can be easily included in the energy equation (Eq. 6) with
minimum disturbance, i.e., in the roughness layer, rCp
and k are calculated using Eqs. 24. Otherwise, r, Cp,
and k are the density, specic heat, and conductivity either of the melt or of the air, respectively.
The viscosity of the non-Newtonian polymer melt, Z,
is modeled by the widely accepted seven-constant CrossWLF model [1, 1618]:

(3)
FIG. 2. Schematic of lling of a disk cavity.

DOI 10.1002/pen

TABLE 2. Properties of the POM melt, air, and mold material.


Material
POM melt
Air
Mold (steel)

ZT; g ; P

r (kg/m3)

K (W/mK)

Cp (J/kgK)

1153
1.0
7800

0.14
0.037
29

2101
1.0
460

Z0 T; P

1 Z0 T; P  g =t 1n
Z0 T; P D1  exp

A1  T  T 
A2  T  T 


7

where,

FIG. 3. The analytical and simulation results of ow front development


in the disk cavity with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K, Ra 0.

T P D2 D3 P
A2 A~2 D3 P
n; t ; D1 ; D2 ; D3 ; A1 ; A~2 are material dependent constants:
To track the movement of the melt front, a scalar variable j which is the level set function, is used to identify
the interface between the melt and the air. The zero-level
set of j indicates the melt front, which is transported by:
qj
~
v  rj 0;
qt

jrjj 1

(8)

Around the melt-air interface, the physical properties


such as density and viscosity are calculated using:
a 1  ha1 ha2

(9)

where, a1 and a2 stand for the properties of melt and air,


respectively, and where h is the heaviside function [19].
The initial and boundary conditions are given as:
u Uin ; T Tmelt at the inlet; T Twall on the mold wall
(10)
where, Uin, Tmelt, and Twall are velocity at the inlet, melt
temperature at the inlet, and the mold temperature, which
are assumed to be constant during the cavity lling.
j x  Xini at t 0

qu
0;
qx

qT
0 at the axisymmetric axis
qx

(13)

The governing equations (Eqs. 46) and the level set


equation (Eq. 8) may be expressed in the general form:


q
q
q
qf

rf 
ruj f
C
S
qt
qx
qxj
qxj

(14)

For the continuity equation (Eq. 4), 


r; f; C and S correspond to 1, 1, 0, and 0, respectively; for the momentum
equation (Eq. 5), they correspond to 1, ui, Z=r, and
qu
 r1 rp qxij qxq j rZ; for the energy equation (Eq. 6), they
correspond to rCP T, k, and Zg2 ; for the level set equation
(Eq. 8), they correspond to 1, j, 0, and 0.
The nite volume method of Partankar [20] is used to
solve the general equation (Eq. 14). A xed and staggered
grid is used, with the scalar variables stored in the centers
of the control volumes, while the velocities are located at
the control volume faces. The coupling between velocity
and pressure is handled by the SIMPLER algorithm, and
the diffusionconvection effect in the momentum equa-

(11)

where, Xini is the initial position of the polymer melt in


the cavity.
u 0 on the melt-wetted wall;
qu
0 on the air-welled wall
qy

12

TABLE 3. The constants in the cross-WLF model for POM.


Material
POM

T* (Pa)

D1 (Pas)

0.382 2.29  105 7.54  1012

DOI 10.1002/pen

D2 (K) D3
223

A1

2 (K)
A

28.5

51.6

FIG. 4. Cross-sectional proles of the radial velocity component at


r 2.0 mm with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K.

POLYMER ENGINEERING AND SCIENCE-2007 2015

FIG. 5. Cross-sectional temperature proles with different roughness at


r 2.0 mm with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K.

FIG. 7. Cross-sectional shear rate proles with different roughness at


r 2.0 mm with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K.

tions is modeled by the power-law scheme. The fully


implicit scheme is used to discretize the transient term.
The upwind scheme is used to model the convection of
the level set equations. The resulting algebraic equations
are solved using the TriDiagonal Maxtrix Algorithm.

Table 2 gives the density, thermal conductivity, and


specic heat used in the simulation for POM, air, and the
mold material, respectively. To enhance the convergence
of the numerical solution but without sacricing accuracy,
the viscosity of air was assumed to be 1.0 Pa s. For
POM, the constants in the cross-WLF model are given in
Table 3.
The rst validation of the numerical model is based on
the mass conservation principle. The development of the
melt front in the radial direction of the disk cavity can be
obtained analytically as:

Numerical Results and Discussion


The developed numerical procedure was applied to
polymer lling (in this case Polyoxymethylene, POM) of
a microdisk cavity, taking into consideration the surface
roughness effects. Figure 2 shows the schematic of lling
of a disk cavity. The diameters of the injection inlet and
the disk are d1 0.4 mm and d2 8 mm, respectively.
The upper wall is assumed to be smooth and the lower
wall is rough. The roughness prole of the lower wall is
assumed to be the same as the third prole as shown in
Table 1 except explicit mentioned otherwise. Polymer
melt enters with a constant and uniform velocity, i.e.,
constant volume ow rate, pushing air out through the
outlet. It is assumed that the melt ow is axisymmetric
during the lling of the disk cavity, therefore only a twodimensional axisymmetric analysis is performed. 40 
200 grids in the x and radial direction respectively as
shown in Fig. 2 are used to discretize the computational
domain.

FIG. 6. Cross-sectional viscosity proles with different roughness at r


2.0 mm with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K.

2016 POLYMER ENGINEERING AND SCIENCE-2007

s
2H
2
din
eff d1 Uin t
rt
4Heff

(15)

where, dini is the initial position of the melt front, which


is 0.6 mm, and d1 is the diameter of the inlet, which is
0.4 mm. Figure 3 shows the analytical and simulated development of the radial ow length of the molded disk,
where the inlet velocity, Uin, is 2 m/s, the melt temperature (Tmelt) 453 K, the wall temperature (Twall) 323 K,
and the average roughness (Ra) 0. It can be observed that
the analytical and the simulation results are in good
agreement as required.

FIG. 8. Pressure development with different roughness for lling the


disk cavity with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K.

DOI 10.1002/pen

FIG. 9. Pressure for lling the cavities with different roughness proles
but the same Ra (40 mm) with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K,
Twall 323 K.

To investigate roughness effect on heat transfer, the


melt and the mold temperature are set at 453 and 323 K,
respectively, Heff 0.4 mm, Uin 2 m/s, and the average roughness (Ra) is varied. Figures 47 show the proles of the velocity, temperature, viscosity, and shear rate
of the melt at the radial position of 2.0 mm for a lling
volume of 13 cm3. Figure 4 shows that the higher the
roughness, the larger is the maximum velocity at the vicinity of the center of the cross-section. This is because
for the same effective thickness, the higher the surface
roughness will result in a higher effective conductivity of
the roughness layer. This will lead to a lower temperature
in the roughness layer, and thus lower melt temperature
(see Fig. 5) and higher melt viscosity (see Fig. 6) next to
the wall. It is therefore also expected that the higher surface roughness will result in a higher maximum shear rate
next to the surface roughness layer (see Fig. 7). Figure 8
shows the pressure development at the center of the bottom surface of the cavity as a function of surface roughness. As expected, with the same effective thickness and
the same lling volume, higher pressure is required to ll
the cavity with higher surface roughness.
On the basis of the proposed model, surface roughness
effect is a function of the roughness prole and the equiv-

FIG. 10. Pressure for lling the cavities with different proles but the
same d (80 mm) with Heff 0.4 mm, Uin 2 m/s, Tmelt 453 K, Twall
323 K.
DOI 10.1002/pen

FIG. 11. The surface of the disk insert.

alent height of the roughness layer. Figure 9 shows the


predicted pressures for lling the cavities with the triangle
prole (i.e., the rst prole shown in Table 1) and the
rectangle prole (i.e., the third prole shown in Table 1).
The two proles have the same Ra (i.e., 40 mm). It can be
observed that higher pressure is required for lling the
cavity with the triangle prole. This is because for the
same Ra, the equivalent height of the roughness layer (d)
of the cavity with triangle prole is twice that of the rectangle prole. Similarly, if cavities have the same d (i.e.,
80 mm) but different roughness proles, the roughness
effect may also be different. Figure 10 shows the predicted pressure for lling the cavities with the same d but
with different proles (i.e., the triangle prole and the
rectangle prole). It can be observed that the difference in
pressure for lling the cavities with these two different
roughness proles is very small. Therefore, the surface
roughness effect is signicantly dependent on the equivalent height of the roughness layer and less dependent on
the specic roughness prole.

EXPERIMENT
To verify the effectiveness of the proposed roughness
model and the simulation results, experiments were performed on a 5-ton microinjection molding machine
(JMV-015S-5t) using a three-plate mold. A microdisk
insert of 8-mm diameter that provides the lower wall of
the disk cavity was employed. The disk insert was
machined using EDM such that each half of the disk has
different surface roughness but with the same mean line

FIG. 12. Measured surface roughness prole of the disk insert, i.e., the
lower wall of the cavity.

POLYMER ENGINEERING AND SCIENCE-2007 2017

FIG. 13. Measured surface roughness prole of the upper wall of the
cavity.

(see Fig. 11). On one half, Ra is 5.1 mm and on the other


half, Ra is 1.8 mm. Figure 12 shows the surface roughness
prole of the insert measured along the A-B direction as
shown in Fig. 11 using a Talyscan 150 dual gauge system. It can be observed that, although the two halves have
different surface roughness (e.g. Ra), they have almost
the same mean line. The measured roughness prole of
the upper wall of the cavity is shown in Fig. 13. It can be
observed that the upper wall is very smooth (i.e., Ra
0.1 mm) and its roughness can be neglected as compared
to the lower wall. Therefore, the disk cavity employed in
the experiment is equivalent to that in the simulation (see
Fig. 2). The effective cavity thickness of 250 mm was
employed. The material employed was POM (Ultraform
W2320 003). Four cartridge heaters were used to heat up
the micromold and the mold temperature was monitored
using two thermocouples. The radial ow length on each
half of the molded part was measured using a ROI OMIS
II series optical microscope.
Figure 14 shows the molded parts of different sizes
obtained by varying injection rate with the mold temperature of 323 K and the melt temperature of 453 K. As
expected, the melt ows faster in the smoother half of the
cavity than it does in the rougher half.
In the simulation, for simplicity, it was judged that the
rectangle prole (i.e., the third prole as shown in Table
1) can be used to describe the roughness prole measured
as shown in Fig. 12. The effective height of the roughness
layer (d) is 4 mm on one half, and 11 mm on the other
half. The injection rate, mold temperature, and melt temperature were set as the same as those in the experiment.
160  200 grids in the x and radial direction respectively
are used to discretize the computational domain.
Figure 15 shows the volumes of material on each half
of the cavity versus the total volumes lled for both experimental and simulation results. It can be observed that
the simulation results are in good agreement with the experimental results. The volume differences between the

FIG. 15. Comparison between the simulation and the experimental


results.

two halves predicted by the simulation are not signicantly different from those obtained from experiment.
Therefore, the proposed roughness model can effectively
model the effects of surface roughness on micro injection
molding.
CONCLUSION
The mold surface roughness effects on microinjection
molding were investigated. Its effects on the effective
dimensions of the mold cavity and heat transfer were
modeled. Subsequently, by employing the nite volume
method and the level set method, a numerical procedure
incorporating the surface roughness model proposed was
implemented. Finally, simulation and experiment were
conducted for molding of a microdisk. In summary, the
following conclusions were obtained:
Mold surface roughness has a signicant effect on the
volume of the mold cavity. Ignoring mold surface
roughness may lead to signicant inaccuracy in predicting the melt front position and lling pressure. Therefore, the effective dimension, which is a function of
mold surface roughness, should be used in the analysis
of microinjection molding.
The surface roughness has a signicant effect on heat
transfer between the melt and the mold, and thus inuences the proles of velocity, temperature, viscosity,
and shear rate of the melt, as well as the lling pressure.
Surface roughness effect is signicantly dependent on
the equivalent height of the roughness layer and less
dependent on the specic roughness prole.
The simulation results and the experimental results are
in good agreement. The proposed model could model
the mold surface roughness effects on microinjection
molding.

ACKNOWLEDGMENTS
FIG. 14. The molded parts of different sizes obtained by varying injection rate with Tmold 323 K and Tmelt 453 K.
2018 POLYMER ENGINEERING AND SCIENCE-2007

The authors would like to thank MOLDFLOW Pty.


Ltd. for their cooperation.
DOI 10.1002/pen

REFERENCES
1. L.Y. Yu, C.G. Koh, L.J. Lee, and D.W. Koelling, Int. Commun. Heat Mass Transfer, 42(5), 871 (2002).
2. J. Zhao, R.H. Mayes, G. Chen, H. Xie, and P.S. Chan,
Polym. Eng. Sci., 43(9), 1542 (2003).
3. Y.K. Shen and W.Y. Wu, Int. Commun. Heat Mass Transfer, 29(3), 423 (2002).
4. N.S. Ong and Y.H. Koh, Mater. Manuf. Process, 20, 1 (2005).
5. C. Kleinstreuer and J. Koo, J. Fluids Eng., 126, 1 (2004).
6. K.M. Awati, Y. Park, E. Weisser, and M.E. Mackay,
J. Non-Newtonian Fluid, 89, 117 (2000).
7. R.D. Chien, W.R. Jong, and S.C. Chen, J. Micromech.
Microeng., 15, 1389 (2005).
8. C.A. Grifths, S.S. Dimov, E.B. Brousseau, and R.T. Hoyle,
J. Mater. Process. Technol., 189, 418 (2007).
9. Y.D. Hu, C. Werner, and D.Q. Li, J. Fluids Eng., 125, 871
(2003).
10. G. Croce and P.D. Agaro, Superlattices Microstruct., 35,
601 (2004).

DOI 10.1002/pen

11. J. Koo and C. Kleinstreuer, J. Micromech. Microeng., 13,


568 (2003).
12. J. Koo and C. Kleinstreuer, Int. J. Heat Mass Transfer, 48,
2625 (2005).
13. W.L. Qu, G.H. Mala, and D.Q. Li, Int. J. Heat Mass Transfer, 43, 3925 (2000).
14. G.M. Mala and D. Li, Int. J. Heat Fluid Flow, 20(2), 142
(1999).
15. S.G. Kandlikar, S. Joshi, and S.R. Tian, Heat Transfer Eng.,
24(3), 4 (2003).
16. D.G. Yao and B. Kim, J. Micromech. Microeng., 12, 604
(2002).
17. G. Xu, L. Yu, L.J. Lee, and K.W. Koelling, Polym. Eng.
Sci., 45(6), 866 (2005).
18. L. Yu, L.J. Leee, and K.W. Koelling, Polym. Eng. Sci.,
44(10), 1866 (2004).
19. Y.C. Chang, T.Y. Hou, B. Merriman, and S. Osher,
J. Comput. Phys., 124, 449 (1996).
20. S.V. Patankar, Numerical Heat Transfer and Fluid Flow,
Hemisphere, Washington, DC (1980).

POLYMER ENGINEERING AND SCIENCE-2007 2019

You might also like