Numerical Simulation of Ram Extrusion Process for Ceramic Materials

Mingyang Li1, Lie Tang1, Fei Xue2, Robert Landers1

1
Missouri University of Science and Technology
2
Shanghai Dianji University
REVIEWED, August 17 2011

Abstract

The freeze–form extrusion process for aqueous–based ceramic paste is complex due to the non–
Newtonian behavior of the paste. In this paper the process is studied numerically using a
developed mathematical model. The ceramic paste viscosity is characterized by the Herschel–
Bulkley model. The relationship between plunger velocity and extrusion force is computed
numerically. The influence of air, which is mixed with the paste during the loading process, is
also examined. Due to the compressibility introduced by the trapped air, the plunger force
dynamic response is typically dominated by a first order response. It is also shown that the
extrusion plunger force depends on the volume of air in the extruder. Good agreement is
obtained between the simulation results and experimental data.

Keywords: Ceramic paste extrusion; Numerical simulation; Non–Newtonian flow

Nomenclature
A: Cross–sectional Area ( m 2 )
C,D: Constants
F: Force ( N )
k: Consistency index ( kg m ⋅ s )
l: Length ( m )
n: Power law index
p: Pressure ( Pa )
p& : Pressure drop rate ( Pa m )
r ,θ , z : Coordinate ( m, rad , m )
T: Temperature ( K )
t: Time ( s )
ur , uθ , uz : Velocity in r, θ and z directions ( m s )

290
ubulk
: Paste bulk velocity ( m s )
up
: Plunger velocity ( m s )
V: Volume ( m 3 )
V& : Volume flow rate ( m 3 s )
ρ: Density ( kg m3 )
η: Non
Newtonian viscosity ( Pa ⋅ s )
τ: Shear stress ( Pa )
τ0 : Yield stress threshold ( Pa )
γ: Shear rate ( s −1 )
γc : Critical shear rate ( s −1 )

1. Introduction

Due to their high heat resistance and hardness, ceramic materials are widely used in automotive,
aerospace and other modern industries. However, traditional technologies for ceramic material
processing are often expensive and time consuming, especially when complex geometries are
involved. In recently years, several Solid Freeform Fabrication (SFF) processes have been
developed to fabricate ceramic components from three–dimensional (3D) CAD models,
including Fused Deposition of Ceramics [1], Fused Deposition Modeling [2], Extrusion
Freeform Fabrication [3], 3D Printing [4], Selective Laser Sintering [5,6], Shape Deposition
Manufacturing [7], and Robocasting [8, 9].

Freeze–form Extrusion Fabrication (FEF) is a novel, environmentally friendly SFF method for
ceramic parts fabrication [10–12]. It is designed to use ceramic pastes with a high solids loading
up to 50 vol.% and only trace amounts of organic binder (vol.1–4%). During the process,
aqueous based colloidal paste is extruded from one or multiple extruders to fabricate a ceramic
component in a layer–by–layer manner. The low temperature environment will help the part to
maintain its shape by freezing the water present in the paste.

The behavior of the paste flow and the paste characteristics (e.g., viscosity and compressibility)
are crucial for ceramic component fabrication using extrusion methods. During paste
preparation and loading, air bubbles are trapped in the paste, resulting in paste compressibility.
Paste compressibility has typically been ignored by previous research studies concerning paste
flow behavior and characteristics, e.g., Benbow–Bridgwater equation [13], viscoplasticity
method [14], upper bound techniques [15], finite element method [16,17], and artificial neural
network [18].

291
So far, few studies have been conducted in analysing the dynamic response of extrusion forces
when extruding non Newtonian pastes. An analytical model is first developed to describe the
relationship between plunger velocity and extrusion force when paste compressi1bility is
considered. The simulation results are then compared with experimental results.

2. Geometry and Boundary Conditions

2.1 Geometry

A ram extruder for viscosity measurement is used to study the relationship between extrusion
force and plunger velocity. As shown in Fig.1, the ram extruder includes a barrel (part 2) to hold
the paste, a die (part 4) with a small diameter, a plunger (parts 5 and 7) and seals (parts 1, 3, and
6). The geometry model and symbols of the extruder are shown in Fig. 2.

Fig. 1: Ram extruder schematic.

Fig. 2: Ram extruder model.

292
The dimensions of this ram extruder are lb = 60 mm, ld = 20 mm, and rb = 4.775 mm. Two
different diameter dies, die 1 (rd = 0.548 mm) and die 2 (rd = 1.048 mm), are used in this study.

2.2 Assumptions

The following assumptions are made:

(1) Since the velocity is small and the viscosity is large, the Reynold Number will be small;
therefore, the flow is taken as laminar.
(2) The material is homogenous.
(3) The barrel is completely filled by paste, except for studies where air is present.
(4) If otherwise stated, the paste is considered to be incompressible.
(5) The temperature is constant. Therefore, viscosity is only a function of shear rate and density
is only a function of pressure.
(6) The paste does not experience slip at the wall.

2.3 Boundary Conditions

The following boundary conditions are used:

(1) Since the paste does not experience slip at the wall, the relative velocity of the paste at the
wall in each direction is zero.
(2) The pressure at the extruder outlet is 101325 Pa. Therefore, the pressure of the paste at
boundary 5 shown in Fig. 2 is 101325 Pa.
(3) The plunger is modeled as a moving wall. Thus, the velocity of fluid at boundary 1 shown in
Fig. 2 is equal to u p .

3. Mathematical Modeling

3.1 Governing equations

For paste flowing in a pipe, its continuity and momentum equations are given by the Navier–
Stokes equations in cylindrical coordinates. Since there is no source or velocity in the θ
direction, the terms involve θ can be neglected. The continuity equation becomes

1 ∂ ∂u
( rur ) + z = 0 (1)
r ∂r ∂z

The momentum equation in the radial direction becomes

 ∂ur ∂u ∂u  ∂p 1 ∂ τ ∂τ
ρ + ur r + u z z  = ρ g r − + ( rτ rr ) − rr + rz (2)
 ∂t ∂r ∂z  ∂r r ∂r r ∂z

293
The momentum equation in the θ direction is neglected and the momentum equation in the z
direction becomes

 ∂u z ∂u ∂u  ∂p 1 ∂ ∂τ
ρ + ur z + u z z  = ρ g z − + ( rτ rz ) + zz (3)
 ∂t ∂r ∂z  ∂z r ∂r ∂z

3.2 Viscosity model

The paste viscosity is described by Herschel
Bulkley model [19]

 τ0 γ 
n −1

 +k  if γ > γc
 γ  γc 
η= (4)
  2τ 0   k ( n − 1) τ 0 
  γ + k ( 2 − n ) +  γ − 2 γ
γc 
if γ ≤ γc
 c   c

3.3 Pressure drop rate

∂u z ∂u ∂τ zz
For fully developed flow, = 0 , ur = 0 , z = 0 , and = 0 . In this case equation (2) can be
∂t ∂z ∂z
neglected, and equation (3) can be rewritten as

∂p 1 ∂
ρ gz − + ( rτ rz ) = 0 (5)
∂z r ∂r

Rearranging equation (5) and integrating with respect to r

1  ∂p 
rτ rz =  − ρ g z  r 2 + D0 (6)
2  ∂z 

When r = 0 , τ rz must be finite. Therefore, D0 = 0 and

1  ∂p 
τ rz =  − ρ g z  r (7)
2  ∂z 

∂u z
When fluid flows in the negative z direction (i.e., > 0 ) the value of τ rz should be positive.
∂r
From equation (4) and using this fact

294
  k  ∂u z 
n
∂u z
 τ + if ≥ γc

0
γ cn −1  ∂r  ∂r
τ rz =  2
(8)
  2τ 0 + k 2 − n   ∂uz  +  k ( n − 1) − τ 0   ∂u z  ∂u z
 γ ( )     if < γc
  c   ∂r   γ c γ c2   ∂r  ∂r

∂u z
From equation (8), when ≥ γc
∂r
n
k  ∂u  1  ∂p 
τ 0 + n −1  z  =  − ρ g z  r (9)
γ c  ∂r  2  ∂z 

Rearranging equation (9)

1
 n −1  ∂p n
 γ c  − ρ g z  n −1 
∂u z du z
= =  ∂z  r − γc τ0  (10)
∂r dr  2k k 
 

n −1 n −1
 ∂p  γ γ τ
Letting p& =  − ρ g z  , C1 = c , and C2 = c 0 , equation (10) can be rewritten as
 ∂z  2k k
1
du z
= ( C1 pr
& − C2 ) n (11)
dr

Solving equation (11)

n +1
n 1
uz = ⋅ ( C1 pr
& − C2 ) n + D1 (12)
n + 1 C1 p&

∂u z
From equation (8) with < γc
∂r

  k ( n − 1) τ 0   ∂uz  1  ∂p
2
 2τ 0   ∂u z 
 & + k ( 2 − n )  + − 2   =  − ρ gz  r (13)
 γc   ∂r   γc γ c   ∂r  2  ∂z 

2τ 0 k ( n − 1) τ 0
Letting C3 = + k ( 2 − n ) and C4 = − 2 , equation (13) can be rewritten as
γc γc γc

295
 2
 ∂u   ∂u  1
C4  z  + C3  z & =0
 − pr (14)
 ∂r   ∂r  2

Solving equation (14)

2 1
& − C3
∂uz du z ± C3 + 2C4 pr 1 C3
∂r
=
dr
=
2C4
=±
2C4
( C3
2
+ 2C4
&
pr ) 2 −
2C4
(15)

Since τ 0 > 0 , γ c > 0 , k > 0 , p& > 0 , r ≥ 0 , and 0 < n < 1 , C3 > 0 and C4 < 0 . Therefore, if
1
du z 1 C3
dr
=−
2C4
( C3
2
+ 2C4
&
pr ) 2 −
2C4
, when r = 0

2τ 0
+ k ( 2 − n)
du z C3 γc 2τ + γ k ( 2 − n )
=− =− = 0 c ⋅ γ > 2γ c (16)
dr C4 k ( n − 1) τ 0 τ 0 + γ c k (1 − n ) c
− 2
γc γc

∂u z
This is contrary to the assumption < γ c . Therefore
∂r
1
du z 1 C
=
dr 2C4
( & )2 − 3
C32 + 2C4 pr
2C4
(17)

Solving equation (17)

3
1 C3
uz = 2
6C4 C0
( C3
2
+ 2C 4
&
pr ) 2 −
2C4
r + D2 (18)

Therefore, the paste velocity in the z direction is

 n 1 n +1
∂u z
⋅
 n + 1 C p& ( C1
&
pr − C 2 ) n +D
1 if ≥ γc
 1 ∂r
uz =  3
(19)
 1 ( C 2 + 2C pr C3 ∂u z
 6C4 p&
2 3 4
& ) 2 −
2C4
r + D2 if
∂r
< γc

∂u z ∂u
There are two possible conditions at r = r0 : (1) ≤ γ c and (2) z > γ c . For condition (1), the
∂r ∂r
governing equations are

296
 3
1 C
uz = 2
6C4 p&
( 2
& ) 2 − 3 r + D2
C3 + 2C4 pr
2C4
(20)

r0

∫ u dA = 2π ∫ u rdr = V& = π r u
2
z z 0 bulk (21)
A 0

The boundary condition is

uz r = r0
=0 (22)

Substituting equation (22) into equation (20)

3
1 C
0= 2
6C4 p&
( 2
& 0 ) 2 − 3 r0 + D2
C3 + 2C4 pr
2C4
(23)

Solving equation (23) for D2

3
C3 1
D2 = r0 − 2 ( C32 + 2C4 pr
& 0 )2 (24)
2C4 6C4 p&

Substituting equation (24) into equation (20)

 1 3
C   1 3
C 
u z =  2 ( C32 + 2C4 pr
& )2 − 3 r  −  2 ( C32 + 2C4 pr
& 0 ) 2 − 3 r0  (25)
 6C4 p& 2C4   6C4 p& 2C4 

Substituting equation (25) into equation (21)

r0
  1 3
  1 3
  u r2
∫0  6C42 p& ( 3
C 2
+ 2C 4
&
pr ) 2 − 2CC3 r  −  2 ( C32 + 2C4 pr
C
& 0 ) 2 − 3 r0   rdr = bulk 0
2C4  
(26)
 4   6C4 p& 2

Integrating equation (26)

5 7 3
r0 ( C32 + 2C4 pr
& 0 )2 C37 − ( C32 + 2C4 pr
& 0 )2 Cr
3 0
3
( C32 + 2C4 pr
& 0 ) 2 r02 ubulk r02
+ + − − =0 (27)
30C43 p& 2 210C44 p& 3 12C4 12C42 p& 2

In equation (27), the parameters C1 , C2 , C3 , and C4 are functions of the material property, and
the parameter r0 is determined by the ram geometry. Then, given a value for ubulk , p& can be
computed numerically by equation (27). For condition (2), the governing equations are given by
equation (21) and

297
  n 1 n +1
∂u z
⋅
 n + 1 C p& ( C1
&
pr − C 2 ) n +D
1 if ≥ γc
 1 ∂r
uz =  3
(28)
 1 ( C 2 + 2C pr
 6C42 p& 3 4
& ) 2 − 2CC3 r + D2 if
∂u z
∂r
< γc
4

∂u z
The boundary conditions are given by equation (22) and when = γ c , uz is continuous.
∂r
∂u z
Substituting equation (21) into equation (28) for ≥ γc
∂r
n +1
n 1
⋅ ( C1 pr
& 0 − C2 ) n + D1 = 0 (29)
n + 1 C1 p&

Solving equation (29) for D1

n +1
n 1
D1 = − ⋅ ( C1 pr
& 0 − C2 ) n (30)
n + 1 C1 p&

∂u z
Substituting equation (30) into equation (28) for ≥ γc
∂r
n +1 n +1
n 1 n 1
uz = ⋅ ( C1 pr
& − C2 ) n − ⋅ ( C1 pr
& 0 − C2 ) n (31)
n + 1 C1 p& n + 1 C1 p&

∂u z
Therefore, when = γ c , the radius at which the critical shear rate occurs is
∂r

γ cn + C2
rc = (32)
C1 p&

∂u z
Since uz is continuous when = γc
∂r
3
n 1  n +1 n +1
 1 C
⋅  ( C1 prc − C2 ) − ( C1 pr0 − C2 )  = 2 ( C3 + 2C4 prc ) 2 − 3 rc + D2
& n & n
2
& (33)
n + 1 C1 p&   6C4 p& 2C4

Substituting equation (32) into equation (33)

3

 C3 ( γ c + C2 ) 1  2 2C4 ( γ c + C2 ) 
n n 2
n 1  n +1 n +1
D2 = ⋅ γ c − ( C1 pr
& 0 − C2 ) n + − 2 C3 +  (34)
n + 1 C1 p&   2C4C1 p& 6C4 p&  C1 

9

298
 ∂u z
Substituting equation (34) into equation (28) for < γc
∂r
3
1 C3 n 1  n +1 n +1

uz = 2
6C4 p&
( C3
2
+ 2C 4
&
pr ) 2 −
2C4
r+ ⋅ 
n + 1 C1 p& 
γ c − ( C1
&
pr0 − C 2 ) n

3 (35)
C3 ( γ + C2 ) 1  2 2C4 ( γ c + C2 ) 
n n 2
c
+ − 2 C3 + 
2C4C1 p& 6C4 p&  C1 


From equation (21)

r0 rc r0
ubulk r02
∫ uz rdr = ∫ uz rdr + ∫ uz rdr =
0 0 rc
2
(36)

Substituting equation (31) and (35) into equation (36)

5 7
 2C4 ( γ cn + C2 )  2  2C4 ( γ cn + C2 )  2
(γ c + C2 ) C3 +
n 2
 C3 − C3 +
7 2
 3
 C1   C1 
 C3  γ cn + C2 
+ −  
30C43C1 p& 3 210C44 p& 3 6C4  C1 p& 

1  n  C3 ( γ c + C2 )
n
1  n +1 n +1
+  ⋅ γ − ( 1 0 2) 
C &
pr − C n +
2  n + 1 C1 p& 
c
  2C4C1 p&
3

1  2C 4 ( γ n
c + C 2 )  2
  γ cn + C2 
2

− 2 C3 + 2
  
6C4 p&  C1    C1 p& 
 

n2 1  2 n +1  γ cn + C2  2 n +1 
+ ⋅ 2 2 r (
0 1 0C &
pr − C 2 ) n −
 γ c  (37)
( n + 1)( 2n + 1) C1 p&   C1
&
p  
n 3
1  3 n +1

− & 0 − C2 ) n − γ c3 n +1 
⋅ 3 3 ( C1 pr
( n + 1)( 2n + 1)( 3n + 1) C1 p&  
n +1   u r2
2
n 1  γ cn + C2 
− ⋅ ( 1 0 2) 0 
C &
pr − C n  r 2
−   − bulk 0
2 ( n + 1) C1 p&   C1 p&   2
=0

In equation (37), γ c , C1 , C2 , C3 , and C4 are given or functions of material properties, and r0 is

determined by the ram geometry. Given a value of ubulk , p& can be solved numerically by

10

299
equation (37). Then, for a specific ram geometry and ubulk , rc is calculated from equation (32)
and p& is solved from equation (27) or equation (37).

3.4 Dynamic response

Consider a particle of compressible material in an incompressible paste. The mixture flows in a

pipe, as shown in Fig. 3.

Fig. 3: Schematic of a particle of compressible material in an incompressible paste flowing

in a pipe.

Applying mass conservation to the incompressible paste

dV ( t )
∫ ua dA − ∫ u p dA =
Aa Ap
dt
(38)

The volume of the particle of compressible material is a function of pressure

V = V ( pt ) (39)

&d
From Section 3.3 and p = pl

p = p ( z , ua ) (40)

When the ram geometry consists of a large radius barrel connected with a small radius needle, it
can be shown that the pressure drop in the barrel can be neglected. This situation can be modeled
as a layer of compressible material (either pure compressible material or a mixture of
11

300
compressible material and incompressible paste) above incompressible paste in the barrel. This
situation is illustrated in Fig. 4.

Fig. 4: Schematic of a situation with a layer of compressible material above incompressible

paste in the barrel.

In this model, p is only a function of ua . In this case, equations (38) and (40), respectively, can
be written as

dV ( t )
(u a − u p ) Ap =
dt
(41)

p = p ( ua ) (42)

If the compressible material in Fig. 4 is pure air with initial layer thickness l0 , equation (39)
becomes

p ( 0 )V ( 0 )
V (t ) = (43)
p (t )

Integrate equation (41) with respect to time and noting that V ( 0 ) = l0 Ap

V ( t ) =  za − z p ( t )  Ap (44)

Combining equations (43) and (44), and solving for p ( t )

p ( 0 ) l0
p (t ) = (45)
za − z p ( t )

12

301
 dza
Substituting equation (45) into equation (42) with the expression ua =
dt

 dz  p ( 0 ) l0
p (t ) = p  a = (46)
 dt  za − z p ( t )

This equation can be rewritten as the following Initial Value Problem

dza  p ( 0 ) l0 
= p −1   (47)
dt  za − z p ( t ) 

where za ( 0 ) = l0 . In this case, za can be obtained by solving equation (47), and ua and Fram ,
dza
respectively, can be obtained from ua = and Fram = p ( ua ) Ap .
dt

4. Results and Discussion

In order to predict the extrusion force, a paste viscosity is obtained by conducting a set of
experiments using die 1 (i.e., rd = 0.548 mm). After determining the paste viscosity model, a
series of experiments are conducted using die 2 (rd = 1.048 mm) to validate the viscosity model.
The dynamic response data is also recorded in these experiments.

Each experiment consists of three extrusion experiments, two without paste and one with paste.
The first experiment is conducted without paste at a ram velocity 0.25 mm/s to test whether the
barrel and plunger have been well lubricated. The second experiment is conducted without paste
using the same ram velocity that will be used in the third experiment, which will be conducted
with paste. The second experiment is used to determine the friction between the barrel and
plunger.

4.1 Viscosity Model

The paste viscosity is tested using the ram extruder depicted in Fig. 1 with die 1. The Herschel–
Bulkley model parameters are identified using steady state forces for different plunger velocities.
The viscosity model parameters are listed in Table 1. The identified model is then compared with
experimental results in Fig. 5. The goodness of fit is validated by a sum of squares due to the
error of 2.78 N2, a correlation coefficient of 0.993, and a root mean squared error of 1.18 N.

13

302
 Fig. 5. Viscosity model test results.

Table 1. Material properties of alumina paste.

Viscosity model
Consistency index k ( kg m ⋅ s ) 102.20
Power law index n 0.4626
Yield stress threshold τ 0 ( Pa ) 454.23
Critical shear rate γ c (1 s ) 10.042

4.2 Steady State

After determining the paste viscosity, tests are conducted using die 2 to experimentally verify the
viscosity model. The viscosity model in equation (4) is used in the proposed mathematical model
to predict the extrusion force. The comparison between validation experiment and analytical data
is shown in Fig. 6.

14

303
Fig. 6: Comparison of experimental and theoretical steady state extrusion forces using die 2.

It can be observed that good agreement is obtained between experiment and simulation results
especially in low velocity area (see Table 2). The maximum absolute percent error is 27.1%. The
difference in high velocity area may be caused by friction between barrel and plunger. Although
the barrel is always well lubricated before each extrusion experiment, when the barrel is loaded
with paste, the lubrication conditions may change. The error bars in Figure 6 are derived from a
series of experiments with a variety of constant plunger velocities where paste was not extruded.
The range of friction for each velocity was determined via the recorded ram forces in these
experiments.

Table 2. Errors for experiments in Figure 6.

Ram Velocity (mm/s) 0.05 0.1 0.15 0.25 0.5 0.75 1

Error (%)| 11.3 4.65 4.57 4.94 18.7 22.5 27.1

15

304
4.3 Dynamic Response

To predict the extrusion force using the developed analytical model, the air volume fraction must
be known. However, the air volume fraction in the paste is difficult to measure in practice and
varies significantly between different batches of paste. To identify the air volume fraction, the
extrusion process is given a set of constant velocities as shown in Figure 7. According to the
dynamic model derived in Section 3.4, the dynamic response time constant is only a function of
initial air layer thickness. Therefore, the time constant of first transient phase (in this case
between 288 and 381 s) is be used to identify the air volume fraction. With the identified air
volume fraction, the extrusion force dynamic response is calculated for different plunger
velocities and compared with the experimental results, as shown in Fig. 7. In this case, according
to the transient section between time 288 to 381 s, the initial air layer thickness is identified as
8.5 mm. The friction is assumed to be a constant 2 N based on the average of all of the friction
data that was collected (not shown here). The friction varies tremendously from experiment to
experiment. The results in Fig. 7 are plotted with a constant friction of 1.5 N. In this case the
simulation shows much better correlation with the experiment.

Fig. 7: Experimental and simulation dynamic extrusion responses with friction = 2 N.

16

305
 Fig. 8: Experimental and simulation dynamic extrusion responses with friction = 1.5 N.

It can be observed that

(1) Good agreement is obtained between the experimental and simulated steady state forces.

(2) Before the large air bubble release, good agreement is obtained between the experimental
and simulated transient responses.

(3) After the large air bubble release, the transient response has a much smaller time constant,
which means the compressibility is mainly caused by air trapped in paste.

(4) The difference of dynamic response at the beginning of extrusion may be due to the fact
that the die is not completely filled by paste.

17

306
5. Summary and Conclusions

An analytical dynamic model was developed to describe the relationship between the ram
velocity and extrusion force in the extrusion of aqueous–based ceramic materials. Based on the
experimental and simulated results, the following conclusions can be drawn:

1. The mathematical model developed in this paper can be used to predict the steady state
extrusion force and dynamic response in ram extrusion process of aqueous–based ceramic
materials.

2. Since the paste viscosity is non–linear, the relationship between extrusion force and plunger
velocity is also highly non–linear.

3. The large time constant of the dynamic response is mainly caused by the air trapped in the
paste during both the paste preparation and loading procedures.

6. Acknowledgement

This project is supported by National Science Foundation (CMMI 0856419) and Center for
Aerospace Manufacturing Technologies at Missouri University of Science and Technology.

7. References

[1] A. Bandyopadhyay, P. Panda, M. Agarwala, S. Danforth, and A. Safari, “Processing of

Piezocomposites by Fused Deposition Technique,” Journal of the American Ceramic Society,
Vol. 80, No. 6, pp. 1366–1372, 2000.
[2] S. Crump, “Appartus and Method for Ceramic Three–Dimensional Objects,” U.S.Patent No.
5121329, 1992.
[3] G. Hilmas, J. Lombardi, and R. Hoffman, “Advances in the Fabrication of Functional Graded
Materials using Extrusion Freeform Fabrication,” Functionally Solid Freeform Fabrication
Symposium, Austin, TX, pp. 319–324, 1996.
[4] M. Cima, M. Oliveira, H. Wang, E. Sachs, and R. Holman, “Slurry–Based 3DP and Fine
Ceramic Components,” Solid Freeform Fabrication Symposium, Austin, TX, 2001.
[5] J. Kruth, P. Mercelis, L. Froyen, and M. Rombouts, “Binding Mechanisms in Selective Laser
Sintering and Selective Laser Melting,” Solid Freeform Fabrication Symposium, Austin, TX,
2004.

18

307
[6] M. Leu, E. Adamek, T. Huang, G. Hilmas, and F. Dogan, “Freeform Fabrication of
Zirconium Diboride Parts Using Selective Laser Sintering,” Solid Freeform Fabrication
Symposium, Austin, TX , 2008.
[7] J. Stampfl, A. Cooper, R. Leitgeb, Y. Cheng, and F. Prinz, “Shape Deposition Manufacturing
of Microscopic Ceramic and Metallic Parts Using Silicon Molds,” U.S. Patent, No. 6242163,
2001.
[8] J. Cesarano III, R. Segalmen, and P. Calvert, “Robocasting Provides Moldless Fabrication
from Slurry Deposition,” Ceramics Industry, Vol. 148, pp. 94–102, 1998.
[9] G. He, D. Hirschfeld, J. Cesarano III, J. Stuecker, “Processing of Silicon Nitride–Tungsten
Prototypes,” Ceramic Transactions, Vol. 114, pp. 325–332, 2000.
[10] T. Huang, M. S. Mason, G. E. Hilmas, and M. C. Leu, “Freeze–Form Extrusion Fabrication
of Ceramic Parts,” Virtual and Physical Prototyping, Vol. 1, No. 2, pp. 93–100, 2006.
[11] M. S. Mason, T. Huang, R. G. Landers, M. C. Leu, and G. E. Hilmas, “Aqueous–based
Extrusion of High Solids Loading Ceramic Pastes: Process Modeling and Control,” Journal of
Materials Processing Technology, Vol. 209, pp. 2946–2957, 2009.
[12] X. Zhao, R. G. Landers, and M. C. Leu, “Adaptive Extrusion Force Control of Freeze–form
Extrusion Fabrication Processes,” ASME Journal of Manufacturing Science and Engineering,
Vol. 132, No. 6, 2010.
[13] J.J. Benbow, T.A. Lawson, E.W. Oxley and J. Bridgwater. “Prediction of Paste Extrusion
Pressure,” American Ceramic Society Bulletin, Vol.68, pp.1821–1824, 1989.
[14] M. Padmanabhan and M. Bhattacharya, “Analysis of Pressure Drop in Extruder Dies,”
Journal of Food Science, Vol. 54, pp. 709–713, 1989.
[15] T. Shepard, E. Nisaratanaporn and H.B. McShane, “Material Flow and Pressure Prediction
when Extruding through Bridge Dies,” Zeitschrift fuer Metallkunde Materials Research and
Advanced Techniques, Vol.89, pp. 327–337, 1998.
[16] D.J. Horrobin and R.M. Nedderman, “Die Entry Pressure Drops in Paste Extrusion,”
Chemical Engineering Science, Vol. 53, No. 18, pp.3215–3225, 1998.
[17] U. Lang and W. Michaeli, “Development of a Mathematical Model for the Calculation of
the Pressure Drop in Extrusion Dies,” Journal of Reinforced Plastics and Composites, Vol. 17,
pp. 1110–1118, 1998.
[18] Y. Y. Li and J. Bridgwater, “Prediction of Extrusion Pressure using an Artificial Neural
Network,” Powder Technology, Vol.108, No.1, pp. 65–73, 2000.
[19] W. Herschel and R. Bulkley. “Measurement of Consistency as Applied to Rubber Benzene
Solutions,” ASTM Part II, Vol. 26, No. 82: 621–629, 1926.

19

308