700 Radl a Noveldesignforhot-meltextrusionpelletizers

8/3/2019 700 Radl a Noveldesignforhot-meltextrusionpelletizers

1/13

A novel design for hot-melt extrusion pelletizers

Stefan Radl a, Thomas Tritthart b, Johannes G. Khinast a,

a Institute for Process and Particle Engineering, Graz University of Technology, A-8010 Graz, Austriab mnadis Melt Extrusion Technologies, Austria

a r t i c l e i n f o

Article history:

Received 8 January 2009

Received in revised form

17 November 2009Accepted 20 November 2009Available online 3 December 2009

Keywords:

Hot-melt extrusion

Pellets

Pharmaceuticals

Simulation

Product processing

Non-Newtonian fluids

a b s t r a c t

In this work we investigated a novel die design for the scale-up of hot melt extrusion (HME) devices for

direct pelletization of pharmaceutics. Therefore we analyzed the temperature distribution in a lab- and

production-scale die as well as melt flow through the die. Finally we explored the possibilities of an

inner rotating knife for stabilizing melt flow. The work was based on computational fluid dynamics for

simulating non-Newtonian melt flow and corresponding temperature fields.

The results show that a tight temperature control of the die material is necessary to guarantee a safe

scale-up of the process. Even in lab-scale applications temperature inhomogeneities have been

observed both experimentally as well as in the simulation. These inhomogeneities act as an trigger to

destabilize melt flow and hence could lead to a shutdown of the process. The proposed inner rotating

knife acts as a pulsating device and consequently is able to enhance process stability. However, due to

heat dissipation in the small gap between rotor and stator, this device has to be fitted with a separate

low-speed drive and cannot be coupled directly to the main extruder shaft.

& 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Melt extrusion processes have been used in industrial

applications for many years, the production of thin films being

only one of the most prominent example. Starting from the

polymer and plastic industry, hot-melt extrusion (HME) has also

found numerous applications in pharmaceutical manufacturing

practice (Breitenbach, 2002; Crowley et al., 2004, 2007; Repka

et al., 2007). By use of melt extrusion, various dosage forms can be

manufactured, ranging from pellets, over granules to tablets and

transdermal drug delivery systems. Compared to other pharma-

ceutical production processes, HME has the benefit of being a

solvent free, environmental friendly and cost-efficient technology.

Furthermore, by HME it is possible to improve bioavailability of

difficult actives by the formation of solid dispersions and solid

solutions. This is relevant for poorly-soluble pharmaceuticallyactive substances, frequently encountered among novel active

ingredients (Doelker et al., 2005; Klein et al., 2007; Miller et al.,

2007). Such benefits have led to an increased interest of HME

technology in recent years.

A typical HME process consists of a feeding system, an

extruder with conveying, mixing and melting section, a die

section as well as further downstream processing units.

A schematic diagram of a HME process is depicted in Fig. 1.

The extruder is divided into a feed, transition and metering zone.

Pitch and helix angle are different in each of these zones anddesigned to allow mixing, compression, melting, and plastification

of the feed material. Finally, the metering zone ensures a constant

flow rate of the melt. Often co-rotating twin-screw extruders are

used due to their superior mixing characteristics. In these

extruders two parallel mounted shafts are driven by the

gearbox and the screw flights are typically fully intermeshing,

i.e., each flight wipes both the element on the adjacent shaft and

the internal surface of the mixing chamber. This setup eliminates

stagnation areas within the extruder and ensures a narrow and

well-defined residence time distribution. The residence time is

typically in the order of 2 min. Thus, thermal stress of heat-

sensitive compounds is minimized and heat-sensitive materials

can be processes without significant reduction of drug activity.

After forming the melt in the die, the thermoplastic strand isforced between calendar rolls to produce films, or is fed into

another device to form pharmaceuticals directly, e.g. pellets or

tablets.

The flow of the melt in extruders has been discussed by

various authors with experimental, theoretical and computational

methods (Bertrand et al., 2003; Carneiro et al., 2004; Khalifeh and

Clermont, 2005; Rauwendaal, 2006). Hence, there has been an

immense interest in melt flow in the extruder section. Also, there

have been numerous studies addressing the details of flow

through the die hole and phenomena like die swelling (Carneiro

et al., 2001; Tome et al., 2007) or shark skinning (Kulikov and

Hornung, 2001; Migler et al., 2002; Molenaar et al., 1998).

ARTICLE IN PRESS

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/ces

Chemical Engineering Science

0009-2509/$- see front matter & 2009 Elsevier Ltd. All rights reserved.

doi:10.1016/j.ces.2009.11.034

Corresponding author. Tel.: + 43316 8737978; fax: +43 316873 7963.

E-mail address: [email protected] (J.G. Khinast).

Chemical Engineering Science 65 (2010) 19761988
http://-/?-http://www.elsevier.com/ceshttp://dx.doi.org/10.1016/j.ces.2009.11.034mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.ces.2009.11.034http://www.elsevier.com/ceshttp://-/?-


2/13

ARTICLE IN PRESS

However, relatively low attention has been paid to the combined

effect of temperature distribution in the die material and the melt

(Lin and Jaluria, 1998; Pittman and Sander, 1994). Specifically in

the area of melt extrusion for pharmaceuticals the literature is

very scarce. Up to now, no study exists that focuses on the flow of

HME drug products through the extrusion head and the die.

For the development of solid dosage forms, the production of

spherical pellets via HME is of interest due to their use incontrolled-release drug delivery systems. These spherical pellets

produced via HME offer additional flexibility for further modifica-

tions, e.g., by coating. Significant contributions in the area of

pelletization have been made by Follonier and co-workers in the

mid-90s (Follonier et al., 1994, 1995). A recent report on the

production of pellets by HME has been published (Young et al.,

2002). Furthermore, various patents for this technology exist

(see for example Rein, 2005). Today, pelletization using HME can

be seen as a promising technology that may have a commercial

breakthrough in the near future.

However, compared to the extrusion of polymers, HME in

pharmaceutical applications is significantly more demanding

as the dosage forms are a mixture of active pharmaceutical

ingredients (API), matrix carriers and other excipients. These

ingredients have to be adjusted to give both excellent therapeutic

as well as adjustable processing properties of the formulation.

Furthermore, the regulatory bodies are increasingly demanding

an enhanced process and product understanding, in line with the

Quality by Design (QbD) initiative. Hence, along with the

knowledge to efficiently manufacture a drug product, also insight

into the process has to be provided. This paradigm shift from trial-

and-error methods to an science-based process design provides

rational connection of process parameters and product quality

attributes.

In this study our aim is the rational design of pharmaceutical

production processes. Specifically, we focus on the design of a

novel die including a direct pelletization step.

1.1. Background

Rheological data and models relevant to hot melt extrusion can

be found in various sources. For example, the effect of drug

content changes on the thermophysical and rheological para-

meters of the formulation has been analyzed recently (Chokshi

et al., 2005). In their work they used the Cross model for

quantifying the shear thinning effect of their melt. Temperature

effects were modeled using the Arrhenius equation. The effect of

different polymers (Eudragit EPO, different polyvinylpyrrolidones

and Poloxamer 188) on a formulation of indomethacin was

studied and they observed a zero-shear viscosity ranging from

5 (formulation with Poloxamer 188 at 60 1C) to over 27,000 Pa s

(formulation with Eudragit EPO and 1201C). The normalized

activation energy Ea/R was in the order of 100010,000 K.

In summary, it was found that the zero-shear viscosity is strongly

influenced by the drug-to-polymer ratio. Hence, viscosity within a

melt can vary significantly, if local composition gradients exist,

which underlines the need for good mixing of the formulation in

HME devices.

Rouilly et al., investigated the shear-thinning behavior of

thermoplastic sugar beet pulp, a material consisting mainly of

polysaccharides (Rouilly et al., 2006). The power law model wasused to quantify the melts rheological behavior between 110 and

130 1C and different moisture contents.

Grosvenor and Staniforth investigated the effect of molecular

weight on the rheological and tensile properties of poly

(e-caprolactone) (PCL) (Grosvenor and Staniforth, 1996). Thissubstance has found widespread use in the pharmaceutical

industry, e.g., as a release agent. They also used the Arrhenius

law to describe the temperature effect of melt viscosity. Viscosity

was in the range of 10150Pa s and the normalized activation

energy Ea/R was in the range of 38004700K. These literature

data show that melt viscosity in extrusion processes spans a wide

range and that the flow is non-Newtonian. Furthermore, local

composition of the mixture significantly changes the rheological

behavior and consequently impacts the flow pattern. Thus,

equipment design must take into account these facts. However,

only little attention has been devoted to these problems in the

literature. Specifically, the rational design of dies for HME drug

extrusion has not been analyzed critically up to now.

The scale-up of HME devices is influenced by many considera-

tions. Clearly, the temperature distribution in the die and the

melt, the mechanical strength of the die and the distribution of

the melt within the device may be of central importance. This is

especially true for large-scale production systems. Equipment for

high-throughput wet mass extruders in the pharmaceutical

industry are known for many years (Erkoboni, 2003). In contrast,

HME equipment for pellet production is designed nearly exclu-

sively for lab-scale applications, because the quality of the

products may be very sensitive to variations in process para-

meters, e.g., the melt temperature, which is difficult to predict.

Clearly, there is a lack of knowledge in the rational design of

large-scale HMEs.

While transport, mixing and energy dissipation are of central

importance for the design of the screw section, the design of the

die impacts strongly the quality, shape and uniformity of the

pellets. Thus, an optimal design of the extrusion die is extremely

important to achieve the desired shape and dimension of the

extrudate. The fundamentals of optimal die design can be found in

Kostic and Reifschneider (2007), as well as in Ghebre-Sellassie

and Martin (2003) or Rauwendaal (2006). Relevant geometries of

the die for HME drug extrusion include flat dies for the production

of films, as well as profile extrusion dies, e.g. for spaghetti-like

products. The flow of the melt through the die will be influenced

by the melts rheological behavior, the channel geometry and theoperating conditions, including flow rate and local temperature.

For this reasons it is virtually impossible to obtain a flow channel

geometry that can be used for a wide range of different products.

Consequently adjustment capabilities are build into the extrusion

die system. This often includes a variable geometry, e.g. by using

so-called choker bars or valves, or a device for controlling the die

temperature, e.g., by using cartridge heaters (Rauwendaal, 2006).

Referring to the former possibility, uniformity of the melt flow is

achieved by the use of choker bars or flex-lips together with a

special design of the manifold for melt distribution. These devices

for controlling the melt flow are located at numerous points along

the width of the die (e.g., in case of a sheet die). The adjustment of

these devices is controlled by scanning the thickness of the

extrudate in case of film extrusion.

gear,motor and

bearing

down-stream

processing

feedingsystem

extruder die

Fig. 1. Schematic diagram of a hot melt extrusion system.

S. Radl et al. / Chemical Engineering Science 65 (2010) 19761988 1977


3/13

ARTICLE IN PRESS

In addition to mechanical design strategies, there have been

several efforts to prevent extrusion dies from freezing using

controlled heat transfer. In underwater pelletization of polymers

the die has been insulated on the exit side to reduce heat transfer

to the liquid media (Jackson et al., 2007). Alternatively, the liquid

media can be heated. However, this has the drawback of increased

energy consumption. Furthermore, the melt temperature can be

controlled using different die metals (Bertolotti, 1989).

In summary, an ideal die design will: (Kostic and Reifschneider,2007)

balance the melt flow to provide a uniform exit velocity acrossthe entire die exit,

achieve a minimal pressure drop, avoid abrupt changes that may cause stagnation areas and

thermal degradation,

allow adjustment during production by inclusion of flow-control devices for optimization of the flow distribution,

have a modular design for better manufacturability, assembly,cleaning and convenient modification and

have a so-called die land, i.e., a region upstream of the die,which has a length of at least 10 times the product thickness in

order to facilitate polymer melt stress relaxation.

A wide range of designs for extrusion dies can be found in

literature. However, their application to pharmaceutical manu-

facturing processes is limited.

1.2. Objectives

The objective of this study is to develop a method to

computationally assess the feasibility of different complex die

designs, which we will apply to a new HME drug extrusion

process with direct granulation. Direct granulation refers to a

novel process, where the hot, still molten extrudate, is cut directly

after exiting the die. Cutting of the molten strand is achieved with

a rapidly rotating cutter knife. By doing so, the granules will formperfectly shaped pellets (micro pellets) immediately due to the

action of surface tension and the shrinkage due to solidification.

Hence, there is no need for a subsequent spheronization step. This

process has the advantage of not requiring a further melting of the

product, thus lowering equipment costs and reducing energy

demand. It also allows effective integration in a continuous

manufacturing environment.

In order to study in detail the flow characteristics in the

extrusion die device, a computational tool has been developed.

The momentum and mass conservation equations, as well as the

energy equation, have been solved to calculate the velocity,

pressure and temperature field in the melt as well as in the

extrusion die. Melt rheology was described by a shear-rate and

temperature-dependent viscosity. Viscoelastic effects have beenexcluded from this work.

The die design studied in this work is based on a cylindrical

shape of the extrusion device, i.e., the melt is entering the die in

axial direction and exiting in radial direction. To facilitate melt

flow, the effect of an additional rotor that is located inside the

extrusion die was investigated. This rotor works as a pulsation

device to clear plugged die holes. Consequently, the robustness of

the process can be increased and a continuous operation can be

ensured.

Requirements for the process design include:

Melt pressure and temperature have to be constant for eachindividual die hole to ensure a uniform melt distribution and

pellet diameter.

The design must allow a tight temperature control of the diematerial.

Pressure loss across the die should be minimized to allowa high throughput, i.e., the thickness of the die should be

minimized.

The residence time in the extrusion device should be short(12 min) and the residence time distribution should be

narrow.

This paper is structured as follows: first we provide some

background information on the computational method used in

this work. The available rheological models and data for

pharmaceutically relevant melts are highlighted. In the results

section we first discuss the temperature distribution in a lab-scale

and the novel production-scale extrusion device. Finally, we focus

on the production scale extruder and investigate the effect of a

knife rotating inside the extrusion die.

2. Materials and methods

2.1. Materials

The polyol D-Mannitol (CPharmMannidex 16700), sorbitol

(CPharmSorbidex S 16606) and the polysaccharide maltodextrin

(CPharmDry 01980 Maltodextrin DE 8) were purchased from

Cerestar Austria Handelsgesellschaft m.b.H. Glucono-d-lactone

(F2500) was supplied by Jungbunzlauer Austria AG.

2.2. Model substance

The model substance used within this work consisted of 37 w%

D-Mannitol, 38 w% Glucono-d-Lactone, 20 w% maltodextrin and

5 w% sorbitol. The rheological behavior of the melt was deter-

mined at different temperatures using a high-pressure capillary

rheometer (Rheograph 2002, Gottfert GmbH, Germany) according

to DIN 53014. The density of the melt was measured using aPVT-100 (SWO Polymertechnik GmbH, Germany). The heat

conductivity was determined with a K-System II (Advanced CAE

Technology Inc., USA) according to ASTM D5930-97. All measure-

ments were performed at the University of Leoben (Schuschnigg

et al., 2007). The models used for the description of the melt

rheology as well as the numerical values for the physical

properties of the melt are described at the end of Section 2.3 of

this paper.

2.3. Computational method

In order to compute the time-dependent velocity and

temperature distribution of the melt, the finite volume method

was used to solve the underlying equations for mass, momentumand energy conservation. Assuming an incompressible media, the

continuity equation can be written as:

r ~u 0: 1

where ~u denotes the velocity vector. The momentum equation

can be written as

@r ~u@t

~u rr~u rp r m~x r~u rT~u: 2

Here m denotes the dynamic viscosity that depends on the shearrate and the local temperature. r is the density and p is thepressure.

In our study a rotating knife has been considered within the

cylindrical die. Therefore, it was necessary to introduce as second,

rotating reference frame. In this rotating reference frame,

S. Radl et al. / Chemical Engineering Science 65 (2010) 197619881978


4/13

ARTICLE IN PRESS

characterized by the angular velocity vector ~O, the centrifugal

force Fcent r ~O ~O ~r and the Coriolis force Fcor 2 r ~O ~u have to be considered. In this case, the momentum

equation can be written as: (Brenn, 2004)

@r ~u

@t ~u rr~u rp r m~x r~u rT~ur ~O

~O ~r2 r ~O ~u; 3

By making this equation dimensionless the relative impact of

these additional forces becomes apparent:

Re @~u

@t~u

r~u

rp r2~u

1

Ek Ro ~O

~O

~r

2 Re

Ro~O

~u

; 4

where Ro=U/(O L) is the Rossby number (i.e., inertial to Coriolis

forces). Ek=m/(r O L2) is the Ekmann number (i.e., viscous toCoriolis forces). In the current framework Re will be very small, Ro

will be O(1), because U will be in the order ofO L (note that the

rotating motion is fast compared to the axial fluid flow in

the extrusion head). The Ekmann number will be in the order of

1/Re, i.e., it will be high. Hence, centrifugal and Coriolis forces canbe safely neglected in the current work. This allows a more

efficient implementation in the simulation software.

Following Bird et al., the energy equation for an incompressible

fluid with constant heat capacity cp and thermal conductivity l is:

(Bird et al., 2002)

r @h

@tr ~u rh

l

cp r2ht : r~u: 5

In Eq. (5) the term t : r~u denotes the heat of dissipation, where tis the stress tensor. The enthalpy h is related to the temperature

field, i.e.,

h cp TT0: 6

In Eq. (5) the stress tensor t can be calculated from the rheologicalproperties of the melt and the velocity field. In the solid domain(we solve both for the die material and the die flow) the velocity

vector ~u in Eq. (5) was set to zero, which resulted in the Laplace

equation in case of the steady-state cases. The enthalpy equation

was solved simultaneously with the fluid dynamics, as the

rheology is strongly dependent on the temperature. As the

temperature field in the fluid and solid domain was calculated

separately, an iterative technique was employed to couple these

two regions. This was done by imposing the calculated tempera-

ture boundary field from the solid domain on the fluid domain. To

guarantee the heat balance at the boundary, the heat fluxes in the

fluid and solid domain must be equal, i.e.

ls @Ts

@~n

lf @Tf

@~n

; 7

where the index s stands for the solid domain, frefers to the fluid

domain and ~n is the unit normal vector of the boundary. With

these boundary conditions the temperature field in the fluid

domain can be calculated. Subsequently, the temperature bound-

ary conditions were updated in the solid region and the cycle was

repeated until the difference in the calculated heat fluxes was

below a certain threshold.

The thermal conductivity of the melt was assumed to be

constant at lf=0.205W/mK, which was supported by the

measurements. The density was measured to be approximately

constant and 1480 kg/m3. The heat capacity was estimated to be

1260J/kgK=m2/s2/K and was based on that of saccharose at room

temperature. A modification of the Carreau model was used to

account for the effect of shear rate _g 1=s and temperature T (K)

on the local viscosity, i.e., m m _g; T, i.e.,

m _g; T aT m0minf 1 B aT _gCD1=C; 8

where the dimensionless quantity aT is the temperature correc-

tion function defined as

aT expEaR

1

T

1

T0

: 9

Eq. (9) is based on the Arrhenius equation with the normalized

activation energy Ea/R. The constants describing the fluid under

investigation where given as:

m0=1680Pasminf=0P asB=2.50 103 s

C=3.88

D=9.9 102

Ea/R=2.13 104 K

T0=378.15 K

As simulation software the open-source CFD package OpenFOAM has

been used. This software enables easy modification of the governing

equations and is numerically efficient. For mesh generation Open-

FOAMs internal mesh generator blockMesh as well as CUBIT have

been used (Blacker, 2007). The computational mesh was designed

such that the flow field and the temperature gradients were captured

well. To obtain a mesh independent solution, the mesh has been

locally refined. The largest meshes (for 3D simulations) consisted of

approximately 280,000 cells, whereas for the 2D simulations the

maximum cell number was around 60,000 cells.

The solver has been verified against an analytical solution for

isothermal non-Newtonian flow. For this case the solution for the

steady-state flow of a fluid in a straight pipe with circular cross

section can be evaluated from: (Bohme, 2000)

Qp

8

d3

t3w

Ztw

0

t2 _gt dt 10

In Eq. (10) Q denotes the flow rate (m3/s), d is the pipe diameter

(m), tw is the wall shear stress (Pa) and _gt is the inverse

Fig. 2. Results for isothermal non-Newtonian pipe flow (line: analytical solution,

symbols: numerical solution; 1 mm pipe diameter, 3 mm pipe length, 378.15K

melt temperature, viscosity according to Eqs. (8) and (9)).



5/13

ARTICLE IN PRESS

viscosity function, i.e., the function describing the rheological

behavior of the fluid as given by Eq. (8). The wall shear stress twdepends linearly on the pressure drop Dp over a pipe with length

Dl and hence is known. A simple force balance yields:

tw Dp

Dl

d

4: 11

The comparison of the simulation results with the analytical

solution is presented in Fig. 2. As can be seen, an excellent

agreement is obtained.

3. Results

In order to asses the novel concept for the die, numerous

simulations of the non-isothermal melt flow in the novel

cylindrical die were performed.

3.1. Flow through the die hole

The simulations of the non-isothermal melt flow through the

die hole showed a local temperature maximum near the wall. This

is a phenomena well known in the literature (Ghebre-Sellassie

and Martin, 2003) and will not be discussed in more detail. Thepressure drop over the final die hole (diameter=1 mm, length= 3

mm) was in the order of 120 bar at the design melt temperature of

95 1C and a flow rate of 0.25 kg/h. The maximum temperature rise

was about 7 K. A typical result for the temperature distribution is

shown in Fig. 3.

3.2. Temperature distribution in a lab-scale die

First, the temperature distribution in the planar lab-scale die

plate (refer to Fig. 4) was analyzed. The die plate consists of a cone

that acts as a distributor for the melt (Fig. 4 right) and a planar

plate with 16 die holes. To reduce pressure drop, the die holes

consist of a 5 mm diameter pilot bore and a 1 mm diameter exit

bore.

The die plate is mounted directly to the extruder head via sixbolts (Fig. 4 left) and is in thermal contact with the extruder,

which was temperature controlled (386.15 K). An isometric view

that shows the die plate assembled with the extruder head and

the melt flow is presented in Fig. 5. Because of the symmetry of

the plate, only 1/6 of the plate was modeled and subsequently

simulated.

Since the extruder head, the die plate and the melt are in

thermal contact, it is important to understand the heat exchange

between these three regions. Hence, it was necessary to simulate

heat and melt flow simultaneously. Furthermore, the rotating

cutter knife causes a turbulent air flow at the front side of the die

plate. Hence, it is essential to take into account the convective

heat transfer from the die plate to the surrounding air. Thus, the

boundary conditions for this situation were chosen as:

Convective heat transfer from the front side (melt exit, lightregion in Fig. 6) of the die plate to the surrounding air

(T=293.15 K). The heat transfer coefficient a to thesurrounding air was obtained from an idealized assumption

of air flow over a flat plate. This analysis showed thata is in therange between 50 and 80 W/m2 K under operating conditions.

The surfaces in contact with the screws have the temperatureof the extruder head (dark region in Fig. 6).

The melt entering the extrusion device has the sametemperature as the extruder head (368 K).

The extruder head surface temperature is constant on the fulllength up to the die plate.

The results for the temperature distribution are shown in Fig. 7 for

both heat transfer coefficients of 50 (left) and 80 W/m2 K (right).

The figure shows an axial cross section through the extrusion

head, the melt region and the die plate. As can be seen, the

temperature distribution especially in the die plate is very

inhomogeneous. This is true for both heat transfer coefficients

studied, but is more pronounced in the case of a=80W/m2 K asexpected. Clearly, the die plate is virtually insulated from the

extrusion head by the melt channel. Consequently, heat flow to

the conical distributor of the die is limited and this part is

significantly cooler. As the flow rate of the melt is relatively low

(in our case 2 kg/h), the hot melt cannot heat the conical

distributor. In the contrary, the melt is cooled to some extend at

the inner surface of the channel. These results are in goodagreement with experimental observations that showed

significantly lower temperatures at the center of the die plate

(Tritthart, 2007, personal communication). The thermal situation

also affects the flow in the melt channel, i.e., the velocity profile in

the melt distributor. This is due to lower melt velocities at the

cooler side of the channel, which is the consequence of the higher

melt viscosity (results not shown). Also the pressure drop in the

melt distributor was increased by 34% (!) as the heat transfer

coefficient was changed from 50 to 80 W/m2 K. Hence, a small

change in the external heat transfer to the surrounding air causes

significant variability in the melt flow, indicating the sensitivity of

the process to environmental characteristics.

Above mentioned computations are rather expensive, as the

grid resolution in the fluid region (the melt) must be fine to

Fig. 3. Temperature distribution in a capillary die hole (flow is entering from the

top, 1 mm hole diameter, 3 mm length).



6/13

ARTICLE IN PRESS

capture the details of the temperature field. Therefore, we have

also tested what happens when we treat the melt as a static

insulation layer. The extrusion head temperature for this situation

was assumed to be constant on the outer side of the melt channel,

which is in fairly good agreement with the fully coupled

simulation (see Fig. 7). With these boundary conditions, we have

calculated the temperature field in the die plate only. In Fig. 8 weshow the results for this simplified case. Clearly, we observe the

same characteristics of the temperature distribution as in the fully

coupled simulation. Due to the assumption of a constant

extrusion head temperature the minimal temperature for the

simplified case is 34K higher. This is usually an acceptable

deviation, as the uncertainty introduced by the assumed heat

transfer coefficient is also significant.

3.3. Temperature distribution in a production-scale die

In this chapter we focus on a new design of a HME device

for use at the production scale. Specifically, we are first inte-

rested in the temperature distribution in such a device. Fig. 9

shows an exploded view of this new design consisting of a

cylindrical die together with an inner rotor. This rotor has thefunction to create a pressure pulse that facilitates melt flow

through the die holes (a detailed discussion on this is provided in

the next chapter).

As the investigations related to the lab-scale die showed that a

homogenous die temperature is critical, we first focused on the

effect of heating channels in the die. These channels consist of

axial bores in the die (see Fig. 9) which are thermo-regulated by

means of liquid flow through the bores. The flow rate through the

channels was designed such that the wall temperature of the axial

bores can be assumed as constant. Consequently, the only factor

that effects the die temperature is the number and arrangement

of the channels in the die.

In Fig. 10 temperature contour plots for two different

arrangements of the heating channels are shown. It was

Fig. 4. Area in contact with the melt for the lab-scale planar extrusion head (left: view onto the outlet side; right: view in melt flow direction).

melt inflowfrom extruder

die plate

die hole extrusionhead

meltchannel

Fig. 5. Schematics of the die plate, the extruder head and the melt (isometric

view).

Fig. 6. Boundary conditions at the front side of the die plate, black: constant

temperature (melt temperature), grey: convective heat transfer to the surrounding

air.



7/13

ARTICLE IN PRESS

assumed that the rotor is in thermal contact with the cylindrical

die. Hence, this can lead to an unwanted cooling of the melt, as

can be seen in Fig. 10 (left).

The calculations show that heating channels are necessary

next to each row of die holes to allow a tight control of the

temperature (Fig. 10, right). As can be seen from Fig. 10 (right) the

lowest temperature in this case is located near the outlet of

the melt. Here the die temperature has a local minimum that is

about 1.5 K below the melt temperature. A more precise tempe-

rature control does not seem feasible due to limited manufactur-

ability and the fact that the cylinder cannot be insulated to thesurrounding air.

If an insufficient number of heating channels is used and the

end plate is not temperature controlled (Fig. 10, left), a significant

temperature gradient over the die exit is observed. This could

lead to partial solidification as observed during the lab-scale

experiments.

3.4. Pressure distribution around an inner rotating knife

The new cylindrical design for a HME device incorporates a

rotating knife that aids melt flow through the die holes. This is

realized by means of a pressure pulse that periodically increases

the pressure in front of certain die holes. Consequently the flow

extrusionhead

melt

dieplate

flowdirection

flowdirection

Fig. 7. Temperature distribution in the die plate and the extrusion head (coupled simulation; left: a=50 W/m2 K, right: a=80W/m2 K, temperature contour lines areseparated by 0.5 K).

Fig. 8. Temperature distribution for the simplified case (left: a=50W/m2 K, right: a=80W/m2 K, temperature contour lines are separated by 0.5 K).

cylindrical

die

rotor

melt

inlet

melt

discharge

heating

channels

Fig. 9. 3D view of the cylindrical die with inner rotating knife.



8/13

ARTICLE IN PRESS

through these holes will be higher for a short period, thus spilling

unwanted accumulations from the system.

The rotor itself consists of a cylinder and multiple straight

knives located on the perimeter of the cylinder, covering the full

axial length. A two-dimensional sketch of the rotor geometry is

provided in Fig. 11.

To asses the impact of the rotors geometry, simulations have

been conduced. Here the main focus was the pressure profile of

the pulsating stream induced by the rotor. In the simulations the

flow in slices perpendicular to the cylinder axis was analyzed. The

geometrical parameters are summarized in Table 1. The boundary

conditions were chosen to mimic a die that has a constant

temperature. Because the simulations were conducted in 2D only,

rotor

cylindricaldie

contact line

temperature-controlled end

plate

Fig. 10. Temperature distribution in the production-scale die with one heating channel per two die row (left) as well as one heating channel per die row and a

temperature-controlled end plate (right). Rotor and cylindrical die have been assumed to be in thermal contact along the full contact line.

Fig. 11. Geometry of the rotor, the knife and the cylindrical die for the 2D

simulations.

Table 1

Geometrical parameters for the 2D simulations.

Gap (distance knifecylinder) 13 mm

Knife angle 30901

Rotor diameter 8095 mm

Stator diameter 100 mm

Fig. 12. Pressure variation versus angular position for different gap widths (2D

simulation, 601 knife angle, 80 mm rotor diameter, 60 min1).

Fig. 13. Pressure variation versus angular position for different gap widths (2D

simulation, 601 knife angle, 95 mm rotor diameter, 60 min1).



9/13

ARTICLE IN PRESS

the effects of temperature build-up along the cylinder axis could

not be investigated. Hence, the 2D simulations mimic a situation

where only radial heat transfer can take place. The rotational

speed was varied between 20 and 60 min1. The rotor and stator

diameter, the knife angle and the gap as well as the angular

position relative to the knifes edge are shown in Fig. 11. Results

for a knife angle of 601 are shown in Figs. 12 and 13 for two

different rotor diameters. In these figures the pressure variation

refers to the difference between the pressure at the inner shell ofthe cylindrical die and the mean pressure in the die.

The four curves in Figs. 12 and 13 indicate four different rotor-

to-stator gaps. The angular position is relative to the knife edge in

circumferential direction (see Fig. 11). As can be seen from Fig. 12,

when the gap decreases, the pressure peak becomes more and

more localized. Hence, only the die hole near the knife edge will

experience a short pressure variation while all others are nearly

unaffected by the rotating knife. This is beneficial, since the pellet

size distribution will be more uniform in this case with only a few

slightly larger particles. A simple analysis of the simulated

pressure curve shows, that when the pressure varies as in Figs. 12

and 13, the maximal deviation from the mean pellet diameter is

73% for all gap sizes investigated.

In case the inner rotor diameter is increased while all other

geometrical parameters are held constant, this situation changes.As can be seen from Fig. 13 the shape of the pressure profile

changes to a more flat one. This is because the height of the rotor

knife relative to the rotor-to-stator gap is decreasing with an

increasing rotor inner diameter. In this situation there will be no

sharp pressure peak and all die holes on the perimeter will be

influenced by the rotating knife. In addition, for the cases of a

Fig. 14. Temperature (left, with streamlines) and pressure (right) distribution near the knife (2D simulation, 60 1 knife angle, 80mm rotor diameter, 0.75 mm knife gap,

60min1).

Fig. 15. Temperature (left, with streamlines) and pressure (right) distribution near the knife (2D simulation, 60 1 knife angle, 95mm rotor diameter, top: 1 mm knife gap,

bottom: 1.75mm knife gap, 60 min1

).



10/13

ARTICLE IN PRESS

rotor-to-stator gap of 1.5 and 1.75 mm and a rotor diameter of

95mm, we observe that the pressure is negative in front and

behind the blade (see Fig. 13). This is because in both cases, the

high pressure zone is localized near the knife edge and the zone

near the cylindrical die is nearly unaffected by the knifes

movement (see Fig. 15). The pressure build-up in circumferential

direction is very small and the maximum (positive) pressure is

observed between two consecutive knifes at the perimeter (i.e., at

an angular position of 901

in case of two knifes, data not shown).This peculiar behavior is caused by the extremely small height of

the rotor knife relative to the rotor-to-stator gap.

In Figs. 14 and 15 we compare the temperature and pressure

distribution in the two-dimensional plane for different rotor

diameters and knife gaps. In addition, we show the effect of the

rotor-to-stator gap on the flow field in Fig. 15. As can be seen in all

of these figures, the maximal melt temperature is about 30 K

above the wall temperature. Also in the case of a larger rotor

diameter, i.e., 95 mm, the temperature rise is significant and

above 20 K. Hence a significant amount of heat is generated at

the relevant rotational speed, which cannot be removed with the

proposed rotor diameters. In Fig. 14 it can be seen that if the gap is

sufficiently small, a local pressure peak is built up in the vicinity

of the knife. However, for a given gap size this localized pressure

peak vanishes if the rotor diameter is increased, while the overall

pressure drop over the knife is similar to the case of a smaller

rotor diameter (compare Figs. 14 and 15). Furthermore, if we

increase the rotor-to-stator gap (see Fig. 15), the flow pattern

changes and the recirculation zone near the knife edge is no

longer observed. Thus, the melt flow is relatively unaffected by

the knife and is not pushed against the cylindrical die. Such an

arrangement causes a very localized pressure near the knife edge

but not at the cylindrical die. Hence, excessively large rotor-to-

stator gaps are ineffective for the generation of the desired

pressure pulse.

3.5. 3D simulation results

The results of the two-dimensional simulations showed that a

knife angle of 601 and a rotor-to-stator gap smaller than 1 mm is

necessary to create the desired pressure peak needed to remove

plugs from the die hole and to ensure smooth and continuous

operation. However, to assess the temperature distribution of the

melt in axial direction, full three-dimensional simulations are

necessary. The parameters for these 3D simulations are summar-ized in Table 2.

As an additional parameter for the geometry of the internal

rotor, the pitch p of a helically-shaped knife has been studied. The

pitch is the axial distance of a (hypothetical) point traveled during

a single revolution of a helix. p is inversely proportional to the

angle f between the axis of the cylindrical device and the knife

edge. The relationship between pitch and the axis-knife-edge

angle is

cotf p

p D12

where D is the outer knife edge diameter.

The boundary conditions for the 3D simulations consist of a

uniform pressure and temperature at the melt inlet. The velocity

gradient has been set to zero at the inlet. While the melt flow has

been specified to be normal to the cylindrical outlet surface, the

velocities in tangential and axial direction at this surface have

been set to zero. At all other surfaces the no-slip boundary

condition has been applied. This is supported by our experimental

results that showed generally smooth surfaces of the product, that

indicate that a stick-slip transition has not occurred in the die.

Also, uniform wall temperatures have been used in the simula-

tions.

The results for a rotor speed of 60 min1 are shown in Fig. 16

for a straight knife. As can be seen, the velocity vectors of the melt

(Fig. 16, left) are essentially perpendicular to the knife, i.e., the

main flow consists of a circular motion. This is because the mean

flow velocity in axial direction is very slow. Only at the inlet of the

melt, where a uniform pressure boundary condition has beenused in the simulation, the local pressure before the knife leads to

an outflow. At the backside of the knife melt is locally sucked into

the extrusion device.

The temperature distribution (Fig. 16, middle) shows a

significant temperature gradient in axial direction. The maximal

Table 2

Geometrical parameters for the 3D simulations.

Gap (distance knifecylinder) 1 mm

Knife angle 30 1C

Die diameter 45100 mm

Pitch mm 1000 to +3000mm

Fig. 16. Results for straight knife at 60 rpm (left: velocity vectors and pressure contour, middle: temperature, right: pressure).



11/13

ARTICLE IN PRESS

temperature is more than 10 K below the prediction of the two-

dimensional simulation (see Fig. 15). However, the temperature

heterogeneities inside the device are still significant and would

lead to considerable differences of melt outlet velocity. Also, the

temperature maximum is now closer to the wall compared with

the two-dimensional case. This is because there is an additional

convective energy transport in radial direction that was not

included in the 2D simulation. Furthermore, the pressure

distribution (see Fig. 16, right) indicates that the pressure lossin axial direction is negligible. In the vicinity of the knife we

observe that the local pressure maximum is becoming smaller in

axial direction. This is due to the change in melt viscosity as a

consequence of the increasing temperature. Thus, the higher

temperature leads to a lower melt viscosity and a lower pressure

build up near the knife edge.

Furthermore, the flow and temperature distribution has been

investigated at a lower rotor speed of 20 min1 for a pitched knife

(Fig. 17). The results show that the temperature difference over

the full length for this situation is about 11 K. As can be seen from

Fig. 17 (right), this has the positive effect of a more uniform

distribution of the pressure maximum in front of the knife.

Interestingly, the pressure maximum is nearly the same as for a

rotor speed of 60 min1.

The influence of the knife pitch has been investigated for the

case of a rotor speed of 20 min1. However, the simulations

showed that the differences in general are small. The only

observation that can be made is that for the smallest pitch

(Fig. 17) the peak pressure is more uniform along the axis. The

disadvantage in this case is the increasing in- and outflow at the

melt inlet that may lead to an unwanted oscillatory fluid motion

in the inlet channel upstream of the extrusion device.

The power drawn by the melt and the torque needed to turn

the rotor have been investigated as a function of rotor speed and

pitch. As can be seen from Table 3 the differences in the torque

requirements between the two rotor speeds are small, i.e., a lower

rotor speed does not decrease the torque accordingly. This is due

to the shear-thinning behavior of the melt. The influence of the

pitch is small (o5%) for a rotor speed of 20 min1 and higher

(o50%) for 60min1 because axial pumping increases with rotor

speed.

An alternative design of the rotor has been studied with a

smaller diameter of 60 mm and a rotor speed of 120 min1 (see

Fig. 18). The main idea for this design is to mount the rotor

directly to the extruder, i.e., the rotor does not necessarily need a

separate drive. To keep the throughput identical to the large

stator diameter of 100 mm, the cylindrical die has to be designed

longer. The results show that under this conditions the melt is

heated significantly (about 20K), which is unacceptable to

guarantee a uniform temperature along the cylinder axis. The

results for a stator diameter of 45 mm are shown in Fig. 19. The

torque requirement for this setup was 45 Nm, the power demand

is 565W. Also for this design heat-up of the melt is significant(18 K) along the axis. Hence, it is not possible to mount the rotor

directly to the extruder, even if the diameter of the cylinder is

reduced.

4. Discussion

Pelletization via hot melt extrusion has a significant potential

for becoming one of the primary production process for solid

dosage forms. However, the process scale-up (in addition to

challenges like cleaning in place, CIP) is one of the most important

problems that impedes a breakthrough of this technology.

Within this work the importance of a proper temperature

control of extrusion dies has been highlighted. It could be shown

that in the lab-scale setup the low thermal conductivity of themelt as well as the heat transfer from the die plate lead to

an undesired thermal situation. This can lead to a partial

solidification of the melt which may result in an unstable flow

through individual die holes. Observations during lab-scale

tests supported this speculation. Furthermore, the temperature

Fig. 17. Results for pitched knife (p= +1000 mm) at 20 rpm (left: velocity vectors and pressure contour, middle: temperature, right: pressure).

Table 3

Torque requirements for 100 mm stator diameter.

Case Power demand (W) Torque (Nm)

60 rpm, straight 499 79

60 rpm, pitch+1000 mm 631 100

60 rpm, pitch+2000 mm 593 94

60 rpm, pitch+3000 mm 594 9560rpm, pitch 3000 mm 730 116

20 rpm, straight 202 96

20 rpm, pitch+1000 mm 195 93

20 rpm, pitch+2000 mm 198 95

20 rpm, pitch+3000 mm 198 95

20rpm, pitch 3000 mm 204 97



12/13

ARTICLE IN PRESS

distribution in the die material might become even worse during

scale-up. This is a critical point, since melt rheology is very

sensitive to temperature changes.

The investigation of the concept of a cylindrical die has shown

that this design is a feasible option for increasing the throughput

for highly viscous HME formulations. An inner rotating knife can

be used as a pulsating device for improving melt flow and

consequently for increasing process stability. However, there are

some limitations. Viscous dissipation heats up the melt along theaxis of the cylinder. This can have the effect of worsening the melt

distribution among the die holes. Hence, the rotor speed has to be

decoupled from the speed of the extruder. Second, a proper stator

diameter has to be chosen. This decision is mainly influenced by

considerations on the manufacturability of the die.

Notation

aT parameter of the Carreau viscosity model

B parameter of the Carreau viscosity model, s

cp specific heat capacity, J/kg K

C parameter of the Carreau viscosity model

d diameter, m

dr rotor diameter, m

ds stator diameter, m

D parameter of the Carreau viscosity model

Ea activation energy, J/mol

Ek Ekmann number

g radial gap, m

h specific enthalpy, J/kg

l length, m~n unit normal vector, m

p pressure, Pa

p pitch, m

q heat flux, W/m2

Q volumetric flow rate, m3/s

R gas constant, J/mol K

Re Reynolds number

Ro Rossby number

t time, s

T temperature, K

T0 reference temperature, K~u velocity vector, m/s

U reference velocity, m/s

Greek letters

a knife anglea heat transfer coefficient, W/m2 Kb angular position_g shear rate, s1

l heat conductivity, W/m K

m viscosity, Pa sm0 viscosity at zero shear rate, Pa sminf viscosity at infinite shear rate, Pa sr density, kg/m3

t stress tensor, Patw wall shear stress, Paf angle between the axis of the cylindrical device and the

knife edge~O angular velocity vector, s

1

Acknowledgments

JGK acknowledges partial funding of this work through the EU

Marie Curie Chair program MEXC-CT-2004-006767. Furthermore,

we acknowledge the financial support by mnadis Melt Extrusion

Technologies.

Fig. 19. Temperature distribution for straight knife at 120rpm (45mm stator

diameter, 1mm gap).

Fig. 18. Temperature distribution for straight knife at 120rpm (60 mm stator diameter, left: 1 mm gap, right: 2mm gap).



13/13

ARTICLE IN PRESS

References

Bertolotti, F., 1989. Die for Hot Die Face Cutting Thermoplastic Polymers, EuropeanPatent EP0152844.

Bertrand, F., Thibault, F., Delamare, L., Tanguy, P.A., 2003. Adaptive finite elementsimulations of fluid flow in twin-screw extruders. Computers & ChemicalEngineering 27, 491500.

Bird, R.B., Stewart, W.E., Lightfood, E.N., 2002. Transport Phenomena. Wiley, NewYork.

Blacker, T., 2007. Cubit Users Manual, Sandia National Laboratories.Bohme, G., 2000. Stromungsmechanik nichtnewtonscher Fluide. B.G. Teubner,

Stuttgart.Breitenbach, J., 2002. Melt extrusion: from process to drug delivery technology.

European Journal of Pharmaceutics and Biopharmaceutics 54, 107117.Brenn, G., 2004. Stromungslehre und Warmeubertragung II VT. Graz University of

Technology, Graz.Carneiro, O.S., Covas, J.A., Ferreira, J.A., Cerqueira, M.F., 2004. On-line monitoring of

the residence time distribution along a kneading block of a twin-screwextruder. Polymer Testing 23, 925937.

Carneiro, O.S., Nobrega, J.M., Pinho, F.T., Oliveira, P.J., 2001. Computer aidedrheological design of extrusion dies for profiles. Journal of Materials ProcessingTechnology 114, 7586.

Chokshi, R.J., Sandhu, H.K., Iyer, R.M., Shah, N.H., Malick, A.W., Zia, H., 2005.Characterization of physico-mechanical properties of indomethacin andpolymers to assess their suitability for hot-melt extrusion processs as ameans to manufacture solid dispersion/solution. Journal of PharmaceuticalSciences 94, 24632474.

Crowley, M.M., Fredersdorf, A., Schroeder, B., Kucera, S., Prodduturi, S., Repka, M.A.,McGinity, J.W., 2004. The influence of guaifenesin and ketoprofen on the

properties of hot-melt extruded polyethylene oxide films. European Journal ofPharmaceutical Sciences 22, 409418.Crowley, M.M., Zhang, F., Repka, M.A., Thumma, S., Upadhye, S.B., Battu, S.K.,

McGinity, J.W., Martin, C., 2007. Pharmaceutical applications ofhot-melt extrusion: part I. Drug Development and Industrial Pharmacy 33,909926.

Doelker, E., Bilati, U., Nguyen, C.A., Galindo-Rodriguez, S., Sarraf, A.G., 2005.Processing of polymeric dosage forms for advanced drug delivery: from melt-extrudates to nanoparticles. Chimia 59, 336339.

Erkoboni, D.F., 2003. Extrusion/Spheronization. In: Ghebre-Sellassie, I., Martin, C.(Eds.), Pharmaceutical Extrusion Technology, New York, U.S., pp. 277322.

Follonier, N., Doelker, E., Cole, E.T., 1994. Evaluation of hot-melt extrusion as a newtechnique for the production of polymer-based pellets for sustained-releasecapsules containing high loadings of freely soluble drugs. Drug Developmentand Industrial Pharmacy 20, 13231339.

Follonier, N., Doelker, E., Cole, E.T., 1995. Various ways of modulating the releaseof diltiazem hydrochloride from hot-melt extruded sustained-releasepellets prepared using polymeric materials. Journal of Controlled Release 36,243250.

Ghebre-Sellassie, I., Martin, C., 2003. Pharmaceutical Extrusion Technology. MarcelDekker Inc., New York.

Grosvenor, M.P., Staniforth, J.N., 1996. The effect of molecular weight on therheological and tensile properties of poly (epsilon-caprolactone). International

Journal of Pharmaceutics 135, 103109. Jackson, R.A., Royer, D.J., Waggoner, M.G., 2007. Polymer underwater pelletizer

apparatus and process incorporating same, US Patent 7226553.Khalifeh, A., Clermont, J.R., 2005. Numerical simulations of non-isothermal three-

dimensional flows in an extruder by a finite-volume method. Journal of Non-Newtonian Fluid Mechanics 126, 722.

Klein, C.E., Chiu, Y.L., Awni, W., Zhu, T., Heuser, R.S., Doan, T., Breitenbach, J.,Morris, J.B., Brun, S.C., Hanna, G.J., 2007. The tablet formulation of lopinavir/ritonavir provides similar bioavailability to the soft-gelatin capsule formula-

tion with less pharmacokinetic variability and diminished food effect.JaidsJournal of Acquired Immune Deficiency Syndromes 44, 401410.

Kostic, M.M., Reifschneider, L.G., 2007. Design of extrusion dies. In: Lee, S. (Ed.),Encyclopedia of Chemical Processing, New York, pp. 633649.

Kulikov, O.L., Hornung, K., 2001. A simple geometrical solution to the surfacefracturing problem in extrusion processes. Journal of Non-Newtonian FluidMechanics 98, 107115.

Lin, P., Jaluria, Y., 1998. Conjugate thermal transport in the channel of an extruderfor non-Newtonian fluids. International Journal of Heat and Mass Transfer 41,32393253.

Migler, K.B., Son, Y., Qiao, F., Flynn, K., 2002. Extensional deformation, cohesivefailure, and boundary conditions during sharkskin melt fracture. Journal ofRheology 46, 383400.

Miller, D.A., McConville, J.T., Yang, W., Williams, R.O., McGinity, J.W., 2007. Hot-melt extrusion for enhanced delivery of drug particles. Journal of Pharmaceu-tical Sciences 96, 361376.

Molenaar, J., Koopmans, R.J., den Doelder, C.F.J., 1998. Onset of the sharkskinphenomenon in polymer extrusion. Physical Review E 58, 46834691.

Pittman, J.F.T., Sander, R., 1994. Thermal effects in extrusion

slit dies. Interna-tional Polymer Processing 9, 326345.Rauwendaal, C., 2006. Polymer Extrusion. Hanser, Munich, Germany.Rein, R., 2005. Vorrichtung zur Herstellung gerundeter Pellets. 05002005.6,

European Patent EP1563897.Repka, M.A., Battu, S.K., Upadhye, S.B., Thumma, S., Crowley, M.M., Zhang, F.,

Martin, C., McGinity, J.W., 2007. Pharmaceutical applications of hot-meltextrusion: part II. Drug Development and Industrial Pharmacy 33, 10431057.

Rouilly, A., Jorda, J., Rigal, L., 2006. Thermo-mechanical processing of sugar beetpulp. II. Thermal and rheological properties of thermoplastic SBP. Carbohy-drate Polymers 66, 117125.

Schuschnigg, S., Duretek, I., Fertschej, A., 2007. Stoffdatenbestimmung undSimulation von einer Heiabschlagsduse (in German). Institut fur Kunststoff-verarbeitung, Montanuniversitat Leoben, Leoben, Austria.

Tome, M.F., Grossi, L., Castelo, A., Cuminato, J.A., Mckee, S., Walters, K., 2007. Die-swell, splashing drop and a numerical technique for solving the Oldroyd Bmodel for axisymmetric free surface flows. Journal of Non-Newtonian FluidMechanics 141, 148166.

Young, C.R., Koleng, J.J., McGinity, J.W., 2002. Production of spherical pellets by a

hot-melt extrusion and spheronization process. International Journal ofPharmaceutics 242, 8792.


Documents

700 Radl a Noveldesignforhot-meltextrusionpelletizers